* [PATCH v6 00/12] target/hexagon: introduce idef-parser
@ 2021-07-20 12:30 Alessandro Di Federico via
2021-07-20 12:30 ` [PATCH v6 01/12] target/hexagon: update MAINTAINERS for idef-parser Alessandro Di Federico via
` (11 more replies)
0 siblings, 12 replies; 27+ messages in thread
From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw)
To: qemu-devel
Cc: tsimpson, bcain, babush, nizzo, richard.henderson,
Alessandro Di Federico
From: Alessandro Di Federico <ale@rev.ng>
This patchset introduces the idef-parser for target/hexagon.
It's the sixth iteration of the patchset and includes fixes suggested in
previous iterations.
`idef-parser` is a build-time tool built using flex and bison. Its aim
is to generate a large part of the tiny code generator frontend for
Hexagon. The prototype of idef-parser has been presented at KVM Forum
2019 ("QEMU-Hexagon: Automatic Translation of the ISA Manual Pseudcode
to Tiny Code Instructions"):
https://www.youtube.com/watch?v=3EpnTYBOXCI
`target/hexagon/idef-parser/README.rst` provides an overview of the
parser and its inner working.
A couple of notes:
* Among other things, this commit ensures the signedness is always
properly propagated and that memory safety tools (valgrind and various
sanitizers) report no issues on the parser.
* We now also indent the generated C code using `indent`, if availble.
This has no run-time effect, it's just to ease debugging.
* These commits build successfully on the CI (including using clang),
with some exceptions.
The build on the CI fails for the cross-i386-* jobs. This seems to be
due to the fact that, despite we specified glib as a native
dependency, meson still chooses the i386 bit version of the library,
as opposed to the x86-64.
We also introduced a commit to make sure containers produced by the
developer Docker registry (as opposed to the official one) is
used. Otherwise, the changes we made to the containers would not
take effect.
However, as Daniel pointed out, our solution is an hack.
Since this is a problem only for developers when introducing new
dependencies for the containers, we might consider not merging the
last patch at all. But it's required in order to get some success in
testing on the CI it required.
Also, building debian-hexagon-cross still fails since it's not built
by the CI by default.
Help on these matters is appreciated.
Alessandro Di Federico (5):
target/hexagon: update MAINTAINERS for idef-parser
target/hexagon: import README for idef-parser
target/hexagon: prepare input for the idef-parser
target/hexagon: call idef-parser functions
gitlab-ci: do not use qemu-project Docker registry
Niccolò Izzo (2):
target/hexagon: introduce new helper functions
target/hexagon: import additional tests
Paolo Montesel (5):
target/hexagon: make slot number an unsigned
target/hexagon: make helper functions non-static
target/hexagon: expose next PC in DisasContext
target/hexagon: import lexer for idef-parser
target/hexagon: import parser for idef-parser
target/hexagon/idef-parser/README.rst | 447 +++
target/hexagon/genptr.h | 43 +
target/hexagon/idef-parser/idef-parser.h | 258 ++
target/hexagon/idef-parser/parser-helpers.h | 346 +++
target/hexagon/macros.h | 11 +-
target/hexagon/op_helper.h | 37 +
target/hexagon/translate.h | 1 +
target/hexagon/genptr.c | 176 +-
target/hexagon/idef-parser/parser-helpers.c | 2396 +++++++++++++++++
target/hexagon/op_helper.c | 29 +-
target/hexagon/translate.c | 3 +-
.gitlab-ci.d/container-cross.yml | 2 +-
.gitlab-ci.d/container-template.yml | 4 +-
MAINTAINERS | 8 +
target/hexagon/README | 5 +
target/hexagon/gen_helper_funcs.py | 5 +-
target/hexagon/gen_helper_protos.py | 5 +-
target/hexagon/gen_idef_parser_funcs.py | 114 +
target/hexagon/gen_tcg_funcs.py | 28 +-
target/hexagon/hex_common.py | 10 +
target/hexagon/idef-parser/idef-parser.lex | 642 +++++
target/hexagon/idef-parser/idef-parser.y | 975 +++++++
target/hexagon/idef-parser/macros.inc | 153 ++
target/hexagon/idef-parser/prepare | 24 +
target/hexagon/meson.build | 126 +-
tests/docker/dockerfiles/alpine.docker | 2 +
tests/docker/dockerfiles/centos8.docker | 2 +
tests/docker/dockerfiles/debian-amd64.docker | 2 +
tests/docker/dockerfiles/debian10.docker | 3 +
.../dockerfiles/fedora-i386-cross.docker | 3 +
.../dockerfiles/fedora-win32-cross.docker | 3 +
.../dockerfiles/fedora-win64-cross.docker | 3 +
tests/docker/dockerfiles/fedora.docker | 2 +
tests/docker/dockerfiles/opensuse-leap.docker | 2 +
tests/docker/dockerfiles/ubuntu.docker | 3 +
tests/docker/dockerfiles/ubuntu1804.docker | 2 +
tests/docker/dockerfiles/ubuntu2004.docker | 2 +
tests/tcg/hexagon/Makefile.target | 30 +-
tests/tcg/hexagon/crt.S | 26 +
tests/tcg/hexagon/test_abs.S | 20 +
tests/tcg/hexagon/test_bitcnt.S | 42 +
tests/tcg/hexagon/test_bitsplit.S | 25 +
tests/tcg/hexagon/test_call.S | 60 +
tests/tcg/hexagon/test_clobber.S | 35 +
tests/tcg/hexagon/test_cmp.S | 34 +
tests/tcg/hexagon/test_cmpy.S | 31 +
tests/tcg/hexagon/test_dotnew.S | 36 +
tests/tcg/hexagon/test_ext.S | 16 +
tests/tcg/hexagon/test_fibonacci.S | 33 +
tests/tcg/hexagon/test_hl.S | 19 +
tests/tcg/hexagon/test_hwloops.S | 22 +
tests/tcg/hexagon/test_jmp.S | 25 +
tests/tcg/hexagon/test_lsr.S | 39 +
tests/tcg/hexagon/test_mpyi.S | 20 +
tests/tcg/hexagon/test_overflow.S | 63 +
tests/tcg/hexagon/test_packet.S | 26 +
tests/tcg/hexagon/test_reorder.S | 29 +
tests/tcg/hexagon/test_round.S | 31 +
tests/tcg/hexagon/test_vavgw.S | 33 +
tests/tcg/hexagon/test_vcmpb.S | 32 +
tests/tcg/hexagon/test_vcmpw.S | 29 +
tests/tcg/hexagon/test_vcmpy.S | 50 +
tests/tcg/hexagon/test_vlsrw.S | 23 +
tests/tcg/hexagon/test_vmaxh.S | 37 +
tests/tcg/hexagon/test_vminh.S | 37 +
tests/tcg/hexagon/test_vpmpyh.S | 30 +
tests/tcg/hexagon/test_vspliceb.S | 33 +
67 files changed, 6753 insertions(+), 90 deletions(-)
create mode 100644 target/hexagon/idef-parser/README.rst
create mode 100644 target/hexagon/idef-parser/idef-parser.h
create mode 100644 target/hexagon/idef-parser/parser-helpers.h
create mode 100644 target/hexagon/op_helper.h
create mode 100644 target/hexagon/idef-parser/parser-helpers.c
create mode 100644 target/hexagon/gen_idef_parser_funcs.py
create mode 100644 target/hexagon/idef-parser/idef-parser.lex
create mode 100644 target/hexagon/idef-parser/idef-parser.y
create mode 100644 target/hexagon/idef-parser/macros.inc
create mode 100755 target/hexagon/idef-parser/prepare
create mode 100644 tests/tcg/hexagon/crt.S
create mode 100644 tests/tcg/hexagon/test_abs.S
create mode 100644 tests/tcg/hexagon/test_bitcnt.S
create mode 100644 tests/tcg/hexagon/test_bitsplit.S
create mode 100644 tests/tcg/hexagon/test_call.S
create mode 100644 tests/tcg/hexagon/test_clobber.S
create mode 100644 tests/tcg/hexagon/test_cmp.S
create mode 100644 tests/tcg/hexagon/test_cmpy.S
create mode 100644 tests/tcg/hexagon/test_dotnew.S
create mode 100644 tests/tcg/hexagon/test_ext.S
create mode 100644 tests/tcg/hexagon/test_fibonacci.S
create mode 100644 tests/tcg/hexagon/test_hl.S
create mode 100644 tests/tcg/hexagon/test_hwloops.S
create mode 100644 tests/tcg/hexagon/test_jmp.S
create mode 100644 tests/tcg/hexagon/test_lsr.S
create mode 100644 tests/tcg/hexagon/test_mpyi.S
create mode 100644 tests/tcg/hexagon/test_overflow.S
create mode 100644 tests/tcg/hexagon/test_packet.S
create mode 100644 tests/tcg/hexagon/test_reorder.S
create mode 100644 tests/tcg/hexagon/test_round.S
create mode 100644 tests/tcg/hexagon/test_vavgw.S
create mode 100644 tests/tcg/hexagon/test_vcmpb.S
create mode 100644 tests/tcg/hexagon/test_vcmpw.S
create mode 100644 tests/tcg/hexagon/test_vcmpy.S
create mode 100644 tests/tcg/hexagon/test_vlsrw.S
create mode 100644 tests/tcg/hexagon/test_vmaxh.S
create mode 100644 tests/tcg/hexagon/test_vminh.S
create mode 100644 tests/tcg/hexagon/test_vpmpyh.S
create mode 100644 tests/tcg/hexagon/test_vspliceb.S
--
2.32.0
^ permalink raw reply [flat|nested] 27+ messages in thread* [PATCH v6 01/12] target/hexagon: update MAINTAINERS for idef-parser 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via @ 2021-07-20 12:30 ` Alessandro Di Federico via 2021-09-07 18:09 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 02/12] target/hexagon: import README " Alessandro Di Federico via ` (10 subsequent siblings) 11 siblings, 1 reply; 27+ messages in thread From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw) To: qemu-devel Cc: tsimpson, bcain, babush, nizzo, richard.henderson, Alessandro Di Federico From: Alessandro Di Federico <ale@rev.ng> Signed-off-by: Alessandro Di Federico <ale@rev.ng> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> --- MAINTAINERS | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/MAINTAINERS b/MAINTAINERS index 4256ad1adb..6c825a9eb4 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -194,6 +194,8 @@ Hexagon TCG CPUs M: Taylor Simpson <tsimpson@quicinc.com> S: Supported F: target/hexagon/ +X: target/hexagon/idef-parser/ +X: target/hexagon/gen_idef_parser_funcs.py F: linux-user/hexagon/ F: tests/tcg/hexagon/ F: disas/hexagon.c @@ -201,6 +203,12 @@ F: configs/targets/hexagon-linux-user/default.mak F: docker/dockerfiles/debian-hexagon-cross.docker F: docker/dockerfiles/debian-hexagon-cross.docker.d/build-toolchain.sh +Hexagon idef-parser +M: Alessandro Di Federico <ale@rev.ng> +S: Supported +F: target/hexagon/idef-parser/ +F: target/hexagon/gen_idef_parser_funcs.py + HPPA (PA-RISC) TCG CPUs M: Richard Henderson <richard.henderson@linaro.org> S: Maintained -- 2.32.0 ^ permalink raw reply related [flat|nested] 27+ messages in thread
* RE: [PATCH v6 01/12] target/hexagon: update MAINTAINERS for idef-parser 2021-07-20 12:30 ` [PATCH v6 01/12] target/hexagon: update MAINTAINERS for idef-parser Alessandro Di Federico via @ 2021-09-07 18:09 ` Taylor Simpson 0 siblings, 0 replies; 27+ messages in thread From: Taylor Simpson @ 2021-09-07 18:09 UTC (permalink / raw) To: Alessandro Di Federico, qemu-devel@nongnu.org Cc: babush@rev.ng, Brian Cain, richard.henderson@linaro.org, nizzo@rev.ng, Alessandro Di Federico > -----Original Message----- > From: Alessandro Di Federico <ale.qemu@rev.ng> > Sent: Tuesday, July 20, 2021 7:30 AM > To: qemu-devel@nongnu.org > Cc: Taylor Simpson <tsimpson@quicinc.com>; Brian Cain > <bcain@quicinc.com>; babush@rev.ng; nizzo@rev.ng; > richard.henderson@linaro.org; Alessandro Di Federico <ale@rev.ng> > Subject: [PATCH v6 01/12] target/hexagon: update MAINTAINERS for idef- > parser > > From: Alessandro Di Federico <ale@rev.ng> > > Signed-off-by: Alessandro Di Federico <ale@rev.ng> > Reviewed-by: Richard Henderson <richard.henderson@linaro.org> > --- > MAINTAINERS | 8 ++++++++ > 1 file changed, 8 insertions(+) Reviewed-by: Taylor Simpson <tsimpson@quicinc.com> ^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 02/12] target/hexagon: import README for idef-parser 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 01/12] target/hexagon: update MAINTAINERS for idef-parser Alessandro Di Federico via @ 2021-07-20 12:30 ` Alessandro Di Federico via 2021-09-07 18:09 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 03/12] target/hexagon: make slot number an unsigned Alessandro Di Federico via ` (9 subsequent siblings) 11 siblings, 1 reply; 27+ messages in thread From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw) To: qemu-devel Cc: tsimpson, bcain, babush, nizzo, richard.henderson, Alessandro Di Federico From: Alessandro Di Federico <ale@rev.ng> Signed-off-by: Alessandro Di Federico <ale@rev.ng> --- target/hexagon/idef-parser/README.rst | 447 ++++++++++++++++++++++++++ target/hexagon/README | 5 + 2 files changed, 452 insertions(+) create mode 100644 target/hexagon/idef-parser/README.rst diff --git a/target/hexagon/idef-parser/README.rst b/target/hexagon/idef-parser/README.rst new file mode 100644 index 0000000000..9f20bce36a --- /dev/null +++ b/target/hexagon/idef-parser/README.rst @@ -0,0 +1,447 @@ +Hexagon ISA instruction definitions to tinycode generator compiler +------------------------------------------------------------------ + +idef-parser is a small compiler able to translate the Hexagon ISA description +language into tinycode generator code, that can be easily integrated into QEMU. + +Compilation Example +------------------- + +To better understand the scope of the idef-parser, we'll explore an applicative +example. Let's start by one of the simplest Hexagon instruction: the ``add``. + +The ISA description language represents the ``add`` instruction as +follows: + +.. code:: c + + A2_add(RdV, in RsV, in RtV) { + { RdV=RsV+RtV;} + } + +idef-parser will compile the above code into the following code: + +.. code:: c + + /* A2_add */ + void emit_A2_add(DisasContext *ctx, Insn *insn, Packet *pkt, TCGv_i32 RdV, + TCGv_i32 RsV, TCGv_i32 RtV) + /* { RdV=RsV+RtV;} */ + { + tcg_gen_movi_i32(RdV, 0); + TCGv_i32 tmp_0 = tcg_temp_new_i32(); + tcg_gen_add_i32(tmp_0, RsV, RtV); + tcg_gen_mov_i32(RdV, tmp_0); + tcg_temp_free_i32(tmp_0); + } + +The output of the compilation process will be a function, containing the +tinycode generator code, implementing the correct semantics. That function will +not access any global variable, because all the accessed data structures will be +passed explicitly as function parameters. Among the passed parameters we will +have TCGv (tinycode variables) representing the input and output registers of +the architecture, integers representing the immediates that come from the code, +and other data structures which hold information about the disassemblation +context (``DisasContext`` struct). + +Let's begin by describing the input code. The ``add`` instruction is associated +with a unique identifier, in this case ``A2_add``, which allows to distinguish +variants of the same instruction, and expresses the class to which the +instruction belongs, in this case ``A2`` corresponds to the Hexagon +``ALU32/ALU`` instruction subclass. + +After the instruction identifier, we have a series of parameters that represents +TCG variables that will be passed to the generated function. Parameters marked +with ``in`` are already initialized, while the others are output parameters. + +We will leverage this information to infer several information: + +- Fill in the output function signature with the correct TCGv registers +- Fill in the output function signature with the immediate integers +- Keep track of which registers, among the declared one, have been + initialized + +Let's now observe the actual instruction description code, in this case: + +.. code:: c + + { RdV=RsV+RtV;} + +This code is composed by a subset of the C syntax, and is the result of the +application of some macro definitions contained in the ``macros.h`` file. + +This file is used to reduce the complexity of the input language where complex +variants of similar constructs can be mapped to a unique primitive, so that the +idef-parser has to handle a lower number of computation primitives. + +As you may notice, the description code modifies the registers which have been +declared by the declaration statements. In this case all the three registers +will be declared, ``RsV`` and ``RtV`` will also be read and ``RdV`` will be +written. + +Now let's have a quick look at the generated code, line by line. + +:: + + tcg_gen_movi_i32(RdV, 0); + +This code starts by zero-initializing ``RdV``, since reading from that register +without initialization will cause a segmentation fault by QEMU. This is emitted +since a declaration of the ``RdV`` register was parsed, but we got no indication +that the variable has been initialized by the caller. + +:: + + TCGv_i32 tmp_0 = tcg_temp_new_i32(); + +Then we are declaring a temporary TCGv to hold the result from the sum +operation. + +:: + + tcg_gen_add_i32(tmp_0, RsV, RtV); + +Now we are actually generating the sum tinycode operator between the selected +registers, storing the result in the just declared temporary. + +:: + + tcg_gen_mov_i32(RdV, tmp_0); + +The result of the addition is now stored in the temporary, we move it into the +correct destination register. This might not seem an efficient code, but QEMU +will perform some tinycode optimization, reducing the unnecessary copy. + +:: + + tcg_temp_free_i32(tmp_0); + +Finally, we free the temporary we used to hold the addition result. + +Parser Structure +---------------- + +The idef-parser is built using the ``flex`` and ``bison``. + +``flex`` is used to split the input string into tokens, each described using a +regular expression. The token description is contained in the +``idef-parser.lex`` source file. The flex-generated scanner takes care also to +extract from the input text other meaningful information, e.g., the numerical +value in case of an immediate constant, and decorates the token with the +extracted information. + +``bison`` is used to generate the actual parser, starting from the parsing +description contained in the ``idef-parser.y`` file. The generated parser +executes the ``main`` function at the end of the ``idef-parser.y`` file, which +opens input and output files, creates the parsing context, and eventually calls +the ``yyparse()`` function, which starts the execution of the LALR(1) parser +(see `Wikipedia <https://en.wikipedia.org/wiki/LALR_parser>`__ for more +information about LALR parsing techniques). The LALR(1) parser, whenever it has +to shift a token, calls the ``yylex()`` function, which is defined by the +flex-generated code, and reads the input file returning the next scanned token. + +The tokens are mapped on the source language grammar, defined in the +``idef-parser.y`` file to build a unique syntactic tree, according to the +specified operator precedences and associativity rules. + +The grammar describes the whole file which contains the Hexagon instruction +descriptions, therefore it starts from the ``input`` nonterminal, which is a +list of instructions, each instruction is represented by the following grammar +rule, representing the structure of the input file shown above: + +:: + + instruction : INAME code + + code : LBR decls statements decls RBR + + statements : statements statement + | statement + + statement : control_statement + | rvalue SEMI + | code_block + | SEMI + + code_block : LBR statements RBR + | LBR RBR + +With this initial portion of the grammar we are defining the instruction +statements, which are enclosed by the declarations. Each statement can be a +``control_statement``, a code block, which is just a bracket-enclosed list of +statements, a ``SEMI``, which is a ``nop`` instruction, and an ``rvalue SEMI``. + +Expressions +~~~~~~~~~~~ + +``rvalue`` is the nonterminal representing expressions, which are everything +that could be assigned to a variable. ``rvalue SEMI`` can be a statement on its +own because the assign statement, just as in the C language, is itself an +expression. + +``rvalue``\ s can be registers, immediates, predicates, control registers, +variables, or any combination of other ``rvalue``\ s through operators. An +``rvalue`` can be either an immediate or a TCGv, the actual type is determined +by the ``t_hex_value.type`` field. In case it is an immediate, its combination +with other immediates can be performed at compile-time (constant folding), only +the result will be written into the code. If the ``rvalue`` instead is a TCGv, +the operations performed on it will have to be emitted as tinycode instructions, +therefore their result will be known only at runtime. An immediate can be copied +into a TCGv through the ``rvalue_materialize`` function, which allocates a +temporary TCGv and copies the value of the immediate in it. Each temporary +should be freed after that it is no more used, we usually free both operands of +each operator, in an SSA fashion. + +``lvalue``\ s instead represents all the variables which can be assigned to a +value, and are specialized into registers and variables: + +:: + + lvalue : REG + | VAR + +The effective assignment of ``lvalue``\ s is handled by the ``gen_assign()`` +function. + +Automatic Variables +~~~~~~~~~~~~~~~~~~~ + +The input code can contain implicitly declared automatic variables, which are +initialized with a value and then used. We performed a dedicated handling of +such variables, because they will be matched by a generic ``VARID`` token, which +will feature the variable name as a decoration. Each time that the variable is +found, we have to check if that's the first variable use, in that case we +declare a new automatic variable in the tinycode, which can be considered at all +effects as an immediate. Special care is taken to make sure that each variable +is declared only the first time it is seen. Furthermore the variable might +inherit some characteristics like the signedness and the bit width, which must +be propagated from the initialization of the variable to all the further uses of +the variable. + +The combination of ``rvalue``\ s are handled through the use of the +``gen_bin_op`` and ``gen_bin_cmp`` helper functions. These two functions handle +the appropriate compile-time or run-time emission of operations to perform the +required computation. + +Type System +~~~~~~~~~~~ + +idef-parser features a simple type system which is used to correctly implement +the signedness and bit width of the operations. + +The type of each ``rvalue`` is determined by two attributes: its bit width +(``unsigned bit_width``) and its signedness (``HexSignedness signedness``). + +For each operation, the type of ``rvalue``\ s influence the way in which the +operands are handled and emitted. For example a right shift between signed +operators will be an algebraic shift, while one between unsigned operators will +be a logical shift. If one of the two operands is signed, and the other is +unsigned, the operation will be signed. + +The bit width also influences the outcome of the operations, in particular while +the input languages features a fine granularity type system, with types of 8, +16, 32, 64 (and more for vectorial instructions) bits, the tinycode only +features 32 and 64 bit widths. We propagate as much as possible the fine +granularity type, until the value has to be used inside an operation between +``rvalue``\ s; in that case if one of the two operands is greater than 32 bits +we promote the whole operation to 64 bit, taking care of properly extending the +two operands. Fortunately, the most critical instructions already feature +explicit casts and zero/sign extensions which are properly propagated down to +our parser. + +Control Statements +~~~~~~~~~~~~~~~~~~ + +``control_statement``\ s are all the statements which modify the order of +execution of the generated code according to input parameters. They are expanded +by the following grammar rule: + +:: + + control_statement : frame_check + | cancel_statement + | if_statement + | for_statement + | fpart1_statement + +``if_statement``\ s require the emission of labels and branch instructions which +effectively perform conditional jumps (``tcg_gen_brcondi``) according to the +value of an expression. All the predicated instructions, and in general all the +instructions where there could be alternative values assigned to an ``lvalue``, +like C-style ternary expressions: + +:: + + rvalue : rvalue QMARK rvalue COLON rvalue + +Are handled using the conditional move tinycode instruction +(``tcg_gen_movcond``), which avoids the additional complexity of managing labels +and jumps. + +Instead, regarding the ``for`` loops, exploiting the fact that they always +iterate on immediate values, therefore their iteration ranges are always known +at compile time, we implemented those emitting plain C ``for`` loops. This is +possible because the loops will be executed in the QEMU code, leading to the +consequential unrolling of the for loop, since the tinycode generator +instructions will be executed multiple times, and the respective generated +tinycode will represent the unrolled execution of the loop. + +Parsing Context +~~~~~~~~~~~~~~~ + +All the helper functions in ``idef-parser.y`` carry two fixed parameters, which +are the parsing context ``c`` and the ``YYLLOC`` location information. The +context is explicitly passed to all the functions because the parser we generate +is a reentrant one, meaning that it does not have any global variable, and +therefore the instruction compilation could easily be parallelized in the +future. Finally for each rule we propagate information about the location of the +involved tokens to generate a pretty error reporting, able to highlight the +portion of the input code which generated each error. + +Debugging +--------- + +Developing the idef-parser can lead to two types of errors: compile-time errors +and parsing errors. + +Compile-time errors in Bison-generated parsers are usually due to conflicts in +the described grammar. Conflicts forbid the grammar to produce a unique +derivation tree, thus must be solved (except for the dangling else problem, +which is marked as expected through the ``%expect 1`` Bison option). + +For solving conflicts you need a basic understanding of `shift-reduce conflicts +<https://www.gnu.org/software/Bison/manual/html_node/Shift_002fReduce.html>`__ +and `reduce-reduce conflicts +<https://www.gnu.org/software/Bison/manual/html_node/Reduce_002fReduce.html>`__, +then, if you are using a Bison version > 3.7.1 you can ask Bison to generate +some counterexamples which highlight ambiguous derivations, passing the +``-Wcex`` option to Bison. In general shift/reduce conflicts are solved by +redesigning the grammar in an unambiguous way or by setting the token priority +correctly, while reduce/reduce conflicts are solved by redesigning the +interested part of the grammar. + +Run-time errors can be divided between lexing and parsing errors, lexing errors +are hard to detect, since the ``VAR`` token will catch everything which is not +catched by other tokens, but easy to fix, because most of the time a simple +regex editing will be enough. + +idef-parser features a fancy parsing error reporting scheme, which for each +parsing error reports the fragment of the input text which was involved in the +parsing rule that generated an error. + +Implementing an instruction goes through several sequential steps, here are some +suggestions to make each instruction proceed to the next step. + +- not-emitted + + Means that the parsing of the input code relative to that instruction failed, + this could be due to a lexical error or to some mismatch between the order of + valid tokens and a parser rule. You should check that tokens are correctly + identified and mapped, and that there is a rule matching the token sequence + that you need to parse. + +- emitted + + This instruction class contains all the instruction which are emitted but + fail to compile when included in QEMU. The compilation errors are shown by + the QEMU building process and will lead to fixing the bug. Most common + errors regard the mismatch of parameters for tinycode generator functions, + which boil down to errors in the idef-parser type system. + +- compiled + + These instruction generate valid tinycode generator code, which however fail + the QEMU or the harness tests, these cases must be handled manually by + looking into the failing tests and looking at the generated tinycode + generator instruction and at the generated tinycode itself. Tip: handle the + failing harness tests first, because they usually feature only a single + instruction, thus will require less execution trace navigation. If a + multi-threaded test fail, fixing all the other tests will be the easier + option, hoping that the multi-threaded one will be indirectly fixed. + +- tests-passed + + This is the final goal for each instruction, meaning that the instruction + passes the test suite. + +Another approach to fix QEMU system test, where many instructions might fail, is +to compare the execution trace of your implementation with the reference +implementations already present in QEMU. To do so you should obtain a QEMU build +where the instruction pass the test, and run it with the following command: + +:: + + sudo unshare -p sudo -u <USER> bash -c \ + 'env -i <qemu-hexagon full path> -d cpu <TEST>' + +And do the same for your implementation, the generated execution traces will be +inherently aligned and can be inspected for behavioral differences using the +``diff`` tool. + +Limitations and Future Development +---------------------------------- + +The main limitation of the current parser is given by the syntax-driven nature +of the Bison-generated parsers. This has the severe implication of only being +able to generate code in the order of evaluation of the various rules, without, +in any case, being able to backtrack and alter the generated code. + +An example limitation is highlighted by this statement of the input language: + +:: + + { (PsV==0xff) ? (PdV=0xff) : (PdV=0x00); } + +This ternary assignment, when written in this form requires us to emit some +proper control flow statements, which emit a jump to the first or to the second +code block, whose implementation is extremely convoluted, because when matching +the ternary assignment, the code evaluating the two assignments will be already +generated. + +Instead we pre-process that statement, making it become: + +:: + + { PdV = ((PsV==0xff)) ? 0xff : 0x00; } + +Which can be easily matched by the following parser rules: + +:: + + statement | rvalue SEMI + + rvalue : rvalue QMARK rvalue COLON rvalue + | rvalue EQ rvalue + | LPAR rvalue RPAR + | assign_statement + | IMM + + assign_statement : pre ASSIGN rvalue + +Another example that highlight the limitation of the flex/bison parser can be +found even in the add operation we already saw: + +:: + + TCGv_i32 tmp_0 = tcg_temp_new_i32(); + tcg_gen_add_i32(tmp_0, RsV, RtV); + tcg_gen_mov_i32(RdV, tmp_0); + +The fact that we cannot directly use ``RdV`` as the destination of the sum is a +consequence of the syntax-driven nature of the parser. In fact when we parse the +assignment, the ``rvalue`` token, representing the sum has already been reduced, +and thus its code emitted and unchangeable. We rely on the fact that QEMU will +optimize our code reducing the useless move operations and the relative +temporaries. + +A possible improvement of the parser regards the support for vectorial +instructions and floating point instructions, which will require the extension +of the scanner, the parser, and a partial re-design of the type system, allowing +to build the vectorial semantics over the available vectorial tinycode generator +primitives. + +A more radical improvement will use the parser, not to generate directly the +tinycode generator code, but to generate an intermediate representation like the +LLVM IR, which in turn could be compiled using the clang TCG backend. That code +could be furtherly optimized, overcoming the limitations of the syntax-driven +parsing and could lead to a more optimized generated code. diff --git a/target/hexagon/README b/target/hexagon/README index b0b2435070..2f2814380c 100644 --- a/target/hexagon/README +++ b/target/hexagon/README @@ -23,6 +23,10 @@ Hexagon-specific code are encode*.def Encoding patterns for each instruction iclass.def Instruction class definitions used to determine legal VLIW slots for each instruction + qemu/target/hexagon/idef-parser + Parser that, given the high-level definitions of an instruction, + produces a C function generating equivalent tiny code instructions. + See README.rst. qemu/linux-user/hexagon Helpers for loading the ELF file and making Linux system calls, signals, etc @@ -43,6 +47,7 @@ header files in <BUILD_DIR>/target/hexagon gen_tcg_funcs.py -> tcg_funcs_generated.c.inc gen_tcg_func_table.py -> tcg_func_table_generated.c.inc gen_helper_funcs.py -> helper_funcs_generated.c.inc + gen_idef_parser_funcs.py -> idef_parser_input.h Qemu helper functions have 3 parts DEF_HELPER declaration indicates the signature of the helper -- 2.32.0 ^ permalink raw reply related [flat|nested] 27+ messages in thread
* RE: [PATCH v6 02/12] target/hexagon: import README for idef-parser 2021-07-20 12:30 ` [PATCH v6 02/12] target/hexagon: import README " Alessandro Di Federico via @ 2021-09-07 18:09 ` Taylor Simpson 0 siblings, 0 replies; 27+ messages in thread From: Taylor Simpson @ 2021-09-07 18:09 UTC (permalink / raw) To: Alessandro Di Federico, qemu-devel@nongnu.org Cc: babush@rev.ng, Brian Cain, richard.henderson@linaro.org, nizzo@rev.ng, Alessandro Di Federico > -----Original Message----- > From: Alessandro Di Federico <ale.qemu@rev.ng> > Sent: Tuesday, July 20, 2021 7:30 AM > To: qemu-devel@nongnu.org > Cc: Taylor Simpson <tsimpson@quicinc.com>; Brian Cain > <bcain@quicinc.com>; babush@rev.ng; nizzo@rev.ng; > richard.henderson@linaro.org; Alessandro Di Federico <ale@rev.ng> > Subject: [PATCH v6 02/12] target/hexagon: import README for idef-parser > > From: Alessandro Di Federico <ale@rev.ng> > > Signed-off-by: Alessandro Di Federico <ale@rev.ng> > --- > target/hexagon/idef-parser/README.rst | 447 > +Now let's have a quick look at the generated code, line by line. > + > +:: > + > + tcg_gen_movi_i32(RdV, 0); > + > +This code starts by zero-initializing ``RdV``, since reading from that > +register without initialization will cause a segmentation fault by > +QEMU. This is emitted since a declaration of the ``RdV`` register was > +parsed, but we got no indication that the variable has been initialized by the > caller. You should only need this for predicated instructions. For non-predicated instructions, RdV will always be written. > +Parser Structure > +---------------- Before you describe the parser, you should describe what gen_idef_parser_funcs.py does and the contents of idef_parser_input.h.inc and idef_parser_input.preprocessed.h.inc. > +Expressions > +~~~~~~~~~~~ Rather than describing the grammar in detail here, it would be better to state that we are parsing C language statements and describe what things need to be handled specially (e.g., macros, variables like RdV, etc). > +Debugging > +--------- In this section, you should describe how to narrow down to a specific instruction tag (e.g., A2_add) that is emitting incorrect TCG code when a test failure is identified. In other words, how do you switch back to the C helper function instead of the idef-parser generated version, and how can you binary search to the one emitting incorrect code? ^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 03/12] target/hexagon: make slot number an unsigned 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 01/12] target/hexagon: update MAINTAINERS for idef-parser Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 02/12] target/hexagon: import README " Alessandro Di Federico via @ 2021-07-20 12:30 ` Alessandro Di Federico via 2021-09-07 18:09 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 04/12] target/hexagon: make helper functions non-static Alessandro Di Federico via ` (8 subsequent siblings) 11 siblings, 1 reply; 27+ messages in thread From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw) To: qemu-devel Cc: tsimpson, bcain, babush, nizzo, richard.henderson, Alessandro Di Federico From: Paolo Montesel <babush@rev.ng> Signed-off-by: Alessandro Di Federico <ale@rev.ng> Signed-off-by: Paolo Montesel <babush@rev.ng> Acked-by: Richard Henderson <richard.henderson@linaro.org> --- target/hexagon/macros.h | 2 +- target/hexagon/genptr.c | 24 +++++++++++++----------- 2 files changed, 14 insertions(+), 12 deletions(-) diff --git a/target/hexagon/macros.h b/target/hexagon/macros.h index 094b8dabb5..cd4f878fcf 100644 --- a/target/hexagon/macros.h +++ b/target/hexagon/macros.h @@ -185,7 +185,7 @@ #define LOAD_CANCEL(EA) do { CANCEL; } while (0) #ifdef QEMU_GENERATE -static inline void gen_pred_cancel(TCGv pred, int slot_num) +static inline void gen_pred_cancel(TCGv pred, uint32_t slot_num) { TCGv slot_mask = tcg_const_tl(1 << slot_num); TCGv tmp = tcg_temp_new(); diff --git a/target/hexagon/genptr.c b/target/hexagon/genptr.c index 7333299615..2c7f4136b5 100644 --- a/target/hexagon/genptr.c +++ b/target/hexagon/genptr.c @@ -27,7 +27,8 @@ #undef QEMU_GENERATE #include "gen_tcg.h" -static inline void gen_log_predicated_reg_write(int rnum, TCGv val, int slot) +static inline void gen_log_predicated_reg_write(int rnum, TCGv val, + uint32_t slot) { TCGv zero = tcg_const_tl(0); TCGv slot_mask = tcg_temp_new(); @@ -60,7 +61,8 @@ static inline void gen_log_reg_write(int rnum, TCGv val) } } -static void gen_log_predicated_reg_write_pair(int rnum, TCGv_i64 val, int slot) +static void gen_log_predicated_reg_write_pair(int rnum, TCGv_i64 val, + uint32_t slot) { TCGv val32 = tcg_temp_new(); TCGv zero = tcg_const_tl(0); @@ -379,7 +381,7 @@ static inline void gen_store_conditional8(DisasContext *ctx, tcg_gen_movi_tl(hex_llsc_addr, ~0); } -static inline void gen_store32(TCGv vaddr, TCGv src, int width, int slot) +static inline void gen_store32(TCGv vaddr, TCGv src, int width, uint32_t slot) { tcg_gen_mov_tl(hex_store_addr[slot], vaddr); tcg_gen_movi_tl(hex_store_width[slot], width); @@ -387,14 +389,14 @@ static inline void gen_store32(TCGv vaddr, TCGv src, int width, int slot) } static inline void gen_store1(TCGv_env cpu_env, TCGv vaddr, TCGv src, - DisasContext *ctx, int slot) + DisasContext *ctx, uint32_t slot) { gen_store32(vaddr, src, 1, slot); ctx->store_width[slot] = 1; } static inline void gen_store1i(TCGv_env cpu_env, TCGv vaddr, int32_t src, - DisasContext *ctx, int slot) + DisasContext *ctx, uint32_t slot) { TCGv tmp = tcg_const_tl(src); gen_store1(cpu_env, vaddr, tmp, ctx, slot); @@ -402,14 +404,14 @@ static inline void gen_store1i(TCGv_env cpu_env, TCGv vaddr, int32_t src, } static inline void gen_store2(TCGv_env cpu_env, TCGv vaddr, TCGv src, - DisasContext *ctx, int slot) + DisasContext *ctx, uint32_t slot) { gen_store32(vaddr, src, 2, slot); ctx->store_width[slot] = 2; } static inline void gen_store2i(TCGv_env cpu_env, TCGv vaddr, int32_t src, - DisasContext *ctx, int slot) + DisasContext *ctx, uint32_t slot) { TCGv tmp = tcg_const_tl(src); gen_store2(cpu_env, vaddr, tmp, ctx, slot); @@ -417,14 +419,14 @@ static inline void gen_store2i(TCGv_env cpu_env, TCGv vaddr, int32_t src, } static inline void gen_store4(TCGv_env cpu_env, TCGv vaddr, TCGv src, - DisasContext *ctx, int slot) + DisasContext *ctx, uint32_t slot) { gen_store32(vaddr, src, 4, slot); ctx->store_width[slot] = 4; } static inline void gen_store4i(TCGv_env cpu_env, TCGv vaddr, int32_t src, - DisasContext *ctx, int slot) + DisasContext *ctx, uint32_t slot) { TCGv tmp = tcg_const_tl(src); gen_store4(cpu_env, vaddr, tmp, ctx, slot); @@ -432,7 +434,7 @@ static inline void gen_store4i(TCGv_env cpu_env, TCGv vaddr, int32_t src, } static inline void gen_store8(TCGv_env cpu_env, TCGv vaddr, TCGv_i64 src, - DisasContext *ctx, int slot) + DisasContext *ctx, uint32_t slot) { tcg_gen_mov_tl(hex_store_addr[slot], vaddr); tcg_gen_movi_tl(hex_store_width[slot], 8); @@ -441,7 +443,7 @@ static inline void gen_store8(TCGv_env cpu_env, TCGv vaddr, TCGv_i64 src, } static inline void gen_store8i(TCGv_env cpu_env, TCGv vaddr, int64_t src, - DisasContext *ctx, int slot) + DisasContext *ctx, uint32_t slot) { TCGv_i64 tmp = tcg_const_i64(src); gen_store8(cpu_env, vaddr, tmp, ctx, slot); -- 2.32.0 ^ permalink raw reply related [flat|nested] 27+ messages in thread
* RE: [PATCH v6 03/12] target/hexagon: make slot number an unsigned 2021-07-20 12:30 ` [PATCH v6 03/12] target/hexagon: make slot number an unsigned Alessandro Di Federico via @ 2021-09-07 18:09 ` Taylor Simpson 0 siblings, 0 replies; 27+ messages in thread From: Taylor Simpson @ 2021-09-07 18:09 UTC (permalink / raw) To: Alessandro Di Federico, qemu-devel@nongnu.org Cc: babush@rev.ng, Brian Cain, richard.henderson@linaro.org, nizzo@rev.ng, Alessandro Di Federico > -----Original Message----- > From: Alessandro Di Federico <ale.qemu@rev.ng> > Sent: Tuesday, July 20, 2021 7:30 AM > To: qemu-devel@nongnu.org > Cc: Taylor Simpson <tsimpson@quicinc.com>; Brian Cain > <bcain@quicinc.com>; babush@rev.ng; nizzo@rev.ng; > richard.henderson@linaro.org; Alessandro Di Federico <ale@rev.ng> > Subject: [PATCH v6 03/12] target/hexagon: make slot number an unsigned > > From: Paolo Montesel <babush@rev.ng> > > Signed-off-by: Alessandro Di Federico <ale@rev.ng> > Signed-off-by: Paolo Montesel <babush@rev.ng> > Acked-by: Richard Henderson <richard.henderson@linaro.org> Could you describe why this is needed? I presume it has something to do with the type system in the idef-parser? ^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 04/12] target/hexagon: make helper functions non-static 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via ` (2 preceding siblings ...) 2021-07-20 12:30 ` [PATCH v6 03/12] target/hexagon: make slot number an unsigned Alessandro Di Federico via @ 2021-07-20 12:30 ` Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 05/12] target/hexagon: introduce new helper functions Alessandro Di Federico via ` (7 subsequent siblings) 11 siblings, 0 replies; 27+ messages in thread From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw) To: qemu-devel Cc: tsimpson, bcain, babush, nizzo, richard.henderson, Alessandro Di Federico From: Paolo Montesel <babush@rev.ng> Make certain helper functions non-static, making them available outside genptr.c. These functions are required by code generated by the idef-parser. This commit also makes some op_helper.c non-static in order to avoid having them marked as unused when using the idef-parser generated code. Signed-off-by: Alessandro Di Federico <ale@rev.ng> Signed-off-by: Paolo Montesel <babush@rev.ng> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Taylor Simpson <tsimpson@quicinc.com> --- target/hexagon/genptr.h | 30 +++++++++++++++++++ target/hexagon/op_helper.h | 37 ++++++++++++++++++++++++ target/hexagon/genptr.c | 59 +++++++++++++++++++++----------------- target/hexagon/op_helper.c | 29 +++++++++---------- 4 files changed, 113 insertions(+), 42 deletions(-) create mode 100644 target/hexagon/op_helper.h diff --git a/target/hexagon/genptr.h b/target/hexagon/genptr.h index c158005d2a..d71dd7e1ce 100644 --- a/target/hexagon/genptr.h +++ b/target/hexagon/genptr.h @@ -19,7 +19,37 @@ #define HEXAGON_GENPTR_H #include "insn.h" +#include "tcg/tcg.h" +#include "translate.h" extern const SemanticInsn opcode_genptr[]; +void gen_store32(TCGv vaddr, TCGv src, int width, uint32_t slot); +void gen_store1(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx, + uint32_t slot); +void gen_store2(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx, + uint32_t slot); +void gen_store4(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx, + uint32_t slot); +void gen_store8(TCGv_env cpu_env, TCGv vaddr, TCGv_i64 src, DisasContext *ctx, + uint32_t slot); +void gen_store1i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx, + uint32_t slot); +void gen_store2i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx, + uint32_t slot); +void gen_store4i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx, + uint32_t slot); +void gen_store8i(TCGv_env cpu_env, TCGv vaddr, int64_t src, DisasContext *ctx, + uint32_t slot); +TCGv gen_read_preg(TCGv pred, uint8_t num); +void gen_log_reg_write(int rnum, TCGv val); +void gen_log_pred_write(DisasContext *ctx, int pnum, TCGv val); +TCGv gen_8bitsof(TCGv result, TCGv value); +void gen_set_byte_i64(int N, TCGv_i64 result, TCGv src); +TCGv gen_get_byte(TCGv result, int N, TCGv src, bool sign); +TCGv gen_get_byte_i64(TCGv result, int N, TCGv_i64 src, bool sign); +TCGv gen_get_half(TCGv result, int N, TCGv src, bool sign); +void gen_set_half(int N, TCGv result, TCGv src); +void gen_set_half_i64(int N, TCGv_i64 result, TCGv src); + #endif diff --git a/target/hexagon/op_helper.h b/target/hexagon/op_helper.h new file mode 100644 index 0000000000..2d39637b8f --- /dev/null +++ b/target/hexagon/op_helper.h @@ -0,0 +1,37 @@ +/* + * Copyright(c) 2019-2021 Qualcomm Innovation Center, Inc. All Rights Reserved. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <http://www.gnu.org/licenses/>. + */ + +#ifndef HEXAGON_OP_HELPER_H +#define HEXAGON_OP_HELPER_H + +/* Misc functions */ +void cancel_slot(CPUHexagonState *env, uint32_t slot); +void write_new_pc(CPUHexagonState *env, target_ulong addr); + +uint8_t mem_load1(CPUHexagonState *env, uint32_t slot, target_ulong vaddr); +uint16_t mem_load2(CPUHexagonState *env, uint32_t slot, target_ulong vaddr); +uint32_t mem_load4(CPUHexagonState *env, uint32_t slot, target_ulong vaddr); +uint64_t mem_load8(CPUHexagonState *env, uint32_t slot, target_ulong vaddr); + +void log_reg_write(CPUHexagonState *env, int rnum, target_ulong val, + uint32_t slot); +void log_store64(CPUHexagonState *env, target_ulong addr, int64_t val, + int width, int slot); +void log_store32(CPUHexagonState *env, target_ulong addr, target_ulong val, + int width, int slot); + +#endif diff --git a/target/hexagon/genptr.c b/target/hexagon/genptr.c index 2c7f4136b5..5e7f446657 100644 --- a/target/hexagon/genptr.c +++ b/target/hexagon/genptr.c @@ -26,6 +26,13 @@ #include "macros.h" #undef QEMU_GENERATE #include "gen_tcg.h" +#include "genptr.h" + +TCGv gen_read_preg(TCGv pred, uint8_t num) +{ + tcg_gen_mov_tl(pred, hex_pred[num]); + return pred; +} static inline void gen_log_predicated_reg_write(int rnum, TCGv val, uint32_t slot) @@ -52,7 +59,7 @@ static inline void gen_log_predicated_reg_write(int rnum, TCGv val, tcg_temp_free(slot_mask); } -static inline void gen_log_reg_write(int rnum, TCGv val) +void gen_log_reg_write(int rnum, TCGv val) { tcg_gen_mov_tl(hex_new_value[rnum], val); if (HEX_DEBUG) { @@ -115,7 +122,7 @@ static void gen_log_reg_write_pair(int rnum, TCGv_i64 val) } } -static inline void gen_log_pred_write(DisasContext *ctx, int pnum, TCGv val) +void gen_log_pred_write(DisasContext *ctx, int pnum, TCGv val) { TCGv base_val = tcg_temp_new(); @@ -255,7 +262,7 @@ static inline void gen_write_ctrl_reg_pair(DisasContext *ctx, int reg_num, } } -static TCGv gen_get_byte(TCGv result, int N, TCGv src, bool sign) +TCGv gen_get_byte(TCGv result, int N, TCGv src, bool sign) { if (sign) { tcg_gen_sextract_tl(result, src, N * 8, 8); @@ -265,7 +272,7 @@ static TCGv gen_get_byte(TCGv result, int N, TCGv src, bool sign) return result; } -static TCGv gen_get_byte_i64(TCGv result, int N, TCGv_i64 src, bool sign) +TCGv gen_get_byte_i64(TCGv result, int N, TCGv_i64 src, bool sign) { TCGv_i64 res64 = tcg_temp_new_i64(); if (sign) { @@ -279,7 +286,7 @@ static TCGv gen_get_byte_i64(TCGv result, int N, TCGv_i64 src, bool sign) return result; } -static inline TCGv gen_get_half(TCGv result, int N, TCGv src, bool sign) +TCGv gen_get_half(TCGv result, int N, TCGv src, bool sign) { if (sign) { tcg_gen_sextract_tl(result, src, N * 16, 16); @@ -289,12 +296,12 @@ static inline TCGv gen_get_half(TCGv result, int N, TCGv src, bool sign) return result; } -static inline void gen_set_half(int N, TCGv result, TCGv src) +void gen_set_half(int N, TCGv result, TCGv src) { tcg_gen_deposit_tl(result, result, src, N * 16, 16); } -static inline void gen_set_half_i64(int N, TCGv_i64 result, TCGv src) +void gen_set_half_i64(int N, TCGv_i64 result, TCGv src) { TCGv_i64 src64 = tcg_temp_new_i64(); tcg_gen_extu_i32_i64(src64, src); @@ -302,7 +309,7 @@ static inline void gen_set_half_i64(int N, TCGv_i64 result, TCGv src) tcg_temp_free_i64(src64); } -static void gen_set_byte_i64(int N, TCGv_i64 result, TCGv src) +void gen_set_byte_i64(int N, TCGv_i64 result, TCGv src) { TCGv_i64 src64 = tcg_temp_new_i64(); tcg_gen_extu_i32_i64(src64, src); @@ -381,60 +388,60 @@ static inline void gen_store_conditional8(DisasContext *ctx, tcg_gen_movi_tl(hex_llsc_addr, ~0); } -static inline void gen_store32(TCGv vaddr, TCGv src, int width, uint32_t slot) +void gen_store32(TCGv vaddr, TCGv src, int width, uint32_t slot) { tcg_gen_mov_tl(hex_store_addr[slot], vaddr); tcg_gen_movi_tl(hex_store_width[slot], width); tcg_gen_mov_tl(hex_store_val32[slot], src); } -static inline void gen_store1(TCGv_env cpu_env, TCGv vaddr, TCGv src, - DisasContext *ctx, uint32_t slot) +void gen_store1(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx, + uint32_t slot) { gen_store32(vaddr, src, 1, slot); ctx->store_width[slot] = 1; } -static inline void gen_store1i(TCGv_env cpu_env, TCGv vaddr, int32_t src, - DisasContext *ctx, uint32_t slot) +void gen_store1i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx, + uint32_t slot) { TCGv tmp = tcg_const_tl(src); gen_store1(cpu_env, vaddr, tmp, ctx, slot); tcg_temp_free(tmp); } -static inline void gen_store2(TCGv_env cpu_env, TCGv vaddr, TCGv src, - DisasContext *ctx, uint32_t slot) +void gen_store2(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx, + uint32_t slot) { gen_store32(vaddr, src, 2, slot); ctx->store_width[slot] = 2; } -static inline void gen_store2i(TCGv_env cpu_env, TCGv vaddr, int32_t src, - DisasContext *ctx, uint32_t slot) +void gen_store2i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx, + uint32_t slot) { TCGv tmp = tcg_const_tl(src); gen_store2(cpu_env, vaddr, tmp, ctx, slot); tcg_temp_free(tmp); } -static inline void gen_store4(TCGv_env cpu_env, TCGv vaddr, TCGv src, - DisasContext *ctx, uint32_t slot) +void gen_store4(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx, + uint32_t slot) { gen_store32(vaddr, src, 4, slot); ctx->store_width[slot] = 4; } -static inline void gen_store4i(TCGv_env cpu_env, TCGv vaddr, int32_t src, - DisasContext *ctx, uint32_t slot) +void gen_store4i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx, + uint32_t slot) { TCGv tmp = tcg_const_tl(src); gen_store4(cpu_env, vaddr, tmp, ctx, slot); tcg_temp_free(tmp); } -static inline void gen_store8(TCGv_env cpu_env, TCGv vaddr, TCGv_i64 src, - DisasContext *ctx, uint32_t slot) +void gen_store8(TCGv_env cpu_env, TCGv vaddr, TCGv_i64 src, DisasContext *ctx, + uint32_t slot) { tcg_gen_mov_tl(hex_store_addr[slot], vaddr); tcg_gen_movi_tl(hex_store_width[slot], 8); @@ -442,15 +449,15 @@ static inline void gen_store8(TCGv_env cpu_env, TCGv vaddr, TCGv_i64 src, ctx->store_width[slot] = 8; } -static inline void gen_store8i(TCGv_env cpu_env, TCGv vaddr, int64_t src, - DisasContext *ctx, uint32_t slot) +void gen_store8i(TCGv_env cpu_env, TCGv vaddr, int64_t src, DisasContext *ctx, + uint32_t slot) { TCGv_i64 tmp = tcg_const_i64(src); gen_store8(cpu_env, vaddr, tmp, ctx, slot); tcg_temp_free_i64(tmp); } -static TCGv gen_8bitsof(TCGv result, TCGv value) +TCGv gen_8bitsof(TCGv result, TCGv value) { TCGv zero = tcg_const_tl(0); TCGv ones = tcg_const_tl(0xff); diff --git a/target/hexagon/op_helper.c b/target/hexagon/op_helper.c index 459555966d..a864e68837 100644 --- a/target/hexagon/op_helper.c +++ b/target/hexagon/op_helper.c @@ -25,6 +25,7 @@ #include "arch.h" #include "hex_arch_types.h" #include "fma_emu.h" +#include "op_helper.h" #define SF_BIAS 127 #define SF_MANTBITS 23 @@ -45,8 +46,8 @@ void QEMU_NORETURN HELPER(raise_exception)(CPUHexagonState *env, uint32_t excp) do_raise_exception_err(env, excp, 0); } -static void log_reg_write(CPUHexagonState *env, int rnum, - target_ulong val, uint32_t slot) +void log_reg_write(CPUHexagonState *env, int rnum, target_ulong val, + uint32_t slot) { HEX_DEBUG_LOG("log_reg_write[%d] = " TARGET_FMT_ld " (0x" TARGET_FMT_lx ")", rnum, val, val); @@ -77,8 +78,8 @@ static void log_pred_write(CPUHexagonState *env, int pnum, target_ulong val) } } -static void log_store32(CPUHexagonState *env, target_ulong addr, - target_ulong val, int width, int slot) +void log_store32(CPUHexagonState *env, target_ulong addr, target_ulong val, + int width, int slot) { HEX_DEBUG_LOG("log_store%d(0x" TARGET_FMT_lx ", %" PRId32 " [0x08%" PRIx32 "])\n", @@ -88,8 +89,8 @@ static void log_store32(CPUHexagonState *env, target_ulong addr, env->mem_log_stores[slot].data32 = val; } -static void log_store64(CPUHexagonState *env, target_ulong addr, - int64_t val, int width, int slot) +void log_store64(CPUHexagonState *env, target_ulong addr, int64_t val, + int width, int slot) { HEX_DEBUG_LOG("log_store%d(0x" TARGET_FMT_lx ", %" PRId64 " [0x016%" PRIx64 "])\n", @@ -99,7 +100,7 @@ static void log_store64(CPUHexagonState *env, target_ulong addr, env->mem_log_stores[slot].data64 = val; } -static void write_new_pc(CPUHexagonState *env, target_ulong addr) +void write_new_pc(CPUHexagonState *env, target_ulong addr) { HEX_DEBUG_LOG("write_new_pc(0x" TARGET_FMT_lx ")\n", addr); @@ -390,8 +391,7 @@ static void check_noshuf(CPUHexagonState *env, uint32_t slot) } } -static uint8_t mem_load1(CPUHexagonState *env, uint32_t slot, - target_ulong vaddr) +uint8_t mem_load1(CPUHexagonState *env, uint32_t slot, target_ulong vaddr) { uint8_t retval; check_noshuf(env, slot); @@ -399,8 +399,7 @@ static uint8_t mem_load1(CPUHexagonState *env, uint32_t slot, return retval; } -static uint16_t mem_load2(CPUHexagonState *env, uint32_t slot, - target_ulong vaddr) +uint16_t mem_load2(CPUHexagonState *env, uint32_t slot, target_ulong vaddr) { uint16_t retval; check_noshuf(env, slot); @@ -408,8 +407,7 @@ static uint16_t mem_load2(CPUHexagonState *env, uint32_t slot, return retval; } -static uint32_t mem_load4(CPUHexagonState *env, uint32_t slot, - target_ulong vaddr) +uint32_t mem_load4(CPUHexagonState *env, uint32_t slot, target_ulong vaddr) { uint32_t retval; check_noshuf(env, slot); @@ -417,8 +415,7 @@ static uint32_t mem_load4(CPUHexagonState *env, uint32_t slot, return retval; } -static uint64_t mem_load8(CPUHexagonState *env, uint32_t slot, - target_ulong vaddr) +uint64_t mem_load8(CPUHexagonState *env, uint32_t slot, target_ulong vaddr) { uint64_t retval; check_noshuf(env, slot); @@ -1167,7 +1164,7 @@ float64 HELPER(dfmpyhh)(CPUHexagonState *env, float64 RxxV, return RxxV; } -static void cancel_slot(CPUHexagonState *env, uint32_t slot) +void cancel_slot(CPUHexagonState *env, uint32_t slot) { HEX_DEBUG_LOG("Slot %d cancelled\n", slot); env->slot_cancelled |= (1 << slot); -- 2.32.0 ^ permalink raw reply related [flat|nested] 27+ messages in thread
* [PATCH v6 05/12] target/hexagon: introduce new helper functions 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via ` (3 preceding siblings ...) 2021-07-20 12:30 ` [PATCH v6 04/12] target/hexagon: make helper functions non-static Alessandro Di Federico via @ 2021-07-20 12:30 ` Alessandro Di Federico via 2021-09-07 18:09 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 06/12] target/hexagon: expose next PC in DisasContext Alessandro Di Federico via ` (6 subsequent siblings) 11 siblings, 1 reply; 27+ messages in thread From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw) To: qemu-devel Cc: tsimpson, bcain, babush, nizzo, richard.henderson, Alessandro Di Federico From: Niccolò Izzo <nizzo@rev.ng> These helpers will be employed by the idef-parser generated code. Signed-off-by: Alessandro Di Federico <ale@rev.ng> Signed-off-by: Niccolò Izzo <nizzo@rev.ng> --- target/hexagon/genptr.h | 15 +++++- target/hexagon/macros.h | 9 ++++ target/hexagon/genptr.c | 113 +++++++++++++++++++++++++++++++++++++--- 3 files changed, 129 insertions(+), 8 deletions(-) diff --git a/target/hexagon/genptr.h b/target/hexagon/genptr.h index d71dd7e1ce..6bd54cc7f4 100644 --- a/target/hexagon/genptr.h +++ b/target/hexagon/genptr.h @@ -24,7 +24,8 @@ extern const SemanticInsn opcode_genptr[]; -void gen_store32(TCGv vaddr, TCGv src, int width, uint32_t slot); +void gen_store32(DisasContext *ctx, TCGv vaddr, TCGv src, tcg_target_long width, + uint32_t slot); void gen_store1(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx, uint32_t slot); void gen_store2(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx, @@ -44,6 +45,17 @@ void gen_store8i(TCGv_env cpu_env, TCGv vaddr, int64_t src, DisasContext *ctx, TCGv gen_read_preg(TCGv pred, uint8_t num); void gen_log_reg_write(int rnum, TCGv val); void gen_log_pred_write(DisasContext *ctx, int pnum, TCGv val); +void gen_write_new_pc(TCGv addr); +void gen_set_usr_field(int field, TCGv val); +void gen_set_usr_fieldi(int field, int x); +void gen_sat_i32(TCGv dest, TCGv source, int width); +void gen_sat_i32_ovfl(TCGv ovfl, TCGv dest, TCGv source, int width); +void gen_satu_i32(TCGv dest, TCGv source, int width); +void gen_satu_i32_ovfl(TCGv ovfl, TCGv dest, TCGv source, int width); +void gen_sat_i64(TCGv_i64 dest, TCGv_i64 source, int width); +void gen_sat_i64_ovfl(TCGv ovfl, TCGv_i64 dest, TCGv_i64 source, int width); +void gen_satu_i64(TCGv_i64 dest, TCGv_i64 source, int width); +void gen_satu_i64_ovfl(TCGv ovfl, TCGv_i64 dest, TCGv_i64 source, int width); TCGv gen_8bitsof(TCGv result, TCGv value); void gen_set_byte_i64(int N, TCGv_i64 result, TCGv src); TCGv gen_get_byte(TCGv result, int N, TCGv src, bool sign); @@ -51,5 +63,6 @@ TCGv gen_get_byte_i64(TCGv result, int N, TCGv_i64 src, bool sign); TCGv gen_get_half(TCGv result, int N, TCGv src, bool sign); void gen_set_half(int N, TCGv result, TCGv src); void gen_set_half_i64(int N, TCGv_i64 result, TCGv src); +TCGv gen_read_reg(TCGv result, int num); #endif diff --git a/target/hexagon/macros.h b/target/hexagon/macros.h index cd4f878fcf..2dabdbb49e 100644 --- a/target/hexagon/macros.h +++ b/target/hexagon/macros.h @@ -180,7 +180,16 @@ #define MEM_STORE8(VA, DATA, SLOT) log_store64(env, VA, DATA, 8, SLOT) #endif +#ifdef QEMU_GENERATE +static inline void gen_cancel(uint32_t slot) +{ + tcg_gen_ori_tl(hex_slot_cancelled, hex_slot_cancelled, 1 << slot); +} + +#define CANCEL gen_cancel(slot); +#else #define CANCEL cancel_slot(env, slot) +#endif #define LOAD_CANCEL(EA) do { CANCEL; } while (0) diff --git a/target/hexagon/genptr.c b/target/hexagon/genptr.c index 5e7f446657..b7eca94de7 100644 --- a/target/hexagon/genptr.c +++ b/target/hexagon/genptr.c @@ -28,6 +28,12 @@ #include "gen_tcg.h" #include "genptr.h" +TCGv gen_read_reg(TCGv result, int num) +{ + tcg_gen_mov_tl(result, hex_gpr[num]); + return result; +} + TCGv gen_read_preg(TCGv pred, uint8_t num) { tcg_gen_mov_tl(pred, hex_pred[num]); @@ -388,18 +394,19 @@ static inline void gen_store_conditional8(DisasContext *ctx, tcg_gen_movi_tl(hex_llsc_addr, ~0); } -void gen_store32(TCGv vaddr, TCGv src, int width, uint32_t slot) +void gen_store32(DisasContext *ctx, TCGv vaddr, TCGv src, tcg_target_long width, + uint32_t slot) { tcg_gen_mov_tl(hex_store_addr[slot], vaddr); tcg_gen_movi_tl(hex_store_width[slot], width); tcg_gen_mov_tl(hex_store_val32[slot], src); + ctx->store_width[slot] = width; } void gen_store1(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx, uint32_t slot) { - gen_store32(vaddr, src, 1, slot); - ctx->store_width[slot] = 1; + gen_store32(ctx, vaddr, src, 1, slot); } void gen_store1i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx, @@ -413,8 +420,7 @@ void gen_store1i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx, void gen_store2(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx, uint32_t slot) { - gen_store32(vaddr, src, 2, slot); - ctx->store_width[slot] = 2; + gen_store32(ctx, vaddr, src, 2, slot); } void gen_store2i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx, @@ -428,8 +434,7 @@ void gen_store2i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx, void gen_store4(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext *ctx, uint32_t slot) { - gen_store32(vaddr, src, 4, slot); - ctx->store_width[slot] = 4; + gen_store32(ctx, vaddr, src, 4, slot); } void gen_store4i(TCGv_env cpu_env, TCGv vaddr, int32_t src, DisasContext *ctx, @@ -468,5 +473,99 @@ TCGv gen_8bitsof(TCGv result, TCGv value) return result; } +void gen_set_usr_field(int field, TCGv val) +{ + tcg_gen_deposit_tl(hex_gpr[HEX_REG_USR], hex_gpr[HEX_REG_USR], val, + reg_field_info[field].offset, + reg_field_info[field].width); +} + +void gen_set_usr_fieldi(int field, int x) +{ + TCGv val = tcg_const_tl(x); + gen_set_usr_field(field, val); + tcg_temp_free(val); +} + +void gen_write_new_pc(TCGv addr) +{ + /* If there are multiple branches in a packet, ignore the second one */ + TCGv zero = tcg_const_tl(0); + tcg_gen_movcond_tl(TCG_COND_NE, hex_next_PC, hex_branch_taken, zero, + hex_next_PC, addr); + tcg_gen_movi_tl(hex_branch_taken, 1); + tcg_temp_free(zero); +} + +void gen_sat_i32(TCGv dest, TCGv source, int width) +{ + TCGv max_val = tcg_const_tl((1 << (width - 1)) - 1); + TCGv min_val = tcg_const_tl(-(1 << (width - 1))); + tcg_gen_smin_tl(dest, source, max_val); + tcg_gen_smax_tl(dest, dest, min_val); + tcg_temp_free(max_val); + tcg_temp_free(min_val); +} + +void gen_sat_i32_ovfl(TCGv ovfl, TCGv dest, TCGv source, int width) +{ + gen_sat_i32(dest, source, width); + tcg_gen_setcond_tl(TCG_COND_NE, ovfl, source, dest); +} + +void gen_satu_i32(TCGv dest, TCGv source, int width) +{ + TCGv max_val = tcg_const_tl((1 << width) - 1); + tcg_gen_movcond_tl(TCG_COND_GTU, dest, source, max_val, max_val, source); + TCGv zero = tcg_const_tl(0); + tcg_gen_movcond_tl(TCG_COND_LT, dest, source, zero, zero, dest); + tcg_temp_free(max_val); + tcg_temp_free(zero); +} + +void gen_satu_i32_ovfl(TCGv ovfl, TCGv dest, TCGv source, int width) +{ + gen_satu_i32(dest, source, width); + tcg_gen_setcond_tl(TCG_COND_NE, ovfl, source, dest); +} + +void gen_sat_i64(TCGv_i64 dest, TCGv_i64 source, int width) +{ + TCGv_i64 max_val = tcg_const_i64((1 << (width - 1)) - 1); + TCGv_i64 min_val = tcg_const_i64(-(1 << (width - 1))); + tcg_gen_smin_i64(dest, source, max_val); + tcg_gen_smax_i64(dest, dest, min_val); + tcg_temp_free_i64(max_val); + tcg_temp_free_i64(min_val); +} + +void gen_sat_i64_ovfl(TCGv ovfl, TCGv_i64 dest, TCGv_i64 source, int width) +{ + gen_sat_i64(dest, source, width); + TCGv_i64 ovfl_64 = tcg_temp_new_i64(); + tcg_gen_setcond_i64(TCG_COND_NE, ovfl_64, dest, source); + tcg_gen_trunc_i64_tl(ovfl, ovfl_64); + tcg_temp_free_i64(ovfl_64); +} + +void gen_satu_i64(TCGv_i64 dest, TCGv_i64 source, int width) +{ + TCGv_i64 max_val = tcg_const_i64((1 << width) - 1); + tcg_gen_movcond_i64(TCG_COND_GTU, dest, source, max_val, max_val, source); + TCGv_i64 zero = tcg_const_i64(0); + tcg_gen_movcond_i64(TCG_COND_LT, dest, source, zero, zero, dest); + tcg_temp_free_i64(max_val); + tcg_temp_free_i64(zero); +} + +void gen_satu_i64_ovfl(TCGv ovfl, TCGv_i64 dest, TCGv_i64 source, int width) +{ + gen_sat_i64(dest, source, width); + TCGv_i64 ovfl_64 = tcg_temp_new_i64(); + tcg_gen_setcond_i64(TCG_COND_NE, ovfl_64, dest, source); + tcg_gen_trunc_i64_tl(ovfl, ovfl_64); + tcg_temp_free_i64(ovfl_64); +} + #include "tcg_funcs_generated.c.inc" #include "tcg_func_table_generated.c.inc" -- 2.32.0 ^ permalink raw reply related [flat|nested] 27+ messages in thread
* RE: [PATCH v6 05/12] target/hexagon: introduce new helper functions 2021-07-20 12:30 ` [PATCH v6 05/12] target/hexagon: introduce new helper functions Alessandro Di Federico via @ 2021-09-07 18:09 ` Taylor Simpson 0 siblings, 0 replies; 27+ messages in thread From: Taylor Simpson @ 2021-09-07 18:09 UTC (permalink / raw) To: Alessandro Di Federico, qemu-devel@nongnu.org Cc: babush@rev.ng, Brian Cain, richard.henderson@linaro.org, nizzo@rev.ng, Alessandro Di Federico > -----Original Message----- > From: Alessandro Di Federico <ale.qemu@rev.ng> > Sent: Tuesday, July 20, 2021 7:30 AM > To: qemu-devel@nongnu.org > Cc: Taylor Simpson <tsimpson@quicinc.com>; Brian Cain > <bcain@quicinc.com>; babush@rev.ng; nizzo@rev.ng; > richard.henderson@linaro.org; Alessandro Di Federico <ale@rev.ng> > Subject: [PATCH v6 05/12] target/hexagon: introduce new helper functions > > From: Niccolò Izzo <nizzo@rev.ng> > > These helpers will be employed by the idef-parser generated code. > > Signed-off-by: Alessandro Di Federico <ale@rev.ng> > Signed-off-by: Niccolò Izzo <nizzo@rev.ng> > diff --git a/target/hexagon/genptr.c b/target/hexagon/genptr.c index > 5e7f446657..b7eca94de7 100644 > --- a/target/hexagon/genptr.c > +++ b/target/hexagon/genptr.c > -void gen_store32(TCGv vaddr, TCGv src, int width, uint32_t slot) > +void gen_store32(DisasContext *ctx, TCGv vaddr, TCGv src, tcg_target_long > width, > + uint32_t slot) > { > tcg_gen_mov_tl(hex_store_addr[slot], vaddr); > tcg_gen_movi_tl(hex_store_width[slot], width); > tcg_gen_mov_tl(hex_store_val32[slot], src); > + ctx->store_width[slot] = width; This isn't strictly a new helper function. You are moving the update to the DisasContext here instead of doing it in the functions that call this one. > } > > void gen_store1(TCGv_env cpu_env, TCGv vaddr, TCGv src, DisasContext > *ctx, > uint32_t slot) > { > - gen_store32(vaddr, src, 1, slot); > - ctx->store_width[slot] = 1; Here is where you the code above was previously. I'm not sure if this is better or worse, but it's not a new function. Otherwise Reviewed-by: Taylor Simpson <tsimpson@quicinc.com> ^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 06/12] target/hexagon: expose next PC in DisasContext 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via ` (4 preceding siblings ...) 2021-07-20 12:30 ` [PATCH v6 05/12] target/hexagon: introduce new helper functions Alessandro Di Federico via @ 2021-07-20 12:30 ` Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 07/12] target/hexagon: prepare input for the idef-parser Alessandro Di Federico via ` (5 subsequent siblings) 11 siblings, 0 replies; 27+ messages in thread From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw) To: qemu-devel Cc: tsimpson, bcain, babush, nizzo, richard.henderson, Alessandro Di Federico From: Paolo Montesel <babush@rev.ng> Signed-off-by: Alessandro Di Federico <ale@rev.ng> Signed-off-by: Paolo Montesel <babush@rev.ng> Reviewed-by: Richard Henderson <richard.henderson@linaro.org> Reviewed-by: Taylor Simpson <tsimpson@quicinc.com> --- target/hexagon/translate.h | 1 + target/hexagon/translate.c | 3 ++- 2 files changed, 3 insertions(+), 1 deletion(-) diff --git a/target/hexagon/translate.h b/target/hexagon/translate.h index 703fd1345f..d5a7fece31 100644 --- a/target/hexagon/translate.h +++ b/target/hexagon/translate.h @@ -37,6 +37,7 @@ typedef struct DisasContext { DECLARE_BITMAP(pregs_written, NUM_PREGS); uint8_t store_width[STORES_MAX]; bool s1_store_processed; + uint32_t npc; } DisasContext; static inline void ctx_log_reg_write(DisasContext *ctx, int rnum) diff --git a/target/hexagon/translate.c b/target/hexagon/translate.c index b23d36adf5..63cdf9b90d 100644 --- a/target/hexagon/translate.c +++ b/target/hexagon/translate.c @@ -509,11 +509,12 @@ static void decode_and_translate_packet(CPUHexagonState *env, DisasContext *ctx) if (decode_packet(nwords, words, &pkt, false) > 0) { HEX_DEBUG_PRINT_PKT(&pkt); gen_start_packet(ctx, &pkt); + ctx->npc = ctx->base.pc_next + pkt.encod_pkt_size_in_bytes; for (i = 0; i < pkt.num_insns; i++) { gen_insn(env, ctx, &pkt.insn[i], &pkt); } gen_commit_packet(ctx, &pkt); - ctx->base.pc_next += pkt.encod_pkt_size_in_bytes; + ctx->base.pc_next = ctx->npc; } else { gen_exception_end_tb(ctx, HEX_EXCP_INVALID_PACKET); } -- 2.32.0 ^ permalink raw reply related [flat|nested] 27+ messages in thread
* [PATCH v6 07/12] target/hexagon: prepare input for the idef-parser 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via ` (5 preceding siblings ...) 2021-07-20 12:30 ` [PATCH v6 06/12] target/hexagon: expose next PC in DisasContext Alessandro Di Federico via @ 2021-07-20 12:30 ` Alessandro Di Federico via 2021-09-07 18:09 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 08/12] target/hexagon: import lexer for idef-parser Alessandro Di Federico via ` (4 subsequent siblings) 11 siblings, 1 reply; 27+ messages in thread From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw) To: qemu-devel Cc: tsimpson, bcain, babush, nizzo, richard.henderson, Alessandro Di Federico From: Alessandro Di Federico <ale@rev.ng> Introduce infrastructure necessary to produce a file suitable for being parsed by the idef-parser. Signed-off-by: Alessandro Di Federico <ale@rev.ng> --- target/hexagon/gen_idef_parser_funcs.py | 114 ++++++++++++++++++ target/hexagon/idef-parser/macros.inc | 153 ++++++++++++++++++++++++ target/hexagon/idef-parser/prepare | 24 ++++ target/hexagon/meson.build | 16 +++ 4 files changed, 307 insertions(+) create mode 100644 target/hexagon/gen_idef_parser_funcs.py create mode 100644 target/hexagon/idef-parser/macros.inc create mode 100755 target/hexagon/idef-parser/prepare diff --git a/target/hexagon/gen_idef_parser_funcs.py b/target/hexagon/gen_idef_parser_funcs.py new file mode 100644 index 0000000000..7b8e0f6981 --- /dev/null +++ b/target/hexagon/gen_idef_parser_funcs.py @@ -0,0 +1,114 @@ +#!/usr/bin/env python3 + +## +## Copyright(c) 2019-2021 rev.ng Srls. All Rights Reserved. +## +## This program is free software; you can redistribute it and/or modify +## it under the terms of the GNU General Public License as published by +## the Free Software Foundation; either version 2 of the License, or +## (at your option) any later version. +## +## This program is distributed in the hope that it will be useful, +## but WITHOUT ANY WARRANTY; without even the implied warranty of +## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +## GNU General Public License for more details. +## +## You should have received a copy of the GNU General Public License +## along with this program; if not, see <http://www.gnu.org/licenses/>. +## + +import sys +import re +import string +from io import StringIO + +import hex_common + +## +## Generate code to be fed to the idef_parser +## +## Consider A2_add: +## +## Rd32=add(Rs32,Rt32), { RdV=RsV+RtV;} +## +## We produce: +## +## A2_add(RdV, in RsV, in RtV) { +## { RdV=RsV+RtV;} +## } +## +## A2_add represents the instruction tag. Then we have a list of TCGv +## that the code generated by the parser can expect in input. Some of +## them are inputs ("in" prefix), while some others are outputs. +## +def main(): + hex_common.read_semantics_file(sys.argv[1]) + hex_common.read_attribs_file(sys.argv[2]) + hex_common.read_overrides_file(sys.argv[3]) + hex_common.calculate_attribs() + tagregs = hex_common.get_tagregs() + tagimms = hex_common.get_tagimms() + + with open(sys.argv[4], 'w') as f: + f.write('#include "macros.inc"\n\n') + + for tag in hex_common.tags: + ## Skip the priv instructions + if ( "A_PRIV" in hex_common.attribdict[tag] ) : + continue + ## Skip the guest instructions + if ( "A_GUEST" in hex_common.attribdict[tag] ) : + continue + ## Skip instructions using switch + if ( tag in {'S4_vrcrotate_acc', 'S4_vrcrotate'} ) : + continue + ## Skip trap instructions + if ( tag in {'J2_trap0', 'J2_trap1'} ) : + continue + ## Skip 128-bit instructions + if ( tag in {'A7_croundd_ri', 'A7_croundd_rr'} ) : + continue + ## Skip other unsupported instructions + if ( tag.startswith('S2_cabacdecbin') ) : + continue + if ( tag.startswith('Y') ) : + continue + if ( tag.startswith('V6_') ) : + continue + if ( tag.startswith('F') ) : + continue + if ( tag.endswith('_locked') ) : + continue + + regs = tagregs[tag] + imms = tagimms[tag] + + arguments = [] + if hex_common.need_ea(tag): + arguments.append("EA") + + for regtype,regid,toss,numregs in regs: + prefix = "in " if hex_common.is_read(regid) else "" + + is_pair = hex_common.is_pair(regid) + is_single_old = (hex_common.is_single(regid) + and hex_common.is_old_val(regtype, regid, tag)) + is_single_new = (hex_common.is_single(regid) + and hex_common.is_new_val(regtype, regid, tag)) + + if is_pair or is_single_old: + arguments.append("%s%s%sV" % (prefix, regtype, regid)) + elif is_single_new: + arguments.append("%s%s%sN" % (prefix, regtype, regid)) + else: + print("Bad register parse: ",regtype,regid,toss,numregs) + + for immlett,bits,immshift in imms: + arguments.append(hex_common.imm_name(immlett)) + + f.write("%s(%s) {\n" % (tag, ", ".join(arguments))) + f.write(" %s\n" % hex_common.semdict[tag]) + f.write("}\n\n") + +if __name__ == "__main__": + main() diff --git a/target/hexagon/idef-parser/macros.inc b/target/hexagon/idef-parser/macros.inc new file mode 100644 index 0000000000..20535691e8 --- /dev/null +++ b/target/hexagon/idef-parser/macros.inc @@ -0,0 +1,153 @@ +/* + * Copyright(c) 2019-2021 rev.ng Srls. All Rights Reserved. + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, see <http://www.gnu.org/licenses/>. + */ + +/* Copy rules */ +#define fLSBOLD(VAL) (fGETBIT(0, VAL)) +#define fSATH(VAL) fSATN(16, VAL) +#define fSATUH(VAL) fSATUN(16, VAL) +#define fVSATH(VAL) fVSATN(16, VAL) +#define fVSATUH(VAL) fVSATUN(16, VAL) +#define fSATUB(VAL) fSATUN(8, VAL) +#define fSATB(VAL) fSATN(8, VAL) +#define fVSATUB(VAL) fVSATUN(8, VAL) +#define fVSATB(VAL) fVSATN(8, VAL) +#define fCALL(A) fWRITE_LR(fREAD_NPC()); fWRITE_NPC(A); +#define fCALLR(A) fWRITE_LR(fREAD_NPC()); fWRITE_NPC(A); +#define fCAST2_8s(A) fSXTN(16, 64, A) +#define fCAST2_8u(A) fZXTN(16, 64, A) +#define fCAST8S_16S(A) (fSXTN(64, 128, A)) +#define fCAST16S_8S(A) (fSXTN(128, 64, A)) +#define fVSATW(A) fVSATN(32, fCAST8_8s(A)) +#define fSATW(A) fSATN(32, fCAST8_8s(A)) +#define fVSAT(A) fVSATN(32, A) +#define fSAT(A) fSATN(32, A) + +/* Ease parsing */ +#define f8BITSOF(VAL) ((VAL) ? 0xff : 0x00) +#define fREAD_GP() (Constant_extended ? (0) : GP) +#define fCLIP(DST, SRC, U) (DST = fMIN((1 << U) - 1, fMAX(SRC, -(1 << U)))) +#define fBIDIR_ASHIFTL(SRC, SHAMT, REGSTYPE) \ + ((SHAMT > 0) ? \ + (fCAST##REGSTYPE##s(SRC) << SHAMT) : \ + (fCAST##REGSTYPE##s(SRC) >> -SHAMT)) + +#define fBIDIR_LSHIFTL(SRC, SHAMT, REGSTYPE) \ + ((SHAMT > 0) ? \ + (fCAST##REGSTYPE##u(SRC) << SHAMT) : \ + (fCAST##REGSTYPE##u(SRC) >>> -SHAMT)) + +#define fBIDIR_ASHIFTR(SRC, SHAMT, REGSTYPE) \ + ((SHAMT > 0) ? \ + (fCAST##REGSTYPE##s(SRC) >> SHAMT) : \ + (fCAST##REGSTYPE##s(SRC) << -SHAMT)) + +#define fBIDIR_SHIFTR(SRC, SHAMT, REGSTYPE) \ + (((SHAMT) < 0) ? ((fCAST##REGSTYPE(SRC) << ((-(SHAMT)) - 1)) << 1) \ + : (fCAST##REGSTYPE(SRC) >> (SHAMT))) + +#define fBIDIR_LSHIFTR(SRC, SHAMT, REGSTYPE) \ + fBIDIR_SHIFTR(SRC, SHAMT, REGSTYPE##u) + +#define fSATVALN(N, VAL) \ + fSET_OVERFLOW( \ + ((VAL) < 0) ? (-(1LL << ((N) - 1))) : ((1LL << ((N) - 1)) - 1) \ + ) + +#define fSAT_ORIG_SHL(A, ORIG_REG) \ + (((fCAST4s((fSAT(A)) ^ (fCAST4s(ORIG_REG)))) < 0) \ + ? fSATVALN(32, (fCAST4s(ORIG_REG))) \ + : ((((ORIG_REG) > 0) && ((A) == 0)) ? fSATVALN(32, (ORIG_REG)) \ + : fSAT(A))) + +#define fBIDIR_ASHIFTR_SAT(SRC, SHAMT, REGSTYPE) \ + (((SHAMT) < 0) ? fSAT_ORIG_SHL((fCAST##REGSTYPE##s(SRC) \ + << ((-(SHAMT)) - 1)) << 1, (SRC)) \ + : (fCAST##REGSTYPE##s(SRC) >> (SHAMT))) + +#define fBIDIR_ASHIFTL_SAT(SRC, SHAMT, REGSTYPE) \ + (((SHAMT) < 0) \ + ? ((fCAST##REGSTYPE##s(SRC) >> ((-(SHAMT)) - 1)) >> 1) \ + : fSAT_ORIG_SHL(fCAST##REGSTYPE##s(SRC) << (SHAMT), (SRC))) + +#define fEXTRACTU_BIDIR(INREG, WIDTH, OFFSET) \ + (fZXTN(WIDTH, 32, fBIDIR_LSHIFTR((INREG), (OFFSET), 4_8))) + +#define fADDSAT64(DST, A, B) \ + __a = fCAST8u(A); \ + __b = fCAST8u(B); \ + __sum = __a + __b; \ + __xor = __a ^ __b; \ + __mask = fCAST8s(0x8000000000000000ULL); \ + if (((__a ^ __b) | ~(__a ^ __sum)) & __mask) { \ + DST = __sum; \ + } else { \ + DST = ((__sum & __mask) >> 63) + __mask; \ + fSET_OVERFLOW(); \ + } + +/* Least significant bit operations */ +#define fLSBNEW0 fLSBNEW(P0N) +#define fLSBNEW1 fLSBNEW(P1N) +#define fLSBOLDNOT(VAL) fGETBIT(0, ~VAL) +#define fLSBNEWNOT(PRED) (fLSBNEW(~PRED)) +#define fLSBNEW0NOT fLSBNEW(~P0N) +#define fLSBNEW1NOT fLSBNEW(~P1N) + +/* Assignments */ +#define fPCALIGN(IMM) (IMM = IMM & ~3) +#define fWRITE_LR(A) (LR = A) +#define fWRITE_FP(A) (FP = A) +#define fWRITE_SP(A) (SP = A) +#define fBRANCH(LOC, TYPE) (PC = LOC) +#define fJUMPR(REGNO, TARGET, TYPE) (PC = TARGET) +#define fWRITE_LOOP_REGS0(START, COUNT) SA0 = START; (LC0 = COUNT) +#define fWRITE_LOOP_REGS1(START, COUNT) SA1 = START; (LC1 = COUNT) +#define fWRITE_LC0(VAL) (LC0 = VAL) +#define fWRITE_LC1(VAL) (LC1 = VAL) +#define fSET_LPCFG(VAL) (USR.LPCFG = VAL) +#define fWRITE_P0(VAL) P0 = VAL; +#define fWRITE_P1(VAL) P1 = VAL; +#define fWRITE_P3(VAL) P3 = VAL; +#define fEA_RI(REG, IMM) (EA = REG + IMM) +#define fEA_RRs(REG, REG2, SCALE) (EA = REG + (REG2 << SCALE)) +#define fEA_IRs(IMM, REG, SCALE) (EA = IMM + (REG << SCALE)) +#define fEA_IMM(IMM) (EA = IMM) +#define fEA_REG(REG) (EA = REG) +#define fEA_BREVR(REG) (EA = fbrev(REG)) +#define fEA_GPI(IMM) (EA = fREAD_GP() + IMM) +#define fPM_I(REG, IMM) (REG = REG + IMM) +#define fPM_M(REG, MVAL) (REG = REG + MVAL) +#define fWRITE_NPC(VAL) (PC = VAL) + +/* Unary operators */ +#define fROUND(A) (A + 0x8000) + +/* Binary operators */ +#define fADD128(A, B) (A + B) +#define fSUB128(A, B) (A - B) +#define fSHIFTR128(A, B) (size8s_t) (A >> B) +#define fSHIFTL128(A, B) (A << B) +#define fAND128(A, B) (A & B) +#define fSCALE(N, A) (A << N) +#define fASHIFTR(SRC, SHAMT, REGSTYPE) (SRC >> SHAMT) +#define fLSHIFTR(SRC, SHAMT, REGSTYPE) (SRC >>> SHAMT) +#define fROTL(SRC, SHAMT, REGSTYPE) fROTL(SRC, SHAMT) +#define fASHIFTL(SRC, SHAMT, REGSTYPE) (fCAST##REGSTYPE##s(SRC) << SHAMT) + +/* Purge non-relavant parts */ +#define fHIDE(A) +#define fBRANCH_SPECULATE_STALL(A, B, C, D, E) diff --git a/target/hexagon/idef-parser/prepare b/target/hexagon/idef-parser/prepare new file mode 100755 index 0000000000..c9bff5fb80 --- /dev/null +++ b/target/hexagon/idef-parser/prepare @@ -0,0 +1,24 @@ +#!/bin/bash + +# +# Copyright(c) 2019-2021 rev.ng Srls. All Rights Reserved. +# +# This library is free software; you can redistribute it and/or +# modify it under the terms of the GNU Lesser General Public +# License as published by the Free Software Foundation; either +# version 2 of the License, or (at your option) any later version. +# +# This library is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +# Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public +# License along with this library; if not, see <http://www.gnu.org/licenses/>. +# + +set -e +set -o pipefail + +# Run the preprocessor and drop comments +cpp "$@" diff --git a/target/hexagon/meson.build b/target/hexagon/meson.build index 6fd9360b74..a592662c1b 100644 --- a/target/hexagon/meson.build +++ b/target/hexagon/meson.build @@ -20,6 +20,7 @@ hexagon_ss = ss.source_set() hex_common_py = 'hex_common.py' attribs_def = meson.current_source_dir() / 'attribs_def.h.inc' gen_tcg_h = meson.current_source_dir() / 'gen_tcg.h' +idef_parser_dir = meson.current_source_dir() / 'idef-parser' # # Step 1 @@ -175,4 +176,19 @@ hexagon_ss.add(files( 'fma_emu.c', )) +idef_parser_input_generated = custom_target( + 'idef_parser_input.h.inc', + output: 'idef_parser_input.h.inc', + depends: [semantics_generated], + depend_files: [hex_common_py, gen_tcg_h], + command: [python, files('gen_idef_parser_funcs.py'), semantics_generated, attribs_def, gen_tcg_h, '@OUTPUT@'], +) + +idef_parser_input_generated_prep = custom_target( + 'idef_parser_input.preprocessed.h.inc', + output: 'idef_parser_input.preprocessed.h.inc', + input: idef_parser_input_generated, + command: [idef_parser_dir / 'prepare', '@INPUT@', '-I' + idef_parser_dir, '-o', '@OUTPUT@'], +) + target_arch += {'hexagon': hexagon_ss} -- 2.32.0 ^ permalink raw reply related [flat|nested] 27+ messages in thread
* RE: [PATCH v6 07/12] target/hexagon: prepare input for the idef-parser 2021-07-20 12:30 ` [PATCH v6 07/12] target/hexagon: prepare input for the idef-parser Alessandro Di Federico via @ 2021-09-07 18:09 ` Taylor Simpson 2021-10-18 11:23 ` Anton Johansson via 0 siblings, 1 reply; 27+ messages in thread From: Taylor Simpson @ 2021-09-07 18:09 UTC (permalink / raw) To: Alessandro Di Federico, qemu-devel@nongnu.org Cc: babush@rev.ng, Brian Cain, richard.henderson@linaro.org, nizzo@rev.ng, Alessandro Di Federico > -----Original Message----- > From: Alessandro Di Federico <ale.qemu@rev.ng> > Sent: Tuesday, July 20, 2021 7:30 AM > To: qemu-devel@nongnu.org > Cc: Taylor Simpson <tsimpson@quicinc.com>; Brian Cain > <bcain@quicinc.com>; babush@rev.ng; nizzo@rev.ng; > richard.henderson@linaro.org; Alessandro Di Federico <ale@rev.ng> > Subject: [PATCH v6 07/12] target/hexagon: prepare input for the idef-parser > > From: Alessandro Di Federico <ale@rev.ng> > > Introduce infrastructure necessary to produce a file suitable for being parsed > by the idef-parser. > > Signed-off-by: Alessandro Di Federico <ale@rev.ng> > diff --git a/target/hexagon/gen_idef_parser_funcs.py > b/target/hexagon/gen_idef_parser_funcs.py > new file mode 100644 > index 0000000000..7b8e0f6981 > --- /dev/null > +++ b/target/hexagon/gen_idef_parser_funcs.py > +def main(): > + hex_common.read_semantics_file(sys.argv[1]) > + hex_common.read_attribs_file(sys.argv[2]) > + hex_common.read_overrides_file(sys.argv[3]) Do you really need the overrides file? > diff --git a/target/hexagon/idef-parser/macros.inc b/target/hexagon/idef- > parser/macros.inc > new file mode 100644 > index 0000000000..20535691e8 > --- /dev/null > +++ b/target/hexagon/idef-parser/macros.inc > +/* Ease parsing */ > +#define fADDSAT64(DST, A, B) \ > + __a = fCAST8u(A); \ > + __b = fCAST8u(B); \ > + __sum = __a + __b; \ > + __xor = __a ^ __b; \ > + __mask = fCAST8s(0x8000000000000000ULL); \ > + if (((__a ^ __b) | ~(__a ^ __sum)) & __mask) { \ > + DST = __sum; \ > + } else { \ > + DST = ((__sum & __mask) >> 63) + __mask; \ > + fSET_OVERFLOW(); \ > + } There are a bunch of these with pretty complex semantics. Wouldn't it be easier to recognize them in the parser and call a hand-written helper? > +/* Assignments */ > +#define fPCALIGN(IMM) (IMM = IMM & ~3) > +#define fWRITE_LR(A) (LR = A) > +#define fWRITE_FP(A) (FP = A) > +#define fWRITE_SP(A) (SP = A) > +#define fBRANCH(LOC, TYPE) (PC = LOC) > +#define fJUMPR(REGNO, TARGET, TYPE) (PC = TARGET) This should invoke write_new_pc because there may be more than one in the same packet. > +fWRITE_NPC(VAL) (PC = VAL) Ditto > +/* Binary operators */ > +#define fADD128(A, B) (A + B) > +#define fSUB128(A, B) (A - B) > +#define fSHIFTR128(A, B) (size8s_t) (A >> B) > +#define fSHIFTL128(A, B) (A << B) > +#define fAND128(A, B) (A & B) Will these operate on 128 bit numbers? ^ permalink raw reply [flat|nested] 27+ messages in thread
* Re: [PATCH v6 07/12] target/hexagon: prepare input for the idef-parser 2021-09-07 18:09 ` Taylor Simpson @ 2021-10-18 11:23 ` Anton Johansson via 2021-10-28 15:44 ` Taylor Simpson 0 siblings, 1 reply; 27+ messages in thread From: Anton Johansson via @ 2021-10-18 11:23 UTC (permalink / raw) To: Taylor Simpson, Alessandro Di Federico, qemu-devel@nongnu.org Cc: Brian Cain, babush@rev.ng, nizzo@rev.ng, richard.henderson@linaro.org, Alessandro Di Federico [-- Attachment #1: Type: text/plain, Size: 1464 bytes --] On 9/7/21 18:09, Taylor Simpson wrote: >> +#define fADDSAT64(DST, A, B) \ >> + __a = fCAST8u(A); \ >> + __b = fCAST8u(B); \ >> + __sum = __a + __b; \ >> + __xor = __a ^ __b; \ >> + __mask = fCAST8s(0x8000000000000000ULL); \ >> + if (((__a ^ __b) | ~(__a ^ __sum)) & __mask) { \ >> + DST = __sum; \ >> + } else { \ >> + DST = ((__sum & __mask) >> 63) + __mask; \ >> + fSET_OVERFLOW(); \ >> + } > There are a bunch of these with pretty complex semantics. Wouldn't it be easier to recognize them in the parser and call a hand-written helper? > These macro redefinitions are needed to work with the auto type system in idef-parser. We can drop these specializations in the upcoming patchset where we parse fHIDE declarations. If we still want to resort to helpers here, it's probably better to exclude instructions using fADDSAT64 (and similar) directly, and fallback on helpers/overrides for those instructions. -- Anton Johansson, rev.ng Srls. [-- Attachment #2: Type: text/html, Size: 2018 bytes --] ^ permalink raw reply [flat|nested] 27+ messages in thread
* RE: [PATCH v6 07/12] target/hexagon: prepare input for the idef-parser 2021-10-18 11:23 ` Anton Johansson via @ 2021-10-28 15:44 ` Taylor Simpson 0 siblings, 0 replies; 27+ messages in thread From: Taylor Simpson @ 2021-10-28 15:44 UTC (permalink / raw) To: anjo@rev.ng, Alessandro Di Federico, qemu-devel@nongnu.org Cc: babush@rev.ng, Brian Cain, richard.henderson@linaro.org, nizzo@rev.ng, Alessandro Di Federico > From: Anton Johansson <anjo@rev.ng> > Sent: Monday, October 18, 2021 6:23 AM > To: Taylor Simpson <tsimpson@quicinc.com>; Alessandro Di Federico <ale.qemu@rev.ng>; qemu-devel@nongnu.org > Cc: Brian Cain <bcain@quicinc.com>; babush@rev.ng; nizzo@rev.ng; richard.henderson@linaro.org; Alessandro Di Federico <ale@rev.ng> > Subject: Re: [PATCH v6 07/12] target/hexagon: prepare input for the idef-parser > > On 9/7/21 18:09, Taylor Simpson wrote: > +#define fADDSAT64(DST, A, B) \ > + __a = fCAST8u(A); \ > + __b = fCAST8u(B); \ > + __sum = __a + __b; \ > + __xor = __a ^ __b; \ > + __mask = fCAST8s(0x8000000000000000ULL); \ > + if (((__a ^ __b) | ~(__a ^ __sum)) & __mask) { \ > + DST = __sum; \ > + } else { \ > + DST = ((__sum & __mask) >> 63) + __mask; \ > + fSET_OVERFLOW(); \ > + } > There are a bunch of these with pretty complex semantics. Wouldn't it be easier to recognize them in the parser and call a hand-written helper? > > These macro redefinitions are needed to work with the auto type system in idef-parser. We can drop these specializations in the upcoming patchset where we parse fHIDE declarations. > If we still want to resort to helpers here, it's probably better to exclude instructions using fADDSAT64 (and similar) directly, and fallback on helpers/overrides for those instructions. Sorry, I wasn't clear. I meant a hand-written function to generate the TCG code - not a qemu TCG helper. Thanks, Taylor ^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 08/12] target/hexagon: import lexer for idef-parser 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via ` (6 preceding siblings ...) 2021-07-20 12:30 ` [PATCH v6 07/12] target/hexagon: prepare input for the idef-parser Alessandro Di Federico via @ 2021-07-20 12:30 ` Alessandro Di Federico via 2021-09-07 18:08 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 09/12] target/hexagon: import parser " Alessandro Di Federico via ` (3 subsequent siblings) 11 siblings, 1 reply; 27+ messages in thread From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw) To: qemu-devel Cc: tsimpson, bcain, babush, nizzo, richard.henderson, Alessandro Di Federico From: Paolo Montesel <babush@rev.ng> Signed-off-by: Alessandro Di Federico <ale@rev.ng> Signed-off-by: Paolo Montesel <babush@rev.ng> --- target/hexagon/idef-parser/idef-parser.h | 258 +++++++ target/hexagon/idef-parser/idef-parser.lex | 642 ++++++++++++++++++ target/hexagon/meson.build | 4 + tests/docker/dockerfiles/alpine.docker | 1 + tests/docker/dockerfiles/centos8.docker | 1 + tests/docker/dockerfiles/debian-amd64.docker | 1 + tests/docker/dockerfiles/debian10.docker | 1 + .../dockerfiles/fedora-i386-cross.docker | 1 + .../dockerfiles/fedora-win32-cross.docker | 1 + .../dockerfiles/fedora-win64-cross.docker | 1 + tests/docker/dockerfiles/fedora.docker | 1 + tests/docker/dockerfiles/opensuse-leap.docker | 1 + tests/docker/dockerfiles/ubuntu.docker | 1 + tests/docker/dockerfiles/ubuntu1804.docker | 1 + tests/docker/dockerfiles/ubuntu2004.docker | 2 + 15 files changed, 917 insertions(+) create mode 100644 target/hexagon/idef-parser/idef-parser.h create mode 100644 target/hexagon/idef-parser/idef-parser.lex diff --git a/target/hexagon/idef-parser/idef-parser.h b/target/hexagon/idef-parser/idef-parser.h new file mode 100644 index 0000000000..f2c07793c6 --- /dev/null +++ b/target/hexagon/idef-parser/idef-parser.h @@ -0,0 +1,258 @@ +/* + * Copyright(c) 2019-2021 rev.ng Srls. All Rights Reserved. + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, see <http://www.gnu.org/licenses/>. + */ + +#ifndef IDEF_PARSER_H +#define IDEF_PARSER_H + +#include <inttypes.h> +#include <stdio.h> +#include <stdbool.h> +#include <glib.h> + +#define TCGV_NAME_SIZE 7 +#define MAX_WRITTEN_REGS 32 +#define OFFSET_STR_LEN 32 +#define ALLOC_LIST_LEN 32 +#define ALLOC_NAME_SIZE 32 +#define INIT_LIST_LEN 32 +#define OUT_BUF_LEN (1024 * 1024) +#define SIGNATURE_BUF_LEN (128 * 1024) +#define HEADER_BUF_LEN (128 * 1024) + +/* Variadic macros to wrap the buffer printing functions */ +#define EMIT(c, ...) \ + do { \ + g_string_append_printf((c)->out_str, __VA_ARGS__); \ + } while (0) + +#define EMIT_SIG(c, ...) \ + do { \ + g_string_append_printf((c)->signature_str, __VA_ARGS__); \ + } while (0) + +#define EMIT_HEAD(c, ...) \ + do { \ + g_string_append_printf((c)->header_str, __VA_ARGS__); \ + } while (0) + +/** + * Type of register, assigned to the HexReg.type field + */ +typedef enum {GENERAL_PURPOSE, CONTROL, MODIFIER, DOTNEW} RegType; + +typedef enum { UNKNOWN_SIGNEDNESS, SIGNED, UNSIGNED } HexSignedness; + +/** + * Semantic record of the REG tokens, identifying registers + */ +typedef struct HexReg { + uint8_t id; /**< Identifier of the register */ + RegType type; /**< Type of the register */ + unsigned bit_width; /**< Bit width of the reg, 32 or 64 bits */ +} HexReg; + +/** + * Data structure, identifying a TCGv temporary value + */ +typedef struct HexTmp { + int index; /**< Index of the TCGv temporary value */ +} HexTmp; + +/** + * Enum of the possible immediated, an immediate is a value which is known + * at tinycode generation time, e.g. an integer value, not a TCGv + */ +enum ImmUnionTag { + I, + VARIABLE, + VALUE, + QEMU_TMP, + IMM_PC, + IMM_NPC, + IMM_CONSTEXT, +}; + +/** + * Semantic record of the IMM token, identifying an immediate constant + */ +typedef struct HexImm { + union { + char id; /**< Identifier of the immediate */ + uint64_t value; /**< Immediate value (for VALUE type immediates) */ + uint64_t index; /**< Index of the immediate (for int temp vars) */ + }; + enum ImmUnionTag type; /**< Type of the immediate */ +} HexImm; + +/** + * Semantic record of the PRE token, identifying a predicate + */ +typedef struct HexPre { + char id; /**< Identifier of the predicate */ +} HexPre; + +/** + * Semantic record of the SAT token, identifying the saturate operator + */ +typedef struct HexSat { + bool set_overflow; /**< Set-overflow feature for the sat operator */ + HexSignedness signedness;/**< Unsigned flag for the saturate operator */ +} HexSat; + +/** + * Semantic record of the CAST token, identifying the cast operator + */ +typedef struct HexCast { + int bit_width; /**< Bit width of the cast operator */ + HexSignedness signedness; /**< Unsigned flag for the cast operator */ +} HexCast; + +/** + * Semantic record of the EXTRACT token, identifying the cast operator + */ +typedef struct HexExtract { + int bit_width; /**< Bit width of the extract operator */ + int storage_bit_width; /**< Actual bit width of the extract operator */ + HexSignedness signedness; /**< Unsigned flag for the extract operator */ +} HexExtract; + +/** + * Semantic record of the MPY token, identifying the fMPY multiplication + * operator + */ +typedef struct HexMpy { + /**< Bit width of the first operand of fMPY op */ + int first_bit_width; + /**< Bit width of the second operand of fMPY */ + int second_bit_width; + /**< Signedness flag for the first operand of fMPY */ + HexSignedness first_signedness; + /**< Signedness flag for second operand of fMPY */ + HexSignedness second_signedness; +} HexMpy; + +/** + * Semantic record of the VARID token, identifying automatic variables + * of the input language + */ +typedef struct HexVar { + GString *name; /**< Name of the VARID automatic variable */ +} HexVar; + +/** + * Data structure uniquely identifying an automatic VARID variable, used for + * keeping track of declared variable, so that any variable is declared only + * once, and its properties are propagated through all the subsequent instances + * of that variable + */ +typedef struct Var { + GString *name; /**< Name of the VARID automatic variable */ + uint8_t bit_width; /**< Bit width of the VARID automatic variable */ + HexSignedness signedness; /**< Unsigned flag for the VARID automatic var */ +} Var; + +/** + * Enum of the possible rvalue types, used in the HexValue.type field + */ +typedef enum RvalueUnionTag { + REGISTER, REGISTER_ARG, TEMP, IMMEDIATE, PREDICATE, VARID +} RvalueUnionTag; + +/** + * Semantic record of the rvalue token, identifying any numeric value, + * immediate or register based. The rvalue tokens are combined together + * through the use of several operators, to encode expressions + */ +typedef struct HexValue { + union { + HexReg reg; /**< rvalue of register type */ + HexTmp tmp; /**< rvalue of temporary type */ + HexImm imm; /**< rvalue of immediate type */ + HexPre pre; /**< rvalue of predicate type */ + HexVar var; /**< rvalue of automatic variable type */ + }; + RvalueUnionTag type; /**< Type of the rvalue */ + unsigned bit_width; /**< Bit width of the rvalue */ + HexSignedness signedness; /**< Unsigned flag for the rvalue */ + bool is_dotnew; /**< rvalue of predicate type is dotnew? */ + bool is_manual; /**< Opt out of automatic freeing of params */ +} HexValue; + +/** + * State of ternary operator + */ +typedef enum TernaryState { IN_LEFT, IN_RIGHT } TernaryState; + +/** + * Data structure used to handle side effects inside ternary operators + */ +typedef struct Ternary { + TernaryState state; + HexValue cond; +} Ternary; + +/** + * Operator type, used for referencing the correct operator when calling the + * gen_bin_op() function, which in turn will generate the correct code to + * execute the operation between the two rvalues + */ +typedef enum OpType { + ADD_OP, SUB_OP, MUL_OP, ASL_OP, ASR_OP, LSR_OP, ANDB_OP, ORB_OP, + XORB_OP, ANDL_OP, MINI_OP, MAXI_OP, MOD_OP +} OpType; + +/** + * Data structure including instruction specific information, to be cleared + * out after the compilation of each instruction + */ +typedef struct Inst { + GString *name; /**< Name of the compiled instruction */ + char *code_begin; /**< Beginning of instruction input code */ + char *code_end; /**< End of instruction input code */ + int tmp_count; /**< Index of the last declared TCGv temp */ + int qemu_tmp_count; /**< Index of the last declared int temp */ + int if_count; /**< Index of the last declared if label */ + int error_count; /**< Number of generated errors */ + GArray *allocated; /**< Allocated VARID automatic vars */ + GArray *init_list; /**< List of initialized registers */ + GArray *strings; /**< Strings allocated by the instruction */ +} Inst; + +/** + * Data structure representing the whole translation context, which in a + * reentrant flex/bison parser just like ours is passed between the scanner + * and the parser, holding all the necessary information to perform the + * parsing, this data structure survives between the compilation of different + * instructions + * + */ +typedef struct Context { + void *scanner; /**< Reentrant parser state pointer */ + char *input_buffer; /**< Buffer containing the input code */ + GString *out_str; /**< String containing the output code */ + GString *signature_str; /**< String containing the signatures code */ + GString *header_str; /**< String containing the output code */ + FILE *defines_file; /**< FILE * of the generated header */ + FILE *output_file; /**< FILE * of the C output file */ + FILE *enabled_file; /**< FILE * of the list of enabled inst */ + GArray *ternary; /**< Array to track nesting of ternary ops */ + int total_insn; /**< Number of instructions in input file */ + int implemented_insn; /**< Instruction compiled without errors */ + Inst inst; /**< Parsing data of the current inst */ +} Context; + +#endif /* IDEF_PARSER_H */ diff --git a/target/hexagon/idef-parser/idef-parser.lex b/target/hexagon/idef-parser/idef-parser.lex new file mode 100644 index 0000000000..98486dbaed --- /dev/null +++ b/target/hexagon/idef-parser/idef-parser.lex @@ -0,0 +1,642 @@ +%option noyywrap noinput nounput +%option 8bit reentrant bison-bridge +%option warn nodefault +%option bison-locations + +%{ +/* + * Copyright(c) 2019-2021 rev.ng Srls. All Rights Reserved. + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, see <http://www.gnu.org/licenses/>. + */ + +#include <string.h> +#include <stdbool.h> + +#include "hex_regs.h" + +#include "idef-parser.h" +#include "idef-parser.tab.h" + +/* Keep track of scanner position for error message printout */ +#define YY_USER_ACTION yylloc->first_column = yylloc->last_column; \ + for (int i = 0; yytext[i] != '\0'; i++) { \ + yylloc->last_column++; \ + } + +/* Global Error Counter */ +int error_count; + +%} + +/* Definitions */ +DIGIT [0-9] +LOWER_ID [a-z] +UPPER_ID [A-Z] +ID LOWER_ID|UPPER_ID +INST_NAME [A-Z]+[0-9]_([A-Za-z]|[0-9]|_)+ +HEX_DIGIT [0-9a-fA-F] +REG_ID_32 e|s|d|t|u|v|x|y +REG_ID_64 ee|ss|dd|tt|uu|vv|xx|yy +SYS_ID_32 s|d +SYS_ID_64 ss|dd +LOWER_PRE d|s|t|u|v|e|x|x +IMM_ID r|s|S|u|U +VAR_ID [a-zA-Z_][a-zA-Z0-9_]* +SIGN_ID s|u + +/* Tokens */ +%% + +[ \t\f\v]+ { /* Ignore whitespaces. */ } +[\n\r]+ { /* Ignore newlines. */ } +^#.*$ { /* Ignore linemarkers. */ } + +{INST_NAME} { yylval->string = g_string_new(yytext); + return INAME; } +"fFLOAT" | +"fUNFLOAT" | +"fDOUBLE" | +"fUNDOUBLE" | +"0.0" | +"0x1.0p52" | +"0x1.0p-52" { return FAIL; } +"R"{REG_ID_32}"V" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = GENERAL_PURPOSE; + yylval->rvalue.reg.id = yytext[1]; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = false; + yylval->rvalue.signedness = SIGNED; + return REG; } +"R"{REG_ID_32}"N" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = DOTNEW; + yylval->rvalue.reg.id = yytext[1]; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = true; + yylval->rvalue.signedness = SIGNED; + return REG; } +"R"{REG_ID_64}"V" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = GENERAL_PURPOSE; + yylval->rvalue.reg.id = yytext[1]; + yylval->rvalue.reg.bit_width = 64; + yylval->rvalue.bit_width = 64; + yylval->rvalue.is_dotnew = false; + yylval->rvalue.signedness = SIGNED; + return REG; } +"R"{REG_ID_64}"N" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = DOTNEW; + yylval->rvalue.reg.id = yytext[1]; + yylval->rvalue.reg.bit_width = 64; + yylval->rvalue.bit_width = 64; + yylval->rvalue.is_dotnew = true; + yylval->rvalue.signedness = SIGNED; + return REG; } +"MuV" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = MODIFIER; + yylval->rvalue.reg.id = 'u'; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = SIGNED; + return REG; } +"C"{REG_ID_32}"V" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = CONTROL; + yylval->rvalue.reg.id = yytext[1]; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = false; + yylval->rvalue.signedness = SIGNED; + return REG; } +"C"{REG_ID_64}"V" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = CONTROL; + yylval->rvalue.reg.id = yytext[1]; + yylval->rvalue.reg.bit_width = 64; + yylval->rvalue.bit_width = 64; + yylval->rvalue.is_dotnew = false; + yylval->rvalue.signedness = SIGNED; + return REG; } +{IMM_ID}"iV" { + yylval->rvalue.type = IMMEDIATE; + yylval->rvalue.signedness = SIGNED; + yylval->rvalue.imm.type = VARIABLE; + yylval->rvalue.imm.id = yytext[0]; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = false; + return IMM; } +"P"{LOWER_PRE}"V" { + yylval->rvalue.type = PREDICATE; + yylval->rvalue.pre.id = yytext[1]; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = false; + yylval->rvalue.signedness = SIGNED; + return PRE; } +"P"{LOWER_PRE}"N" { + yylval->rvalue.type = PREDICATE; + yylval->rvalue.pre.id = yytext[1]; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = true; + yylval->rvalue.signedness = SIGNED; + return PRE; } +"in R"{REG_ID_32}"V" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = GENERAL_PURPOSE; + yylval->rvalue.reg.id = yytext[4]; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = false; + yylval->rvalue.signedness = SIGNED; + return RREG; } +"in R"{REG_ID_64}"V" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = GENERAL_PURPOSE; + yylval->rvalue.reg.id = yytext[4]; + yylval->rvalue.reg.bit_width = 64; + yylval->rvalue.bit_width = 64; + yylval->rvalue.is_dotnew = false; + yylval->rvalue.signedness = SIGNED; + return RREG; } +"in N"{REG_ID_32}"N" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = DOTNEW; + yylval->rvalue.reg.id = yytext[4]; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = true; + yylval->rvalue.signedness = SIGNED; + return RREG; } +"in N"{REG_ID_64}"N" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = DOTNEW; + yylval->rvalue.reg.id = yytext[4]; + yylval->rvalue.reg.bit_width = 64; + yylval->rvalue.bit_width = 64; + yylval->rvalue.is_dotnew = true; + yylval->rvalue.signedness = SIGNED; + return RREG; } +"in P"{LOWER_PRE}"V" { + yylval->rvalue.type = PREDICATE; + yylval->rvalue.pre.id = yytext[4]; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = false; + yylval->rvalue.signedness = SIGNED; + return RPRE; } +"in P"{LOWER_PRE}"N" { + yylval->rvalue.type = PREDICATE; + yylval->rvalue.pre.id = yytext[4]; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = true; + yylval->rvalue.signedness = SIGNED; + return RPRE; } +"in MuV" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = MODIFIER; + yylval->rvalue.reg.id = 'u'; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = SIGNED; + return RREG; } +"in C"{REG_ID_32}"V" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = CONTROL; + yylval->rvalue.reg.id = yytext[4]; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = false; + yylval->rvalue.signedness = SIGNED; + return RREG; } +"in C"{REG_ID_64}"V" { + yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = CONTROL; + yylval->rvalue.reg.id = yytext[4]; + yylval->rvalue.reg.bit_width = 64; + yylval->rvalue.bit_width = 64; + yylval->rvalue.is_dotnew = false; + yylval->rvalue.signedness = SIGNED; + return RREG; } +"fGEN_TCG_"{INST_NAME}"(" { return FWRAP; } +"IV1DEAD()" | +"fPAUSE(uiV);" { return ';'; } +"+=" { return INC; } +"-=" { return DEC; } +"++" { return PLUSPLUS; } +"&=" { return ANDA; } +"|=" { return ORA; } +"^=" { return XORA; } +"<<" { return ASL; } +">>" { return ASR; } +">>>" { return LSR; } +"==" { return EQ; } +"!=" { return NEQ; } +"<=" { return LTE; } +">=" { return GTE; } +"&&" { return ANDL; } +"||" { return ORL; } +"else" { return ELSE; } +"for" { return FOR; } +"fREAD_IREG" { return ICIRC; } +"fPART1" { return PART1; } +"if" { return IF; } +"fFRAME_SCRAMBLE" { return FSCR; } +"fFRAME_UNSCRAMBLE" { return FSCR; } +"fFRAMECHECK" { return FCHK; } +"Constant_extended" { return CONSTEXT; } +"fCL1_"{DIGIT} { return LOCNT; } +"fBREV_8" { return BREV_8; } +"fBREV_4" { return BREV_4; } +"fbrev" { return BREV; } +"fSXTN" { return SXT; } +"fZXTN" { return ZXT; } +"fDF_MAX" | +"fSF_MAX" | +"fMAX" { return MAX; } +"fDF_MIN" | +"fSF_MIN" | +"fMIN" { return MIN; } +"fABS" { return ABS; } +"fRNDN" { return ROUND; } +"fCRND" { return CROUND; } +"fCRNDN" { return CROUND; } +"fPM_CIRI" { return CIRCADD; } +"fPM_CIRR" { return CIRCADD; } +"fCOUNTONES_"{DIGIT} { return COUNTONES; } +"fSATN" { yylval->sat.set_overflow = true; + yylval->sat.signedness = SIGNED; + return SAT; } +"fVSATN" { yylval->sat.set_overflow = false; + yylval->sat.signedness = SIGNED; + return SAT; } +"fSATUN" { yylval->sat.set_overflow = false; + yylval->sat.signedness = UNSIGNED; + return SAT; } +"fSE32_64" { yylval->cast.bit_width = 64; + yylval->cast.signedness = SIGNED; + return CAST; } +"fCAST4_4u" { yylval->cast.bit_width = 32; + yylval->cast.signedness = UNSIGNED; + return CAST; } +"fCAST4_8s" { yylval->cast.bit_width = 64; + yylval->cast.signedness = SIGNED; + return CAST; } +"fCAST4_8u" { return CAST4_8U; } +"fCAST4u" { yylval->cast.bit_width = 32; + yylval->cast.signedness = UNSIGNED; + return CAST; } +"fNEWREG" | +"fCAST4s" { yylval->cast.bit_width = 32; + yylval->cast.signedness = SIGNED; + return CAST; } +"fCAST8_8u" { yylval->cast.bit_width = 64; + yylval->cast.signedness = UNSIGNED; + return CAST; } +"fCAST8u" { yylval->cast.bit_width = 64; + yylval->cast.signedness = UNSIGNED; + return CAST; } +"fCAST8s" { yylval->cast.bit_width = 64; + yylval->cast.signedness = SIGNED; + return CAST; } +"fGETBIT" { yylval->extract.bit_width = 1; + yylval->extract.storage_bit_width = 1; + yylval->extract.signedness = UNSIGNED; + return EXTRACT; } +"fGETBYTE" { yylval->extract.bit_width = 8; + yylval->extract.storage_bit_width = 8; + yylval->extract.signedness = SIGNED; + return EXTRACT; } +"fGETUBYTE" { yylval->extract.bit_width = 8; + yylval->extract.storage_bit_width = 8; + yylval->extract.signedness = UNSIGNED; + return EXTRACT; } +"fGETHALF" { yylval->extract.bit_width = 16; + yylval->extract.storage_bit_width = 16; + yylval->extract.signedness = SIGNED; + return EXTRACT; } +"fGETUHALF" { yylval->extract.bit_width = 16; + yylval->extract.storage_bit_width = 16; + yylval->extract.signedness = UNSIGNED; + return EXTRACT; } +"fGETWORD" { yylval->extract.bit_width = 32; + yylval->extract.storage_bit_width = 64; + yylval->extract.signedness = SIGNED; + return EXTRACT; } +"fGETUWORD" { yylval->extract.bit_width = 32; + yylval->extract.storage_bit_width = 64; + yylval->extract.signedness = UNSIGNED; + return EXTRACT; } +"fEXTRACTU_BITS" { return EXTBITS; } +"fEXTRACTU_RANGE" { return EXTRANGE; } +"fSETBIT" { yylval->cast.bit_width = 1; + yylval->cast.signedness = SIGNED; + return DEPOSIT; } +"fSETBYTE" { yylval->cast.bit_width = 8; + yylval->cast.signedness = SIGNED; + return DEPOSIT; } +"fSETHALF" { yylval->cast.bit_width = 16; + yylval->cast.signedness = SIGNED; + return SETHALF; } +"fSETWORD" { yylval->cast.bit_width = 32; + yylval->cast.signedness = SIGNED; + return DEPOSIT; } +"fINSERT_BITS" { return INSBITS; } +"fSETBITS" { return SETBITS; } +"fMPY8UU" { yylval->mpy.first_bit_width = 8; + yylval->mpy.second_bit_width = 8; + yylval->mpy.first_signedness = UNSIGNED; + yylval->mpy.second_signedness = UNSIGNED; + return MPY; } +"fMPY8US" { yylval->mpy.first_bit_width = 8; + yylval->mpy.second_bit_width = 8; + yylval->mpy.first_signedness = UNSIGNED; + yylval->mpy.second_signedness = SIGNED; + return MPY; } +"fMPY8SU" { yylval->mpy.first_bit_width = 8; + yylval->mpy.second_bit_width = 8; + yylval->mpy.first_signedness = SIGNED; + yylval->mpy.second_signedness = UNSIGNED; + return MPY; } +"fMPY8SS" { yylval->mpy.first_bit_width = 8; + yylval->mpy.second_bit_width = 8; + yylval->mpy.first_signedness = SIGNED; + yylval->mpy.second_signedness = SIGNED; + return MPY; } +"fMPY16UU" { yylval->mpy.first_bit_width = 16; + yylval->mpy.second_bit_width = 16; + yylval->mpy.first_signedness = UNSIGNED; + yylval->mpy.second_signedness = UNSIGNED; + return MPY; } +"fMPY16US" { yylval->mpy.first_bit_width = 16; + yylval->mpy.second_bit_width = 16; + yylval->mpy.first_signedness = UNSIGNED; + yylval->mpy.second_signedness = SIGNED; + return MPY; } +"fMPY16SU" { yylval->mpy.first_bit_width = 16; + yylval->mpy.second_bit_width = 16; + yylval->mpy.first_signedness = SIGNED; + yylval->mpy.second_signedness = UNSIGNED; + return MPY; } +"fMPY16SS" { yylval->mpy.first_bit_width = 16; + yylval->mpy.second_bit_width = 16; + yylval->mpy.first_signedness = SIGNED; + yylval->mpy.second_signedness = SIGNED; + return MPY; } +"fMPY32UU" { yylval->mpy.first_bit_width = 32; + yylval->mpy.second_bit_width = 32; + yylval->mpy.first_signedness = UNSIGNED; + yylval->mpy.second_signedness = UNSIGNED; + return MPY; } +"fMPY32US" { yylval->mpy.first_bit_width = 32; + yylval->mpy.second_bit_width = 32; + yylval->mpy.first_signedness = UNSIGNED; + yylval->mpy.second_signedness = SIGNED; + return MPY; } +"fMPY32SU" { yylval->mpy.first_bit_width = 32; + yylval->mpy.second_bit_width = 32; + yylval->mpy.first_signedness = SIGNED; + yylval->mpy.second_signedness = UNSIGNED; + return MPY; } +"fSFMPY" | +"fMPY32SS" { yylval->mpy.first_bit_width = 32; + yylval->mpy.second_bit_width = 32; + yylval->mpy.first_signedness = SIGNED; + yylval->mpy.second_signedness = SIGNED; + return MPY; } +"fMPY3216SS" { yylval->mpy.first_bit_width = 32; + yylval->mpy.second_bit_width = 16; + yylval->mpy.first_signedness = SIGNED; + yylval->mpy.second_signedness = SIGNED; + return MPY; } +"fMPY3216SU" { yylval->mpy.first_bit_width = 32; + yylval->mpy.second_bit_width = 16; + yylval->mpy.first_signedness = SIGNED; + yylval->mpy.second_signedness = UNSIGNED; + return MPY; } +"fNEWREG_ST" | +"fIMMEXT" | +"fMUST_IMMEXT" | +"fCAST2_2s" | +"fCAST2_2u" | +"fCAST4_4s" | +"fCAST8_8s" | +"fZE8_16" | +"fSE8_16" | +"fZE16_32" | +"fSE16_32" | +"fZE32_64" | +"fPASS" | +"fECHO" { return IDENTITY; } +"(size8"[us]"_t)" { yylval->cast.bit_width = 64; + if (yytext[6] == 'u') { + yylval->cast.signedness = UNSIGNED; + } else { + yylval->cast.signedness = SIGNED; + } + return CAST; } +"(int)" { yylval->cast.bit_width = 32; + yylval->cast.signedness = SIGNED; + return CAST; } +"(unsigned int)" { yylval->cast.bit_width = 32; + yylval->cast.signedness = UNSIGNED; + return CAST; } +"fREAD_PC()" | +"PC" { return PC; } +"fREAD_NPC()" | +"NPC" { return NPC; } +"fGET_LPCFG" | +"USR.LPCFG" { return LPCFG; } +"LOAD_CANCEL(EA)" | +"STORE_CANCEL(EA)" | +"CANCEL" { return CANCEL; } +"N"{LOWER_ID} { yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = GENERAL_PURPOSE; + yylval->rvalue.reg.id = yytext[1]; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = UNSIGNED; + return REG; } +"N"{LOWER_ID}"N" { yylval->rvalue.type = REGISTER_ARG; + yylval->rvalue.reg.type = DOTNEW; + yylval->rvalue.reg.id = yytext[1]; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = UNSIGNED; + return REG; } +"fREAD_SP()" | +"SP" { yylval->rvalue.type = REGISTER; + yylval->rvalue.reg.type = GENERAL_PURPOSE; + yylval->rvalue.reg.id = HEX_REG_SP; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = UNSIGNED; + return REG; } +"fREAD_FP()" | +"FP" { yylval->rvalue.type = REGISTER; + yylval->rvalue.reg.type = GENERAL_PURPOSE; + yylval->rvalue.reg.id = HEX_REG_FP; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = UNSIGNED; + return REG; } +"fREAD_LR()" | +"LR" { yylval->rvalue.type = REGISTER; + yylval->rvalue.reg.type = GENERAL_PURPOSE; + yylval->rvalue.reg.id = HEX_REG_LR; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = UNSIGNED; + return REG; } +"fREAD_GP()" | +"GP" { yylval->rvalue.type = REGISTER; + yylval->rvalue.reg.type = CONTROL; + yylval->rvalue.reg.id = HEX_REG_GP; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = UNSIGNED; + return REG; } +"fREAD_LC"[01] { yylval->rvalue.type = REGISTER; + yylval->rvalue.reg.type = CONTROL; + yylval->rvalue.reg.id = HEX_REG_LC0 + + (yytext[8] - '0') * 2; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = UNSIGNED; + return REG; } +"LC"[01] { yylval->rvalue.type = REGISTER; + yylval->rvalue.reg.type = CONTROL; + yylval->rvalue.reg.id = HEX_REG_LC0 + + (yytext[2] - '0') * 2; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = UNSIGNED; + return REG; } +"fREAD_SA"[01] { yylval->rvalue.type = REGISTER; + yylval->rvalue.reg.type = CONTROL; + yylval->rvalue.reg.id = HEX_REG_SA0 + + (yytext[8] - '0') * 2; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = UNSIGNED; + return REG; } +"SA"[01] { yylval->rvalue.type = REGISTER; + yylval->rvalue.reg.type = CONTROL; + yylval->rvalue.reg.id = HEX_REG_SA0 + + (yytext[2] - '0') * 2; + yylval->rvalue.reg.bit_width = 32; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = UNSIGNED; + return REG; } +"MuN" { return MUN; } +"fREAD_P0()" { yylval->rvalue.type = PREDICATE; + yylval->rvalue.pre.id = '0'; + yylval->rvalue.bit_width = 32; + return PRE; } +[pP]{DIGIT} { yylval->rvalue.type = PREDICATE; + yylval->rvalue.pre.id = yytext[1]; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = false; + return PRE; } +[pP]{DIGIT}[nN] { yylval->rvalue.type = PREDICATE; + yylval->rvalue.pre.id = yytext[1]; + yylval->rvalue.bit_width = 32; + yylval->rvalue.is_dotnew = true; + return PRE; } +"fLSBNEW" { return LSBNEW; } +"N" { yylval->rvalue.type = IMMEDIATE; + yylval->rvalue.bit_width = 32; + yylval->rvalue.imm.type = VARIABLE; + yylval->rvalue.imm.id = 'N'; + return IMM; } +"i" { yylval->rvalue.type = IMMEDIATE; + yylval->rvalue.bit_width = 32; + yylval->rvalue.imm.type = I; + return IMM; } +{SIGN_ID} { if (yytext[0] == 'u') { + yylval->signedness = UNSIGNED; + } else { + yylval->signedness = SIGNED; + } + return SIGN; + } +"fSF_BIAS()" { yylval->rvalue.type = IMMEDIATE; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = SIGNED; + yylval->rvalue.imm.type = VALUE; + yylval->rvalue.imm.value = 127; + return IMM; } +{DIGIT}+ { yylval->rvalue.type = IMMEDIATE; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = SIGNED; + yylval->rvalue.imm.type = VALUE; + yylval->rvalue.imm.value = atoi(yytext); + return IMM; } +{DIGIT}+"LL" { yylval->rvalue.type = IMMEDIATE; + yylval->rvalue.bit_width = 64; + yylval->rvalue.signedness = SIGNED; + yylval->rvalue.imm.type = VALUE; + yylval->rvalue.imm.value = strtoll(yytext, NULL, 10); + return IMM; } +{DIGIT}+"ULL" { yylval->rvalue.type = IMMEDIATE; + yylval->rvalue.bit_width = 64; + yylval->rvalue.signedness = UNSIGNED; + yylval->rvalue.imm.type = VALUE; + yylval->rvalue.imm.value = strtoull(yytext, + NULL, + 10); + return IMM; } +"0x"{HEX_DIGIT}+ { yylval->rvalue.type = IMMEDIATE; + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = SIGNED; + yylval->rvalue.imm.type = VALUE; + yylval->rvalue.imm.value = strtoul(yytext, NULL, 16); + return IMM; } +"0x"{HEX_DIGIT}+"LL" { yylval->rvalue.type = IMMEDIATE; + yylval->rvalue.bit_width = 64; + yylval->rvalue.signedness = SIGNED; + yylval->rvalue.imm.type = VALUE; + yylval->rvalue.imm.value = strtoll(yytext, NULL, 16); + return IMM; } +"0x"{HEX_DIGIT}+"ULL" { yylval->rvalue.type = IMMEDIATE; + yylval->rvalue.bit_width = 64; + yylval->rvalue.signedness = UNSIGNED; + yylval->rvalue.imm.type = VALUE; + yylval->rvalue.imm.value = strtoull(yytext, + NULL, + 16); + return IMM; } +"fCONSTLL" { return CONSTLL; } +"fCONSTULL" { return CONSTULL; } +"fLOAD" { return LOAD; } +"fSTORE" { return STORE; } +"fROTL" { return ROTL; } +"fSET_OVERFLOW" { return SETOVF; } +"fDEINTERLEAVE" { return DEINTERLEAVE; } +"fINTERLEAVE" { return INTERLEAVE; } +"fCARRY_FROM_ADD" { return CARRY_FROM_ADD; } +{VAR_ID} { /* Variable name, we adopt the C names convention */ + yylval->rvalue.type = VARID; + yylval->rvalue.var.name = g_string_new(yytext); + /* Default types are int */ + yylval->rvalue.bit_width = 32; + yylval->rvalue.signedness = SIGNED; + return VAR; } +"fHINTJR(RsV)" { /* Emit no token */ } +. { return yytext[0]; } + +%% diff --git a/target/hexagon/meson.build b/target/hexagon/meson.build index a592662c1b..839f965f3c 100644 --- a/target/hexagon/meson.build +++ b/target/hexagon/meson.build @@ -191,4 +191,8 @@ idef_parser_input_generated_prep = custom_target( command: [idef_parser_dir / 'prepare', '@INPUT@', '-I' + idef_parser_dir, '-o', '@OUTPUT@'], ) +flex = generator(find_program('flex'), + output: ['@BASENAME@.yy.c', '@BASENAME@.yy.h'], + arguments: ['-o', '@OUTPUT0@', '--header-file=@OUTPUT1@', '@INPUT@']) + target_arch += {'hexagon': hexagon_ss} diff --git a/tests/docker/dockerfiles/alpine.docker b/tests/docker/dockerfiles/alpine.docker index 7e6997e301..a6e593fb49 100644 --- a/tests/docker/dockerfiles/alpine.docker +++ b/tests/docker/dockerfiles/alpine.docker @@ -12,6 +12,7 @@ ENV PACKAGES \ ccache \ coreutils \ curl-dev \ + flex \ g++ \ gcc \ git \ diff --git a/tests/docker/dockerfiles/centos8.docker b/tests/docker/dockerfiles/centos8.docker index 46398c61ee..7d67cc2397 100644 --- a/tests/docker/dockerfiles/centos8.docker +++ b/tests/docker/dockerfiles/centos8.docker @@ -19,6 +19,7 @@ ENV PACKAGES \ device-mapper-multipath-devel \ diffutils \ findutils \ + flex \ gcc \ gcc-c++ \ genisoimage \ diff --git a/tests/docker/dockerfiles/debian-amd64.docker b/tests/docker/dockerfiles/debian-amd64.docker index ed546edcd6..cd4d8ebe9c 100644 --- a/tests/docker/dockerfiles/debian-amd64.docker +++ b/tests/docker/dockerfiles/debian-amd64.docker @@ -17,6 +17,7 @@ RUN apt update && \ cscope \ genisoimage \ exuberant-ctags \ + flex \ global \ libbz2-dev \ liblzo2-dev \ diff --git a/tests/docker/dockerfiles/debian10.docker b/tests/docker/dockerfiles/debian10.docker index b414af1b9f..8a0bdf39ea 100644 --- a/tests/docker/dockerfiles/debian10.docker +++ b/tests/docker/dockerfiles/debian10.docker @@ -23,6 +23,7 @@ RUN apt update && \ ccache \ clang \ dbus \ + flex \ gdb-multiarch \ gettext \ git \ diff --git a/tests/docker/dockerfiles/fedora-i386-cross.docker b/tests/docker/dockerfiles/fedora-i386-cross.docker index dbb8195eb1..99b68fbd53 100644 --- a/tests/docker/dockerfiles/fedora-i386-cross.docker +++ b/tests/docker/dockerfiles/fedora-i386-cross.docker @@ -4,6 +4,7 @@ ENV PACKAGES \ ccache \ diffutils \ findutils \ + flex \ gcc \ git \ libffi-devel.i686 \ diff --git a/tests/docker/dockerfiles/fedora-win32-cross.docker b/tests/docker/dockerfiles/fedora-win32-cross.docker index 5a03e1af43..e5116adbb3 100644 --- a/tests/docker/dockerfiles/fedora-win32-cross.docker +++ b/tests/docker/dockerfiles/fedora-win32-cross.docker @@ -7,6 +7,7 @@ ENV PACKAGES \ ccache \ diffutils \ findutils \ + flex \ gcc \ gettext \ git \ diff --git a/tests/docker/dockerfiles/fedora-win64-cross.docker b/tests/docker/dockerfiles/fedora-win64-cross.docker index ff706040c4..b564811c41 100644 --- a/tests/docker/dockerfiles/fedora-win64-cross.docker +++ b/tests/docker/dockerfiles/fedora-win64-cross.docker @@ -7,6 +7,7 @@ ENV PACKAGES \ ccache \ diffutils \ findutils \ + flex \ gcc \ gettext \ git \ diff --git a/tests/docker/dockerfiles/fedora.docker b/tests/docker/dockerfiles/fedora.docker index eec1add7f6..e35366f33d 100644 --- a/tests/docker/dockerfiles/fedora.docker +++ b/tests/docker/dockerfiles/fedora.docker @@ -20,6 +20,7 @@ ENV PACKAGES \ device-mapper-multipath-devel \ diffutils \ findutils \ + flex \ gcc \ gcc-c++ \ gcovr \ diff --git a/tests/docker/dockerfiles/opensuse-leap.docker b/tests/docker/dockerfiles/opensuse-leap.docker index 5a8bee0289..5826260fff 100644 --- a/tests/docker/dockerfiles/opensuse-leap.docker +++ b/tests/docker/dockerfiles/opensuse-leap.docker @@ -15,6 +15,7 @@ ENV PACKAGES \ dbus-1 \ diffutils \ findutils \ + flex \ gcc \ gcc-c++ \ gcovr \ diff --git a/tests/docker/dockerfiles/ubuntu.docker b/tests/docker/dockerfiles/ubuntu.docker index f0e0180d21..e1832b4ab3 100644 --- a/tests/docker/dockerfiles/ubuntu.docker +++ b/tests/docker/dockerfiles/ubuntu.docker @@ -14,6 +14,7 @@ ENV PACKAGES \ ccache \ clang \ dbus \ + flex \ gcc \ gettext \ git \ diff --git a/tests/docker/dockerfiles/ubuntu1804.docker b/tests/docker/dockerfiles/ubuntu1804.docker index 0880bf3e29..52dd458e31 100644 --- a/tests/docker/dockerfiles/ubuntu1804.docker +++ b/tests/docker/dockerfiles/ubuntu1804.docker @@ -11,6 +11,7 @@ ENV PACKAGES \ diffutils \ exuberant-ctags \ findutils \ + flex \ g++ \ gcc \ gcovr \ diff --git a/tests/docker/dockerfiles/ubuntu2004.docker b/tests/docker/dockerfiles/ubuntu2004.docker index 39de63d012..b3ebcffac2 100644 --- a/tests/docker/dockerfiles/ubuntu2004.docker +++ b/tests/docker/dockerfiles/ubuntu2004.docker @@ -1,5 +1,6 @@ FROM docker.io/library/ubuntu:20.04 ENV PACKAGES \ + bison \ bc \ bsdmainutils \ bzip2 \ @@ -11,6 +12,7 @@ ENV PACKAGES \ diffutils \ exuberant-ctags \ findutils \ + flex \ g++ \ gcc \ gcovr \ -- 2.32.0 ^ permalink raw reply related [flat|nested] 27+ messages in thread
* RE: [PATCH v6 08/12] target/hexagon: import lexer for idef-parser 2021-07-20 12:30 ` [PATCH v6 08/12] target/hexagon: import lexer for idef-parser Alessandro Di Federico via @ 2021-09-07 18:08 ` Taylor Simpson 2021-10-18 11:37 ` Anton Johansson via 0 siblings, 1 reply; 27+ messages in thread From: Taylor Simpson @ 2021-09-07 18:08 UTC (permalink / raw) To: Alessandro Di Federico, qemu-devel@nongnu.org Cc: babush@rev.ng, Brian Cain, richard.henderson@linaro.org, nizzo@rev.ng, Alessandro Di Federico > -----Original Message----- > From: Alessandro Di Federico <ale.qemu@rev.ng> > Sent: Tuesday, July 20, 2021 7:30 AM > To: qemu-devel@nongnu.org > Cc: Taylor Simpson <tsimpson@quicinc.com>; Brian Cain > <bcain@quicinc.com>; babush@rev.ng; nizzo@rev.ng; > richard.henderson@linaro.org; Alessandro Di Federico <ale@rev.ng> > Subject: [PATCH v6 08/12] target/hexagon: import lexer for idef-parser > > From: Paolo Montesel <babush@rev.ng> > > Signed-off-by: Alessandro Di Federico <ale@rev.ng> > Signed-off-by: Paolo Montesel <babush@rev.ng> > --- > target/hexagon/idef-parser/idef-parser.h | 258 +++++++ > target/hexagon/idef-parser/idef-parser.lex | 642 ++++++++++++++++++ > target/hexagon/meson.build | 4 + > tests/docker/dockerfiles/alpine.docker | 1 + > tests/docker/dockerfiles/centos8.docker | 1 + > tests/docker/dockerfiles/debian-amd64.docker | 1 + > tests/docker/dockerfiles/debian10.docker | 1 + > .../dockerfiles/fedora-i386-cross.docker | 1 + > .../dockerfiles/fedora-win32-cross.docker | 1 + > .../dockerfiles/fedora-win64-cross.docker | 1 + > tests/docker/dockerfiles/fedora.docker | 1 + > tests/docker/dockerfiles/opensuse-leap.docker | 1 + > tests/docker/dockerfiles/ubuntu.docker | 1 + > tests/docker/dockerfiles/ubuntu1804.docker | 1 + > tests/docker/dockerfiles/ubuntu2004.docker | 2 + Here you are adding flex to several docker files, then in the next patch you add bison. Create a single patch with the necessary changes to the docker files together that is separate from the idef lexer for Hexagon. > 15 files changed, 917 insertions(+) > create mode 100644 target/hexagon/idef-parser/idef-parser.h > create mode 100644 target/hexagon/idef-parser/idef-parser.lex > > diff --git a/target/hexagon/idef-parser/idef-parser.h b/target/hexagon/idef- > parser/idef-parser.h > new file mode 100644 > +/** > + * Type of register, assigned to the HexReg.type field > + */ > +typedef enum {GENERAL_PURPOSE, CONTROL, MODIFIER, DOTNEW} > RegType; Should be HexRegType to match the other names in this file (e.g., HexSignedness) > +/** > + * Semantic record of the PRE token, identifying a predicate > + */ > +typedef struct HexPre { > + char id; /**< Identifier of the predicate */ > +} HexPre; Call this HexPred - HexPre sounds like something that comes before ... > diff --git a/target/hexagon/idef-parser/idef-parser.lex > b/target/hexagon/idef-parser/idef-parser.lex > +REG_ID_32 e|s|d|t|u|v|x|y > +REG_ID_64 ee|ss|dd|tt|uu|vv|xx|yy > +SYS_ID_32 s|d > +SYS_ID_64 ss|dd > +LOWER_PRE d|s|t|u|v|e|x|x Call this PRED_ID to be consistent with REG_ID_32, etc > +"R"{REG_ID_32}"V" { > + yylval->rvalue.type = REGISTER_ARG; > + yylval->rvalue.reg.type = GENERAL_PURPOSE; > + yylval->rvalue.reg.id = yytext[1]; > + yylval->rvalue.reg.bit_width = 32; > + yylval->rvalue.bit_width = 32; > + yylval->rvalue.is_dotnew = false; > + yylval->rvalue.signedness = SIGNED; > + return REG; } > +"in R"{REG_ID_32}"V" { > + yylval->rvalue.type = REGISTER_ARG; > + yylval->rvalue.reg.type = GENERAL_PURPOSE; > + yylval->rvalue.reg.id = yytext[4]; > + yylval->rvalue.reg.bit_width = 32; > + yylval->rvalue.bit_width = 32; > + yylval->rvalue.is_dotnew = false; > + yylval->rvalue.signedness = SIGNED; > + return RREG; } Are you dropping the "in" here? This looks the same as the "R"(REG_ID_32}"V" case. Ditto for all the ones below ... > +"in R"{REG_ID_64}"V" { > + yylval->rvalue.type = REGISTER_ARG; > + yylval->rvalue.reg.type = GENERAL_PURPOSE; > + yylval->rvalue.reg.id = yytext[4]; > + yylval->rvalue.reg.bit_width = 64; > + yylval->rvalue.bit_width = 64; > + yylval->rvalue.is_dotnew = false; > + yylval->rvalue.signedness = SIGNED; > + return RREG; } > +"in N"{REG_ID_32}"N" { > + yylval->rvalue.type = REGISTER_ARG; > + yylval->rvalue.reg.type = DOTNEW; > + yylval->rvalue.reg.id = yytext[4]; > + yylval->rvalue.reg.bit_width = 32; > + yylval->rvalue.bit_width = 32; > + yylval->rvalue.is_dotnew = true; > + yylval->rvalue.signedness = SIGNED; > + return RREG; } > +"in N"{REG_ID_64}"N" { > + yylval->rvalue.type = REGISTER_ARG; > + yylval->rvalue.reg.type = DOTNEW; > + yylval->rvalue.reg.id = yytext[4]; > + yylval->rvalue.reg.bit_width = 64; > + yylval->rvalue.bit_width = 64; > + yylval->rvalue.is_dotnew = true; > + yylval->rvalue.signedness = SIGNED; > + return RREG; } > +"in P"{LOWER_PRE}"V" { > + yylval->rvalue.type = PREDICATE; > + yylval->rvalue.pre.id = yytext[4]; > + yylval->rvalue.bit_width = 32; > + yylval->rvalue.is_dotnew = false; > + yylval->rvalue.signedness = SIGNED; > + return RPRE; } > +"in P"{LOWER_PRE}"N" { > + yylval->rvalue.type = PREDICATE; > + yylval->rvalue.pre.id = yytext[4]; > + yylval->rvalue.bit_width = 32; > + yylval->rvalue.is_dotnew = true; > + yylval->rvalue.signedness = SIGNED; > + return RPRE; } > +"in MuV" { > + yylval->rvalue.type = REGISTER_ARG; > + yylval->rvalue.reg.type = MODIFIER; > + yylval->rvalue.reg.id = 'u'; > + yylval->rvalue.reg.bit_width = 32; > + yylval->rvalue.bit_width = 32; > + yylval->rvalue.signedness = SIGNED; > + return RREG; } > +"in C"{REG_ID_32}"V" { > + yylval->rvalue.type = REGISTER_ARG; > + yylval->rvalue.reg.type = CONTROL; > + yylval->rvalue.reg.id = yytext[4]; > + yylval->rvalue.reg.bit_width = 32; > + yylval->rvalue.bit_width = 32; > + yylval->rvalue.is_dotnew = false; > + yylval->rvalue.signedness = SIGNED; > + return RREG; } > +"in C"{REG_ID_64}"V" { > + yylval->rvalue.type = REGISTER_ARG; > + yylval->rvalue.reg.type = CONTROL; > + yylval->rvalue.reg.id = yytext[4]; > + yylval->rvalue.reg.bit_width = 64; > + yylval->rvalue.bit_width = 64; > + yylval->rvalue.is_dotnew = false; > + yylval->rvalue.signedness = SIGNED; > + return RREG; } > +"fGEN_TCG_"{INST_NAME}"(" { return FWRAP; } When would one of these need to be parsed? > +"fNEWREG" | > +"fCAST4s" { yylval->cast.bit_width = 32; > + yylval->cast.signedness = SIGNED; > + return CAST; } This doesn't look right - is fNEWREG the same as fCAST4s? > +"fNEWREG_ST" | > +"fIMMEXT" | > +"fMUST_IMMEXT" | > +"fCAST2_2s" | > +"fCAST2_2u" | > +"fCAST4_4s" | > +"fCAST8_8s" | > +"fZE8_16" | > +"fSE8_16" | > +"fZE16_32" | > +"fSE16_32" | > +"fZE32_64" | > +"fPASS" | > +"fECHO" { return IDENTITY; } The fCAST, fZE, and fSE macros are not the same as identity. > +"fCONSTLL" { return CONSTLL; } > +"fCONSTULL" { return CONSTULL; } These can just be converts. > +"fHINTJR(RsV)" { /* Emit no token */ } Put this in the list of IDENTITY above ^ permalink raw reply [flat|nested] 27+ messages in thread
* Re: [PATCH v6 08/12] target/hexagon: import lexer for idef-parser 2021-09-07 18:08 ` Taylor Simpson @ 2021-10-18 11:37 ` Anton Johansson via 2021-10-28 15:49 ` Taylor Simpson 0 siblings, 1 reply; 27+ messages in thread From: Anton Johansson via @ 2021-10-18 11:37 UTC (permalink / raw) To: Taylor Simpson, Alessandro Di Federico, qemu-devel@nongnu.org Cc: Brian Cain, babush@rev.ng, nizzo@rev.ng, richard.henderson@linaro.org, Alessandro Di Federico [-- Attachment #1: Type: text/plain, Size: 855 bytes --] On 9/7/21 18:08, Taylor Simpson wrote: >> +"fNEWREG" | >> +"fCAST4s" { yylval->cast.bit_width = 32; >> + yylval->cast.signedness = SIGNED; >> + return CAST; } > This doesn't look right - is fNEWREG the same as fCAST4s? We followed the definition of fNEWREG in macros.h where it is given as #define fNEWREG(VAL) ((uint32_t) (VAL)) >> +"fCONSTLL" { return CONSTLL; } >> +"fCONSTULL" { return CONSTULL; } > These can just be converts. What is meant by "converts" here? >> +"fHINTJR(RsV)" { /* Emit no token */ } > Put this in the list of IDENTITY above Same as for fNEWREG. We followed the definition in macros.h as #define fHINTJR(TARGET) { /* Not modelled in qemu */ } where it no-ops. -- Anton Johansson, rev.ng Srls. [-- Attachment #2: Type: text/html, Size: 1875 bytes --] ^ permalink raw reply [flat|nested] 27+ messages in thread
* RE: [PATCH v6 08/12] target/hexagon: import lexer for idef-parser 2021-10-18 11:37 ` Anton Johansson via @ 2021-10-28 15:49 ` Taylor Simpson 0 siblings, 0 replies; 27+ messages in thread From: Taylor Simpson @ 2021-10-28 15:49 UTC (permalink / raw) To: anjo@rev.ng, Alessandro Di Federico, qemu-devel@nongnu.org Cc: babush@rev.ng, Brian Cain, richard.henderson@linaro.org, nizzo@rev.ng, Alessandro Di Federico > From: Anton Johansson <anjo@rev.ng> > Sent: Monday, October 18, 2021 6:37 AM > To: Taylor Simpson <tsimpson@quicinc.com>; Alessandro Di Federico <ale.qemu@rev.ng>; qemu-devel@nongnu.org > Cc: Brian Cain <bcain@quicinc.com>; babush@rev.ng; nizzo@rev.ng; richard.henderson@linaro.org; Alessandro Di Federico <ale@rev.ng> > Subject: Re: [PATCH v6 08/12] target/hexagon: import lexer for idef-parser > > On 9/7/21 18:08, Taylor Simpson wrote: > +"fNEWREG" | > +"fCAST4s" { yylval->cast.bit_width = 32; > + yylval->cast.signedness = SIGNED; > + return CAST; } > This doesn't look right - is fNEWREG the same as fCAST4s? > We followed the definition of fNEWREG in macros.h where it is given as > #define fNEWREG(VAL) ((uint32_t) (VAL)) Well, that's different from fCAST4s. In particular, one is signed and one is unsigned. > +"fCONSTLL" { return CONSTLL; } > +"fCONSTULL" { return CONSTULL; } > These can just be converts. > What is meant by "converts" here? Type conversion to int64_t/uint64_t. > +"fHINTJR(RsV)" { /* Emit no token */ } > Put this in the list of IDENTITY above > Same as for fNEWREG. We followed the definition in macros.h as > #define fHINTJR(TARGET) { /* Not modelled in qemu */ } where it no-ops. OK, as long as it is no-op. Thanks, Taylor ^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 09/12] target/hexagon: import parser for idef-parser 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via ` (7 preceding siblings ...) 2021-07-20 12:30 ` [PATCH v6 08/12] target/hexagon: import lexer for idef-parser Alessandro Di Federico via @ 2021-07-20 12:30 ` Alessandro Di Federico via 2021-09-07 18:08 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 10/12] target/hexagon: call idef-parser functions Alessandro Di Federico via ` (2 subsequent siblings) 11 siblings, 1 reply; 27+ messages in thread From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw) To: qemu-devel Cc: tsimpson, bcain, babush, nizzo, richard.henderson, Alessandro Di Federico From: Paolo Montesel <babush@rev.ng> Signed-off-by: Alessandro Di Federico <ale@rev.ng> Signed-off-by: Paolo Montesel <babush@rev.ng> --- target/hexagon/idef-parser/parser-helpers.h | 346 +++ target/hexagon/idef-parser/parser-helpers.c | 2396 +++++++++++++++++ target/hexagon/idef-parser/idef-parser.y | 975 +++++++ target/hexagon/meson.build | 27 +- tests/docker/dockerfiles/alpine.docker | 1 + tests/docker/dockerfiles/centos8.docker | 1 + tests/docker/dockerfiles/debian-amd64.docker | 1 + tests/docker/dockerfiles/debian10.docker | 2 + .../dockerfiles/fedora-i386-cross.docker | 2 + .../dockerfiles/fedora-win32-cross.docker | 2 + .../dockerfiles/fedora-win64-cross.docker | 2 + tests/docker/dockerfiles/fedora.docker | 1 + tests/docker/dockerfiles/opensuse-leap.docker | 1 + tests/docker/dockerfiles/ubuntu.docker | 2 + tests/docker/dockerfiles/ubuntu1804.docker | 1 + 15 files changed, 3759 insertions(+), 1 deletion(-) create mode 100644 target/hexagon/idef-parser/parser-helpers.h create mode 100644 target/hexagon/idef-parser/parser-helpers.c create mode 100644 target/hexagon/idef-parser/idef-parser.y diff --git a/target/hexagon/idef-parser/parser-helpers.h b/target/hexagon/idef-parser/parser-helpers.h new file mode 100644 index 0000000000..2d1fc5ab07 --- /dev/null +++ b/target/hexagon/idef-parser/parser-helpers.h @@ -0,0 +1,346 @@ +/* + * Copyright(c) 2019-2021 rev.ng Srls. All Rights Reserved. + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, see <http://www.gnu.org/licenses/>. + */ + +#ifndef PARSER_HELPERS_H +#define PARSER_HELPERS_H + +#include <assert.h> +#include <inttypes.h> +#include <stdarg.h> +#include <stdbool.h> +#include <stdint.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <unistd.h> + +#include "tcg/tcg-cond.h" + +#include "idef-parser.tab.h" +#include "idef-parser.yy.h" +#include "idef-parser.h" + +/* Decomment this to disable yyasserts */ +/* #define NDEBUG */ + +#define ERR_LINE_CONTEXT 40 + +#define START_COMMENT "/" "*" +#define END_COMMENT "*" "/" + +void yyerror(YYLTYPE *locp, + yyscan_t scanner __attribute__((unused)), + Context *c, + const char *s); + +#ifndef NDEBUG +#define yyassert(context, locp, condition, msg) \ + if (!(condition)) { \ + yyerror(locp, (context)->scanner, (context), (msg)); \ + } +#endif + +bool is_direct_predicate(HexValue *value); + +/* Print functions */ +void str_print(Context *c, YYLTYPE *locp, const char *string); + +void uint8_print(Context *c, YYLTYPE *locp, uint8_t *num); + +void uint64_print(Context *c, YYLTYPE *locp, uint64_t *num); + +void int_print(Context *c, YYLTYPE *locp, int *num); + +void uint_print(Context *c, YYLTYPE *locp, unsigned *num); + +void tmp_print(Context *c, YYLTYPE *locp, HexTmp *tmp); + +void pre_print(Context *c, YYLTYPE *locp, HexPre *pre, bool is_dotnew); + +void reg_compose(Context *c, YYLTYPE *locp, HexReg *reg, char reg_id[5]); + +void reg_print(Context *c, YYLTYPE *locp, HexReg *reg); + +void imm_print(Context *c, YYLTYPE *locp, HexImm *imm); + +void var_print(Context *c, YYLTYPE *locp, HexVar *var); + +void rvalue_out(Context *c, YYLTYPE *locp, void *pointer); + +void out_assert(Context *c, YYLTYPE *locp, void *dummy); + +/* Copy output code buffer into stdout */ +void commit(Context *c); + +#define OUT_IMPL(c, locp, x) \ + _Generic(*x, \ + char: str_print, \ + uint8_t: uint8_print, \ + uint64_t: uint64_print, \ + int: int_print, \ + unsigned: uint_print, \ + HexValue: rvalue_out, \ + default: out_assert \ + )(c, locp, x); + +/* Make a FOREACH macro */ +#define FE_1(c, locp, WHAT, X) WHAT(c, locp, X) +#define FE_2(c, locp, WHAT, X, ...) \ + WHAT(c, locp, X)FE_1(c, locp, WHAT, __VA_ARGS__) +#define FE_3(c, locp, WHAT, X, ...) \ + WHAT(c, locp, X)FE_2(c, locp, WHAT, __VA_ARGS__) +#define FE_4(c, locp, WHAT, X, ...) \ + WHAT(c, locp, X)FE_3(c, locp, WHAT, __VA_ARGS__) +#define FE_5(c, locp, WHAT, X, ...) \ + WHAT(c, locp, X)FE_4(c, locp, WHAT, __VA_ARGS__) +#define FE_6(c, locp, WHAT, X, ...) \ + WHAT(c, locp, X)FE_5(c, locp, WHAT, __VA_ARGS__) +#define FE_7(c, locp, WHAT, X, ...) \ + WHAT(c, locp, X)FE_6(c, locp, WHAT, __VA_ARGS__) +#define FE_8(c, locp, WHAT, X, ...) \ + WHAT(c, locp, X)FE_7(c, locp, WHAT, __VA_ARGS__) +#define FE_9(c, locp, WHAT, X, ...) \ + WHAT(c, locp, X)FE_8(c, locp, WHAT, __VA_ARGS__) +/* repeat as needed */ + +#define GET_MACRO(_1, _2, _3, _4, _5, _6, _7, _8, _9, NAME, ...) NAME + +#define FOR_EACH(c, locp, action, ...) \ + do { \ + GET_MACRO(__VA_ARGS__, \ + FE_9, \ + FE_8, \ + FE_7, \ + FE_6, \ + FE_5, \ + FE_4, \ + FE_3, \ + FE_2, \ + FE_1)(c, locp, action, \ + __VA_ARGS__) \ + } while (0) + +#define OUT(c, locp, ...) FOR_EACH((c), (locp), OUT_IMPL, __VA_ARGS__) + +const char *cmp_swap(Context *c, YYLTYPE *locp, const char *type); + +/* Temporary values creation */ +HexValue gen_tmp(Context *c, YYLTYPE *locp, unsigned bit_width); + +HexValue gen_tmp_value(Context *c, + YYLTYPE *locp, + const char *value, + unsigned bit_width); + +HexValue gen_imm_value(Context *c __attribute__((unused)), + YYLTYPE *locp, + int value, + unsigned bit_width); + +void gen_rvalue_free(Context *c, YYLTYPE *locp, HexValue *rvalue); + +HexValue rvalue_materialize(Context *c, YYLTYPE *locp, HexValue *rvalue); + +HexValue gen_rvalue_extend(Context *c, YYLTYPE *locp, HexValue *rvalue); + +HexValue gen_rvalue_truncate(Context *c, YYLTYPE *locp, HexValue *rvalue); + +int find_variable(Context *c, YYLTYPE *locp, HexValue *varid); + +void gen_varid_allocate(Context *c, + YYLTYPE *locp, + HexValue *varid, + int width, + HexSignedness signedness); + +HexValue gen_bin_cmp(Context *c, + YYLTYPE *locp, + TCGCond type, + HexValue *op1_ptr, + HexValue *op2_ptr); + +/* Code generation functions */ +HexValue gen_bin_op(Context *c, + YYLTYPE *locp, + OpType type, + HexValue *operand1, + HexValue *operand2); + +HexValue gen_cast_op(Context *c, + YYLTYPE *locp, + HexValue *source, + unsigned target_width); + +HexValue gen_extend_op(Context *c, + YYLTYPE *locp, + HexValue *src_width_ptr, + HexValue *dst_width_ptr, + HexValue *value_ptr, + HexSignedness signedness); + +void gen_rdeposit_op(Context *c, + YYLTYPE *locp, + HexValue *dest, + HexValue *value, + HexValue *begin, + HexValue *width); + +void gen_deposit_op(Context *c, + YYLTYPE *locp, + HexValue *dest, + HexValue *value, + HexValue *index, + HexCast *cast); + +HexValue gen_rextract_op(Context *c, + YYLTYPE *locp, + HexValue *source, + int begin, + int width); + +HexValue gen_extract_op(Context *c, + YYLTYPE *locp, + HexValue *source, + HexValue *index, + HexExtract *extract); + +HexValue gen_read_reg(Context *c, YYLTYPE *locp, HexValue *reg); + +void gen_write_reg(Context *c, YYLTYPE *locp, HexValue *reg, HexValue *value); + +void gen_assign(Context *c, + YYLTYPE *locp, + HexValue *dest, + HexValue *value); + +HexValue gen_convround(Context *c, + YYLTYPE *locp, + HexValue *source); + +HexValue gen_round(Context *c, + YYLTYPE *locp, + HexValue *source, + HexValue *position); + +HexValue gen_convround_n(Context *c, + YYLTYPE *locp, + HexValue *source_ptr, + HexValue *bit_pos_ptr); + +/* Circular addressing mode with auto-increment */ +void gen_circ_op(Context *c, + YYLTYPE *locp, + HexValue *addr, + HexValue *increment, + HexValue *modifier); + +HexValue gen_locnt_op(Context *c, YYLTYPE *locp, HexValue *source); + +HexValue gen_ctpop_op(Context *c, YYLTYPE *locp, HexValue *source); + +HexValue gen_fbrev_4(Context *c, YYLTYPE *locp, HexValue *source); + +HexValue gen_fbrev_8(Context *c, YYLTYPE *locp, HexValue *source); + +HexValue gen_rotl(Context *c, YYLTYPE *locp, HexValue *source, HexValue *n); + +HexValue gen_deinterleave(Context *c, YYLTYPE *locp, HexValue *mixed); + +HexValue gen_interleave(Context *c, + YYLTYPE *locp, + HexValue *odd, + HexValue *even); + +HexValue gen_carry_from_add(Context *c, + YYLTYPE *locp, + HexValue *op1, + HexValue *op2, + HexValue *op3); + +void gen_inst(Context *c, GString *iname); + +void gen_inst_args(Context *c, YYLTYPE *locp); + +void gen_inst_code(Context *c, YYLTYPE *locp); + +void gen_pre_assign(Context *c, YYLTYPE *locp, HexValue *lp, HexValue *rp); + +void gen_load(Context *c, YYLTYPE *locp, HexValue *size, + HexSignedness signedness, HexValue *ea, HexValue *dst); + +void gen_store(Context *c, YYLTYPE *locp, HexValue *size, HexValue *ea, + HexValue *src); + +void gen_sethalf(Context *c, YYLTYPE *locp, HexCast *sh, HexValue *n, + HexValue *dst, HexValue *val); + +void gen_setbits(Context *c, YYLTYPE *locp, HexValue *hi, HexValue *lo, + HexValue *dst, HexValue *val); + +int gen_if_cond(Context *c, YYLTYPE *locp, HexValue *cond); + +int gen_if_else(Context *c, YYLTYPE *locp, int index); + +HexValue gen_rvalue_pre(Context *c, YYLTYPE *locp, HexValue *pre); + +HexValue gen_rvalue_var(Context *c, YYLTYPE *locp, HexValue *var); + +HexValue gen_rvalue_mpy(Context *c, YYLTYPE *locp, HexMpy *mpy, HexValue *a, + HexValue *b); + +HexValue gen_rvalue_not(Context *c, YYLTYPE *locp, HexValue *v); + +HexValue gen_rvalue_notl(Context *c, YYLTYPE *locp, HexValue *v); + +HexValue gen_rvalue_sat(Context *c, YYLTYPE *locp, HexSat *sat, HexValue *n, + HexValue *v); + +HexValue gen_rvalue_ternary(Context *c, YYLTYPE *locp, HexValue *cond, + HexValue *t, HexValue *e); + +HexValue gen_rvalue_fscr(Context *c, YYLTYPE *locp, HexValue *v); + +HexValue gen_rvalue_abs(Context *c, YYLTYPE *locp, HexValue *v); + +HexValue gen_rvalue_neg(Context *c, YYLTYPE *locp, HexValue *v); + +HexValue gen_rvalue_brev(Context *c, YYLTYPE *locp, HexValue *v); + +void gen_set_overflow(Context *c, YYLTYPE *locp, HexValue *vp); + +HexValue gen_rvalue_ternary(Context *c, YYLTYPE *locp, HexValue *cond, + HexValue *t, HexValue *e); + +const char *cond_to_str(TCGCond cond); + +void emit_header(Context *c); + +void emit_arg(Context *c, YYLTYPE *locp, HexValue *arg); + +void emit_footer(Context *c); + +void track_string(Context *c, GString *s); + +void free_variables(Context *c, YYLTYPE *locp); + +void free_instruction(Context *c); + +void assert_signedness(Context *c, + YYLTYPE *locp, + HexSignedness signedness); + +#endif /* PARSER_HELPERS_h */ diff --git a/target/hexagon/idef-parser/parser-helpers.c b/target/hexagon/idef-parser/parser-helpers.c new file mode 100644 index 0000000000..66cad733bb --- /dev/null +++ b/target/hexagon/idef-parser/parser-helpers.c @@ -0,0 +1,2396 @@ +/* + * Copyright(c) 2019-2021 rev.ng Srls. All Rights Reserved. + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, see <http://www.gnu.org/licenses/>. + */ + +#include <assert.h> +#include <inttypes.h> +#include <stdarg.h> +#include <stdbool.h> +#include <stdint.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <unistd.h> + +#include "idef-parser.h" +#include "parser-helpers.h" +#include "idef-parser.tab.h" +#include "idef-parser.yy.h" + +void yyerror(YYLTYPE *locp, + yyscan_t scanner __attribute__((unused)), + Context *c, + const char *s) +{ + const char *code_ptr = c->input_buffer; + + fprintf(stderr, "WARNING (%s): '%s'\n", c->inst.name->str, s); + + fprintf(stderr, "Problematic range: "); + for (int i = locp->first_column; i < locp->last_column; i++) { + if (code_ptr[i] != '\n') { + fprintf(stderr, "%c", code_ptr[i]); + } + } + fprintf(stderr, "\n"); + + for (int i = 0; + i < 80 && + code_ptr[locp->first_column - 10 + i] != '\0' && + code_ptr[locp->first_column - 10 + i] != '\n'; + i++) { + fprintf(stderr, "%c", code_ptr[locp->first_column - 10 + i]); + } + fprintf(stderr, "\n"); + for (int i = 0; i < 9; i++) { + fprintf(stderr, " "); + } + fprintf(stderr, "^"); + for (int i = 0; i < (locp->last_column - locp->first_column) - 1; i++) { + fprintf(stderr, "~"); + } + fprintf(stderr, "\n"); + c->inst.error_count++; +} + +bool is_direct_predicate(HexValue *value) +{ + return value->pre.id >= '0' && value->pre.id <= '3'; + +} + +bool is_inside_ternary(Context *c) +{ + return c->ternary->len > 0; +} + +/* Print functions */ +void str_print(Context *c, YYLTYPE *locp, const char *string) +{ + (void) locp; + EMIT(c, "%s", string); +} + +void uint8_print(Context *c, YYLTYPE *locp, uint8_t *num) +{ + (void) locp; + EMIT(c, "%u", *num); +} + +void uint64_print(Context *c, YYLTYPE *locp, uint64_t *num) +{ + (void) locp; + EMIT(c, "%" PRIu64, *num); +} + +void int_print(Context *c, YYLTYPE *locp, int *num) +{ + (void) locp; + EMIT(c, "%d", *num); +} + +void uint_print(Context *c, YYLTYPE *locp, unsigned *num) +{ + (void) locp; + EMIT(c, "%u", *num); +} + +void tmp_print(Context *c, YYLTYPE *locp, HexTmp *tmp) +{ + (void) locp; + EMIT(c, "tmp_%d", tmp->index); +} + +void pre_print(Context *c, YYLTYPE *locp, HexPre *pre, bool is_dotnew) +{ + (void) locp; + char suffix = is_dotnew ? 'N' : 'V'; + EMIT(c, "P%c%c", pre->id, suffix); +} + +void reg_compose(Context *c, YYLTYPE *locp, HexReg *reg, char reg_id[5]) +{ + switch (reg->type) { + case GENERAL_PURPOSE: + reg_id[0] = 'R'; + break; + case CONTROL: + reg_id[0] = 'C'; + break; + case MODIFIER: + reg_id[0] = 'M'; + break; + case DOTNEW: + reg_id[0] = 'N'; + reg_id[1] = reg->id; + reg_id[2] = 'N'; + return; + } + switch (reg->bit_width) { + case 32: + reg_id[1] = reg->id; + reg_id[2] = 'V'; + break; + case 64: + reg_id[1] = reg->id; + reg_id[2] = reg->id; + reg_id[3] = 'V'; + break; + default: + yyassert(c, locp, false, "Unhandled register bit width!\n"); + } +} + +static void reg_arg_print(Context *c, YYLTYPE *locp, HexReg *reg) +{ + char reg_id[5] = { 0 }; + reg_compose(c, locp, reg, reg_id); + EMIT(c, "%s", reg_id); +} + +void reg_print(Context *c, YYLTYPE *locp, HexReg *reg) +{ + (void) locp; + EMIT(c, "hex_gpr[%u]", reg->id); +} + +void imm_print(Context *c, YYLTYPE *locp, HexImm *imm) +{ + switch (imm->type) { + case I: + EMIT(c, "i"); + break; + case VARIABLE: + EMIT(c, "%ciV", imm->id); + break; + case VALUE: + EMIT(c, "((int64_t) %" PRIu64 "ULL)", (int64_t) imm->value); + break; + case QEMU_TMP: + EMIT(c, "qemu_tmp_%" PRIu64, imm->index); + break; + case IMM_PC: + EMIT(c, "ctx->base.pc_next"); + break; + case IMM_NPC: + EMIT(c, "ctx->npc"); + break; + case IMM_CONSTEXT: + EMIT(c, "insn->extension_valid"); + break; + default: + yyassert(c, locp, false, "Cannot print this expression!"); + } +} + +void var_print(Context *c, YYLTYPE *locp, HexVar *var) +{ + (void) locp; + EMIT(c, "%s", var->name->str); +} + +void rvalue_out(Context *c, YYLTYPE *locp, void *pointer) +{ + HexValue *rvalue = (HexValue *) pointer; + switch (rvalue->type) { + case REGISTER: + reg_print(c, locp, &rvalue->reg); + break; + case REGISTER_ARG: + reg_arg_print(c, locp, &rvalue->reg); + break; + case TEMP: + tmp_print(c, locp, &rvalue->tmp); + break; + case IMMEDIATE: + imm_print(c, locp, &rvalue->imm); + break; + case VARID: + var_print(c, locp, &rvalue->var); + break; + case PREDICATE: + pre_print(c, locp, &rvalue->pre, rvalue->is_dotnew); + break; + default: + yyassert(c, locp, false, "Cannot print this expression!"); + } +} + +void out_assert(Context *c, YYLTYPE *locp, + void *dummy __attribute__((unused))) { + abort(); + yyassert(c, locp, false, "Unhandled print type!"); +} + +/* Copy output code buffer */ +void commit(Context *c) +{ + /* Emit instruction pseudocode */ + EMIT_SIG(c, "\n" START_COMMENT " "); + for (char *x = c->inst.code_begin; x < c->inst.code_end; x++) { + EMIT_SIG(c, "%c", *x); + } + EMIT_SIG(c, " " END_COMMENT "\n"); + + /* Commit instruction code to output file */ + fwrite(c->signature_str->str, sizeof(char), c->signature_str->len, + c->output_file); + fwrite(c->header_str->str, sizeof(char), c->header_str->len, + c->output_file); + fwrite(c->out_str->str, sizeof(char), c->out_str->len, + c->output_file); + + fwrite(c->signature_str->str, sizeof(char), c->signature_str->len, + c->defines_file); + fprintf(c->defines_file, ";\n"); +} + +static HexValue get_ternary_cond(Context *c, YYLTYPE *locp) +{ + yyassert(c, locp, is_inside_ternary(c), "unexisting condition"); + Ternary *t = &g_array_index(c->ternary, Ternary, 0); + HexValue cond = t->cond; + if (t->state == IN_RIGHT) { + cond = gen_rvalue_notl(c, locp, &cond); + } + for (unsigned i = 1; i < c->ternary->len; ++i) { + Ternary *right = &g_array_index(c->ternary, Ternary, i); + HexValue other = right->cond; + /* Invert condition if we are on the right side */ + if (right->state == IN_RIGHT) { + other = gen_rvalue_notl(c, locp, &other); + } + cond = gen_bin_op(c, locp, ANDL_OP, &cond, &other); + } + return cond; +} + +/* Temporary values creation */ +HexValue gen_tmp(Context *c, YYLTYPE *locp, unsigned bit_width) +{ + HexValue rvalue; + memset(&rvalue, 0, sizeof(HexValue)); + rvalue.type = TEMP; + assert(bit_width == 32 || bit_width == 64); + rvalue.bit_width = bit_width; + rvalue.signedness = SIGNED; + rvalue.is_dotnew = false; + rvalue.is_manual = false; + rvalue.tmp.index = c->inst.tmp_count; + OUT(c, locp, "TCGv_i", &bit_width, " tmp_", &c->inst.tmp_count, + " = tcg_temp_new_i", &bit_width, "();\n"); + c->inst.tmp_count++; + return rvalue; +} + +HexValue gen_tmp_value(Context *c, + YYLTYPE *locp, + const char *value, + unsigned bit_width) +{ + HexValue rvalue; + memset(&rvalue, 0, sizeof(HexValue)); + rvalue.type = TEMP; + rvalue.bit_width = bit_width; + rvalue.signedness = SIGNED; + rvalue.is_dotnew = false; + rvalue.is_manual = false; + rvalue.tmp.index = c->inst.tmp_count; + OUT(c, locp, "TCGv_i", &bit_width, " tmp_", &c->inst.tmp_count, + " = tcg_const_i", &bit_width, "(", value, ");\n"); + c->inst.tmp_count++; + return rvalue; +} + +static HexValue gen_tmp_value_from_imm(Context *c, + YYLTYPE *locp, + HexValue *value) +{ + assert(value->type == IMMEDIATE); + HexValue rvalue; + memset(&rvalue, 0, sizeof(HexValue)); + rvalue.type = TEMP; + rvalue.bit_width = value->bit_width; + rvalue.signedness = value->signedness; + rvalue.is_dotnew = false; + rvalue.is_manual = false; + rvalue.tmp.index = c->inst.tmp_count; + OUT(c, locp, "TCGv_i", &rvalue.bit_width, " tmp_", &c->inst.tmp_count); + OUT(c, locp, " = tcg_const_i", &rvalue.bit_width, + "((int", &rvalue.bit_width, "_t) (", value, "));\n"); + c->inst.tmp_count++; + return rvalue; +} + +HexValue gen_imm_value(Context *c __attribute__((unused)), + YYLTYPE *locp, + int value, + unsigned bit_width) +{ + (void) locp; + HexValue rvalue; + memset(&rvalue, 0, sizeof(HexValue)); + rvalue.type = IMMEDIATE; + rvalue.bit_width = bit_width; + rvalue.signedness = SIGNED; + rvalue.is_dotnew = false; + rvalue.is_manual = false; + rvalue.imm.type = VALUE; + rvalue.imm.value = value; + return rvalue; +} + +void gen_rvalue_free(Context *c, YYLTYPE *locp, HexValue *rvalue) +{ + if (rvalue->type == TEMP && !rvalue->is_manual) { + const char *bit_suffix = (rvalue->bit_width == 64) ? "i64" : "i32"; + OUT(c, locp, "tcg_temp_free_", bit_suffix, "(", rvalue, ");\n"); + } +} + +static void gen_rvalue_free_manual(Context *c, YYLTYPE *locp, HexValue *rvalue) +{ + rvalue->is_manual = false; + gen_rvalue_free(c, locp, rvalue); +} + +HexValue rvalue_materialize(Context *c, YYLTYPE *locp, HexValue *rvalue) +{ + if (rvalue->type == IMMEDIATE) { + HexValue res = gen_tmp_value_from_imm(c, locp, rvalue); + gen_rvalue_free(c, locp, rvalue); + return res; + } + return *rvalue; +} + +HexValue gen_rvalue_extend(Context *c, YYLTYPE *locp, HexValue *rvalue) +{ + if (rvalue->type == IMMEDIATE) { + HexValue res = *rvalue; + res.bit_width = 64; + return res; + } else { + if (rvalue->bit_width == 32) { + HexValue res = gen_tmp(c, locp, 64); + assert_signedness(c, locp, rvalue->signedness); + bool is_unsigned = (rvalue->signedness == UNSIGNED); + const char *sign_suffix = is_unsigned ? "u" : ""; + OUT(c, locp, "tcg_gen_ext", sign_suffix, + "_i32_i64(", &res, ", ", rvalue, ");\n"); + gen_rvalue_free(c, locp, rvalue); + return res; + } + } + return *rvalue; +} + +HexValue gen_rvalue_truncate(Context *c, YYLTYPE *locp, HexValue *rvalue) +{ + if (rvalue->type == IMMEDIATE) { + HexValue res = *rvalue; + res.bit_width = 32; + return res; + } else { + if (rvalue->bit_width == 64) { + HexValue res = gen_tmp(c, locp, 32); + OUT(c, locp, "tcg_gen_trunc_i64_tl(", &res, ", ", rvalue, ");\n"); + gen_rvalue_free(c, locp, rvalue); + return res; + } + } + return *rvalue; +} + +int find_variable(Context *c, YYLTYPE *locp, HexValue *varid) +{ + (void) locp; + for (unsigned i = 0; i < c->inst.allocated->len; i++) { + Var *curr = &g_array_index(c->inst.allocated, Var, i); + if (g_string_equal(varid->var.name, curr->name)) { + return i; + } + } + return -1; +} + +void gen_varid_allocate(Context *c, + YYLTYPE *locp, + HexValue *varid, + int width, + HexSignedness signedness) +{ + assert_signedness(c, locp, signedness); + varid->bit_width = width; + const char *bit_suffix = width == 64 ? "64" : "32"; + int index = find_variable(c, locp, varid); + bool found = index != -1; + if (found) { + Var *other = &g_array_index(c->inst.allocated, Var, index); + varid->var.name = other->name; + varid->bit_width = other->bit_width; + varid->signedness = other->signedness; + } else { + EMIT_HEAD(c, "TCGv_i%s %s", bit_suffix, varid->var.name->str); + EMIT_HEAD(c, " = tcg_temp_local_new_i%s();\n", bit_suffix); + Var new_var = { + .name = varid->var.name, + .bit_width = width, + .signedness = signedness, + }; + g_array_append_val(c->inst.allocated, new_var); + } +} + +enum OpTypes { + IMM_IMM = 0, + IMM_REG = 1, + REG_IMM = 2, + REG_REG = 3, +}; + +void gen_operands_extend(Context *c, + YYLTYPE *locp, + HexValue *op1, + HexValue *op2) { + enum OpTypes op_types = (op1->type != IMMEDIATE) << 1 + | (op2->type != IMMEDIATE); + + switch (op_types) { + case IMM_IMM: + break; + case IMM_REG: + *op2 = gen_rvalue_extend(c, locp, op2); + break; + case REG_IMM: + *op1 = gen_rvalue_extend(c, locp, op1); + break; + case REG_REG: + *op1 = gen_rvalue_extend(c, locp, op1); + *op2 = gen_rvalue_extend(c, locp, op2); + break; + } +} + +HexValue gen_bin_cmp(Context *c, + YYLTYPE *locp, + TCGCond type, + HexValue *op1_ptr, + HexValue *op2_ptr) +{ + HexValue op1 = *op1_ptr; + HexValue op2 = *op2_ptr; + enum OpTypes op_types = (op1.type != IMMEDIATE) << 1 + | (op2.type != IMMEDIATE); + + bool op_is64bit = op1.bit_width == 64 || op2.bit_width == 64; + + /* Extend to 64-bits, if required */ + if (op_is64bit) { + gen_operands_extend(c, locp, &op1, &op2); + } + + const char *bit_suffix = op_is64bit ? "i64" : "i32"; + int bit_width = (op_is64bit) ? 64 : 32; + HexValue res = gen_tmp(c, locp, bit_width); + + switch (op_types) { + case IMM_IMM: + OUT(c, locp, "tcg_gen_movi_", bit_suffix, + "(", &res, ", ", &op1, " == ", &op2, ");\n"); + break; + case IMM_REG: + { + HexValue swp = op2; + op2 = op1; + op1 = swp; + /* Swap comparison direction */ + type = tcg_swap_cond(type); + } + /* fallthrough */ + case REG_IMM: + OUT(c, locp, "tcg_gen_setcondi_", bit_suffix, "("); + OUT(c, locp, cond_to_str(type), ", ", &res, ", ", &op1, ", ", &op2, + ");\n"); + break; + case REG_REG: + OUT(c, locp, "tcg_gen_setcond_", bit_suffix, "("); + OUT(c, locp, cond_to_str(type), ", ", &res, ", ", &op1, ", ", &op2, + ");\n"); + break; + default: + fprintf(stderr, "Error in evalutating immediateness!"); + abort(); + } + + /* Free operands */ + gen_rvalue_free(c, locp, &op1); + gen_rvalue_free(c, locp, &op2); + + return res; +} + +static void gen_simple_op(Context *c, YYLTYPE *locp, unsigned bit_width, + const char *bit_suffix, HexValue *res, + enum OpTypes op_types, HexValue *op1, + HexValue *op2, + const char *imm_imm, + const char *imm_reg, + const char *reg_imm, + const char *reg_reg + ) { + switch (op_types) { + case IMM_IMM: + OUT(c, locp, "int", &bit_width, "_t ", res, + " = ", op1, imm_imm, op2, ";\n"); + break; + case IMM_REG: + OUT(c, locp, imm_reg, bit_suffix, + "(", res, ", ", op2, ", ", op1, ");\n"); + break; + case REG_IMM: + OUT(c, locp, reg_imm, bit_suffix, + "(", res, ", ", op1, ", ", op2, ");\n"); + break; + case REG_REG: + OUT(c, locp, reg_reg, bit_suffix, + "(", res, ", ", op1, ", ", op2, ");\n"); + break; + } + gen_rvalue_free(c, locp, op1); + gen_rvalue_free(c, locp, op2); +} + +static void gen_sub_op(Context *c, YYLTYPE *locp, unsigned bit_width, + const char *bit_suffix, HexValue *res, + enum OpTypes op_types, HexValue *op1, HexValue *op2) +{ + switch (op_types) { + case IMM_IMM: + OUT(c, locp, "int", &bit_width, "_t ", res, + " = ", op1, " - ", op2, ";\n"); + break; + case IMM_REG: + OUT(c, locp, "tcg_gen_subfi_", bit_suffix, + "(", res, ", ", op1, ", ", op2, ");\n"); + break; + case REG_IMM: + OUT(c, locp, "tcg_gen_subi_", bit_suffix, + "(", res, ", ", op1, ", ", op2, ");\n"); + break; + case REG_REG: + OUT(c, locp, "tcg_gen_sub_", bit_suffix, + "(", res, ", ", op1, ", ", op2, ");\n"); + break; + } + gen_rvalue_free(c, locp, op1); + gen_rvalue_free(c, locp, op2); +} + +static void gen_asl_op(Context *c, YYLTYPE *locp, unsigned bit_width, + bool op_is64bit, const char *bit_suffix, HexValue *res, + enum OpTypes op_types, HexValue *op1_ptr, + HexValue *op2_ptr) +{ + HexValue op1 = *op1_ptr; + HexValue op2 = *op2_ptr; + switch (op_types) { + case IMM_IMM: + OUT(c, locp, "int", &bit_width, "_t ", res, + " = ", &op1, " << ", &op2, ";\n"); + break; + case REG_IMM: + { + OUT(c, locp, "if (", &op2, " >= ", &bit_width, ") {\n"); + OUT(c, locp, "tcg_gen_movi_", bit_suffix, "(", res, ", 0);\n"); + OUT(c, locp, "} else {\n"); + OUT(c, locp, "tcg_gen_shli_", bit_suffix, + "(", res, ", ", &op1, ", ", &op2, ");\n"); + OUT(c, locp, "}\n"); + } + break; + case IMM_REG: + op1.bit_width = bit_width; + op1 = rvalue_materialize(c, locp, &op1); + /* Fallthrough */ + case REG_REG: + OUT(c, locp, "tcg_gen_shl_", bit_suffix, + "(", res, ", ", &op1, ", ", &op2, ");\n"); + break; + } + if (op_types == IMM_REG || op_types == REG_REG) { + /* Handle left shift by 64/32 which hexagon-sim expects to clear out */ + /* register */ + HexValue edge = gen_imm_value(c, locp, bit_width, bit_width); + edge = rvalue_materialize(c, locp, &edge); + HexValue zero = gen_tmp_value(c, locp, "0", bit_width); + if (op_is64bit) { + op2 = gen_rvalue_extend(c, locp, &op2); + } + op1 = rvalue_materialize(c, locp, &op1); + op2 = rvalue_materialize(c, locp, &op2); + OUT(c, locp, "tcg_gen_movcond_i", &bit_width); + OUT(c, locp, "(TCG_COND_GEU, ", res, ", ", &op2, ", ", &edge); + OUT(c, locp, ", ", &zero, ", ", res, ");\n"); + gen_rvalue_free(c, locp, &edge); + gen_rvalue_free(c, locp, &zero); + } + gen_rvalue_free(c, locp, &op1); + gen_rvalue_free(c, locp, &op2); +} + +static void gen_asr_op(Context *c, YYLTYPE *locp, unsigned bit_width, + bool op_is64bit, const char *bit_suffix, HexValue *res, + enum OpTypes op_types, HexValue *op1_ptr, + HexValue *op2_ptr) +{ + HexValue op1 = *op1_ptr; + HexValue op2 = *op2_ptr; + switch (op_types) { + case IMM_IMM: + OUT(c, locp, "int", &bit_width, "_t ", res, + " = ", &op1, " >> ", &op2, ";\n"); + break; + case REG_IMM: + OUT(c, locp, "{\nint", &bit_width, "_t shift = ", &op2, ";\n"); + OUT(c, locp, "if (", &op2, " >= ", &bit_width, ") {\n"); + OUT(c, locp, " shift = ", &bit_width, " - 1;\n"); + OUT(c, locp, "}\n"); + OUT(c, locp, "tcg_gen_sari_", bit_suffix, + "(", res, ", ", &op1, ", shift);\n}\n"); + break; + case IMM_REG: + op1 = rvalue_materialize(c, locp, &op1); + /* Fallthrough */ + case REG_REG: + OUT(c, locp, "tcg_gen_sar_", bit_suffix, + "(", res, ", ", &op1, ", ", &op2, ");\n"); + break; + } + if (op_types == IMM_REG || op_types == REG_REG) { + /* Handle right shift by values >= bit_width */ + HexValue edge = gen_imm_value(c, locp, bit_width, bit_width); + edge = rvalue_materialize(c, locp, &edge); + HexValue tmp = gen_tmp(c, locp, bit_width); + HexValue zero = gen_tmp_value(c, locp, "0", bit_width); + if (op_is64bit) { + op2 = gen_rvalue_extend(c, locp, &op2); + } + op1 = rvalue_materialize(c, locp, &op1); + op2 = rvalue_materialize(c, locp, &op2); + + const char *offset = op_is64bit ? "63" : "31"; + OUT(c, locp, "tcg_gen_extract_", bit_suffix, "(", + &tmp, ", ", &op1, ", ", offset, ", 1);\n"); + OUT(c, locp, "tcg_gen_sub_", bit_suffix, "(", + &tmp, ", ", &zero, ", ", &tmp, ");\n"); + OUT(c, locp, "tcg_gen_movcond_i", &bit_width); + OUT(c, locp, "(TCG_COND_GEU, ", res, ", ", &op2, ", ", &edge); + OUT(c, locp, ", ", &tmp, ", ", res, ");\n"); + gen_rvalue_free(c, locp, &edge); + gen_rvalue_free(c, locp, &tmp); + gen_rvalue_free(c, locp, &zero); + } + gen_rvalue_free(c, locp, &op1); + gen_rvalue_free(c, locp, &op2); +} + +static void gen_lsr_op(Context *c, YYLTYPE *locp, unsigned bit_width, + bool op_is64bit, const char *bit_suffix, HexValue *res, + enum OpTypes op_types, HexValue *op1_ptr, + HexValue *op2_ptr) +{ + HexValue op1 = *op1_ptr; + HexValue op2 = *op2_ptr; + switch (op_types) { + case IMM_IMM: + OUT(c, locp, "int", &bit_width, "_t ", + res, " = ", &op1, " >> ", &op2, ";\n"); + break; + case REG_IMM: + OUT(c, locp, "if (", &op2, " >= ", &bit_width, ") {\n"); + OUT(c, locp, "tcg_gen_movi_", bit_suffix, "(", res, ", 0);\n"); + OUT(c, locp, "} else {\n"); + OUT(c, locp, "tcg_gen_shri_", bit_suffix, + "(", res, ", ", &op1, ", ", &op2, ");\n"); + OUT(c, locp, "}\n"); + break; + case IMM_REG: + op1 = rvalue_materialize(c, locp, &op1); + /* Fallthrough */ + case REG_REG: + OUT(c, locp, "tcg_gen_shr_", bit_suffix, + "(", res, ", ", &op1, ", ", &op2, ");\n"); + break; + } + if (op_types == IMM_REG || op_types == REG_REG) { + /* Handle right shift by values >= bit_width */ + HexValue edge = gen_imm_value(c, locp, bit_width, bit_width); + edge = rvalue_materialize(c, locp, &edge); + HexValue zero = gen_tmp_value(c, locp, "0", bit_width); + if (op_is64bit) { + op2 = gen_rvalue_extend(c, locp, &op2); + } + op1 = rvalue_materialize(c, locp, &op1); + op2 = rvalue_materialize(c, locp, &op2); + OUT(c, locp, "tcg_gen_movcond_i", &bit_width); + OUT(c, locp, "(TCG_COND_GEU, ", res, ", ", &op2, ", ", &edge); + OUT(c, locp, ", ", &zero, ", ", res, ");\n"); + gen_rvalue_free(c, locp, &edge); + gen_rvalue_free(c, locp, &zero); + } + gen_rvalue_free(c, locp, &op1); + gen_rvalue_free(c, locp, &op2); +} + +static void gen_andl_op(Context *c, YYLTYPE *locp, unsigned bit_width, + const char *bit_suffix, HexValue *res, + enum OpTypes op_types, HexValue *op1, HexValue *op2) +{ + HexValue zero, tmp1, tmp2; + memset(&zero, 0, sizeof(HexValue)); + memset(&tmp1, 0, sizeof(HexValue)); + memset(&tmp2, 0, sizeof(HexValue)); + switch (op_types) { + case IMM_IMM: + OUT(c, locp, "int", &bit_width, "_t ", + res, " = ", op1, " && ", op2, ";\n"); + break; + case IMM_REG: + zero = gen_tmp_value(c, locp, "0", 32); + tmp2 = gen_bin_cmp(c, locp, TCG_COND_NE, op2, &zero); + OUT(c, locp, "tcg_gen_andi_", bit_suffix, + "(", res, ", ", op1, " != 0 , ", &tmp2, ");\n"); + gen_rvalue_free(c, locp, &tmp2); + break; + case REG_IMM: + zero = gen_tmp_value(c, locp, "0", 32); + tmp1 = gen_bin_cmp(c, locp, TCG_COND_NE, op1, &zero); + OUT(c, locp, "tcg_gen_andi_", bit_suffix, + "(", res, ", ", &tmp1, ", ", op2, " != 0);\n"); + gen_rvalue_free(c, locp, &tmp1); + break; + case REG_REG: + zero = gen_tmp_value(c, locp, "0", 32); + zero.is_manual = true; + tmp1 = gen_bin_cmp(c, locp, TCG_COND_NE, op1, &zero); + tmp2 = gen_bin_cmp(c, locp, TCG_COND_NE, op2, &zero); + OUT(c, locp, "tcg_gen_and_", bit_suffix, + "(", res, ", ", &tmp1, ", ", &tmp2, ");\n"); + gen_rvalue_free_manual(c, locp, &zero); + gen_rvalue_free(c, locp, &tmp1); + gen_rvalue_free(c, locp, &tmp2); + break; + } +} + +static void gen_minmax_op(Context *c, YYLTYPE *locp, unsigned bit_width, + HexValue *res, enum OpTypes op_types, + HexValue *op1_ptr, HexValue *op2_ptr, bool minmax) +{ + HexValue op1 = *op1_ptr; + HexValue op2 = *op2_ptr; + const char *mm; + assert_signedness(c, locp, res->signedness); + bool is_unsigned = res->signedness == UNSIGNED; + if (minmax) { + /* Max */ + mm = is_unsigned ? "tcg_gen_umax" : "tcg_gen_smax"; + } else { + /* Min */ + mm = is_unsigned ? "tcg_gen_umin" : "tcg_gen_smin"; + } + switch (op_types) { + case IMM_IMM: + { + const char *s = minmax ? " <= " : " >= "; + OUT(c, locp, "int", &bit_width, "_t ", res, " = (", &op1, s); + OUT(c, locp, &op2, ") ? ", &op1, " : ", &op2, ";\n"); + } + break; + case IMM_REG: + op1.bit_width = bit_width; + op1 = rvalue_materialize(c, locp, &op1); + OUT(c, locp, mm, "_i", &bit_width, "("); + OUT(c, locp, res, ", ", &op1, ", ", &op2, ");\n"); + break; + case REG_IMM: + op2.bit_width = bit_width; + op2 = rvalue_materialize(c, locp, &op2); + /* Fallthrough */ + case REG_REG: + OUT(c, locp, mm, "_i", &bit_width, "("); + OUT(c, locp, res, ", ", &op1, ", ", &op2, ");\n"); + break; + } + gen_rvalue_free(c, locp, &op1); + gen_rvalue_free(c, locp, &op2); +} + +static void gen_mod_op(Context *c, YYLTYPE *locp, HexValue *res, + enum OpTypes op_types, HexValue *op1, HexValue *op2) +{ + switch (op_types) { + case IMM_IMM: + OUT(c, locp, "int64_t ", res, " = ", op1, " % ", op2, ";\n"); + break; + case IMM_REG: + case REG_IMM: + case REG_REG: + OUT(c, locp, "gen_helper_mod(", + res, ", ", op1, ", ", op2, ");\n"); + break; + } + gen_rvalue_free(c, locp, op1); + gen_rvalue_free(c, locp, op2); +} + +/* Code generation functions */ +HexValue gen_bin_op(Context *c, + YYLTYPE *locp, + OpType type, + HexValue *operand1, + HexValue *operand2) +{ + /* Replicate operands to avoid side effects */ + HexValue op1 = *operand1; + HexValue op2 = *operand2; + + /* Enforce variables' size */ + if (op1.type == VARID) { + int index = find_variable(c, locp, &op1); + yyassert(c, locp, c->inst.allocated->len > 0, + "Variable in bin_op must exist!\n"); + op1.bit_width = g_array_index(c->inst.allocated, Var, index).bit_width; + } + if (op2.type == VARID) { + int index = find_variable(c, locp, &op2); + yyassert(c, locp, c->inst.allocated->len > 0, + "Variable in bin_op must exist!\n"); + op2.bit_width = g_array_index(c->inst.allocated, Var, index).bit_width; + } + + enum OpTypes op_types = (op1.type != IMMEDIATE) << 1 + | (op2.type != IMMEDIATE); + + bool op_is64bit = op1.bit_width == 64 || op2.bit_width == 64; + + /* Shift greater than 32 are 64 bits wide */ + if (type == ASL_OP && op2.type == IMMEDIATE && + op2.imm.type == VALUE && op2.imm.value >= 32) + op_is64bit = true; + + /* Extend to 64-bits, if required */ + if (op_is64bit) { + gen_operands_extend(c, locp, &op1, &op2); + } + + HexValue res; + memset(&res, 0, sizeof(HexValue)); + const char *bit_suffix = op_is64bit ? "i64" : "i32"; + int bit_width = (op_is64bit) ? 64 : 32; + if (op_types != IMM_IMM) { + res = gen_tmp(c, locp, bit_width); + } else { + res.type = IMMEDIATE; + res.is_dotnew = false; + res.is_manual = false; + res.imm.type = QEMU_TMP; + res.imm.index = c->inst.qemu_tmp_count; + res.bit_width = bit_width; + } + /* Handle signedness, if both unsigned -> result is unsigned, else signed */ + assert_signedness(c, locp, op1.signedness); + assert_signedness(c, locp, op2.signedness); + res.signedness = (op1.signedness == UNSIGNED + && op2.signedness == UNSIGNED) ? UNSIGNED : SIGNED; + + switch (type) { + case ADD_OP: + gen_simple_op(c, locp, bit_width, bit_suffix, &res, + op_types, &op1, &op2, + " + ", + "tcg_gen_addi_", + "tcg_gen_addi_", + "tcg_gen_add_"); + break; + case SUB_OP: + gen_sub_op(c, locp, bit_width, bit_suffix, &res, op_types, &op1, &op2); + break; + case MUL_OP: + gen_simple_op(c, locp, bit_width, bit_suffix, &res, + op_types, &op1, &op2, + " * ", + "tcg_gen_muli_", + "tcg_gen_muli_", + "tcg_gen_mul_"); + break; + case ASL_OP: + gen_asl_op(c, locp, bit_width, op_is64bit, bit_suffix, &res, op_types, + &op1, &op2); + break; + case ASR_OP: + gen_asr_op(c, locp, bit_width, op_is64bit, bit_suffix, &res, op_types, + &op1, &op2); + break; + case LSR_OP: + gen_lsr_op(c, locp, bit_width, op_is64bit, bit_suffix, &res, op_types, + &op1, &op2); + break; + case ANDB_OP: + gen_simple_op(c, locp, bit_width, bit_suffix, &res, + op_types, &op1, &op2, + " & ", + "tcg_gen_andi_", + "tcg_gen_andi_", + "tcg_gen_and_"); + break; + case ORB_OP: + gen_simple_op(c, locp, bit_width, bit_suffix, &res, + op_types, &op1, &op2, + " | ", + "tcg_gen_ori_", + "tcg_gen_ori_", + "tcg_gen_or_"); + break; + case XORB_OP: + gen_simple_op(c, locp, bit_width, bit_suffix, &res, + op_types, &op1, &op2, + " ^ ", + "tcg_gen_xori_", + "tcg_gen_xori_", + "tcg_gen_xor_"); + break; + case ANDL_OP: + gen_andl_op(c, locp, bit_width, bit_suffix, &res, op_types, &op1, &op2); + break; + case MINI_OP: + gen_minmax_op(c, locp, bit_width, &res, op_types, &op1, &op2, false); + break; + case MAXI_OP: + gen_minmax_op(c, locp, bit_width, &res, op_types, &op1, &op2, true); + break; + case MOD_OP: + gen_mod_op(c, locp, &res, op_types, &op1, &op2); + break; + } + if (op_types == IMM_IMM) { + c->inst.qemu_tmp_count++; + } + return res; +} + +HexValue gen_cast_op(Context *c, + YYLTYPE *locp, + HexValue *source, + unsigned target_width) { + if (source->bit_width == target_width) { + return *source; + } else if (source->type == IMMEDIATE) { + HexValue res = *source; + res.bit_width = target_width; + return res; + } else { + HexValue res = gen_tmp(c, locp, target_width); + /* Truncate */ + if (source->bit_width > target_width) { + OUT(c, locp, "tcg_gen_trunc_i64_tl(", &res, ", ", source, ");\n"); + } else { + assert_signedness(c, locp, source->signedness); + if (source->signedness == UNSIGNED) { + /* Extend unsigned */ + OUT(c, locp, "tcg_gen_extu_i32_i64(", + &res, ", ", source, ");\n"); + } else { + /* Extend signed */ + OUT(c, locp, "tcg_gen_ext_i32_i64(", + &res, ", ", source, ");\n"); + } + } + res.signedness = source->signedness; + gen_rvalue_free(c, locp, source); + return res; + } +} + +HexValue gen_extend_op(Context *c, + YYLTYPE *locp, + HexValue *src_width_ptr, + HexValue *dst_width_ptr, + HexValue *value_ptr, + HexSignedness signedness) { + HexValue src_width = *src_width_ptr; + HexValue dst_width = *dst_width_ptr; + HexValue value = *value_ptr; + src_width = gen_rvalue_extend(c, locp, &src_width); + value = gen_rvalue_extend(c, locp, &value); + src_width = rvalue_materialize(c, locp, &src_width); + value = rvalue_materialize(c, locp, &value); + + HexValue res = gen_tmp(c, locp, 64); + HexValue shift = gen_tmp_value(c, locp, "64", 64); + HexValue zero = gen_tmp_value(c, locp, "0", 64); + OUT(c, locp, "tcg_gen_sub_i64(", + &shift, ", ", &shift, ", ", &src_width, ");\n"); + assert_signedness(c, locp, signedness); + if (signedness == UNSIGNED) { + HexValue mask = gen_tmp_value(c, locp, "0xffffffffffffffff", 64); + OUT(c, locp, "tcg_gen_shr_i64(", + &mask, ", ", &mask, ", ", &shift, ");\n"); + OUT(c, locp, "tcg_gen_and_i64(", + &res, ", ", &value, ", ", &mask, ");\n"); + gen_rvalue_free(c, locp, &mask); + } else { + OUT(c, locp, "tcg_gen_shl_i64(", + &res, ", ", &value, ", ", &shift, ");\n"); + OUT(c, locp, "tcg_gen_sar_i64(", + &res, ", ", &res, ", ", &shift, ");\n"); + } + OUT(c, locp, "tcg_gen_movcond_i64(TCG_COND_EQ, ", &res, ", "); + OUT(c, locp, &src_width, ", ", &zero, ", ", &zero, ", ", &res, ");\n"); + + gen_rvalue_free(c, locp, &src_width); + gen_rvalue_free(c, locp, &dst_width); + gen_rvalue_free(c, locp, &value); + gen_rvalue_free(c, locp, &shift); + gen_rvalue_free(c, locp, &zero); + + res.signedness = signedness; + return res; +} + +void gen_rdeposit_op(Context *c, + YYLTYPE *locp, + HexValue *dest, + HexValue *value, + HexValue *begin, + HexValue *width) +{ + HexValue dest_m = *dest; + dest_m.is_manual = true; + + HexValue value_m = gen_rvalue_extend(c, locp, value); + HexValue begin_m = gen_rvalue_extend(c, locp, begin); + HexValue width_orig = *width; + width_orig.is_manual = true; + HexValue width_m = gen_rvalue_extend(c, locp, &width_orig); + width_m = rvalue_materialize(c, locp, &width_m); + + HexValue mask = gen_tmp_value(c, locp, "0xffffffffffffffffUL", 64); + mask.signedness = UNSIGNED; + HexValue k64 = gen_tmp_value(c, locp, "64", 64); + k64 = gen_bin_op(c, locp, SUB_OP, &k64, &width_m); + mask = gen_bin_op(c, locp, LSR_OP, &mask, &k64); + begin_m.is_manual = true; + mask = gen_bin_op(c, locp, ASL_OP, &mask, &begin_m); + mask.is_manual = true; + value_m = gen_bin_op(c, locp, ASL_OP, &value_m, &begin_m); + value_m = gen_bin_op(c, locp, ANDB_OP, &value_m, &mask); + + OUT(c, locp, "tcg_gen_not_i64(", &mask, ", ", &mask, ");\n"); + mask.is_manual = false; + HexValue res = gen_bin_op(c, locp, ANDB_OP, &dest_m, &mask); + res = gen_bin_op(c, locp, ORB_OP, &res, &value_m); + + if (dest->bit_width != res.bit_width) { + res = gen_rvalue_truncate(c, locp, &res); + } + + HexValue zero = gen_tmp_value(c, locp, "0", res.bit_width); + OUT(c, locp, "tcg_gen_movcond_i", &res.bit_width, "(TCG_COND_NE, ", dest); + OUT(c, locp, ", ", &width_orig, ", ", &zero, ", ", &res, ", ", dest, + ");\n"); + + gen_rvalue_free(c, locp, &zero); + gen_rvalue_free(c, locp, width); + gen_rvalue_free(c, locp, &res); +} + +void gen_deposit_op(Context *c, + YYLTYPE *locp, + HexValue *dest, + HexValue *value, + HexValue *index, + HexCast *cast) +{ + yyassert(c, locp, index->type == IMMEDIATE, + "Deposit index must be immediate!\n"); + HexValue value_m = *value; + int bit_width = (dest->bit_width == 64) ? 64 : 32; + int width = cast->bit_width; + /* If the destination value is 32, truncate the value, otherwise extend */ + if (dest->bit_width != value->bit_width) { + if (bit_width == 32) { + value_m = gen_rvalue_truncate(c, locp, &value_m); + } else { + value_m = gen_rvalue_extend(c, locp, &value_m); + } + } + value_m = rvalue_materialize(c, locp, &value_m); + OUT(c, locp, "tcg_gen_deposit_i", &bit_width, "(", dest, ", ", dest, ", "); + OUT(c, locp, &value_m, ", ", index, " * ", &width, ", ", &width, ");\n"); + gen_rvalue_free(c, locp, index); + gen_rvalue_free(c, locp, &value_m); +} + +HexValue gen_rextract_op(Context *c, + YYLTYPE *locp, + HexValue *source, + int begin, + int width) { + int bit_width = (source->bit_width == 64) ? 64 : 32; + HexValue res = gen_tmp(c, locp, bit_width); + OUT(c, locp, "tcg_gen_extract_i", &bit_width, "(", &res); + OUT(c, locp, ", ", source, ", ", &begin, ", ", &width, ");\n"); + gen_rvalue_free(c, locp, source); + return res; +} + +HexValue gen_extract_op(Context *c, + YYLTYPE *locp, + HexValue *source, + HexValue *index, + HexExtract *extract) { + yyassert(c, locp, index->type == IMMEDIATE, + "Extract index must be immediate!\n"); + int bit_width = (source->bit_width == 64) ? 64 : 32; + assert_signedness(c, locp, extract->signedness); + const char *sign_prefix = (extract->signedness == UNSIGNED) ? "" : "s"; + int width = extract->bit_width; + HexValue res = gen_tmp(c, locp, bit_width); + res.signedness = extract->signedness; + OUT(c, locp, "tcg_gen_", sign_prefix, "extract_i", &bit_width, + "(", &res, ", ", source); + OUT(c, locp, ", ", index, " * ", &width, ", ", &width, ");\n"); + + /* Some extract operations have bit_width != storage_bit_width */ + if (extract->storage_bit_width > bit_width) { + HexValue tmp = gen_tmp(c, locp, extract->storage_bit_width); + tmp.signedness = extract->signedness; + if (extract->signedness == UNSIGNED) { + /* Extend unsigned */ + OUT(c, locp, "tcg_gen_extu_i32_i64(", + &tmp, ", ", &res, ");\n"); + } else { + /* Extend signed */ + OUT(c, locp, "tcg_gen_ext_i32_i64(", + &tmp, ", ", &res, ");\n"); + } + gen_rvalue_free(c, locp, &res); + res = tmp; + } + + gen_rvalue_free(c, locp, source); + gen_rvalue_free(c, locp, index); + return res; +} + +HexValue gen_read_reg(Context *c, YYLTYPE *locp, HexValue *reg) +{ + yyassert(c, locp, reg->type == REGISTER_ARG || reg->type == REGISTER, + "reg must be a register arg or register!"); + if (reg->type == REGISTER) { + HexValue tmp = gen_tmp(c, locp, 32); + tmp.signedness = reg->signedness; + OUT(c, locp, "gen_read_reg(", &tmp, ", ", ®->reg.id, ");\n"); + gen_rvalue_free(c, locp, reg); + return tmp; + } + return *reg; +} + +void gen_write_reg(Context *c, YYLTYPE *locp, HexValue *reg, HexValue *value) +{ + yyassert(c, locp, reg->type == REGISTER, "reg must be a register!"); + HexValue value_m = *value; + value_m = gen_rvalue_truncate(c, locp, &value_m); + value_m = rvalue_materialize(c, locp, &value_m); + OUT(c, + locp, + "gen_log_reg_write(", ®->reg.id, ", ", + &value_m, ");\n"); + OUT(c, + locp, + "ctx_log_reg_write(ctx, ", ®->reg.id, + ");\n"); + gen_rvalue_free(c, locp, reg); + gen_rvalue_free(c, locp, &value_m); +} + +void gen_assign(Context *c, + YYLTYPE *locp, + HexValue *dest, + HexValue *value) +{ + yyassert(c, + locp, + !is_inside_ternary(c) + || !(dest->type == REGISTER), + "register assign in ternary"); + + HexValue value_m = *value; + if (dest->type == REGISTER) { + gen_write_reg(c, locp, dest, &value_m); + return; + } + /* Create (if not present) and assign to temporary variable */ + if (dest->type == VARID) { + assert_signedness(c, locp, value_m.signedness); + gen_varid_allocate(c, locp, dest, value_m.bit_width, + value_m.signedness); + } + unsigned bit_width = dest->bit_width == 64 ? 64 : 32; + if (bit_width != value_m.bit_width) { + if (bit_width == 64) { + value_m = gen_rvalue_extend(c, locp, &value_m); + } else { + value_m = gen_rvalue_truncate(c, locp, &value_m); + } + } + if (is_inside_ternary(c)) { + value_m = rvalue_materialize(c, locp, &value_m); + HexValue cond = get_ternary_cond(c, locp); + if (cond.bit_width != bit_width) { + if (cond.bit_width == 64) { + cond = gen_rvalue_truncate(c, locp, &cond); + } else { + cond = gen_rvalue_extend(c, locp, &cond); + } + } + HexValue zero = gen_tmp_value(c, locp, "0", bit_width); + OUT(c, locp, "tcg_gen_movcond_i", &bit_width, "(TCG_COND_NE, ", dest); + OUT(c, locp, ", ", &cond, ", ", &zero, ", "); + OUT(c, locp, &value_m, ", ", dest, ");\n"); + gen_rvalue_free(c, locp, &cond); + gen_rvalue_free(c, locp, &zero); + } else { + if (value_m.type == IMMEDIATE) { + OUT(c, locp, "tcg_gen_movi_i", &bit_width, + "(", dest, ", ", &value_m, ");\n"); + } else { + OUT(c, locp, "tcg_gen_mov_i", &bit_width, + "(", dest, ", ", &value_m, ");\n"); + } + } + gen_rvalue_free(c, locp, &value_m); +} + +HexValue gen_convround(Context *c, + YYLTYPE *locp, + HexValue *source) +{ + HexValue src = *source; + src.is_manual = true; + + unsigned bit_width = src.bit_width; + const char *size = (bit_width == 32) ? "32" : "64"; + HexValue res = gen_tmp(c, locp, bit_width); + HexValue mask = gen_tmp_value(c, locp, "0x3", bit_width); + mask.is_manual = true; + HexValue and = gen_bin_op(c, locp, ANDB_OP, &src, &mask); + HexValue one = gen_tmp_value(c, locp, "1", bit_width); + HexValue src_p1 = gen_bin_op(c, locp, ADD_OP, &src, &one); + + OUT(c, locp, "tcg_gen_movcond_i", size, "(TCG_COND_EQ, ", &res); + OUT(c, locp, ", ", &and, ", ", &mask, ", "); + OUT(c, locp, &src_p1, ", ", &src, ");\n"); + + /* Free src but use the original `is_manual` value */ + gen_rvalue_free(c, locp, source); + + /* Free the rest of the values */ + gen_rvalue_free_manual(c, locp, &mask); + gen_rvalue_free(c, locp, &and); + gen_rvalue_free(c, locp, &src_p1); + + return res; +} + +static HexValue gen_convround_n_a(Context *c, + YYLTYPE *locp, + HexValue *a, + HexValue *n) +{ + (void) n; + HexValue res = gen_tmp(c, locp, 64); + OUT(c, locp, "tcg_gen_ext_i32_i64(", &res, ", ", a, ");\n"); + return res; +} + +static HexValue gen_convround_n_b(Context *c, + YYLTYPE *locp, + HexValue *a, + HexValue *n) +{ + HexValue res = gen_tmp(c, locp, 64); + OUT(c, locp, "tcg_gen_ext_i32_i64(", &res, ", ", a, ");\n"); + + HexValue one = gen_tmp_value(c, locp, "1", 32); + HexValue tmp = gen_tmp(c, locp, 32); + HexValue tmp_64 = gen_tmp(c, locp, 64); + + OUT(c, locp, "tcg_gen_shl_i32(", &tmp); + OUT(c, locp, ", ", &one, ", ", n, ");\n"); + OUT(c, locp, "tcg_gen_and_i32(", &tmp); + OUT(c, locp, ", ", &tmp, ", ", a, ");\n"); + OUT(c, locp, "tcg_gen_shri_i32(", &tmp); + OUT(c, locp, ", ", &tmp, ", 1);\n"); + OUT(c, locp, "tcg_gen_ext_i32_i64(", &tmp_64, ", ", &tmp, ");\n"); + OUT(c, locp, "tcg_gen_add_i64(", &res); + OUT(c, locp, ", ", &res, ", ", &tmp_64, ");\n"); + + gen_rvalue_free(c, locp, &one); + gen_rvalue_free(c, locp, &tmp); + gen_rvalue_free(c, locp, &tmp_64); + + return res; +} + +static HexValue gen_convround_n_c(Context *c, + YYLTYPE *locp, + HexValue *a, + HexValue *n) +{ + HexValue res = gen_tmp(c, locp, 64); + OUT(c, locp, "tcg_gen_ext_i32_i64(", &res, ", ", a, ");\n"); + + HexValue one = gen_tmp_value(c, locp, "1", 32); + HexValue tmp = gen_tmp(c, locp, 32); + HexValue tmp_64 = gen_tmp(c, locp, 64); + + OUT(c, locp, "tcg_gen_subi_i32(", &tmp); + OUT(c, locp, ", ", n, ", 1);\n"); + OUT(c, locp, "tcg_gen_shl_i32(", &tmp); + OUT(c, locp, ", ", &one, ", ", &tmp, ");\n"); + OUT(c, locp, "tcg_gen_ext_i32_i64(", &tmp_64, ", ", &tmp, ");\n"); + OUT(c, locp, "tcg_gen_add_i64(", &res); + OUT(c, locp, ", ", &res, ", ", &tmp_64, ");\n"); + + gen_rvalue_free(c, locp, &one); + gen_rvalue_free(c, locp, &tmp); + gen_rvalue_free(c, locp, &tmp_64); + + return res; +} + +HexValue gen_convround_n(Context *c, + YYLTYPE *locp, + HexValue *source_ptr, + HexValue *bit_pos_ptr) +{ + /* If input is 64 bit cast it to 32 */ + HexValue source = gen_cast_op(c, locp, source_ptr, 32); + HexValue bit_pos = gen_cast_op(c, locp, bit_pos_ptr, 32); + + source = rvalue_materialize(c, locp, &source); + bit_pos = rvalue_materialize(c, locp, &bit_pos); + + HexValue r1 = gen_convround_n_a(c, locp, &source, &bit_pos); + HexValue r2 = gen_convround_n_b(c, locp, &source, &bit_pos); + HexValue r3 = gen_convround_n_c(c, locp, &source, &bit_pos); + + HexValue l_32 = gen_tmp_value(c, locp, "1", 32); + + HexValue cond = gen_tmp(c, locp, 32); + HexValue cond_64 = gen_tmp(c, locp, 64); + HexValue mask = gen_tmp(c, locp, 32); + HexValue n_64 = gen_tmp(c, locp, 64); + HexValue res = gen_tmp(c, locp, 64); + HexValue zero = gen_tmp_value(c, locp, "0", 64); + + OUT(c, locp, "tcg_gen_sub_i32(", &mask); + OUT(c, locp, ", ", &bit_pos, ", ", &l_32, ");\n"); + OUT(c, locp, "tcg_gen_shl_i32(", &mask); + OUT(c, locp, ", ", &l_32, ", ", &mask, ");\n"); + OUT(c, locp, "tcg_gen_sub_i32(", &mask); + OUT(c, locp, ", ", &mask, ", ", &l_32, ");\n"); + OUT(c, locp, "tcg_gen_and_i32(", &cond); + OUT(c, locp, ", ", &source, ", ", &mask, ");\n"); + OUT(c, locp, "tcg_gen_extu_i32_i64(", &cond_64, ", ", &cond, ");\n"); + OUT(c, locp, "tcg_gen_ext_i32_i64(", &n_64, ", ", &bit_pos, ");\n"); + + OUT(c, locp, "tcg_gen_movcond_i64"); + OUT(c, locp, "(TCG_COND_EQ, ", &res, ", ", &cond_64, ", ", &zero); + OUT(c, locp, ", ", &r2, ", ", &r3, ");\n"); + + OUT(c, locp, "tcg_gen_movcond_i64"); + OUT(c, locp, "(TCG_COND_EQ, ", &res, ", ", &n_64, ", ", &zero); + OUT(c, locp, ", ", &r1, ", ", &res, ");\n"); + + OUT(c, locp, "tcg_gen_shr_i64(", &res); + OUT(c, locp, ", ", &res, ", ", &n_64, ");\n"); + + gen_rvalue_free(c, locp, &source); + gen_rvalue_free(c, locp, &bit_pos); + + gen_rvalue_free(c, locp, &r1); + gen_rvalue_free(c, locp, &r2); + gen_rvalue_free(c, locp, &r3); + + gen_rvalue_free(c, locp, &cond); + gen_rvalue_free(c, locp, &cond_64); + gen_rvalue_free(c, locp, &l_32); + gen_rvalue_free(c, locp, &mask); + gen_rvalue_free(c, locp, &n_64); + gen_rvalue_free(c, locp, &zero); + + res = gen_rvalue_truncate(c, locp, &res); + return res; +} + +HexValue gen_round(Context *c, + YYLTYPE *locp, + HexValue *source, + HexValue *position) { + yyassert(c, locp, source->bit_width <= 32, + "fRNDN not implemented for bit widths > 32!"); + + HexValue src = *source; + HexValue pos = *position; + + HexValue src_width = gen_imm_value(c, locp, src.bit_width, 32); + HexValue dst_width = gen_imm_value(c, locp, 64, 32); + HexValue a = gen_extend_op(c, locp, &src_width, &dst_width, &src, SIGNED); + + src_width = gen_imm_value(c, locp, 5, 32); + dst_width = gen_imm_value(c, locp, 64, 32); + HexValue b = gen_extend_op(c, locp, &src_width, &dst_width, &pos, UNSIGNED); + + /* Disable auto-free of values used more than once */ + a.is_manual = true; + b.is_manual = true; + + HexValue res = gen_tmp(c, locp, 64); + + HexValue one = gen_tmp_value(c, locp, "1", 64); + HexValue n_m1 = gen_bin_op(c, locp, SUB_OP, &b, &one); + one = gen_tmp_value(c, locp, "1", 64); + HexValue shifted = gen_bin_op(c, locp, ASL_OP, &one, &n_m1); + HexValue sum = gen_bin_op(c, locp, ADD_OP, &shifted, &a); + + HexValue zero = gen_tmp_value(c, locp, "0", 64); + OUT(c, locp, "tcg_gen_movcond_i64"); + OUT(c, locp, "(TCG_COND_EQ, ", &res, ", ", &b, ", ", &zero); + OUT(c, locp, ", ", &a, ", ", &sum, ");\n"); + + gen_rvalue_free_manual(c, locp, &a); + gen_rvalue_free_manual(c, locp, &b); + gen_rvalue_free(c, locp, &zero); + gen_rvalue_free(c, locp, &sum); + + return res; +} + +/* Circular addressing mode with auto-increment */ +void gen_circ_op(Context *c, + YYLTYPE *locp, + HexValue *addr, + HexValue *increment, + HexValue *modifier) { + HexValue increment_m = *increment; + HexValue cs = gen_tmp(c, locp, 32); + increment_m = rvalue_materialize(c, locp, &increment_m); + OUT(c, locp, "gen_read_reg(", &cs, ", HEX_REG_CS0 + MuN);\n"); + OUT(c, + locp, + "gen_helper_fcircadd(", + addr, + ", ", + addr, + ", ", + &increment_m, + ", ", + modifier); + OUT(c, locp, ", ", &cs, ");\n"); + gen_rvalue_free(c, locp, &increment_m); + gen_rvalue_free(c, locp, modifier); + gen_rvalue_free(c, locp, &cs); +} + +HexValue gen_locnt_op(Context *c, YYLTYPE *locp, HexValue *source) +{ + HexValue source_m = *source; + const char *bit_suffix = source->bit_width == 64 ? "64" : "32"; + HexValue res = gen_tmp(c, locp, source->bit_width == 64 ? 64 : 32); + res.type = TEMP; + source_m = rvalue_materialize(c, locp, &source_m); + OUT(c, locp, "tcg_gen_not_i", bit_suffix, "(", + &res, ", ", &source_m, ");\n"); + OUT(c, locp, "tcg_gen_clzi_i", bit_suffix, "(", &res, ", ", &res, ", "); + OUT(c, locp, bit_suffix, ");\n"); + gen_rvalue_free(c, locp, &source_m); + return res; +} + +HexValue gen_ctpop_op(Context *c, YYLTYPE *locp, HexValue *source) +{ + HexValue source_m = *source; + const char *bit_suffix = source_m.bit_width == 64 ? "64" : "32"; + HexValue res = gen_tmp(c, locp, source_m.bit_width == 64 ? 64 : 32); + res.type = TEMP; + source_m = rvalue_materialize(c, locp, &source_m); + OUT(c, locp, "tcg_gen_ctpop_i", bit_suffix, + "(", &res, ", ", &source_m, ");\n"); + gen_rvalue_free(c, locp, &source_m); + return res; +} + +HexValue gen_fbrev_4(Context *c, YYLTYPE *locp, HexValue *source) +{ + HexValue source_m = *source; + + HexValue res = gen_tmp(c, locp, 32); + HexValue tmp1 = gen_tmp(c, locp, 32); + HexValue tmp2 = gen_tmp(c, locp, 32); + + source_m = rvalue_materialize(c, locp, &source_m); + source_m = gen_rvalue_truncate(c, locp, &source_m); + + OUT(c, locp, "tcg_gen_mov_tl(", &res, ", ", &source_m, ");\n"); + OUT(c, locp, "tcg_gen_andi_tl(", &tmp1, ", ", &res, ", 0xaaaaaaaa);\n"); + OUT(c, locp, "tcg_gen_shri_tl(", &tmp1, ", ", &tmp1, ", 1);\n"); + OUT(c, locp, "tcg_gen_andi_tl(", &tmp2, ", ", &res, ", 0x55555555);\n"); + OUT(c, locp, "tcg_gen_shli_tl(", &tmp2, ", ", &tmp2, ", 1);\n"); + OUT(c, locp, "tcg_gen_or_tl(", &res, ", ", &tmp1, ", ", &tmp2, ");\n"); + OUT(c, locp, "tcg_gen_andi_tl(", &tmp1, ", ", &res, ", 0xcccccccc);\n"); + OUT(c, locp, "tcg_gen_shri_tl(", &tmp1, ", ", &tmp1, ", 2);\n"); + OUT(c, locp, "tcg_gen_andi_tl(", &tmp2, ", ", &res, ", 0x33333333);\n"); + OUT(c, locp, "tcg_gen_shli_tl(", &tmp2, ", ", &tmp2, ", 2);\n"); + OUT(c, locp, "tcg_gen_or_tl(", &res, ", ", &tmp1, ", ", &tmp2, ");\n"); + OUT(c, locp, "tcg_gen_andi_tl(", &tmp1, ", ", &res, ", 0xf0f0f0f0);\n"); + OUT(c, locp, "tcg_gen_shri_tl(", &tmp1, ", ", &tmp1, ", 4);\n"); + OUT(c, locp, "tcg_gen_andi_tl(", &tmp2, ", ", &res, ", 0x0f0f0f0f);\n"); + OUT(c, locp, "tcg_gen_shli_tl(", &tmp2, ", ", &tmp2, ", 4);\n"); + OUT(c, locp, "tcg_gen_or_tl(", &res, ", ", &tmp1, ", ", &tmp2, ");\n"); + OUT(c, locp, "tcg_gen_bswap32_tl(", &res, ", ", &res, ", TCG_BSWAP_IZ);\n"); + + gen_rvalue_free(c, locp, &tmp1); + gen_rvalue_free(c, locp, &tmp2); + gen_rvalue_free(c, locp, &source_m); + + return res; +} + +HexValue gen_fbrev_8(Context *c, YYLTYPE *locp, HexValue *source) +{ + HexValue source_m = *source; + + source_m = gen_rvalue_extend(c, locp, &source_m); + source_m = rvalue_materialize(c, locp, &source_m); + + HexValue res = gen_tmp(c, locp, 64); + HexValue tmp1 = gen_tmp(c, locp, 64); + HexValue tmp2 = gen_tmp(c, locp, 64); + + OUT(c, locp, "tcg_gen_mov_i64(", + &res, ", ", &source_m, ");\n"); + OUT(c, locp, "tcg_gen_andi_i64(", + &tmp1, ", ", &res, ", 0xaaaaaaaaaaaaaaaa);\n"); + OUT(c, locp, "tcg_gen_shri_i64(", + &tmp1, ", ", &tmp1, ", 1);\n"); + OUT(c, locp, "tcg_gen_andi_i64(", + &tmp2, ", ", &res, ", 0x5555555555555555);\n"); + OUT(c, locp, "tcg_gen_shli_i64(", + &tmp2, ", ", &tmp2, ", 1);\n"); + OUT(c, locp, "tcg_gen_or_i64(", &res, ", ", &tmp1, ", ", &tmp2, ");\n"); + OUT(c, locp, "tcg_gen_andi_i64(", + &tmp1, ", ", &res, ", 0xcccccccccccccccc);\n"); + OUT(c, locp, "tcg_gen_shri_i64(", + &tmp1, ", ", &tmp1, ", 2);\n"); + OUT(c, locp, "tcg_gen_andi_i64(", + &tmp2, ", ", &res, ", 0x3333333333333333);\n"); + OUT(c, locp, "tcg_gen_shli_i64(", + &tmp2, ", ", &tmp2, ", 2);\n"); + OUT(c, locp, "tcg_gen_or_i64(", &res, ", ", &tmp1, ", ", &tmp2, ");\n"); + OUT(c, locp, "tcg_gen_andi_i64(", + &tmp1, ", ", &res, ", 0xf0f0f0f0f0f0f0f0);\n"); + OUT(c, locp, "tcg_gen_shri_i64(", + &tmp1, ", ", &tmp1, ", 4);\n"); + OUT(c, locp, "tcg_gen_andi_i64(", + &tmp2, ", ", &res, ", 0x0f0f0f0f0f0f0f0f);\n"); + OUT(c, locp, "tcg_gen_shli_i64(", + &tmp2, ", ", &tmp2, ", 4);\n"); + OUT(c, locp, "tcg_gen_or_i64(", &res, ", ", &tmp1, ", ", &tmp2, ");\n"); + OUT(c, locp, "tcg_gen_andi_i64(", + &tmp1, ", ", &res, ", 0xff00ff00ff00ff00);\n"); + OUT(c, locp, "tcg_gen_shri_i64(", + &tmp1, ", ", &tmp1, ", 8);\n"); + OUT(c, locp, "tcg_gen_andi_i64(", + &tmp2, ", ", &res, ", 0x00ff00ff00ff00ff);\n"); + OUT(c, locp, "tcg_gen_shli_i64(", + &tmp2, ", ", &tmp2, ", 8);\n"); + OUT(c, locp, "tcg_gen_or_i64(", &res, ", ", &tmp1, ", ", &tmp2, ");\n"); + OUT(c, locp, "tcg_gen_andi_i64(", + &tmp1, ", ", &res, ", 0xffff0000ffff0000);\n"); + OUT(c, locp, "tcg_gen_shri_i64(", + &tmp1, ", ", &tmp1, ", 16);\n"); + OUT(c, locp, "tcg_gen_andi_i64(", + &tmp2, ", ", &res, ", 0x0000ffff0000ffff);\n"); + OUT(c, locp, "tcg_gen_shli_i64(", + &tmp2, ", ", &tmp2, ", 16);\n"); + OUT(c, locp, "tcg_gen_or_i64(", &res, ", ", &tmp1, ", ", &tmp2, ");\n"); + OUT(c, locp, "tcg_gen_shri_i64(", &tmp1, ", ", &res, ", 32);\n"); + OUT(c, locp, "tcg_gen_shli_i64(", &tmp2, ", ", &res, ", 32);\n"); + OUT(c, locp, "tcg_gen_or_i64(", &res, ", ", &tmp1, ", ", &tmp2, ");\n"); + + gen_rvalue_free(c, locp, &tmp1); + gen_rvalue_free(c, locp, &tmp2); + gen_rvalue_free(c, locp, &source_m); + + return res; +} + +HexValue gen_rotl(Context *c, YYLTYPE *locp, HexValue *source, HexValue *n) +{ + HexValue amount = *n; + if (amount.bit_width < source->bit_width) { + amount = gen_rvalue_extend(c, locp, &amount); + } else { + amount = gen_rvalue_truncate(c, locp, &amount); + } + amount = rvalue_materialize(c, locp, &amount); + const char *suffix = source->bit_width == 64 ? "i64" : "i32"; + + HexValue res = gen_tmp(c, locp, source->bit_width); + res.signedness = source->signedness; + OUT(c, locp, "tcg_gen_rotl_", suffix, "(", + &res, ", ", source, ", ", &amount, ");\n"); + gen_rvalue_free(c, locp, source); + gen_rvalue_free(c, locp, &amount); + + return res; +} + +const char *INTERLEAVE_MASKS[6] = { + "0x5555555555555555ULL", + "0x3333333333333333ULL", + "0x0f0f0f0f0f0f0f0fULL", + "0x00ff00ff00ff00ffULL", + "0x0000ffff0000ffffULL", + "0x00000000ffffffffULL", +}; + +HexValue gen_deinterleave(Context *c, YYLTYPE *locp, HexValue *mixed) +{ + HexValue src = gen_rvalue_extend(c, locp, mixed); + + HexValue a = gen_tmp(c, locp, 64); + a.signedness = UNSIGNED; + HexValue b = gen_tmp(c, locp, 64); + b.signedness = UNSIGNED; + + const char **masks = INTERLEAVE_MASKS; + + OUT(c, locp, "tcg_gen_shri_i64(", &a, ", ", &src, ", 1);\n"); + OUT(c, locp, "tcg_gen_andi_i64(", &a, ", ", &a, ", ", masks[0], ");\n"); + OUT(c, locp, "tcg_gen_andi_i64(", &b, ", ", &src, ", ", masks[0], ");\n"); + + HexValue res = gen_tmp(c, locp, 64); + res.signedness = UNSIGNED; + + unsigned shift = 1; + for (unsigned i = 1; i < 6; ++i) { + OUT(c, locp, "tcg_gen_shri_i64(", &res, ", ", &b, ", ", &shift, ");\n"); + OUT(c, locp, "tcg_gen_or_i64(", &b, ", ", &res, ", ", &b, ");\n"); + OUT(c, locp, "tcg_gen_andi_i64(", &b, ", ", &b, ", ", masks[i], ");\n"); + OUT(c, locp, "tcg_gen_shri_i64(", &res, ", ", &a, ", ", &shift, ");\n"); + OUT(c, locp, "tcg_gen_or_i64(", &a, ", ", &res, ", ", &a, ");\n"); + OUT(c, locp, "tcg_gen_andi_i64(", &a, ", ", &a, ", ", masks[i], ");\n"); + shift <<= 1; + } + + OUT(c, locp, "tcg_gen_shli_i64(", &a, ", ", &a, ", 32);\n"); + OUT(c, locp, "tcg_gen_or_i64(", &res, ", ", &a, ", ", &b, ");\n"); + + gen_rvalue_free(c, locp, &a); + gen_rvalue_free(c, locp, &b); + + return res; +} + +HexValue gen_interleave(Context *c, + YYLTYPE *locp, + HexValue *odd, + HexValue *even) +{ + HexValue a = gen_rvalue_truncate(c, locp, odd); + a.signedness = UNSIGNED; + HexValue b = gen_rvalue_truncate(c, locp, even); + a.signedness = UNSIGNED; + + a = gen_rvalue_extend(c, locp, &a); + b = gen_rvalue_extend(c, locp, &b); + + HexValue res = gen_tmp(c, locp, 64); + res.signedness = UNSIGNED; + + const char **masks = INTERLEAVE_MASKS; + + unsigned shift = 16; + for (int i = 4; i >= 0; --i) { + OUT(c, locp, "tcg_gen_shli_i64(", &res, ", ", &a, ", ", &shift, ");\n"); + OUT(c, locp, "tcg_gen_or_i64(", &a, ", ", &res, ", ", &a, ");\n"); + OUT(c, locp, "tcg_gen_andi_i64(", &a, ", ", &a, ", ", masks[i], ");\n"); + OUT(c, locp, "tcg_gen_shli_i64(", &res, ", ", &b, ", ", &shift, ");\n"); + OUT(c, locp, "tcg_gen_or_i64(", &b, ", ", &res, ", ", &b, ");\n"); + OUT(c, locp, "tcg_gen_andi_i64(", &b, ", ", &b, ", ", masks[i], ");\n"); + shift >>= 1; + } + + OUT(c, locp, "tcg_gen_shli_i64(", &a, ", ", &a, ", 1);\n"); + OUT(c, locp, "tcg_gen_or_i64(", &res, ", ", &a, ", ", &b, ");\n"); + + gen_rvalue_free(c, locp, &a); + gen_rvalue_free(c, locp, &b); + + return res; +} + +HexValue gen_carry_from_add(Context *c, + YYLTYPE *locp, + HexValue *op1, + HexValue *op2, + HexValue *op3) +{ + HexValue opa = rvalue_materialize(c, locp, op1); + HexValue opb = rvalue_materialize(c, locp, op2); + HexValue opc = rvalue_materialize(c, locp, op3); + opc = gen_rvalue_extend(c, locp, &opc); + + HexValue zero = gen_tmp_value(c, locp, "0", 64); + HexValue res = gen_tmp(c, locp, 64); + HexValue cf = gen_tmp(c, locp, 64); + OUT(c, locp, "tcg_gen_add2_i64(", &res, ", ", &cf, ", ", &opa, ", ", &zero); + OUT(c, locp, ", ", &opc, ", ", &zero, ");\n"); + OUT(c, locp, "tcg_gen_add2_i64(", &res, ", ", &cf, ", ", &res, ", ", &cf); + OUT(c, locp, ", ", &opb, ", ", &zero, ");\n"); + + gen_rvalue_free(c, locp, &opa); + gen_rvalue_free(c, locp, &opb); + gen_rvalue_free(c, locp, &opc); + gen_rvalue_free(c, locp, &zero); + gen_rvalue_free(c, locp, &res); + return cf; +} + +void gen_inst(Context *c, GString *iname) +{ + c->total_insn++; + c->inst.name = iname; + c->inst.allocated = g_array_new(FALSE, FALSE, sizeof(Var)); + c->inst.init_list = g_array_new(FALSE, FALSE, sizeof(HexValue)); + c->inst.strings = g_array_new(FALSE, FALSE, sizeof(GString *)); + EMIT_SIG(c, "void emit_%s(DisasContext *ctx, Insn *insn, Packet *pkt", + c->inst.name->str); +} + +void gen_inst_args(Context *c, YYLTYPE *locp) +{ + EMIT_SIG(c, ")"); + EMIT_HEAD(c, "{\n"); + + /* Initialize declared but uninitialized registers, but only for */ + /* non-conditional instructions */ + for (unsigned i = 0; i < c->inst.init_list->len; i++) { + HexValue *reg = &g_array_index(c->inst.init_list, HexValue, i); + if (reg->type == REGISTER + || reg->type == REGISTER_ARG + || reg->type == PREDICATE) { + OUT(c, locp, "tcg_gen_movi_i", ®->bit_width, "(", + reg, ", 0);\n"); + } + } +} + +void gen_inst_code(Context *c, YYLTYPE *locp) +{ + if (c->inst.error_count != 0) { + fprintf(stderr, + "Parsing of instruction %s generated %d errors!\n", + c->inst.name->str, + c->inst.error_count); + } else { + free_variables(c, locp); + c->implemented_insn++; + fprintf(c->enabled_file, "%s\n", c->inst.name->str); + emit_footer(c); + commit(c); + } + free_instruction(c); +} + +void gen_pre_assign(Context *c, YYLTYPE *locp, HexValue *lp, HexValue *rp) +{ + bool is_direct = is_direct_predicate(lp); + char pre_id[2] = " "; + pre_id[0] = lp->pre.id; + /* Extract predicate TCGv */ + if (is_direct) { + *lp = gen_tmp_value(c, locp, "0", 32); + } + HexValue r = rvalue_materialize(c, locp, rp); + r = gen_rvalue_truncate(c, locp, &r); + /* Extract first 8 bits, and store new predicate value */ + if (is_inside_ternary(c)) { + yyassert(c, locp, !is_direct, "direct pre assign inside ternary op"); + HexValue tmp = gen_tmp(c, locp, r.bit_width); + HexValue cond = get_ternary_cond(c, locp); + HexValue zero = gen_tmp_value(c, locp, "0", r.bit_width); + OUT(c, locp, "tcg_gen_mov_i32(", &tmp, ", ", &r, ");\n"); + OUT(c, locp, "tcg_gen_andi_i32(", &tmp, ", ", &tmp, ", 0xff);\n"); + OUT(c, locp, "tcg_gen_movcond_i", &r.bit_width); + OUT(c, locp, "(TCG_COND_NE, ", lp, ", ", &cond, ", ", &zero); + OUT(c, locp, ", ", &tmp, ", ", lp, ");\n"); + gen_rvalue_free(c, locp, &tmp); + gen_rvalue_free(c, locp, &cond); + gen_rvalue_free(c, locp, &zero); + } else { + OUT(c, locp, "tcg_gen_mov_i32(", lp, ", ", &r, ");\n"); + OUT(c, locp, "tcg_gen_andi_i32(", lp, ", ", lp, ", 0xff);\n"); + } + if (is_direct) { + OUT(c, locp, "gen_log_pred_write(ctx, ", pre_id, ", ", lp, ");\n"); + OUT(c, locp, "ctx_log_pred_write(ctx, ", pre_id, ");\n"); + gen_rvalue_free(c, locp, lp); + } + gen_rvalue_free(c, locp, &r); /* Free temporary value */ +} + +void gen_load(Context *c, YYLTYPE *locp, HexValue *size, + HexSignedness signedness, HexValue *ea, HexValue *dst) +{ + /* Memop width is specified in the load macro */ + int bit_width = (size->imm.value > 4) ? 64 : 32; + assert_signedness(c, locp, signedness); + const char *sign_suffix = (size->imm.value > 4) + ? "" + : ((signedness == UNSIGNED) ? "u" : "s"); + char size_suffix[4] = { 0 }; + /* Create temporary variable (if not present) */ + if (dst->type == VARID) { + /* TODO: this is a common pattern, the parser should be varid-aware. */ + gen_varid_allocate(c, locp, dst, bit_width, signedness); + } + snprintf(size_suffix, 4, "%" PRIu64, size->imm.value * 8); + int var_id = find_variable(c, locp, ea); + yyassert(c, locp, var_id != -1, "Load variable must exist!\n"); + /* We need to enforce the variable size */ + ea->bit_width = g_array_index(c->inst.allocated, Var, var_id).bit_width; + OUT(c, locp, "if (insn->slot == 0 && pkt->pkt_has_store_s1) {\n"); + OUT(c, locp, "process_store(ctx, pkt, 1);\n"); + OUT(c, locp, "}\n"); + OUT(c, locp, "tcg_gen_qemu_ld", size_suffix, sign_suffix); + OUT(c, locp, "(", dst, ", ", ea, ", 0);\n"); + /* If the var in EA was truncated it is now a tmp HexValue, so free it. */ + gen_rvalue_free(c, locp, ea); +} + +void gen_store(Context *c, YYLTYPE *locp, HexValue *size, HexValue *ea, + HexValue *src) +{ + HexValue src_m = *src; + /* Memop width is specified in the store macro */ + int mem_width = size->imm.value; + assert(ea->type == VARID); + int var_id = find_variable(c, locp, ea); + yyassert(c, locp, var_id != -1, "Load variable must exist!\n"); + /* We need to enforce the variable size */ + ea->bit_width = g_array_index(c->inst.allocated, Var, var_id).bit_width; + src_m = rvalue_materialize(c, locp, &src_m); + OUT(c, locp, "gen_store", &mem_width, "(cpu_env, ", ea, ", ", &src_m); + OUT(c, locp, ", ctx, insn->slot);\n"); + gen_rvalue_free(c, locp, &src_m); + /* If the var in ea was truncated it is now a tmp HexValue, so free it. */ + gen_rvalue_free(c, locp, ea); +} + +void gen_sethalf(Context *c, YYLTYPE *locp, HexCast *sh, HexValue *n, + HexValue *dst, HexValue *val) +{ + yyassert(c, locp, n->type == IMMEDIATE, + "Deposit index must be immediate!\n"); + if (dst->type == VARID) { + int var_id = find_variable(c, locp, dst); + if (var_id == -1) { + HexValue zero = gen_imm_value(c, locp, 0, 64); + zero.signedness = UNSIGNED; + dst->bit_width = 64; + gen_assign(c, locp, dst, &zero); + } else { + /* We need to enforce the variable size (default is 32) */ + dst->bit_width = g_array_index(c->inst.allocated, + Var, + var_id).bit_width; + } + } + gen_deposit_op(c, locp, dst, val, n, sh); +} + +void gen_setbits(Context *c, YYLTYPE *locp, HexValue *hi, HexValue *lo, + HexValue *dst, HexValue *val) +{ + yyassert(c, locp, hi->type == IMMEDIATE && + hi->imm.type == VALUE && + lo->type == IMMEDIATE && + lo->imm.type == VALUE, + "Range deposit needs immediate values!\n"); + + *val = gen_rvalue_truncate(c, locp, val); + unsigned len = hi->imm.value + 1 - lo->imm.value; + HexValue tmp = gen_tmp(c, locp, 32); + OUT(c, locp, "tcg_gen_neg_i32(", &tmp, ", ", val, ");\n"); + OUT(c, locp, "tcg_gen_deposit_i32(", dst, ", ", dst, ", ", &tmp, ", "); + OUT(c, locp, lo, ", ", &len, ");\n"); + + gen_rvalue_free(c, locp, &tmp); + gen_rvalue_free(c, locp, hi); + gen_rvalue_free(c, locp, lo); + gen_rvalue_free(c, locp, val); +} + +int gen_if_cond(Context *c, YYLTYPE *locp, HexValue *cond) +{ + /* Generate an end label, if false branch to that label */ + OUT(c, locp, "TCGLabel *if_label_", &c->inst.if_count, + " = gen_new_label();\n"); + *cond = rvalue_materialize(c, locp, cond); + const char *bit_suffix = (cond->bit_width == 64) ? "i64" : "i32"; + OUT(c, locp, "tcg_gen_brcondi_", bit_suffix, "(TCG_COND_EQ, ", cond, + ", 0, if_label_", &c->inst.if_count, ");\n"); + gen_rvalue_free(c, locp, cond); + return c->inst.if_count++; +} + +int gen_if_else(Context *c, YYLTYPE *locp, int index) +{ + /* Generate label to jump if else is not verified */ + OUT(c, locp, "TCGLabel *if_label_", &c->inst.if_count, + " = gen_new_label();\n"); + int if_index = c->inst.if_count; + c->inst.if_count++; + /* Jump out of the else statement */ + OUT(c, locp, "tcg_gen_br(if_label_", &if_index, ");\n"); + /* Fix the else label */ + OUT(c, locp, "gen_set_label(if_label_", &index, ");\n"); + return if_index; +} + +HexValue gen_rvalue_pre(Context *c, YYLTYPE *locp, HexValue *pre) +{ + if (is_direct_predicate(pre)) { + bool is_dotnew = pre->is_dotnew; + char predicate_id[2] = { pre->pre.id, '\0' }; + char *pre_str = (char *) &predicate_id; + *pre = gen_tmp_value(c, locp, "0", 32); + if (is_dotnew) { + OUT(c, locp, "tcg_gen_mov_i32(", pre, ", hex_new_pred_value["); + OUT(c, locp, pre_str, "]);\n"); + } else { + OUT(c, locp, "gen_read_preg(", pre, ", ", pre_str, ");\n"); + } + } + return *pre; +} + +HexValue gen_rvalue_var(Context *c, YYLTYPE *locp, HexValue *var) +{ + /* Assign correct bit width and signedness */ + bool found = false; + for (unsigned i = 0; i < c->inst.allocated->len; i++) { + Var *other = &g_array_index(c->inst.allocated, Var, i); + if (g_string_equal(var->var.name, other->name)) { + found = true; + other->name = var->var.name; + var->bit_width = other->bit_width; + var->signedness = other->signedness; + break; + } + } + yyassert(c, locp, found, "Undefined symbol!\n"); + return *var; +} + +HexValue gen_rvalue_mpy(Context *c, YYLTYPE *locp, HexMpy *mpy, HexValue *a, + HexValue *b) +{ + assert_signedness(c, locp, mpy->first_signedness); + assert_signedness(c, locp, mpy->second_signedness); + a->signedness = mpy->first_signedness; + b->signedness = mpy->second_signedness; + *a = gen_cast_op(c, locp, a, mpy->first_bit_width * 2); + /* Handle fMPTY3216.. */ + if (mpy->first_bit_width == 32) { + *b = gen_cast_op(c, locp, b, 64); + } else { + *b = gen_cast_op(c, locp, b, mpy->second_bit_width * 2); + } + HexValue ret = gen_bin_op(c, locp, MUL_OP, a, b); + /* Handle special cases required by the language */ + if (mpy->first_bit_width == 16 && mpy->second_bit_width == 16) { + HexValue src_width = gen_imm_value(c, locp, 32, 32); + HexValue dst_width = gen_imm_value(c, locp, 64, 32); + assert_signedness(c, locp, mpy->first_signedness); + assert_signedness(c, locp, mpy->second_signedness); + bool is_unsigned = (mpy->first_signedness == UNSIGNED + && mpy->second_signedness == UNSIGNED); + HexSignedness signedness = is_unsigned ? UNSIGNED : SIGNED; + ret = gen_extend_op(c, locp, &src_width, &dst_width, &ret, + signedness); + } + return ret; +} + +HexValue gen_rvalue_not(Context *c, YYLTYPE *locp, HexValue *v) +{ + const char *bit_suffix = (v->bit_width == 64) ? "i64" : "i32"; + int bit_width = (v->bit_width == 64) ? 64 : 32; + HexValue res; + memset(&res, 0, sizeof(HexValue)); + res.signedness = v->signedness; + res.is_dotnew = false; + res.is_manual = false; + if (v->type == IMMEDIATE) { + res.type = IMMEDIATE; + res.imm.type = QEMU_TMP; + res.imm.index = c->inst.qemu_tmp_count; + OUT(c, locp, "int", &bit_width, "_t ", &res, " = ~", v, ";\n"); + c->inst.qemu_tmp_count++; + } else { + res = gen_tmp(c, locp, bit_width); + OUT(c, locp, "tcg_gen_not_", bit_suffix, "(", &res, + ", ", v, ");\n"); + gen_rvalue_free(c, locp, v); + } + return res; +} + +HexValue gen_rvalue_notl(Context *c, YYLTYPE *locp, HexValue *v) +{ + const char *bit_suffix = (v->bit_width == 64) ? "i64" : "i32"; + int bit_width = (v->bit_width == 64) ? 64 : 32; + HexValue res; + memset(&res, 0, sizeof(HexValue)); + res.signedness = v->signedness; + res.is_dotnew = false; + res.is_manual = false; + if (v->type == IMMEDIATE) { + res.type = IMMEDIATE; + res.imm.type = QEMU_TMP; + res.imm.index = c->inst.qemu_tmp_count; + OUT(c, locp, "int", &bit_width, "_t ", &res, " = !", v, ";\n"); + c->inst.qemu_tmp_count++; + } else { + res = gen_tmp(c, locp, bit_width); + HexValue zero = gen_tmp_value(c, locp, "0", bit_width); + HexValue one = gen_tmp_value(c, locp, "0xff", bit_width); + OUT(c, locp, "tcg_gen_movcond_", bit_suffix); + OUT(c, locp, "(TCG_COND_EQ, ", &res, ", ", v, ", ", &zero); + OUT(c, locp, ", ", &one, ", ", &zero, ");\n"); + gen_rvalue_free(c, locp, v); + gen_rvalue_free(c, locp, &zero); + gen_rvalue_free(c, locp, &one); + } + return res; +} + +void gen_set_overflow(Context *c, YYLTYPE *locp, HexValue *vp) +{ + HexValue v = *vp; + + HexValue ovfl = gen_tmp(c, locp, 32); + ovfl.signedness = UNSIGNED; + OUT(c, locp, "GET_USR_FIELD(USR_OVF, ", &ovfl, ");\n"); + + if (is_inside_ternary(c)) { + /* Inside ternary operator, need to take care of the side-effect */ + HexValue cond = get_ternary_cond(c, locp); + v = rvalue_materialize(c, locp, &v); + bool is_64bit = cond.bit_width == 64; + unsigned bit_width = cond.bit_width; + if (is_64bit) { + ovfl = gen_rvalue_extend(c, locp, &ovfl); + v = gen_rvalue_extend(c, locp, &v); + } + HexValue tmp = gen_tmp_value(c, locp, "0", cond.bit_width); + OUT(c, locp, "tcg_gen_movcond_i", &bit_width, + "(TCG_COND_NE, ", &tmp, ", ", &cond); + OUT(c, locp, ", ", &tmp, ", ", &v, ", ", &tmp, ");\n"); + OUT(c, locp, "tcg_gen_or_i", &bit_width, "(", + &ovfl, ", ", &ovfl, ", ", &tmp, ");\n"); + if (is_64bit) { + ovfl = gen_rvalue_truncate(c, locp, &ovfl); + } + gen_rvalue_free(c, locp, &tmp); + gen_rvalue_free(c, locp, &cond); + } else { + if (v.type == IMMEDIATE) { + OUT(c, locp, "tcg_gen_ori_i32(", + &ovfl, ", ", &ovfl, ", ", &v, ");\n"); + } else { + OUT(c, locp, "tcg_gen_or_i32(", + &ovfl, ", ", &ovfl, ", ", &v, ");\n"); + } + } + + OUT(c, locp, "SET_USR_FIELD(USR_OVF, ", &ovfl, ");\n"); + + gen_rvalue_free(c, locp, &ovfl); + gen_rvalue_free(c, locp, &v); +} + +HexValue gen_rvalue_sat(Context *c, YYLTYPE *locp, HexSat *sat, HexValue *n, + HexValue *v) +{ + if (sat->set_overflow) { + yyassert(c, locp, n->imm.value < v->bit_width, "To compute overflow, " + "source width must be greater than saturation width!"); + } + HexValue res = gen_tmp(c, locp, v->bit_width); + const char *bit_suffix = (v->bit_width == 64) ? "i64" : "i32"; + assert_signedness(c, locp, sat->signedness); + const char *unsigned_str = (sat->signedness == UNSIGNED) ? "u" : ""; + if (sat->set_overflow) { + HexValue ovfl = gen_tmp(c, locp, 32); + OUT(c, locp, "gen_sat", unsigned_str, "_", bit_suffix, "_ovfl("); + OUT(c, locp, &ovfl, ", ", &res, ", ", v, ", ", &n->imm.value, ");\n"); + gen_set_overflow(c, locp, &ovfl); + } else { + OUT(c, locp, "gen_sat", unsigned_str, "_", bit_suffix, "(", &res, ", "); + OUT(c, locp, v, ", ", &n->imm.value, ");\n"); + } + res.signedness = sat->signedness; + gen_rvalue_free(c, locp, v); + return res; +} + +HexValue gen_rvalue_fscr(Context *c, YYLTYPE *locp, HexValue *v) +{ + HexValue key = gen_tmp(c, locp, 64); + HexValue res = gen_tmp(c, locp, 64); + *v = gen_rvalue_extend(c, locp, v); + HexValue frame_key = gen_tmp(c, locp, 32); + OUT(c, locp, "gen_read_reg(", &frame_key, ", HEX_REG_FRAMEKEY);\n"); + OUT(c, locp, "tcg_gen_concat_i32_i64(", + &key, ", ", &frame_key, ", ", &frame_key, ");\n"); + OUT(c, locp, "tcg_gen_xor_i64(", &res, ", ", v, ", ", &key, ");\n"); + gen_rvalue_free(c, locp, &key); + gen_rvalue_free(c, locp, &frame_key); + gen_rvalue_free(c, locp, v); + return res; +} + +HexValue gen_rvalue_abs(Context *c, YYLTYPE *locp, HexValue *v) +{ + int bit_width = (v->bit_width == 64) ? 64 : 32; + HexValue res; + memset(&res, 0, sizeof(HexValue)); + res.signedness = v->signedness; + res.is_dotnew = false; + res.is_manual = false; + if (v->type == IMMEDIATE) { + res.type = IMMEDIATE; + res.imm.type = QEMU_TMP; + res.imm.index = c->inst.qemu_tmp_count; + OUT(c, locp, "int", &bit_width, "_t ", &res, " = abs(", v, ");\n"); + c->inst.qemu_tmp_count++; + } else { + res = gen_tmp(c, locp, bit_width); + OUT(c, locp, "tcg_gen_abs_i", &bit_width, "(", &res, ", ", v, ");\n"); + gen_rvalue_free(c, locp, v); + } + return res; +} + +HexValue gen_rvalue_neg(Context *c, YYLTYPE *locp, HexValue *v) +{ + const char *bit_suffix = (v->bit_width == 64) ? "i64" : "i32"; + int bit_width = (v->bit_width == 64) ? 64 : 32; + HexValue res; + memset(&res, 0, sizeof(HexValue)); + res.signedness = v->signedness; + res.is_dotnew = false; + res.is_manual = false; + if (v->type == IMMEDIATE) { + res.type = IMMEDIATE; + res.imm.type = QEMU_TMP; + res.imm.index = c->inst.qemu_tmp_count; + OUT(c, locp, "int", &bit_width, "_t ", &res, " = -", v, ";\n"); + c->inst.qemu_tmp_count++; + } else { + res = gen_tmp(c, locp, bit_width); + OUT(c, locp, "tcg_gen_neg_", bit_suffix, "(", &res, ", ", v, ");\n"); + gen_rvalue_free(c, locp, v); + } + return res; +} + +HexValue gen_rvalue_brev(Context *c, YYLTYPE *locp, HexValue *v) +{ + yyassert(c, locp, v->bit_width <= 32, + "fbrev not implemented for 64-bit integers!"); + HexValue res = gen_tmp(c, locp, v->bit_width); + *v = rvalue_materialize(c, locp, v); + OUT(c, locp, "gen_helper_fbrev(", &res, ", ", v, ");\n"); + gen_rvalue_free(c, locp, v); + return res; +} + +HexValue gen_rvalue_ternary(Context *c, YYLTYPE *locp, HexValue *cond, + HexValue *t, HexValue *e) { + bool is_64bit = (t->bit_width == 64) || (e->bit_width == 64); + int bit_width = (is_64bit) ? 64 : 32; + if (is_64bit) { + *cond = gen_rvalue_extend(c, locp, cond); + *t = gen_rvalue_extend(c, locp, t); + *e = gen_rvalue_extend(c, locp, e); + } else { + *cond = gen_rvalue_truncate(c, locp, cond); + } + *cond = rvalue_materialize(c, locp, cond); + *t = rvalue_materialize(c, locp, t); + *e = rvalue_materialize(c, locp, e); + HexValue res = gen_tmp(c, locp, bit_width); + HexValue zero = gen_tmp_value(c, locp, "0", bit_width); + + OUT(c, locp, "tcg_gen_movcond_i", &bit_width); + OUT(c, locp, "(TCG_COND_NE, ", &res, ", ", cond, ", ", &zero); + OUT(c, locp, ", ", t, ", ", e, ");\n"); + + assert(c->ternary->len > 0); + Ternary *ternary = &g_array_index(c->ternary, Ternary, c->ternary->len - 1); + gen_rvalue_free_manual(c, locp, &ternary->cond); + g_array_remove_index(c->ternary, c->ternary->len - 1); + + gen_rvalue_free(c, locp, &zero); + gen_rvalue_free(c, locp, cond); + gen_rvalue_free(c, locp, t); + gen_rvalue_free(c, locp, e); + return res; +} + +const char *cond_to_str(TCGCond cond) +{ + switch (cond) { + case TCG_COND_NEVER: + return "TCG_COND_NEVER"; + case TCG_COND_ALWAYS: + return "TCG_COND_ALWAYS"; + case TCG_COND_EQ: + return "TCG_COND_EQ"; + case TCG_COND_NE: + return "TCG_COND_NE"; + case TCG_COND_LT: + return "TCG_COND_LT"; + case TCG_COND_GE: + return "TCG_COND_GE"; + case TCG_COND_LE: + return "TCG_COND_LE"; + case TCG_COND_GT: + return "TCG_COND_GT"; + case TCG_COND_LTU: + return "TCG_COND_LTU"; + case TCG_COND_GEU: + return "TCG_COND_GEU"; + case TCG_COND_LEU: + return "TCG_COND_LEU"; + case TCG_COND_GTU: + return "TCG_COND_GTU"; + default: + abort(); + } +} + +void emit_arg(Context *c, YYLTYPE *locp, HexValue *arg) +{ + switch (arg->type) { + case REGISTER_ARG: + if (arg->reg.type == DOTNEW) { + EMIT_SIG(c, ", TCGv N%cN", arg->reg.id); + } else { + bool is64 = (arg->bit_width == 64); + const char *type = is64 ? "TCGv_i64" : "TCGv_i32"; + char reg_id[5] = { 0 }; + reg_compose(c, locp, &(arg->reg), reg_id); + EMIT_SIG(c, ", %s %s", type, reg_id); + /* MuV register requires also MuN to provide its index */ + if (arg->reg.type == MODIFIER) { + EMIT_SIG(c, ", int MuN"); + } + } + break; + case PREDICATE: + { + char suffix = arg->is_dotnew ? 'N' : 'V'; + EMIT_SIG(c, ", TCGv P%c%c", arg->pre.id, suffix); + } + break; + default: + { + fprintf(stderr, "emit_arg got unsupported argument!"); + abort(); + } + } +} + +void emit_footer(Context *c) +{ + EMIT(c, "}\n"); + EMIT(c, "\n"); +} + +void track_string(Context *c, GString *s) +{ + g_array_append_val(c->inst.strings, s); +} + +void free_variables(Context *c, YYLTYPE *locp) +{ + for (unsigned i = 0; i < c->inst.allocated->len; ++i) { + Var *var = &g_array_index(c->inst.allocated, Var, i); + const char *suffix = var->bit_width == 64 ? "i64" : "i32"; + OUT(c, locp, "tcg_temp_free_", suffix, "(", var->name->str, ");\n"); + } +} + +void free_instruction(Context *c) +{ + assert(c->ternary->len == 0); + /* Free the strings */ + g_string_truncate(c->signature_str, 0); + g_string_truncate(c->out_str, 0); + g_string_truncate(c->header_str, 0); + /* Free strings allocated by the instruction */ + for (unsigned i = 0; i < c->inst.strings->len; i++) { + g_string_free(g_array_index(c->inst.strings, GString*, i), TRUE); + } + g_array_free(c->inst.strings, TRUE); + /* Free INAME token value */ + g_string_free(c->inst.name, TRUE); + /* Free variables and registers */ + g_array_free(c->inst.allocated, TRUE); + g_array_free(c->inst.init_list, TRUE); + /* Initialize instruction-specific portion of the context */ + memset(&(c->inst), 0, sizeof(Inst)); +} + +void assert_signedness(Context *c, + YYLTYPE *locp, + HexSignedness signedness) { + yyassert(c, + locp, + signedness != UNKNOWN_SIGNEDNESS, + "Unspecified signedness"); +} diff --git a/target/hexagon/idef-parser/idef-parser.y b/target/hexagon/idef-parser/idef-parser.y new file mode 100644 index 0000000000..fe13d83f51 --- /dev/null +++ b/target/hexagon/idef-parser/idef-parser.y @@ -0,0 +1,975 @@ +%{ +/* + * Copyright(c) 2019-2021 rev.ng Srls. All Rights Reserved. + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, see <http://www.gnu.org/licenses/>. + */ + +#include "idef-parser.h" +#include "parser-helpers.h" +#include "idef-parser.tab.h" +#include "idef-parser.yy.h" + +/* Uncomment this to disable yyasserts */ +/* #define NDEBUG */ + +#define ERR_LINE_CONTEXT 40 + +%} + +%lex-param {void *scanner} +%parse-param {void *scanner} +%parse-param {Context *c} + +%define parse.error verbose +%define parse.lac full +%define api.pure full + +%locations + +%union { + GString *string; + HexValue rvalue; + HexSat sat; + HexCast cast; + HexExtract extract; + HexMpy mpy; + HexSignedness signedness; + int index; +} + +/* Tokens */ +%start input + +%expect 1 + +%token INAME DREG DIMM DPRE DEA RREG WREG FREG FIMM RPRE WPRE FPRE FWRAP FEA VAR +%token ABS CROUND ROUND CIRCADD COUNTONES INC DEC ANDA ORA XORA PLUSPLUS ASL +%token ASR LSR EQ NEQ LTE GTE MIN MAX ANDL ORL FOR ICIRC IF MUN FSCR FCHK SXT +%token ZXT CONSTEXT LOCNT BREV SIGN LOAD STORE CONSTLL CONSTULL PC NPC LPCFG +%token CANCEL IDENTITY PART1 BREV_4 BREV_8 ROTL INSBITS SETBITS EXTBITS EXTRANGE +%token CAST4_8U SETOVF FAIL DEINTERLEAVE INTERLEAVE CARRY_FROM_ADD LSBNEW + +%token <rvalue> REG IMM PRE +%token <index> ELSE +%token <mpy> MPY +%token <sat> SAT +%token <cast> CAST DEPOSIT SETHALF +%token <extract> EXTRACT +%type <string> INAME +%type <rvalue> rvalue lvalue VAR assign_statement var +%type <rvalue> DREG DIMM DPRE RREG RPRE FAIL +%type <index> if_stmt IF +%type <signedness> SIGN + +/* Operator Precedences */ +%left MIN MAX +%left '(' +%left ',' +%left '=' +%right CIRCADD +%right INC DEC ANDA ORA XORA +%left '?' ':' +%left ORL +%left ANDL +%left '|' +%left '^' ANDOR +%left '&' +%left EQ NEQ +%left '<' '>' LTE GTE +%left ASL ASR LSR +%right ABS +%left '-' '+' +%left '*' '/' '%' MPY +%right '~' '!' +%left '[' +%right CAST +%right LOCNT BREV + +/* Bison Grammar */ +%% + +/* Input file containing the description of each hexagon instruction */ +input : instructions + { + YYACCEPT; + } + ; + +instructions : instruction instructions + | %empty + ; + +instruction : INAME + { + gen_inst(c, $1); + } + arguments + { + gen_inst_args(c, &@1); + } + code + { + gen_inst_code(c, &@1); + } + | error /* Recover gracefully after instruction compilation error */ + { + free_instruction(c); + } + ; + +arguments : '(' ')' + | '(' argument_list ')'; + +argument_list : decl ',' argument_list + | decl + ; + +var : VAR + { + track_string(c, $1.var.name); + $$ = $1; + } + ; + +/* Return the modified registers list */ +code : '{' statements '}' + { + c->inst.code_begin = c->input_buffer + @2.first_column; + c->inst.code_end = c->input_buffer + @2.last_column - 1; + } + | '{' + { + /* Nop */ + } + '}' + ; + +decl : REG + { + emit_arg(c, &@1, &$1); + /* Enqueue register into initialization list */ + g_array_append_val(c->inst.init_list, $1); + } + | IMM + { + EMIT_SIG(c, ", int %ciV", $1.imm.id); + } + | PRE + { + emit_arg(c, &@1, &$1); + /* Enqueue predicate into initialization list */ + g_array_append_val(c->inst.init_list, $1); + } + | var + { + yyassert(c, &@1, !strcmp($1.var.name->str, "EA"), + "Unknown argument variable!"); + } + | RREG + { + emit_arg(c, &@1, &$1); + } + | WREG + | FREG + | FIMM + | RPRE + { + emit_arg(c, &@1, &$1); + } + | WPRE + | FPRE + | FEA + ; + +code_block : '{' statements '}' + | '{' '}' + ; + +/* A list of one or more statements */ +statements : statements statement + | statement + ; + +/* Statements can be assignment (rvalue ';'), control or memory statements */ +statement : control_statement + | rvalue ';' + { + gen_rvalue_free(c, &@1, &$1); + } + | code_block + | ';' + ; + +assign_statement : lvalue '=' rvalue + { + @1.last_column = @3.last_column; + gen_assign(c, &@1, &$1, &$3); + $$ = $1; + } + | lvalue INC rvalue + { + @1.last_column = @3.last_column; + HexValue tmp = gen_bin_op(c, &@1, ADD_OP, &$1, &$3); + gen_assign(c, &@1, &$1, &tmp); + $$ = $1; + } + | lvalue DEC rvalue + { + @1.last_column = @3.last_column; + HexValue tmp = gen_bin_op(c, &@1, SUB_OP, &$1, &$3); + gen_assign(c, &@1, &$1, &tmp); + $$ = $1; + } + | lvalue ANDA rvalue + { + @1.last_column = @3.last_column; + HexValue tmp = gen_bin_op(c, &@1, ANDB_OP, &$1, &$3); + gen_assign(c, &@1, &$1, &tmp); + $$ = $1; + } + | lvalue ORA rvalue + { + @1.last_column = @3.last_column; + HexValue tmp = gen_bin_op(c, &@1, ORB_OP, &$1, &$3); + gen_assign(c, &@1, &$1, &tmp); + $$ = $1; + } + | lvalue XORA rvalue + { + @1.last_column = @3.last_column; + HexValue tmp = gen_bin_op(c, &@1, XORB_OP, &$1, &$3); + gen_assign(c, &@1, &$1, &tmp); + $$ = $1; + } + | PRE '=' rvalue + { + @1.last_column = @3.last_column; + gen_pre_assign(c, &@1, &$1, &$3); + } + | IMM '=' rvalue + { + @1.last_column = @3.last_column; + yyassert(c, &@1, $3.type == IMMEDIATE, + "Cannot assign non-immediate to immediate!"); + yyassert(c, &@1, $1.imm.type == VARIABLE, + "Cannot assign to non-variable!"); + /* Assign to the function argument */ + OUT(c, &@1, &$1, " = ", &$3, ";\n"); + $$ = $1; + } + | PC '=' rvalue + { + @1.last_column = @3.last_column; + yyassert(c, &@1, c->ternary->len == 0, + "Assignment side-effect not modeled!"); + $3 = gen_rvalue_truncate(c, &@1, &$3); + $3 = rvalue_materialize(c, &@1, &$3); + OUT(c, &@1, "gen_write_new_pc(", &$3, ");\n"); + gen_rvalue_free(c, &@1, &$3); /* Free temporary value */ + } + | LOAD '(' IMM ',' IMM ',' SIGN ',' var ',' lvalue ')' + { + @1.last_column = @12.last_column; + yyassert(c, &@1, c->ternary->len == 0, + "Assignment side-effect not modeled!"); + yyassert(c, &@1, $3.imm.value == 1, + "LOAD of arrays not supported!"); + gen_load(c, &@1, &$5, $7, &$9, &$11); + } + | STORE '(' IMM ',' IMM ',' var ',' rvalue ')' + /* Store primitive */ + { + @1.last_column = @10.last_column; + yyassert(c, &@1, c->ternary->len == 0, + "Assignment side-effect not modeled!"); + yyassert(c, &@1, $3.imm.value == 1, + "STORE of arrays not supported!"); + gen_store(c, &@1, &$5, &$7, &$9); + } + | LPCFG '=' rvalue + { + @1.last_column = @3.last_column; + yyassert(c, &@1, c->ternary->len == 0, + "Assignment side-effect not modeled!"); + $3 = gen_rvalue_truncate(c, &@1, &$3); + $3 = rvalue_materialize(c, &@1, &$3); + OUT(c, &@1, "SET_USR_FIELD(USR_LPCFG, ", &$3, ");\n"); + gen_rvalue_free(c, &@1, &$3); + } + | DEPOSIT '(' rvalue ',' rvalue ',' rvalue ')' + { + @1.last_column = @8.last_column; + yyassert(c, &@1, c->ternary->len == 0, + "Assignment side-effect not modeled!"); + gen_deposit_op(c, &@1, &$5, &$7, &$3, &$1); + } + | SETHALF '(' rvalue ',' lvalue ',' rvalue ')' + { + @1.last_column = @8.last_column; + yyassert(c, &@1, c->ternary->len == 0, + "Assignment side-effect not modeled!"); + gen_sethalf(c, &@1, &$1, &$3, &$5, &$7); + } + | SETBITS '(' rvalue ',' rvalue ',' rvalue ',' rvalue ')' + { + @1.last_column = @10.last_column; + yyassert(c, &@1, c->ternary->len == 0, + "Assignment side-effect not modeled!"); + gen_setbits(c, &@1, &$3, &$5, &$7, &$9); + } + | INSBITS '(' lvalue ',' rvalue ',' rvalue ',' rvalue ')' + { + @1.last_column = @10.last_column; + yyassert(c, &@1, c->ternary->len == 0, + "Assignment side-effect not modeled!"); + gen_rdeposit_op(c, &@1, &$3, &$9, &$7, &$5); + } + | IDENTITY '(' rvalue ')' + { + @1.last_column = @4.last_column; + $$ = $3; + } + ; + +control_statement : frame_check + | cancel_statement + | if_statement + | for_statement + | fpart1_statement + ; + +frame_check : FCHK '(' rvalue ',' rvalue ')' ';' + { + gen_rvalue_free(c, &@1, &$3); + gen_rvalue_free(c, &@1, &$5); + } + ; + +cancel_statement : CANCEL + { + OUT(c, &@1, "gen_cancel(insn->slot);\n"); + } + ; + +if_statement : if_stmt + { + /* Fix else label */ + OUT(c, &@1, "gen_set_label(if_label_", &$1, ");\n"); + } + | if_stmt ELSE + { + @1.last_column = @2.last_column; + $2 = gen_if_else(c, &@1, $1); + } + statement + { + OUT(c, &@1, "gen_set_label(if_label_", &$2, ");\n"); + } + ; + +for_statement : FOR '(' IMM '=' IMM ';' IMM '<' IMM ';' IMM PLUSPLUS ')' + { + @1.last_column = @13.last_column; + OUT(c, &@1, "for (int ", &$3, " = ", &$5, "; ", + &$7, " < ", &$9); + OUT(c, &@1, "; ", &$11, "++) {\n"); + } + code_block + { + OUT(c, &@1, "}\n"); + } + | FOR '(' IMM '=' IMM ';' IMM '<' IMM ';' IMM INC IMM ')' + { + @1.last_column = @14.last_column; + OUT(c, &@1, "for (int ", &$3, " = ", &$5, "; ", + &$7, " < ", &$9); + OUT(c, &@1, "; ", &$11, " += ", &$13, ") {\n"); + } + code_block + { + OUT(c, &@1, "}\n"); + } + ; + +fpart1_statement : PART1 + { + OUT(c, &@1, "if (insn->part1) {\n"); + } + '(' statements ')' + { + @1.last_column = @3.last_column; + OUT(c, &@1, "return; }\n"); + } + ; + +if_stmt : IF '(' rvalue ')' + { + @1.last_column = @3.last_column; + $1 = gen_if_cond(c, &@1, &$3); + } + statement + { + $$ = $1; + } + ; + +rvalue : FAIL + { + yyassert(c, &@1, false, "Encountered a FAIL token as rvalue.\n"); + } + | assign_statement + | REG + { + if ($1.type == REGISTER_ARG) { + $$ = gen_read_reg(c, &@1, &$1); + } else { + $$ = $1; + } + } + | IMM + { + $$ = $1; + } + | CONSTLL '(' IMM ')' + { + $3.signedness = SIGNED; + $3.bit_width = 64; + $$ = $3; + } + | CONSTULL '(' IMM ')' + { + $3.signedness = UNSIGNED; + $3.bit_width = 64; + $$ = $3; + } + | PRE + { + $$ = gen_rvalue_pre(c, &@1, &$1); + } + | PC + { + /* Read PC from the CR */ + HexValue rvalue; + memset(&rvalue, 0, sizeof(HexValue)); + rvalue.type = IMMEDIATE; + rvalue.imm.type = IMM_PC; + rvalue.bit_width = 32; + rvalue.signedness = UNSIGNED; + $$ = rvalue; + } + | NPC + { + /* NPC is only read from CALLs, so we can hardcode it + at translation time */ + HexValue rvalue; + memset(&rvalue, 0, sizeof(HexValue)); + rvalue.type = IMMEDIATE; + rvalue.imm.type = IMM_NPC; + rvalue.bit_width = 32; + rvalue.signedness = UNSIGNED; + $$ = rvalue; + } + | CONSTEXT + { + HexValue rvalue; + memset(&rvalue, 0, sizeof(HexValue)); + rvalue.type = IMMEDIATE; + rvalue.imm.type = IMM_CONSTEXT; + rvalue.signedness = UNSIGNED; + rvalue.is_dotnew = false; + rvalue.is_manual = false; + $$ = rvalue; + } + | var + { + $$ = gen_rvalue_var(c, &@1, &$1); + } + | MPY '(' rvalue ',' rvalue ')' + { + @1.last_column = @6.last_column; + $$ = gen_rvalue_mpy(c, &@1, &$1, &$3, &$5); + } + | rvalue '+' rvalue + { + @1.last_column = @3.last_column; + $$ = gen_bin_op(c, &@1, ADD_OP, &$1, &$3); + } + | rvalue '-' rvalue + { + @1.last_column = @3.last_column; + $$ = gen_bin_op(c, &@1, SUB_OP, &$1, &$3); + } + | rvalue '*' rvalue + { + @1.last_column = @3.last_column; + $$ = gen_bin_op(c, &@1, MUL_OP, &$1, &$3); + } + | rvalue '%' rvalue + { + @1.last_column = @3.last_column; + $$ = gen_bin_op(c, &@1, MOD_OP, &$1, &$3); + } + | rvalue ASL rvalue + { + @1.last_column = @3.last_column; + $$ = gen_bin_op(c, &@1, ASL_OP, &$1, &$3); + } + | rvalue ASR rvalue + { + @1.last_column = @3.last_column; + assert_signedness(c, &@1, $1.signedness); + if ($1.signedness == UNSIGNED) { + $$ = gen_bin_op(c, &@1, LSR_OP, &$1, &$3); + } else if ($1.signedness == SIGNED) { + $$ = gen_bin_op(c, &@1, ASR_OP, &$1, &$3); + } + } + | rvalue LSR rvalue + { + @1.last_column = @3.last_column; + $$ = gen_bin_op(c, &@1, LSR_OP, &$1, &$3); + } + | rvalue '&' rvalue + { + @1.last_column = @3.last_column; + $$ = gen_bin_op(c, &@1, ANDB_OP, &$1, &$3); + } + | rvalue '|' rvalue + { + @1.last_column = @3.last_column; + $$ = gen_bin_op(c, &@1, ORB_OP, &$1, &$3); + } + | rvalue '^' rvalue + { + @1.last_column = @3.last_column; + $$ = gen_bin_op(c, &@1, XORB_OP, &$1, &$3); + } + | rvalue ANDL rvalue + { + @1.last_column = @3.last_column; + $$ = gen_bin_op(c, &@1, ANDL_OP, &$1, &$3); + } + | MIN '(' rvalue ',' rvalue ')' + { + @1.last_column = @3.last_column; + $$ = gen_bin_op(c, &@1, MINI_OP, &$3, &$5); + } + | MAX '(' rvalue ',' rvalue ')' + { + @1.last_column = @3.last_column; + $$ = gen_bin_op(c, &@1, MAXI_OP, &$3, &$5); + } + | '~' rvalue + { + @1.last_column = @2.last_column; + $$ = gen_rvalue_not(c, &@1, &$2); + } + | '!' rvalue + { + @1.last_column = @2.last_column; + $$ = gen_rvalue_notl(c, &@1, &$2); + } + | SAT '(' IMM ',' rvalue ')' + { + @1.last_column = @6.last_column; + $$ = gen_rvalue_sat(c, &@1, &$1, &$3, &$5); + } + | CAST rvalue + { + @1.last_column = @2.last_column; + /* Assign target signedness */ + $2.signedness = $1.signedness; + $$ = gen_cast_op(c, &@1, &$2, $1.bit_width); + $$.signedness = $1.signedness; + } + | rvalue '[' rvalue ']' + { + @1.last_column = @4.last_column; + if ($3.type == IMMEDIATE) { + $$ = gen_tmp(c, &@1, $1.bit_width); + OUT(c, &@1, "tcg_gen_extract_i", &$$.bit_width, "("); + OUT(c, &@1, &$$, ", ", &$1, ", ", &$3, ", 1);\n"); + } else { + HexValue one = gen_imm_value(c, &@1, 1, $3.bit_width); + HexValue tmp = gen_bin_op(c, &@1, ASR_OP, &$1, &$3); + $$ = gen_bin_op(c, &@1, ANDB_OP, &tmp, &one); + } + } + | rvalue EQ rvalue + { + @1.last_column = @3.last_column; + $$ = gen_bin_cmp(c, &@1, TCG_COND_EQ, &$1, &$3); + } + | rvalue NEQ rvalue + { + @1.last_column = @3.last_column; + $$ = gen_bin_cmp(c, &@1, TCG_COND_NE, &$1, &$3); + } + | rvalue '<' rvalue + { + @1.last_column = @3.last_column; + + assert_signedness(c, &@1, $1.signedness); + assert_signedness(c, &@1, $3.signedness); + if ($1.signedness == UNSIGNED || $3.signedness == UNSIGNED) { + $$ = gen_bin_cmp(c, &@1, TCG_COND_LTU, &$1, &$3); + } else { + $$ = gen_bin_cmp(c, &@1, TCG_COND_LT, &$1, &$3); + } + } + | rvalue '>' rvalue + { + @1.last_column = @3.last_column; + + assert_signedness(c, &@1, $1.signedness); + assert_signedness(c, &@1, $3.signedness); + if ($1.signedness == UNSIGNED || $3.signedness == UNSIGNED) { + $$ = gen_bin_cmp(c, &@1, TCG_COND_GTU, &$1, &$3); + } else { + $$ = gen_bin_cmp(c, &@1, TCG_COND_GT, &$1, &$3); + } + } + | rvalue LTE rvalue + { + @1.last_column = @3.last_column; + + assert_signedness(c, &@1, $1.signedness); + assert_signedness(c, &@1, $3.signedness); + if ($1.signedness == UNSIGNED || $3.signedness == UNSIGNED) { + $$ = gen_bin_cmp(c, &@1, TCG_COND_LEU, &$1, &$3); + } else { + $$ = gen_bin_cmp(c, &@1, TCG_COND_LE, &$1, &$3); + } + } + | rvalue GTE rvalue + { + @1.last_column = @3.last_column; + + assert_signedness(c, &@1, $1.signedness); + assert_signedness(c, &@1, $3.signedness); + if ($1.signedness == UNSIGNED || $3.signedness == UNSIGNED) { + $$ = gen_bin_cmp(c, &@1, TCG_COND_GEU, &$1, &$3); + } else { + $$ = gen_bin_cmp(c, &@1, TCG_COND_GE, &$1, &$3); + } + } + | rvalue '?' + { + $1.is_manual = true; + Ternary t = { 0 }; + t.state = IN_LEFT; + t.cond = $1; + g_array_append_val(c->ternary, t); + } + rvalue ':' + { + Ternary *t = &g_array_index(c->ternary, Ternary, + c->ternary->len - 1); + t->state = IN_RIGHT; + } + rvalue + { + @1.last_column = @5.last_column; + $$ = gen_rvalue_ternary(c, &@1, &$1, &$4, &$7); + } + | FSCR '(' rvalue ')' + { + @1.last_column = @4.last_column; + $$ = gen_rvalue_fscr(c, &@1, &$3); + } + | SXT '(' rvalue ',' IMM ',' rvalue ')' + { + @1.last_column = @8.last_column; + yyassert(c, &@1, $5.type == IMMEDIATE && + $5.imm.type == VALUE, + "SXT expects immediate values\n"); + $5.imm.value = 64; + $$ = gen_extend_op(c, &@1, &$3, &$5, &$7, SIGNED); + } + | ZXT '(' rvalue ',' IMM ',' rvalue ')' + { + @1.last_column = @8.last_column; + yyassert(c, &@1, $5.type == IMMEDIATE && + $5.imm.type == VALUE, + "ZXT expects immediate values\n"); + $$ = gen_extend_op(c, &@1, &$3, &$5, &$7, UNSIGNED); + } + | '(' rvalue ')' + { + $$ = $2; + } + | ABS rvalue + { + @1.last_column = @2.last_column; + $$ = gen_rvalue_abs(c, &@1, &$2); + } + | CROUND '(' rvalue ',' rvalue ')' + { + @1.last_column = @6.last_column; + $$ = gen_convround_n(c, &@1, &$3, &$5); + } + | CROUND '(' rvalue ')' + { + @1.last_column = @4.last_column; + $$ = gen_convround(c, &@1, &$3); + } + | ROUND '(' rvalue ',' rvalue ')' + { + @1.last_column = @6.last_column; + $$ = gen_round(c, &@1, &$3, &$5); + } + | '-' rvalue + { + @1.last_column = @2.last_column; + $$ = gen_rvalue_neg(c, &@1, &$2); + } + | ICIRC '(' rvalue ')' ASL IMM + { + @1.last_column = @6.last_column; + $$ = gen_tmp(c, &@1, 32); + OUT(c, &@1, "gen_read_ireg(", &$$, ", ", &$3, ", ", &$6, ");\n"); + gen_rvalue_free(c, &@1, &$3); + } + | CIRCADD '(' rvalue ',' rvalue ',' rvalue ')' + { + @1.last_column = @8.last_column; + gen_circ_op(c, &@1, &$3, &$5, &$7); + } + | LOCNT '(' rvalue ')' + { + @1.last_column = @4.last_column; + /* Leading ones count */ + $$ = gen_locnt_op(c, &@1, &$3); + } + | COUNTONES '(' rvalue ')' + { + @1.last_column = @4.last_column; + /* Ones count */ + $$ = gen_ctpop_op(c, &@1, &$3); + } + | LPCFG + { + $$ = gen_tmp_value(c, &@1, "0", 32); + OUT(c, &@1, "tcg_gen_extract_tl(", &$$, + ", hex_gpr[HEX_REG_USR], "); + OUT(c, &@1, "reg_field_info[USR_LPCFG].offset, "); + OUT(c, &@1, "reg_field_info[USR_LPCFG].width);\n"); + } + | EXTRACT '(' rvalue ',' rvalue ')' + { + @1.last_column = @6.last_column; + $$ = gen_extract_op(c, &@1, &$5, &$3, &$1); + } + | EXTBITS '(' rvalue ',' rvalue ',' rvalue ')' + { + @1.last_column = @8.last_column; + yyassert(c, &@1, $5.type == IMMEDIATE && + $5.imm.type == VALUE && + $7.type == IMMEDIATE && + $7.imm.type == VALUE, + "Range extract needs immediate values!\n"); + $$ = gen_rextract_op(c, &@1, &$3, $7.imm.value, $5.imm.value); + } + | EXTRANGE '(' rvalue ',' rvalue ',' rvalue ')' + { + @1.last_column = @8.last_column; + yyassert(c, &@1, $5.type == IMMEDIATE && + $5.imm.type == VALUE && + $7.type == IMMEDIATE && + $7.imm.type == VALUE, + "Range extract needs immediate values!\n"); + $$ = gen_rextract_op(c, + &@1, + &$3, + $7.imm.value, + $5.imm.value - $7.imm.value + 1); + } + | CAST4_8U '(' rvalue ')' + { + @1.last_column = @4.last_column; + $$ = gen_rvalue_truncate(c, &@1, &$3); + $$.signedness = UNSIGNED; + $$ = rvalue_materialize(c, &@1, &$$); + $$ = gen_rvalue_extend(c, &@1, &$$); + } + | BREV '(' rvalue ')' + { + @1.last_column = @4.last_column; + $$ = gen_rvalue_brev(c, &@1, &$3); + } + | BREV_4 '(' rvalue ')' + { + @1.last_column = @4.last_column; + $$ = gen_fbrev_4(c, &@1, &$3); + } + | BREV_8 '(' rvalue ')' + { + @1.last_column = @4.last_column; + $$ = gen_fbrev_8(c, &@1, &$3); + } + | ROTL '(' rvalue ',' rvalue ')' + { + @1.last_column = @6.last_column; + $$ = gen_rotl(c, &@1, &$3, &$5); + } + | SETOVF '(' ')' + { + @1.last_column = @3.last_column; + HexValue ovfl = gen_imm_value(c, &@1, 1, 32); + gen_set_overflow(c, &@1, &ovfl); + } + | SETOVF '(' rvalue ')' + { + /* Convenience fSET_OVERFLOW with pass-through */ + @1.last_column = @3.last_column; + HexValue ovfl = gen_imm_value(c, &@1, 1, 32); + gen_set_overflow(c, &@1, &ovfl); + $$ = $3; + } + | DEINTERLEAVE '(' rvalue ')' + { + @1.last_column = @4.last_column; + $$ = gen_deinterleave(c, &@1, &$3); + } + | INTERLEAVE '(' rvalue ',' rvalue ')' + { + @1.last_column = @6.last_column; + $$ = gen_interleave(c, &@1, &$3, &$5); + } + | CARRY_FROM_ADD '(' rvalue ',' rvalue ',' rvalue ')' + { + @1.last_column = @8.last_column; + $$ = gen_carry_from_add(c, &@1, &$3, &$5, &$7); + } + | LSBNEW '(' rvalue ')' + { + @1.last_column = @4.last_column; + HexValue one = gen_imm_value(c, &@1, 1, 32); + $$ = gen_bin_op(c, &@1, ANDB_OP, &$3, &one); + } + ; + +lvalue : FAIL + { + @1.last_column = @1.last_column; + yyassert(c, &@1, false, "Encountered a FAIL token as lvalue.\n"); + } + | REG + { + $$ = $1; + } + | var + { + $$ = $1; + } + ; + +%% + +int main(int argc, char **argv) +{ + if (argc != 5) { + fprintf(stderr, + "Semantics: Hexagon ISA to tinycode generator compiler\n\n"); + fprintf(stderr, + "Usage: ./semantics IDEFS EMITTER_C EMITTER_H " + "ENABLED_INSTRUCTIONS_LIST\n"); + return 1; + } + + enum { + ARG_INDEX_ARGV0 = 0, + ARG_INDEX_IDEFS, + ARG_INDEX_EMITTER_C, + ARG_INDEX_EMITTER_H, + ARG_INDEX_ENABLED_INSTRUCTIONS_LIST + }; + + FILE *enabled_file = fopen(argv[ARG_INDEX_ENABLED_INSTRUCTIONS_LIST], "w"); + + FILE *output_file = fopen(argv[ARG_INDEX_EMITTER_C], "w"); + fputs("#include \"qemu/osdep.h\"\n", output_file); + fputs("#include \"qemu/log.h\"\n", output_file); + fputs("#include \"cpu.h\"\n", output_file); + fputs("#include \"internal.h\"\n", output_file); + fputs("#include \"tcg/tcg-op.h\"\n", output_file); + fputs("#include \"insn.h\"\n", output_file); + fputs("#include \"opcodes.h\"\n", output_file); + fputs("#include \"translate.h\"\n", output_file); + fputs("#define QEMU_GENERATE\n", output_file); + fputs("#include \"genptr.h\"\n", output_file); + fputs("#include \"tcg/tcg.h\"\n", output_file); + fputs("#include \"macros.h\"\n", output_file); + fprintf(output_file, "#include \"%s\"\n", argv[ARG_INDEX_EMITTER_H]); + + FILE *defines_file = fopen(argv[ARG_INDEX_EMITTER_H], "w"); + assert(defines_file != NULL); + fputs("#ifndef HEX_EMITTER_H\n", defines_file); + fputs("#define HEX_EMITTER_H\n", defines_file); + fputs("\n", defines_file); + fputs("#include \"insn.h\"\n\n", defines_file); + + /* Parser input file */ + Context context = { 0 }; + context.defines_file = defines_file; + context.output_file = output_file; + context.enabled_file = enabled_file; + /* Initialize buffers */ + context.out_str = g_string_new(NULL); + context.signature_str = g_string_new(NULL); + context.header_str = g_string_new(NULL); + context.ternary = g_array_new(FALSE, TRUE, sizeof(Ternary)); + /* Read input file */ + FILE *input_file = fopen(argv[ARG_INDEX_IDEFS], "r"); + fseek(input_file, 0L, SEEK_END); + long input_size = ftell(input_file); + context.input_buffer = (char *) calloc(input_size + 1, sizeof(char)); + fseek(input_file, 0L, SEEK_SET); + size_t read_chars = fread(context.input_buffer, + sizeof(char), + input_size, + input_file); + if (read_chars != (size_t) input_size) { + fprintf(stderr, "Error: an error occurred while reading input file!\n"); + return -1; + } + yylex_init(&context.scanner); + YY_BUFFER_STATE buffer; + buffer = yy_scan_string(context.input_buffer, context.scanner); + /* Start the parsing procedure */ + yyparse(context.scanner, &context); + if (context.implemented_insn != context.total_insn) { + fprintf(stderr, + "Warning: %d/%d meta instructions have been implemented!\n", + context.implemented_insn, + context.total_insn); + } + fputs("#endif " START_COMMENT " HEX_EMITTER_h " END_COMMENT "\n", + defines_file); + /* Cleanup */ + yy_delete_buffer(buffer, context.scanner); + yylex_destroy(context.scanner); + free(context.input_buffer); + g_string_free(context.out_str, TRUE); + g_string_free(context.signature_str, TRUE); + g_string_free(context.header_str, TRUE); + g_array_free(context.ternary, TRUE); + fclose(output_file); + fclose(input_file); + fclose(defines_file); + fclose(enabled_file); + + return 0; +} diff --git a/target/hexagon/meson.build b/target/hexagon/meson.build index 839f965f3c..c064186b5c 100644 --- a/target/hexagon/meson.build +++ b/target/hexagon/meson.build @@ -184,7 +184,7 @@ idef_parser_input_generated = custom_target( command: [python, files('gen_idef_parser_funcs.py'), semantics_generated, attribs_def, gen_tcg_h, '@OUTPUT@'], ) -idef_parser_input_generated_prep = custom_target( +preprocessed_idef_parser_input_generated = custom_target( 'idef_parser_input.preprocessed.h.inc', output: 'idef_parser_input.preprocessed.h.inc', input: idef_parser_input_generated, @@ -195,4 +195,29 @@ flex = generator(find_program('flex'), output: ['@BASENAME@.yy.c', '@BASENAME@.yy.h'], arguments: ['-o', '@OUTPUT0@', '--header-file=@OUTPUT1@', '@INPUT@']) +bison = generator(find_program('bison'), + output: ['@BASENAME@.tab.c', '@BASENAME@.tab.h'], + arguments: ['@INPUT@', '--defines=@OUTPUT1@', '--output=@OUTPUT0@']) + +glib_dep = dependency('glib-2.0', native: true) + +idef_parser = executable('idef-parser', + [flex.process(idef_parser_dir / 'idef-parser.lex'), + bison.process(idef_parser_dir / 'idef-parser.y'), + idef_parser_dir / 'parser-helpers.c'], + include_directories: ['idef-parser', '../../include/'], + dependencies: [glib_dep], + c_args: ['-Wextra'], + native: true) + +idef_generated_tcg = custom_target( + 'idef-generated-tcg', + output: ['idef-generated-emitter.c', + 'idef-generated-emitter.h.inc', + 'idef-generated-enabled-instructions'], + input: preprocessed_idef_parser_input_generated, + depend_files: [hex_common_py], + command: [idef_parser, '@INPUT@', '@OUTPUT0@', '@OUTPUT1@', '@OUTPUT2@'], +) + target_arch += {'hexagon': hexagon_ss} diff --git a/tests/docker/dockerfiles/alpine.docker b/tests/docker/dockerfiles/alpine.docker index a6e593fb49..45db55ba55 100644 --- a/tests/docker/dockerfiles/alpine.docker +++ b/tests/docker/dockerfiles/alpine.docker @@ -9,6 +9,7 @@ ENV PACKAGES \ alsa-lib-dev \ bash \ binutils \ + bison \ ccache \ coreutils \ curl-dev \ diff --git a/tests/docker/dockerfiles/centos8.docker b/tests/docker/dockerfiles/centos8.docker index 7d67cc2397..43db384b38 100644 --- a/tests/docker/dockerfiles/centos8.docker +++ b/tests/docker/dockerfiles/centos8.docker @@ -5,6 +5,7 @@ ENV PACKAGES \ SDL2-devel \ alsa-lib-devel \ bc \ + bison \ brlapi-devel \ bzip2 \ bzip2-devel \ diff --git a/tests/docker/dockerfiles/debian-amd64.docker b/tests/docker/dockerfiles/debian-amd64.docker index cd4d8ebe9c..805fd6f981 100644 --- a/tests/docker/dockerfiles/debian-amd64.docker +++ b/tests/docker/dockerfiles/debian-amd64.docker @@ -14,6 +14,7 @@ RUN apt update && \ RUN apt update && \ DEBIAN_FRONTEND=noninteractive eatmydata \ apt install -y --no-install-recommends \ + bison \ cscope \ genisoimage \ exuberant-ctags \ diff --git a/tests/docker/dockerfiles/debian10.docker b/tests/docker/dockerfiles/debian10.docker index 8a0bdf39ea..06b8d9d6b0 100644 --- a/tests/docker/dockerfiles/debian10.docker +++ b/tests/docker/dockerfiles/debian10.docker @@ -17,6 +17,7 @@ RUN apt update && \ DEBIAN_FRONTEND=noninteractive apt install -yy eatmydata && \ DEBIAN_FRONTEND=noninteractive eatmydata \ apt install -y --no-install-recommends \ + bison \ bc \ build-essential \ ca-certificates \ @@ -28,6 +29,7 @@ RUN apt update && \ gettext \ git \ libffi-dev \ + libglib2.0-dev \ libncurses5-dev \ ninja-build \ pkg-config \ diff --git a/tests/docker/dockerfiles/fedora-i386-cross.docker b/tests/docker/dockerfiles/fedora-i386-cross.docker index 99b68fbd53..ab974e6dce 100644 --- a/tests/docker/dockerfiles/fedora-i386-cross.docker +++ b/tests/docker/dockerfiles/fedora-i386-cross.docker @@ -1,5 +1,6 @@ FROM registry.fedoraproject.org/fedora:33 ENV PACKAGES \ + bison \ bzip2 \ ccache \ diffutils \ @@ -7,6 +8,7 @@ ENV PACKAGES \ flex \ gcc \ git \ + glib2-devel \ libffi-devel.i686 \ libtasn1-devel.i686 \ libzstd-devel.i686 \ diff --git a/tests/docker/dockerfiles/fedora-win32-cross.docker b/tests/docker/dockerfiles/fedora-win32-cross.docker index e5116adbb3..bcdae720c6 100644 --- a/tests/docker/dockerfiles/fedora-win32-cross.docker +++ b/tests/docker/dockerfiles/fedora-win32-cross.docker @@ -2,6 +2,7 @@ FROM registry.fedoraproject.org/fedora:33 # Please keep this list sorted alphabetically ENV PACKAGES \ + bison \ bc \ bzip2 \ ccache \ @@ -11,6 +12,7 @@ ENV PACKAGES \ gcc \ gettext \ git \ + glib2-devel \ hostname \ make \ meson \ diff --git a/tests/docker/dockerfiles/fedora-win64-cross.docker b/tests/docker/dockerfiles/fedora-win64-cross.docker index b564811c41..dee3555cc0 100644 --- a/tests/docker/dockerfiles/fedora-win64-cross.docker +++ b/tests/docker/dockerfiles/fedora-win64-cross.docker @@ -2,6 +2,7 @@ FROM registry.fedoraproject.org/fedora:33 # Please keep this list sorted alphabetically ENV PACKAGES \ + bison \ bc \ bzip2 \ ccache \ @@ -10,6 +11,7 @@ ENV PACKAGES \ flex \ gcc \ gettext \ + glib2-devel \ git \ hostname \ make \ diff --git a/tests/docker/dockerfiles/fedora.docker b/tests/docker/dockerfiles/fedora.docker index e35366f33d..a1f5904181 100644 --- a/tests/docker/dockerfiles/fedora.docker +++ b/tests/docker/dockerfiles/fedora.docker @@ -2,6 +2,7 @@ FROM registry.fedoraproject.org/fedora:33 # Please keep this list sorted alphabetically ENV PACKAGES \ + bison \ SDL2-devel \ SDL2_image-devel \ alsa-lib-devel \ diff --git a/tests/docker/dockerfiles/opensuse-leap.docker b/tests/docker/dockerfiles/opensuse-leap.docker index 5826260fff..175dfdc119 100644 --- a/tests/docker/dockerfiles/opensuse-leap.docker +++ b/tests/docker/dockerfiles/opensuse-leap.docker @@ -5,6 +5,7 @@ ENV PACKAGES \ Mesa-devel \ alsa-lib-devel \ bc \ + bison \ brlapi-devel \ bzip2 \ ca-certificates \ diff --git a/tests/docker/dockerfiles/ubuntu.docker b/tests/docker/dockerfiles/ubuntu.docker index e1832b4ab3..5667ef592c 100644 --- a/tests/docker/dockerfiles/ubuntu.docker +++ b/tests/docker/dockerfiles/ubuntu.docker @@ -11,6 +11,7 @@ FROM docker.io/library/ubuntu:20.04 ENV PACKAGES \ + bison \ ccache \ clang \ dbus \ @@ -31,6 +32,7 @@ ENV PACKAGES \ libfdt-dev \ libffi-dev \ libgbm-dev \ + libglib2.0-dev \ libgnutls28-dev \ libgtk-3-dev \ libibverbs-dev \ diff --git a/tests/docker/dockerfiles/ubuntu1804.docker b/tests/docker/dockerfiles/ubuntu1804.docker index 52dd458e31..8f24137a34 100644 --- a/tests/docker/dockerfiles/ubuntu1804.docker +++ b/tests/docker/dockerfiles/ubuntu1804.docker @@ -1,6 +1,7 @@ FROM docker.io/library/ubuntu:18.04 ENV PACKAGES \ bc \ + bison \ bsdmainutils \ bzip2 \ ca-certificates \ -- 2.32.0 ^ permalink raw reply related [flat|nested] 27+ messages in thread
* RE: [PATCH v6 09/12] target/hexagon: import parser for idef-parser 2021-07-20 12:30 ` [PATCH v6 09/12] target/hexagon: import parser " Alessandro Di Federico via @ 2021-09-07 18:08 ` Taylor Simpson 2021-09-08 6:35 ` Richard Henderson 2021-10-18 11:42 ` Anton Johansson via 0 siblings, 2 replies; 27+ messages in thread From: Taylor Simpson @ 2021-09-07 18:08 UTC (permalink / raw) To: Alessandro Di Federico, qemu-devel@nongnu.org Cc: babush@rev.ng, Brian Cain, richard.henderson@linaro.org, nizzo@rev.ng, Alessandro Di Federico > -----Original Message----- > From: Alessandro Di Federico <ale.qemu@rev.ng> > Sent: Tuesday, July 20, 2021 7:30 AM > To: qemu-devel@nongnu.org > Cc: Taylor Simpson <tsimpson@quicinc.com>; Brian Cain > <bcain@quicinc.com>; babush@rev.ng; nizzo@rev.ng; > richard.henderson@linaro.org; Alessandro Di Federico <ale@rev.ng> > Subject: [PATCH v6 09/12] target/hexagon: import parser for idef-parser > > From: Paolo Montesel <babush@rev.ng> > > Signed-off-by: Alessandro Di Federico <ale@rev.ng> > Signed-off-by: Paolo Montesel <babush@rev.ng> > diff --git a/target/hexagon/idef-parser/parser-helpers.h > b/target/hexagon/idef-parser/parser-helpers.h > +#define OUT_IMPL(c, locp, x) \ > + _Generic(*x, \ > + char: str_print, \ > + uint8_t: uint8_print, \ > + uint64_t: uint64_print, \ > + int: int_print, \ > + unsigned: uint_print, \ > + HexValue: rvalue_out, \ > + default: out_assert \ > + )(c, locp, x); Unless something has changed, qemu requires building with GCC 4.8 which is old enough that it doesn't support C11 _Generic. > +HexValue gen_bin_cmp(Context *c, > + YYLTYPE *locp, > + TCGCond type, > + HexValue *op1_ptr, > + HexValue *op2_ptr); > + > +/* Code generation functions */ Move this comment above the previous function > +HexValue gen_bin_op(Context *c, > + YYLTYPE *locp, > + OpType type, > + HexValue *operand1, > + HexValue *operand2); > + > diff --git a/target/hexagon/idef-parser/parser-helpers.c > b/target/hexagon/idef-parser/parser-helpers.c > new file mode 100644 > +void pre_print(Context *c, YYLTYPE *locp, HexPre *pre, bool is_dotnew) > +{ > + (void) locp; > + char suffix = is_dotnew ? 'N' : 'V'; > + EMIT(c, "P%c%c", pre->id, suffix); > +} Call this pred_print > + > +void reg_compose(Context *c, YYLTYPE *locp, HexReg *reg, char reg_id[5]) > +{ > + switch (reg->type) { > + case GENERAL_PURPOSE: > + reg_id[0] = 'R'; > + break; > + case CONTROL: > + reg_id[0] = 'C'; > + break; > + case MODIFIER: > + reg_id[0] = 'M'; > + break; > + case DOTNEW: > + reg_id[0] = 'N'; > + reg_id[1] = reg->id; > + reg_id[2] = 'N'; > + return; > + } > + switch (reg->bit_width) { > + case 32: > + reg_id[1] = reg->id; > + reg_id[2] = 'V'; > + break; > + case 64: > + reg_id[1] = reg->id; > + reg_id[2] = reg->id; > + reg_id[3] = 'V'; > + break; > + default: > + yyassert(c, locp, false, "Unhandled register bit width!\n"); > + } > +} It would be better to zero out the result in the above function than having the caller do it. > + > +static void reg_arg_print(Context *c, YYLTYPE *locp, HexReg *reg) > +{ > + char reg_id[5] = { 0 }; This is where the caller zero's out the result ... > + reg_compose(c, locp, reg, reg_id); > + EMIT(c, "%s", reg_id); > +} > +/* Temporary values creation */ > +HexValue gen_tmp(Context *c, YYLTYPE *locp, unsigned bit_width) You should pass the signedness as an argument rather than assuming it's signed. > +{ > + HexValue rvalue; > + memset(&rvalue, 0, sizeof(HexValue)); > + rvalue.type = TEMP; > + assert(bit_width == 32 || bit_width == 64); > + rvalue.bit_width = bit_width; > + rvalue.signedness = SIGNED; > + rvalue.is_dotnew = false; > + rvalue.is_manual = false; > + rvalue.tmp.index = c->inst.tmp_count; > + OUT(c, locp, "TCGv_i", &bit_width, " tmp_", &c->inst.tmp_count, > + " = tcg_temp_new_i", &bit_width, "();\n"); > + c->inst.tmp_count++; > + return rvalue; > +} > + > +HexValue gen_tmp_value(Context *c, > + YYLTYPE *locp, > + const char *value, > + unsigned bit_width) You should pass the signedness as an argument rather than assuming it's signed. > +{ > + HexValue rvalue; > + memset(&rvalue, 0, sizeof(HexValue)); > + rvalue.type = TEMP; > + rvalue.bit_width = bit_width; > + rvalue.signedness = SIGNED; > + rvalue.is_dotnew = false; > + rvalue.is_manual = false; > + rvalue.tmp.index = c->inst.tmp_count; > + OUT(c, locp, "TCGv_i", &bit_width, " tmp_", &c->inst.tmp_count, > + " = tcg_const_i", &bit_width, "(", value, ");\n"); > + c->inst.tmp_count++; > + return rvalue; > +} > + > +static HexValue gen_tmp_value_from_imm(Context *c, > + YYLTYPE *locp, > + HexValue *value) > +{ > + assert(value->type == IMMEDIATE); > + HexValue rvalue; > + memset(&rvalue, 0, sizeof(HexValue)); > + rvalue.type = TEMP; > + rvalue.bit_width = value->bit_width; > + rvalue.signedness = value->signedness; > + rvalue.is_dotnew = false; > + rvalue.is_manual = false; > + rvalue.tmp.index = c->inst.tmp_count; > + OUT(c, locp, "TCGv_i", &rvalue.bit_width, " tmp_", &c->inst.tmp_count); > + OUT(c, locp, " = tcg_const_i", &rvalue.bit_width, > + "((int", &rvalue.bit_width, "_t) (", value, "));\n"); Is the type conversion necessary? You've copied the bit_width and the signedness from the value. > + c->inst.tmp_count++; > + return rvalue; > +} > + > +HexValue gen_imm_value(Context *c __attribute__((unused)), > + YYLTYPE *locp, > + int value, > + unsigned bit_width) You should pass the signedness as an argument rather than assuming it's signed. > +{ > + (void) locp; > + HexValue rvalue; > + memset(&rvalue, 0, sizeof(HexValue)); > + rvalue.type = IMMEDIATE; > + rvalue.bit_width = bit_width; > + rvalue.signedness = SIGNED; > + rvalue.is_dotnew = false; > + rvalue.is_manual = false; > + rvalue.imm.type = VALUE; > + rvalue.imm.value = value; > + return rvalue; > +} > +void gen_varid_allocate(Context *c, > + YYLTYPE *locp, > + HexValue *varid, > + int width, > + HexSignedness signedness) > +{ > + assert_signedness(c, locp, signedness); > + varid->bit_width = width; > + const char *bit_suffix = width == 64 ? "64" : "32"; > + int index = find_variable(c, locp, varid); > + bool found = index != -1; > + if (found) { > + Var *other = &g_array_index(c->inst.allocated, Var, index); > + varid->var.name = other->name; > + varid->bit_width = other->bit_width; > + varid->signedness = other->signedness; Do you need to check that "other" has the same bit_width and signedness? > + } else { > + EMIT_HEAD(c, "TCGv_i%s %s", bit_suffix, varid->var.name->str); > + EMIT_HEAD(c, " = tcg_temp_local_new_i%s();\n", bit_suffix); > + Var new_var = { > + .name = varid->var.name, > + .bit_width = width, > + .signedness = signedness, > + }; > + g_array_append_val(c->inst.allocated, new_var); > + } > +} > + > +void gen_operands_extend(Context *c, > + YYLTYPE *locp, > + HexValue *op1, > + HexValue *op2) { > + enum OpTypes op_types = (op1->type != IMMEDIATE) << 1 > + | (op2->type != IMMEDIATE); > + > + switch (op_types) { > + case IMM_IMM: > + break; > + case IMM_REG: > + *op2 = gen_rvalue_extend(c, locp, op2); > + break; > + case REG_IMM: > + *op1 = gen_rvalue_extend(c, locp, op1); > + break; > + case REG_REG: > + *op1 = gen_rvalue_extend(c, locp, op1); > + *op2 = gen_rvalue_extend(c, locp, op2); > + break; > + } > +} Why are you only extending the REG operands? Will you ever need to extend the IMM operands (e.g., 32-bit immediate and 64-bit register)? > +static void gen_andl_op(Context *c, YYLTYPE *locp, unsigned bit_width, > + const char *bit_suffix, HexValue *res, > + enum OpTypes op_types, HexValue *op1, HexValue *op2) > +{ > + HexValue zero, tmp1, tmp2; > + memset(&zero, 0, sizeof(HexValue)); > + memset(&tmp1, 0, sizeof(HexValue)); > + memset(&tmp2, 0, sizeof(HexValue)); > + switch (op_types) { > + case IMM_IMM: > + OUT(c, locp, "int", &bit_width, "_t ", > + res, " = ", op1, " && ", op2, ";\n"); > + break; > + case IMM_REG: > + zero = gen_tmp_value(c, locp, "0", 32); > + tmp2 = gen_bin_cmp(c, locp, TCG_COND_NE, op2, &zero); > + OUT(c, locp, "tcg_gen_andi_", bit_suffix, > + "(", res, ", ", op1, " != 0 , ", &tmp2, ");\n"); > + gen_rvalue_free(c, locp, &tmp2); > + break; > + case REG_IMM: > + zero = gen_tmp_value(c, locp, "0", 32); > + tmp1 = gen_bin_cmp(c, locp, TCG_COND_NE, op1, &zero); > + OUT(c, locp, "tcg_gen_andi_", bit_suffix, > + "(", res, ", ", &tmp1, ", ", op2, " != 0);\n"); > + gen_rvalue_free(c, locp, &tmp1); > + break; > + case REG_REG: > + zero = gen_tmp_value(c, locp, "0", 32); > + zero.is_manual = true; > + tmp1 = gen_bin_cmp(c, locp, TCG_COND_NE, op1, &zero); > + tmp2 = gen_bin_cmp(c, locp, TCG_COND_NE, op2, &zero); > + OUT(c, locp, "tcg_gen_and_", bit_suffix, > + "(", res, ", ", &tmp1, ", ", &tmp2, ");\n"); > + gen_rvalue_free_manual(c, locp, &zero); > + gen_rvalue_free(c, locp, &tmp1); > + gen_rvalue_free(c, locp, &tmp2); > + break; > + } > +} This might be OK for everything in the Hexagon semantics files. However, it doesn't match the C semantics of the && operator. In C, the expression X && Y only evaluates Y after we check that X is true. > +/* Code generation functions */ > +HexValue gen_bin_op(Context *c, > + YYLTYPE *locp, > + OpType type, > + HexValue *operand1, > + HexValue *operand2) > +{ > + /* Replicate operands to avoid side effects */ > + HexValue op1 = *operand1; > + HexValue op2 = *operand2; > + > + /* Enforce variables' size */ > + if (op1.type == VARID) { > + int index = find_variable(c, locp, &op1); > + yyassert(c, locp, c->inst.allocated->len > 0, > + "Variable in bin_op must exist!\n"); > + op1.bit_width = g_array_index(c->inst.allocated, Var, index).bit_width; > + } > + if (op2.type == VARID) { > + int index = find_variable(c, locp, &op2); > + yyassert(c, locp, c->inst.allocated->len > 0, > + "Variable in bin_op must exist!\n"); > + op2.bit_width = g_array_index(c->inst.allocated, Var, index).bit_width; > + } It's not clear what these checks are doing - please add a comment. > + > + enum OpTypes op_types = (op1.type != IMMEDIATE) << 1 > + | (op2.type != IMMEDIATE); > + > + bool op_is64bit = op1.bit_width == 64 || op2.bit_width == 64; > + > + /* Shift greater than 32 are 64 bits wide */ > + if (type == ASL_OP && op2.type == IMMEDIATE && > + op2.imm.type == VALUE && op2.imm.value >= 32) > + op_is64bit = true; Need curly braces per qemu coding conventions > + > + /* Extend to 64-bits, if required */ > + if (op_is64bit) { > + gen_operands_extend(c, locp, &op1, &op2); > + } > + > + HexValue res; > + memset(&res, 0, sizeof(HexValue)); > + const char *bit_suffix = op_is64bit ? "i64" : "i32"; > + int bit_width = (op_is64bit) ? 64 : 32; > + if (op_types != IMM_IMM) { > + res = gen_tmp(c, locp, bit_width); > + } else { > + res.type = IMMEDIATE; > + res.is_dotnew = false; > + res.is_manual = false; > + res.imm.type = QEMU_TMP; > + res.imm.index = c->inst.qemu_tmp_count; > + res.bit_width = bit_width; > + } > + /* Handle signedness, if both unsigned -> result is unsigned, else signed > */ > + assert_signedness(c, locp, op1.signedness); > + assert_signedness(c, locp, op2.signedness); > + res.signedness = (op1.signedness == UNSIGNED > + && op2.signedness == UNSIGNED) ? UNSIGNED : SIGNED; This doesn't match C semantics. You also have to consider the bit_width of the original two operands. For example, If op1 is unsigned 32 and op2 is signed 64, the result is signed > +HexValue gen_extend_op(Context *c, > + YYLTYPE *locp, > + HexValue *src_width_ptr, > + HexValue *dst_width_ptr, > + HexValue *value_ptr, > + HexSignedness signedness) { > + HexValue src_width = *src_width_ptr; > + HexValue dst_width = *dst_width_ptr; > + HexValue value = *value_ptr; > + src_width = gen_rvalue_extend(c, locp, &src_width); > + value = gen_rvalue_extend(c, locp, &value); > + src_width = rvalue_materialize(c, locp, &src_width); > + value = rvalue_materialize(c, locp, &value); > + > + HexValue res = gen_tmp(c, locp, 64); > + HexValue shift = gen_tmp_value(c, locp, "64", 64); > + HexValue zero = gen_tmp_value(c, locp, "0", 64); > + OUT(c, locp, "tcg_gen_sub_i64(", > + &shift, ", ", &shift, ", ", &src_width, ");\n"); > + assert_signedness(c, locp, signedness); > + if (signedness == UNSIGNED) { > + HexValue mask = gen_tmp_value(c, locp, "0xffffffffffffffff", 64); > + OUT(c, locp, "tcg_gen_shr_i64(", > + &mask, ", ", &mask, ", ", &shift, ");\n"); > + OUT(c, locp, "tcg_gen_and_i64(", > + &res, ", ", &value, ", ", &mask, ");\n"); > + gen_rvalue_free(c, locp, &mask); > + } else { > + OUT(c, locp, "tcg_gen_shl_i64(", > + &res, ", ", &value, ", ", &shift, ");\n"); > + OUT(c, locp, "tcg_gen_sar_i64(", > + &res, ", ", &res, ", ", &shift, ");\n"); > + } This seems strange that you are doing shifts. Why not use tcg_gen_extract_* or tcg_gen_sextract_*? > + OUT(c, locp, "tcg_gen_movcond_i64(TCG_COND_EQ, ", &res, ", "); > + OUT(c, locp, &src_width, ", ", &zero, ", ", &zero, ", ", &res, ");\n"); > + > + gen_rvalue_free(c, locp, &src_width); > + gen_rvalue_free(c, locp, &dst_width); > + gen_rvalue_free(c, locp, &value); > + gen_rvalue_free(c, locp, &shift); > + gen_rvalue_free(c, locp, &zero); > + > + res.signedness = signedness; > + return res; > +} > + > +void gen_rdeposit_op(Context *c, > + YYLTYPE *locp, > + HexValue *dest, > + HexValue *value, > + HexValue *begin, > + HexValue *width) > +{ > + HexValue dest_m = *dest; > + dest_m.is_manual = true; > + > + HexValue value_m = gen_rvalue_extend(c, locp, value); > + HexValue begin_m = gen_rvalue_extend(c, locp, begin); > + HexValue width_orig = *width; > + width_orig.is_manual = true; > + HexValue width_m = gen_rvalue_extend(c, locp, &width_orig); > + width_m = rvalue_materialize(c, locp, &width_m); > + > + HexValue mask = gen_tmp_value(c, locp, "0xffffffffffffffffUL", 64); > + mask.signedness = UNSIGNED; > + HexValue k64 = gen_tmp_value(c, locp, "64", 64); > + k64 = gen_bin_op(c, locp, SUB_OP, &k64, &width_m); > + mask = gen_bin_op(c, locp, LSR_OP, &mask, &k64); > + begin_m.is_manual = true; > + mask = gen_bin_op(c, locp, ASL_OP, &mask, &begin_m); > + mask.is_manual = true; > + value_m = gen_bin_op(c, locp, ASL_OP, &value_m, &begin_m); > + value_m = gen_bin_op(c, locp, ANDB_OP, &value_m, &mask); Why not use tcg_gen_deposit_*? > + > + OUT(c, locp, "tcg_gen_not_i64(", &mask, ", ", &mask, ");\n"); > + mask.is_manual = false; > + HexValue res = gen_bin_op(c, locp, ANDB_OP, &dest_m, &mask); > + res = gen_bin_op(c, locp, ORB_OP, &res, &value_m); > + > + if (dest->bit_width != res.bit_width) { > + res = gen_rvalue_truncate(c, locp, &res); > + } > + > + HexValue zero = gen_tmp_value(c, locp, "0", res.bit_width); > + OUT(c, locp, "tcg_gen_movcond_i", &res.bit_width, "(TCG_COND_NE, ", > dest); > + OUT(c, locp, ", ", &width_orig, ", ", &zero, ", ", &res, ", ", dest, > + ");\n"); Not sure what this is for? > + > + gen_rvalue_free(c, locp, &zero); > + gen_rvalue_free(c, locp, width); > + gen_rvalue_free(c, locp, &res); > +} > +static HexValue gen_convround_n_a(Context *c, > + YYLTYPE *locp, > + HexValue *a, > + HexValue *n) > +{ > + (void) n; > + HexValue res = gen_tmp(c, locp, 64); > + OUT(c, locp, "tcg_gen_ext_i32_i64(", &res, ", ", a, ");\n"); > + return res; > +} This looks like a simple extend, why is it a separate function? > + > +static HexValue gen_convround_n_b(Context *c, > + YYLTYPE *locp, > + HexValue *a, > + HexValue *n) > +{ > + HexValue res = gen_tmp(c, locp, 64); > + OUT(c, locp, "tcg_gen_ext_i32_i64(", &res, ", ", a, ");\n"); > + > + HexValue one = gen_tmp_value(c, locp, "1", 32); > + HexValue tmp = gen_tmp(c, locp, 32); > + HexValue tmp_64 = gen_tmp(c, locp, 64); > + > + OUT(c, locp, "tcg_gen_shl_i32(", &tmp); > + OUT(c, locp, ", ", &one, ", ", n, ");\n"); > + OUT(c, locp, "tcg_gen_and_i32(", &tmp); > + OUT(c, locp, ", ", &tmp, ", ", a, ");\n"); > + OUT(c, locp, "tcg_gen_shri_i32(", &tmp); > + OUT(c, locp, ", ", &tmp, ", 1);\n"); > + OUT(c, locp, "tcg_gen_ext_i32_i64(", &tmp_64, ", ", &tmp, ");\n"); > + OUT(c, locp, "tcg_gen_add_i64(", &res); > + OUT(c, locp, ", ", &res, ", ", &tmp_64, ");\n"); Why not use tcg_gen_sextract_*? > + > + gen_rvalue_free(c, locp, &one); > + gen_rvalue_free(c, locp, &tmp); > + gen_rvalue_free(c, locp, &tmp_64); > + > + return res; > +} > + > +static HexValue gen_convround_n_c(Context *c, > + YYLTYPE *locp, > + HexValue *a, > + HexValue *n) > +{ > + HexValue res = gen_tmp(c, locp, 64); > + OUT(c, locp, "tcg_gen_ext_i32_i64(", &res, ", ", a, ");\n"); > + > + HexValue one = gen_tmp_value(c, locp, "1", 32); > + HexValue tmp = gen_tmp(c, locp, 32); > + HexValue tmp_64 = gen_tmp(c, locp, 64); > + > + OUT(c, locp, "tcg_gen_subi_i32(", &tmp); > + OUT(c, locp, ", ", n, ", 1);\n"); > + OUT(c, locp, "tcg_gen_shl_i32(", &tmp); > + OUT(c, locp, ", ", &one, ", ", &tmp, ");\n"); > + OUT(c, locp, "tcg_gen_ext_i32_i64(", &tmp_64, ", ", &tmp, ");\n"); > + OUT(c, locp, "tcg_gen_add_i64(", &res); > + OUT(c, locp, ", ", &res, ", ", &tmp_64, ");\n"); > + > + gen_rvalue_free(c, locp, &one); > + gen_rvalue_free(c, locp, &tmp); > + gen_rvalue_free(c, locp, &tmp_64); > + > + return res; > +} > + > +HexValue gen_convround_n(Context *c, > + YYLTYPE *locp, > + HexValue *source_ptr, > + HexValue *bit_pos_ptr) > +{ > + /* If input is 64 bit cast it to 32 */ > + HexValue source = gen_cast_op(c, locp, source_ptr, 32); > + HexValue bit_pos = gen_cast_op(c, locp, bit_pos_ptr, 32); > + > + source = rvalue_materialize(c, locp, &source); > + bit_pos = rvalue_materialize(c, locp, &bit_pos); > + > + HexValue r1 = gen_convround_n_a(c, locp, &source, &bit_pos); > + HexValue r2 = gen_convround_n_b(c, locp, &source, &bit_pos); > + HexValue r3 = gen_convround_n_c(c, locp, &source, &bit_pos); > + > + HexValue l_32 = gen_tmp_value(c, locp, "1", 32); > + > + HexValue cond = gen_tmp(c, locp, 32); > + HexValue cond_64 = gen_tmp(c, locp, 64); > + HexValue mask = gen_tmp(c, locp, 32); > + HexValue n_64 = gen_tmp(c, locp, 64); > + HexValue res = gen_tmp(c, locp, 64); > + HexValue zero = gen_tmp_value(c, locp, "0", 64); > + > + OUT(c, locp, "tcg_gen_sub_i32(", &mask); > + OUT(c, locp, ", ", &bit_pos, ", ", &l_32, ");\n"); > + OUT(c, locp, "tcg_gen_shl_i32(", &mask); > + OUT(c, locp, ", ", &l_32, ", ", &mask, ");\n"); > + OUT(c, locp, "tcg_gen_sub_i32(", &mask); > + OUT(c, locp, ", ", &mask, ", ", &l_32, ");\n"); > + OUT(c, locp, "tcg_gen_and_i32(", &cond); > + OUT(c, locp, ", ", &source, ", ", &mask, ");\n"); > + OUT(c, locp, "tcg_gen_extu_i32_i64(", &cond_64, ", ", &cond, ");\n"); > + OUT(c, locp, "tcg_gen_ext_i32_i64(", &n_64, ", ", &bit_pos, ");\n"); Some comments here would be helpful ... > + > + OUT(c, locp, "tcg_gen_movcond_i64"); > + OUT(c, locp, "(TCG_COND_EQ, ", &res, ", ", &cond_64, ", ", &zero); > + OUT(c, locp, ", ", &r2, ", ", &r3, ");\n"); > + > + OUT(c, locp, "tcg_gen_movcond_i64"); > + OUT(c, locp, "(TCG_COND_EQ, ", &res, ", ", &n_64, ", ", &zero); > + OUT(c, locp, ", ", &r1, ", ", &res, ");\n"); > + > + OUT(c, locp, "tcg_gen_shr_i64(", &res); > + OUT(c, locp, ", ", &res, ", ", &n_64, ");\n"); > + > + gen_rvalue_free(c, locp, &source); > + gen_rvalue_free(c, locp, &bit_pos); > + > + gen_rvalue_free(c, locp, &r1); > + gen_rvalue_free(c, locp, &r2); > + gen_rvalue_free(c, locp, &r3); > + > + gen_rvalue_free(c, locp, &cond); > + gen_rvalue_free(c, locp, &cond_64); > + gen_rvalue_free(c, locp, &l_32); > + gen_rvalue_free(c, locp, &mask); > + gen_rvalue_free(c, locp, &n_64); > + gen_rvalue_free(c, locp, &zero); > + > + res = gen_rvalue_truncate(c, locp, &res); > + return res; > +} > + > +HexValue gen_round(Context *c, > + YYLTYPE *locp, > + HexValue *source, > + HexValue *position) { > + yyassert(c, locp, source->bit_width <= 32, > + "fRNDN not implemented for bit widths > 32!"); > + > + HexValue src = *source; > + HexValue pos = *position; > + > + HexValue src_width = gen_imm_value(c, locp, src.bit_width, 32); > + HexValue dst_width = gen_imm_value(c, locp, 64, 32); > + HexValue a = gen_extend_op(c, locp, &src_width, &dst_width, &src, > SIGNED); Are you sure extending is the right thing to do here? > +HexValue gen_fbrev_4(Context *c, YYLTYPE *locp, HexValue *source) > +{ > + HexValue source_m = *source; > + > + HexValue res = gen_tmp(c, locp, 32); > + HexValue tmp1 = gen_tmp(c, locp, 32); > + HexValue tmp2 = gen_tmp(c, locp, 32); > + > + source_m = rvalue_materialize(c, locp, &source_m); > + source_m = gen_rvalue_truncate(c, locp, &source_m); > + > + OUT(c, locp, "tcg_gen_mov_tl(", &res, ", ", &source_m, ");\n"); > + OUT(c, locp, "tcg_gen_andi_tl(", &tmp1, ", ", &res, ", 0xaaaaaaaa);\n"); > + OUT(c, locp, "tcg_gen_shri_tl(", &tmp1, ", ", &tmp1, ", 1);\n"); > + OUT(c, locp, "tcg_gen_andi_tl(", &tmp2, ", ", &res, ", 0x55555555);\n"); > + OUT(c, locp, "tcg_gen_shli_tl(", &tmp2, ", ", &tmp2, ", 1);\n"); > + OUT(c, locp, "tcg_gen_or_tl(", &res, ", ", &tmp1, ", ", &tmp2, ");\n"); > + OUT(c, locp, "tcg_gen_andi_tl(", &tmp1, ", ", &res, ", 0xcccccccc);\n"); > + OUT(c, locp, "tcg_gen_shri_tl(", &tmp1, ", ", &tmp1, ", 2);\n"); > + OUT(c, locp, "tcg_gen_andi_tl(", &tmp2, ", ", &res, ", 0x33333333);\n"); > + OUT(c, locp, "tcg_gen_shli_tl(", &tmp2, ", ", &tmp2, ", 2);\n"); > + OUT(c, locp, "tcg_gen_or_tl(", &res, ", ", &tmp1, ", ", &tmp2, ");\n"); > + OUT(c, locp, "tcg_gen_andi_tl(", &tmp1, ", ", &res, ", 0xf0f0f0f0);\n"); > + OUT(c, locp, "tcg_gen_shri_tl(", &tmp1, ", ", &tmp1, ", 4);\n"); > + OUT(c, locp, "tcg_gen_andi_tl(", &tmp2, ", ", &res, ", 0x0f0f0f0f);\n"); > + OUT(c, locp, "tcg_gen_shli_tl(", &tmp2, ", ", &tmp2, ", 4);\n"); > + OUT(c, locp, "tcg_gen_or_tl(", &res, ", ", &tmp1, ", ", &tmp2, ");\n"); > + OUT(c, locp, "tcg_gen_bswap32_tl(", &res, ", ", &res, ", > TCG_BSWAP_IZ);\n"); This would be better as a helper that just does revbit32 - or maybe just skip any instructions that reference this. > +HexValue gen_fbrev_8(Context *c, YYLTYPE *locp, HexValue *source) Ditto > +HexValue gen_deinterleave(Context *c, YYLTYPE *locp, HexValue *mixed) Ditto > + > +HexValue gen_interleave(Context *c, > + YYLTYPE *locp, > + HexValue *odd, > + HexValue *even) Ditto > +void gen_pre_assign(Context *c, YYLTYPE *locp, HexValue *lp, HexValue > *rp) Change to gen_pred_assign > +void gen_load(Context *c, YYLTYPE *locp, HexValue *size, > + HexSignedness signedness, HexValue *ea, HexValue *dst) > +{ > + /* Memop width is specified in the load macro */ > + int bit_width = (size->imm.value > 4) ? 64 : 32; > + assert_signedness(c, locp, signedness); > + const char *sign_suffix = (size->imm.value > 4) > + ? "" > + : ((signedness == UNSIGNED) ? "u" : "s"); > + char size_suffix[4] = { 0 }; > + /* Create temporary variable (if not present) */ > + if (dst->type == VARID) { > + /* TODO: this is a common pattern, the parser should be varid-aware. > */ > + gen_varid_allocate(c, locp, dst, bit_width, signedness); > + } > + snprintf(size_suffix, 4, "%" PRIu64, size->imm.value * 8); > + int var_id = find_variable(c, locp, ea); > + yyassert(c, locp, var_id != -1, "Load variable must exist!\n"); > + /* We need to enforce the variable size */ > + ea->bit_width = g_array_index(c->inst.allocated, Var, var_id).bit_width; > + OUT(c, locp, "if (insn->slot == 0 && pkt->pkt_has_store_s1) {\n"); > + OUT(c, locp, "process_store(ctx, pkt, 1);\n"); > + OUT(c, locp, "}\n"); > + OUT(c, locp, "tcg_gen_qemu_ld", size_suffix, sign_suffix); > + OUT(c, locp, "(", dst, ", ", ea, ", 0);\n"); Instead of 0, pass ctx->mem_idx > + /* If the var in EA was truncated it is now a tmp HexValue, so free it. */ > + gen_rvalue_free(c, locp, ea); > +} > +void gen_setbits(Context *c, YYLTYPE *locp, HexValue *hi, HexValue *lo, > + HexValue *dst, HexValue *val) > +{ > + yyassert(c, locp, hi->type == IMMEDIATE && > + hi->imm.type == VALUE && > + lo->type == IMMEDIATE && > + lo->imm.type == VALUE, > + "Range deposit needs immediate values!\n"); > + > + *val = gen_rvalue_truncate(c, locp, val); > + unsigned len = hi->imm.value + 1 - lo->imm.value; > + HexValue tmp = gen_tmp(c, locp, 32); > + OUT(c, locp, "tcg_gen_neg_i32(", &tmp, ", ", val, ");\n"); Taking the neg only works if val is 0 or 1. Today, this is always invoked with 0 or 1, but this implementation isn't future-proof. Do an andi with 1 first, then it will be correct for possible inputs. > + OUT(c, locp, "tcg_gen_deposit_i32(", dst, ", ", dst, ", ", &tmp, ", "); > + OUT(c, locp, lo, ", ", &len, ");\n"); > + > + gen_rvalue_free(c, locp, &tmp); > + gen_rvalue_free(c, locp, hi); > + gen_rvalue_free(c, locp, lo); > + gen_rvalue_free(c, locp, val); > +} > +HexValue gen_rvalue_pre(Context *c, YYLTYPE *locp, HexValue *pre) Call this gen_rvalue_pred and name the argument pred > +HexValue gen_rvalue_var(Context *c, YYLTYPE *locp, HexValue *var) > +{ > + /* Assign correct bit width and signedness */ > + bool found = false; > + for (unsigned i = 0; i < c->inst.allocated->len; i++) { > + Var *other = &g_array_index(c->inst.allocated, Var, i); > + if (g_string_equal(var->var.name, other->name)) { Use find_variable function here? > + found = true; > + other->name = var->var.name; > + var->bit_width = other->bit_width; > + var->signedness = other->signedness; > + break; > + } > + } > + yyassert(c, locp, found, "Undefined symbol!\n"); > + return *var; > +} > +HexValue gen_rvalue_not(Context *c, YYLTYPE *locp, HexValue *v) > +{ > + const char *bit_suffix = (v->bit_width == 64) ? "i64" : "i32"; > + int bit_width = (v->bit_width == 64) ? 64 : 32; > + HexValue res; > + memset(&res, 0, sizeof(HexValue)); > + res.signedness = v->signedness; > + res.is_dotnew = false; > + res.is_manual = false; > + if (v->type == IMMEDIATE) { > + res.type = IMMEDIATE; > + res.imm.type = QEMU_TMP; > + res.imm.index = c->inst.qemu_tmp_count; > + OUT(c, locp, "int", &bit_width, "_t ", &res, " = ~", v, ";\n"); I see a lot of these casts to "int", &bit_width, "_t". Shouldn't we also be checking the signedness and decide if it should be uint... instead? > + c->inst.qemu_tmp_count++; > + } else { > + res = gen_tmp(c, locp, bit_width); > + OUT(c, locp, "tcg_gen_not_", bit_suffix, "(", &res, > + ", ", v, ");\n"); > + gen_rvalue_free(c, locp, v); > + } > + return res; > +} > +HexValue gen_rvalue_abs(Context *c, YYLTYPE *locp, HexValue *v) > +{ > + int bit_width = (v->bit_width == 64) ? 64 : 32; > + HexValue res; > + memset(&res, 0, sizeof(HexValue)); > + res.signedness = v->signedness; > + res.is_dotnew = false; > + res.is_manual = false; > + if (v->type == IMMEDIATE) { > + res.type = IMMEDIATE; > + res.imm.type = QEMU_TMP; > + res.imm.index = c->inst.qemu_tmp_count; > + OUT(c, locp, "int", &bit_width, "_t ", &res, " = abs(", v, ");\n"); > + c->inst.qemu_tmp_count++; > + } else { > + res = gen_tmp(c, locp, bit_width); > + OUT(c, locp, "tcg_gen_abs_i", &bit_width, "(", &res, ", ", v, ");\n"); > + gen_rvalue_free(c, locp, v); > + } > + return res; > +} > + > +HexValue gen_rvalue_neg(Context *c, YYLTYPE *locp, HexValue *v) > +{ > + const char *bit_suffix = (v->bit_width == 64) ? "i64" : "i32"; > + int bit_width = (v->bit_width == 64) ? 64 : 32; > + HexValue res; > + memset(&res, 0, sizeof(HexValue)); > + res.signedness = v->signedness; > + res.is_dotnew = false; > + res.is_manual = false; > + if (v->type == IMMEDIATE) { > + res.type = IMMEDIATE; > + res.imm.type = QEMU_TMP; > + res.imm.index = c->inst.qemu_tmp_count; > + OUT(c, locp, "int", &bit_width, "_t ", &res, " = -", v, ";\n"); > + c->inst.qemu_tmp_count++; > + } else { > + res = gen_tmp(c, locp, bit_width); > + OUT(c, locp, "tcg_gen_neg_", bit_suffix, "(", &res, ", ", v, ");\n"); > + gen_rvalue_free(c, locp, v); > + } > + return res; > +} There's too much copy&paste across these. I'd suggest a helper function that takes the C code (e.g., "-") and the TCG function (e.g., tcg_gen_neg_) as arguments. > diff --git a/target/hexagon/idef-parser/idef-parser.y b/target/hexagon/idef- > parser/idef-parser.y > new file mode 100644 > + | CONSTEXT > + { > + HexValue rvalue; > + memset(&rvalue, 0, sizeof(HexValue)); > + rvalue.type = IMMEDIATE; > + rvalue.imm.type = IMM_CONSTEXT; > + rvalue.signedness = UNSIGNED; > + rvalue.is_dotnew = false; > + rvalue.is_manual = false; > + $$ = rvalue; > + } Aren't the constant extenders already taken care of during decoding? > + | CAST rvalue > + { > + @1.last_column = @2.last_column; > + /* Assign target signedness */ > + $2.signedness = $1.signedness; > + $$ = gen_cast_op(c, &@1, &$2, $1.bit_width); > + $$.signedness = $1.signedness; Pass $1.signedness to gen_cast_op rather than setting it here > + | rvalue '[' rvalue ']' > + { > + @1.last_column = @4.last_column; > + if ($3.type == IMMEDIATE) { > + $$ = gen_tmp(c, &@1, $1.bit_width); > + OUT(c, &@1, "tcg_gen_extract_i", &$$.bit_width, "("); > + OUT(c, &@1, &$$, ", ", &$1, ", ", &$3, ", 1);\n"); > + } else { > + HexValue one = gen_imm_value(c, &@1, 1, $3.bit_width); > + HexValue tmp = gen_bin_op(c, &@1, ASR_OP, &$1, &$3); > + $$ = gen_bin_op(c, &@1, ANDB_OP, &tmp, &one); > + } You're assuming $1 is a bitfield. How do you assure it's not a normal array? ^ permalink raw reply [flat|nested] 27+ messages in thread
* Re: [PATCH v6 09/12] target/hexagon: import parser for idef-parser 2021-09-07 18:08 ` Taylor Simpson @ 2021-09-08 6:35 ` Richard Henderson 2021-10-18 11:42 ` Anton Johansson via 1 sibling, 0 replies; 27+ messages in thread From: Richard Henderson @ 2021-09-08 6:35 UTC (permalink / raw) To: Taylor Simpson, Alessandro Di Federico, qemu-devel@nongnu.org Cc: babush@rev.ng, Brian Cain, nizzo@rev.ng, Alessandro Di Federico On 9/7/21 8:08 PM, Taylor Simpson wrote: >> diff --git a/target/hexagon/idef-parser/parser-helpers.h >> b/target/hexagon/idef-parser/parser-helpers.h >> +#define OUT_IMPL(c, locp, x) \ >> + _Generic(*x, \ >> + char: str_print, \ >> + uint8_t: uint8_print, \ >> + uint64_t: uint64_print, \ >> + int: int_print, \ >> + unsigned: uint_print, \ >> + HexValue: rvalue_out, \ >> + default: out_assert \ >> + )(c, locp, x); > > Unless something has changed, qemu requires building with GCC 4.8 which is old enough that it doesn't support C11 _Generic. Something has changed. The minimum is now 7.5, and we use -std=gnu11 now. We also use _Generic in several other places. r~ ^ permalink raw reply [flat|nested] 27+ messages in thread
* Re: [PATCH v6 09/12] target/hexagon: import parser for idef-parser 2021-09-07 18:08 ` Taylor Simpson 2021-09-08 6:35 ` Richard Henderson @ 2021-10-18 11:42 ` Anton Johansson via 2021-10-28 15:56 ` Taylor Simpson 1 sibling, 1 reply; 27+ messages in thread From: Anton Johansson via @ 2021-10-18 11:42 UTC (permalink / raw) To: Taylor Simpson, Alessandro Di Federico, qemu-devel@nongnu.org Cc: Brian Cain, babush@rev.ng, nizzo@rev.ng, richard.henderson@linaro.org, Alessandro Di Federico [-- Attachment #1: Type: text/plain, Size: 852 bytes --] On 9/7/21 18:08, Taylor Simpson wrote: >> +HexValue gen_round(Context *c, >> + YYLTYPE *locp, >> + HexValue *source, >> + HexValue *position) { >> + yyassert(c, locp, source->bit_width <= 32, >> + "fRNDN not implemented for bit widths > 32!"); >> + >> + HexValue src = *source; >> + HexValue pos = *position; >> + >> + HexValue src_width = gen_imm_value(c, locp, src.bit_width, 32); >> + HexValue dst_width = gen_imm_value(c, locp, 64, 32); >> + HexValue a = gen_extend_op(c, locp, &src_width, &dst_width, &src, >> SIGNED); > Are you sure extending is the right thing to do here? I believe so, the fRNDN definition in macros.h also extends here #define fRNDN(A, N) ((((N) == 0) ? (A) : (((fSE32_64(A)) + (1 << ((N) - 1)))))) -- Anton Johansson, rev.ng Srls. [-- Attachment #2: Type: text/html, Size: 1377 bytes --] ^ permalink raw reply [flat|nested] 27+ messages in thread
* RE: [PATCH v6 09/12] target/hexagon: import parser for idef-parser 2021-10-18 11:42 ` Anton Johansson via @ 2021-10-28 15:56 ` Taylor Simpson 0 siblings, 0 replies; 27+ messages in thread From: Taylor Simpson @ 2021-10-28 15:56 UTC (permalink / raw) To: anjo@rev.ng, Alessandro Di Federico, qemu-devel@nongnu.org Cc: babush@rev.ng, Brian Cain, richard.henderson@linaro.org, nizzo@rev.ng, Alessandro Di Federico > From: Anton Johansson <anjo@rev.ng> > Sent: Monday, October 18, 2021 6:42 AM > To: Taylor Simpson <tsimpson@quicinc.com>; Alessandro Di Federico <ale.qemu@rev.ng>; qemu-devel@nongnu.org > Cc: Brian Cain <bcain@quicinc.com>; babush@rev.ng; nizzo@rev.ng; richard.henderson@linaro.org; Alessandro Di Federico <ale@rev.ng> > Subject: Re: [PATCH v6 09/12] target/hexagon: import parser for idef-parser > > On 9/7/21 18:08, Taylor Simpson wrote: > +HexValue gen_round(Context *c, > + YYLTYPE *locp, > + HexValue *source, > + HexValue *position) { > + yyassert(c, locp, source->bit_width <= 32, > + "fRNDN not implemented for bit widths > 32!"); > + > + HexValue src = *source; > + HexValue pos = *position; > + > + HexValue src_width = gen_imm_value(c, locp, src.bit_width, 32); > + HexValue dst_width = gen_imm_value(c, locp, 64, 32); > + HexValue a = gen_extend_op(c, locp, &src_width, &dst_width, &src, SIGNED); > Are you sure extending is the right thing to do here? > I believe so, the fRNDN definition in macros.h also extends here > #define fRNDN(A, N) ((((N) == 0) ? (A) : (((fSE32_64(A)) + (1 << ((N) - 1)))))) OK Thanks, Taylor ^ permalink raw reply [flat|nested] 27+ messages in thread
* [PATCH v6 10/12] target/hexagon: call idef-parser functions 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via ` (8 preceding siblings ...) 2021-07-20 12:30 ` [PATCH v6 09/12] target/hexagon: import parser " Alessandro Di Federico via @ 2021-07-20 12:30 ` Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 11/12] target/hexagon: import additional tests Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 12/12] gitlab-ci: do not use qemu-project Docker registry Alessandro Di Federico via 11 siblings, 0 replies; 27+ messages in thread From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw) To: qemu-devel Cc: tsimpson, bcain, babush, nizzo, richard.henderson, Alessandro Di Federico From: Alessandro Di Federico <ale@rev.ng> Extend gen_tcg_funcs.py in order to emit calls to the functions emitted by the idef-parser, if available. Signed-off-by: Alessandro Di Federico <ale@rev.ng> Reviewed-by: Taylor Simpson <tsimpson@quicinc.com> --- target/hexagon/gen_helper_funcs.py | 5 +- target/hexagon/gen_helper_protos.py | 5 +- target/hexagon/gen_tcg_funcs.py | 28 +++++++++- target/hexagon/hex_common.py | 10 ++++ target/hexagon/meson.build | 81 ++++++++++++++++++----------- 5 files changed, 95 insertions(+), 34 deletions(-) diff --git a/target/hexagon/gen_helper_funcs.py b/target/hexagon/gen_helper_funcs.py index 2b1c5d8e3e..40fcc0269a 100755 --- a/target/hexagon/gen_helper_funcs.py +++ b/target/hexagon/gen_helper_funcs.py @@ -192,11 +192,12 @@ def main(): hex_common.read_semantics_file(sys.argv[1]) hex_common.read_attribs_file(sys.argv[2]) hex_common.read_overrides_file(sys.argv[3]) + hex_common.read_idef_parser_enabled_file(sys.argv[4]) hex_common.calculate_attribs() tagregs = hex_common.get_tagregs() tagimms = hex_common.get_tagimms() - with open(sys.argv[4], 'w') as f: + with open(sys.argv[5], 'w') as f: for tag in hex_common.tags: ## Skip the priv instructions if ( "A_PRIV" in hex_common.attribdict[tag] ) : @@ -213,6 +214,8 @@ def main(): continue if ( hex_common.skip_qemu_helper(tag) ): continue + if ( hex_common.is_idef_parser_enabled(tag) ): + continue gen_helper_function(f, tag, tagregs, tagimms) diff --git a/target/hexagon/gen_helper_protos.py b/target/hexagon/gen_helper_protos.py index ea41007ec9..6aee7229cb 100755 --- a/target/hexagon/gen_helper_protos.py +++ b/target/hexagon/gen_helper_protos.py @@ -121,11 +121,12 @@ def main(): hex_common.read_semantics_file(sys.argv[1]) hex_common.read_attribs_file(sys.argv[2]) hex_common.read_overrides_file(sys.argv[3]) + hex_common.read_idef_parser_enabled_file(sys.argv[4]) hex_common.calculate_attribs() tagregs = hex_common.get_tagregs() tagimms = hex_common.get_tagimms() - with open(sys.argv[4], 'w') as f: + with open(sys.argv[5], 'w') as f: for tag in hex_common.tags: ## Skip the priv instructions if ( "A_PRIV" in hex_common.attribdict[tag] ) : @@ -143,6 +144,8 @@ def main(): if ( hex_common.skip_qemu_helper(tag) ): continue + if ( hex_common.is_idef_parser_enabled(tag) ): + continue gen_helper_prototype(f, tag, tagregs, tagimms) diff --git a/target/hexagon/gen_tcg_funcs.py b/target/hexagon/gen_tcg_funcs.py index 7ceb25b5f6..e12f6b7cfc 100755 --- a/target/hexagon/gen_tcg_funcs.py +++ b/target/hexagon/gen_tcg_funcs.py @@ -394,7 +394,29 @@ def gen_tcg_func(f, tag, regs, imms): if (hex_common.is_read(regid)): genptr_src_read_opn(f,regtype,regid,tag) - if ( hex_common.skip_qemu_helper(tag) ): + if hex_common.is_idef_parser_enabled(tag): + declared = [] + ## Handle registers + for regtype,regid,toss,numregs in regs: + if (hex_common.is_pair(regid) + or (hex_common.is_single(regid) + and hex_common.is_old_val(regtype, regid, tag))): + declared.append("%s%sV" % (regtype, regid)) + if regtype == "M": + declared.append("%s%sN" % (regtype, regid)) + elif hex_common.is_new_val(regtype, regid, tag): + declared.append("%s%sN" % (regtype,regid)) + else: + print("Bad register parse: ",regtype,regid,toss,numregs) + + ## Handle immediates + for immlett,bits,immshift in imms: + declared.append(hex_common.imm_name(immlett)) + + arguments = ", ".join(["ctx", "insn", "pkt"] + declared) + f.write(" emit_%s(%s);\n" % (tag, arguments)) + + elif ( hex_common.skip_qemu_helper(tag) ): f.write(" fGEN_TCG_%s(%s);\n" % (tag, hex_common.semdict[tag])) else: ## Generate the call to the helper @@ -455,12 +477,14 @@ def main(): hex_common.read_attribs_file(sys.argv[2]) hex_common.read_overrides_file(sys.argv[3]) hex_common.calculate_attribs() + hex_common.read_idef_parser_enabled_file(sys.argv[4]) tagregs = hex_common.get_tagregs() tagimms = hex_common.get_tagimms() - with open(sys.argv[4], 'w') as f: + with open(sys.argv[5], 'w') as f: f.write("#ifndef HEXAGON_TCG_FUNCS_H\n") f.write("#define HEXAGON_TCG_FUNCS_H\n\n") + f.write("#include \"idef-generated-emitter.h.inc\"\n\n") for tag in hex_common.tags: ## Skip the priv instructions diff --git a/target/hexagon/hex_common.py b/target/hexagon/hex_common.py index b3b534057d..648ad29e94 100755 --- a/target/hexagon/hex_common.py +++ b/target/hexagon/hex_common.py @@ -28,6 +28,7 @@ attribinfo = {} # Register information and misc tags = [] # list of all tags overrides = {} # tags with helper overrides +idef_parser_enabled = {} # tags enabled for idef-parser # We should do this as a hash for performance, # but to keep order let's keep it as a list. @@ -201,6 +202,9 @@ def need_ea(tag): def skip_qemu_helper(tag): return tag in overrides.keys() +def is_idef_parser_enabled(tag): + return tag in idef_parser_enabled + def imm_name(immlett): return "%siV" % immlett @@ -232,3 +236,9 @@ def read_overrides_file(name): continue tag = overridere.findall(line)[0] overrides[tag] = True + +def read_idef_parser_enabled_file(name): + global idef_parser_enabled + with open(name, "r") as idef_parser_enabled_file: + lines = idef_parser_enabled_file.read().strip().split("\n") + idef_parser_enabled = set(lines) diff --git a/target/hexagon/meson.build b/target/hexagon/meson.build index c064186b5c..92caf9ff3e 100644 --- a/target/hexagon/meson.build +++ b/target/hexagon/meson.build @@ -42,10 +42,7 @@ hexagon_ss.add(semantics_generated) # Step 2 # We use Python scripts to generate the following files # shortcode_generated.h.inc -# helper_protos_generated.h.inc -# tcg_funcs_generated.c.inc # tcg_func_table_generated.c.inc -# helper_funcs_generated.c.inc # printinsn_generated.h.inc # op_regs_generated.h.inc # op_attribs_generated.h.inc @@ -60,24 +57,6 @@ shortcode_generated = custom_target( ) hexagon_ss.add(shortcode_generated) -helper_protos_generated = custom_target( - 'helper_protos_generated.h.inc', - output: 'helper_protos_generated.h.inc', - depends: [semantics_generated], - depend_files: [hex_common_py, attribs_def, gen_tcg_h], - command: [python, files('gen_helper_protos.py'), semantics_generated, attribs_def, gen_tcg_h, '@OUTPUT@'], -) -hexagon_ss.add(helper_protos_generated) - -tcg_funcs_generated = custom_target( - 'tcg_funcs_generated.c.inc', - output: 'tcg_funcs_generated.c.inc', - depends: [semantics_generated], - depend_files: [hex_common_py, attribs_def, gen_tcg_h], - command: [python, files('gen_tcg_funcs.py'), semantics_generated, attribs_def, gen_tcg_h, '@OUTPUT@'], -) -hexagon_ss.add(tcg_funcs_generated) - tcg_func_table_generated = custom_target( 'tcg_func_table_generated.c.inc', output: 'tcg_func_table_generated.c.inc', @@ -87,15 +66,6 @@ tcg_func_table_generated = custom_target( ) hexagon_ss.add(tcg_func_table_generated) -helper_funcs_generated = custom_target( - 'helper_funcs_generated.c.inc', - output: 'helper_funcs_generated.c.inc', - depends: [semantics_generated], - depend_files: [hex_common_py, attribs_def, gen_tcg_h], - command: [python, files('gen_helper_funcs.py'), semantics_generated, attribs_def, gen_tcg_h, '@OUTPUT@'], -) -hexagon_ss.add(helper_funcs_generated) - printinsn_generated = custom_target( 'printinsn_generated.h.inc', output: 'printinsn_generated.h.inc', @@ -220,4 +190,55 @@ idef_generated_tcg = custom_target( command: [idef_parser, '@INPUT@', '@OUTPUT0@', '@OUTPUT1@', '@OUTPUT2@'], ) +indent = find_program('indent', required: false) +if indent.found() + idef_generated_tcg_c = custom_target('asd', + input: idef_generated_tcg[0], + output: 'idef-generated-emitter.indented.c', + command: [indent, '-linux', '@INPUT@', '-o', '@OUTPUT@']) +else + idef_generated_tcg_c = custom_target('asd', + input: idef_generated_tcg[0], + output: 'idef-generated-emitter.indented.c', + command: ['cp', '@INPUT@', '@OUTPUT@']) +endif + +idef_generated_list = idef_generated_tcg[2].full_path() + +hexagon_ss.add(idef_generated_tcg_c) + +# +# Step 5 +# We use Python scripts to generate the following files +# helper_protos_generated.h.inc +# helper_funcs_generated.c.inc +# tcg_funcs_generated.c.inc +# +helper_protos_generated = custom_target( + 'helper_protos_generated.h.inc', + output: 'helper_protos_generated.h.inc', + depends: [semantics_generated, idef_generated_tcg_c, idef_generated_tcg], + depend_files: [hex_common_py, attribs_def, gen_tcg_h], + command: [python, files('gen_helper_protos.py'), semantics_generated, attribs_def, gen_tcg_h, idef_generated_list, '@OUTPUT@'], +) +hexagon_ss.add(helper_protos_generated) + +helper_funcs_generated = custom_target( + 'helper_funcs_generated.c.inc', + output: 'helper_funcs_generated.c.inc', + depends: [semantics_generated, idef_generated_tcg_c, idef_generated_tcg], + depend_files: [hex_common_py, attribs_def, gen_tcg_h], + command: [python, files('gen_helper_funcs.py'), semantics_generated, attribs_def, gen_tcg_h, idef_generated_list, '@OUTPUT@'], +) +hexagon_ss.add(helper_funcs_generated) + +tcg_funcs_generated = custom_target( + 'tcg_funcs_generated.c.inc', + output: 'tcg_funcs_generated.c.inc', + depends: [semantics_generated, idef_generated_tcg_c, idef_generated_tcg], + depend_files: [hex_common_py, attribs_def, gen_tcg_h], + command: [python, files('gen_tcg_funcs.py'), semantics_generated, attribs_def, gen_tcg_h, idef_generated_list, '@OUTPUT@'], +) +hexagon_ss.add(tcg_funcs_generated) + target_arch += {'hexagon': hexagon_ss} -- 2.32.0 ^ permalink raw reply related [flat|nested] 27+ messages in thread
* [PATCH v6 11/12] target/hexagon: import additional tests 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via ` (9 preceding siblings ...) 2021-07-20 12:30 ` [PATCH v6 10/12] target/hexagon: call idef-parser functions Alessandro Di Federico via @ 2021-07-20 12:30 ` Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 12/12] gitlab-ci: do not use qemu-project Docker registry Alessandro Di Federico via 11 siblings, 0 replies; 27+ messages in thread From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw) To: qemu-devel Cc: tsimpson, bcain, babush, nizzo, richard.henderson, Alessandro Di Federico From: Niccolò Izzo <nizzo@rev.ng> Signed-off-by: Alessandro Di Federico <ale@rev.ng> Signed-off-by: Niccolò Izzo <nizzo@rev.ng> --- tests/tcg/hexagon/Makefile.target | 30 +++++++++++++- tests/tcg/hexagon/crt.S | 26 ++++++++++++ tests/tcg/hexagon/test_abs.S | 20 ++++++++++ tests/tcg/hexagon/test_bitcnt.S | 42 ++++++++++++++++++++ tests/tcg/hexagon/test_bitsplit.S | 25 ++++++++++++ tests/tcg/hexagon/test_call.S | 60 ++++++++++++++++++++++++++++ tests/tcg/hexagon/test_clobber.S | 35 +++++++++++++++++ tests/tcg/hexagon/test_cmp.S | 34 ++++++++++++++++ tests/tcg/hexagon/test_cmpy.S | 31 +++++++++++++++ tests/tcg/hexagon/test_dotnew.S | 36 +++++++++++++++++ tests/tcg/hexagon/test_ext.S | 16 ++++++++ tests/tcg/hexagon/test_fibonacci.S | 33 ++++++++++++++++ tests/tcg/hexagon/test_hl.S | 19 +++++++++ tests/tcg/hexagon/test_hwloops.S | 22 +++++++++++ tests/tcg/hexagon/test_jmp.S | 25 ++++++++++++ tests/tcg/hexagon/test_lsr.S | 39 ++++++++++++++++++ tests/tcg/hexagon/test_mpyi.S | 20 ++++++++++ tests/tcg/hexagon/test_overflow.S | 63 ++++++++++++++++++++++++++++++ tests/tcg/hexagon/test_packet.S | 26 ++++++++++++ tests/tcg/hexagon/test_reorder.S | 29 ++++++++++++++ tests/tcg/hexagon/test_round.S | 31 +++++++++++++++ tests/tcg/hexagon/test_vavgw.S | 33 ++++++++++++++++ tests/tcg/hexagon/test_vcmpb.S | 32 +++++++++++++++ tests/tcg/hexagon/test_vcmpw.S | 29 ++++++++++++++ tests/tcg/hexagon/test_vcmpy.S | 50 ++++++++++++++++++++++++ tests/tcg/hexagon/test_vlsrw.S | 23 +++++++++++ tests/tcg/hexagon/test_vmaxh.S | 37 ++++++++++++++++++ tests/tcg/hexagon/test_vminh.S | 37 ++++++++++++++++++ tests/tcg/hexagon/test_vpmpyh.S | 30 ++++++++++++++ tests/tcg/hexagon/test_vspliceb.S | 33 ++++++++++++++++ 30 files changed, 965 insertions(+), 1 deletion(-) create mode 100644 tests/tcg/hexagon/crt.S create mode 100644 tests/tcg/hexagon/test_abs.S create mode 100644 tests/tcg/hexagon/test_bitcnt.S create mode 100644 tests/tcg/hexagon/test_bitsplit.S create mode 100644 tests/tcg/hexagon/test_call.S create mode 100644 tests/tcg/hexagon/test_clobber.S create mode 100644 tests/tcg/hexagon/test_cmp.S create mode 100644 tests/tcg/hexagon/test_cmpy.S create mode 100644 tests/tcg/hexagon/test_dotnew.S create mode 100644 tests/tcg/hexagon/test_ext.S create mode 100644 tests/tcg/hexagon/test_fibonacci.S create mode 100644 tests/tcg/hexagon/test_hl.S create mode 100644 tests/tcg/hexagon/test_hwloops.S create mode 100644 tests/tcg/hexagon/test_jmp.S create mode 100644 tests/tcg/hexagon/test_lsr.S create mode 100644 tests/tcg/hexagon/test_mpyi.S create mode 100644 tests/tcg/hexagon/test_overflow.S create mode 100644 tests/tcg/hexagon/test_packet.S create mode 100644 tests/tcg/hexagon/test_reorder.S create mode 100644 tests/tcg/hexagon/test_round.S create mode 100644 tests/tcg/hexagon/test_vavgw.S create mode 100644 tests/tcg/hexagon/test_vcmpb.S create mode 100644 tests/tcg/hexagon/test_vcmpw.S create mode 100644 tests/tcg/hexagon/test_vcmpy.S create mode 100644 tests/tcg/hexagon/test_vlsrw.S create mode 100644 tests/tcg/hexagon/test_vmaxh.S create mode 100644 tests/tcg/hexagon/test_vminh.S create mode 100644 tests/tcg/hexagon/test_vpmpyh.S create mode 100644 tests/tcg/hexagon/test_vspliceb.S diff --git a/tests/tcg/hexagon/Makefile.target b/tests/tcg/hexagon/Makefile.target index 050cd61c1a..6709433a52 100644 --- a/tests/tcg/hexagon/Makefile.target +++ b/tests/tcg/hexagon/Makefile.target @@ -24,7 +24,7 @@ CFLAGS += -fno-unroll-loops HEX_SRC=$(SRC_PATH)/tests/tcg/hexagon VPATH += $(HEX_SRC) -first: $(HEX_SRC)/first.S +%: $(HEX_SRC)/%.S $(HEX_SRC)/crt.S $(CC) -static -mv67 -nostdlib $^ -o $@ HEX_TESTS = first @@ -40,4 +40,32 @@ HEX_TESTS += load_align HEX_TESTS += atomics HEX_TESTS += fpstuff +HEX_TESTS += test_abs +HEX_TESTS += test_bitcnt +HEX_TESTS += test_bitsplit +HEX_TESTS += test_clobber +HEX_TESTS += test_cmp +HEX_TESTS += test_cmpy +HEX_TESTS += test_dotnew +HEX_TESTS += test_ext +HEX_TESTS += test_fibonacci +HEX_TESTS += test_hl +HEX_TESTS += test_hwloops +HEX_TESTS += test_jmp +HEX_TESTS += test_lsr +HEX_TESTS += test_mpyi +HEX_TESTS += test_overflow +HEX_TESTS += test_packet +HEX_TESTS += test_reorder +HEX_TESTS += test_round +HEX_TESTS += test_vavgw +HEX_TESTS += test_vcmpb +HEX_TESTS += test_vcmpw +HEX_TESTS += test_vcmpy +HEX_TESTS += test_vlsrw +HEX_TESTS += test_vmaxh +HEX_TESTS += test_vminh +HEX_TESTS += test_vpmpyh +HEX_TESTS += test_vspliceb + TESTS += $(HEX_TESTS) diff --git a/tests/tcg/hexagon/crt.S b/tests/tcg/hexagon/crt.S new file mode 100644 index 0000000000..cb119ae407 --- /dev/null +++ b/tests/tcg/hexagon/crt.S @@ -0,0 +1,26 @@ +#define SYS_exit_group 94 + + .text + .globl init +init: + { + allocframe(r29,#0):raw + } + { + r0=#256 + } + { + dealloc_return + } + + .globl pass +pass: + r0 = #0 + r6 = #SYS_exit_group + trap0(#1) + + .globl fail +fail: + r0 = #1 + r6 = #SYS_exit_group + trap0(#1) diff --git a/tests/tcg/hexagon/test_abs.S b/tests/tcg/hexagon/test_abs.S new file mode 100644 index 0000000000..880b2886b5 --- /dev/null +++ b/tests/tcg/hexagon/test_abs.S @@ -0,0 +1,20 @@ +/* Purpose: test example, verify the soundness of the abs operation */ + + .text + .globl _start + +_start: + { + call init + } + { + r1=#-2 + r2=#2 + } + { + r3=abs(r1) + } + { + p0 = cmp.eq(r3, r2); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_bitcnt.S b/tests/tcg/hexagon/test_bitcnt.S new file mode 100644 index 0000000000..df77fe61e2 --- /dev/null +++ b/tests/tcg/hexagon/test_bitcnt.S @@ -0,0 +1,42 @@ +/* Purpose: test example, verify the soundness of the cl[01] operations + * + * the number 0x000001aa has 23 leading zeroes + * they become 55 when considered as 64 bit register + * and it has 1 trailing zero + */ + .text + .globl _start + +_start: + { + call init + } + { + r0=#426 + r1=#0 + } + { + r2=cl0(r0) + } + { + p0 = cmp.eq(r2, #23); if (p0.new) jump:t test2 + jump fail + } + +test2: + { + r2=cl0(r1:0) + } + { + p0 = cmp.eq(r2, #55); if (p0.new) jump:t test3 + jump fail + } + +test3: + { + r2=ct0(r0) + } + { + p0 = cmp.eq(r2, #1); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_bitsplit.S b/tests/tcg/hexagon/test_bitsplit.S new file mode 100644 index 0000000000..787cce72e4 --- /dev/null +++ b/tests/tcg/hexagon/test_bitsplit.S @@ -0,0 +1,25 @@ +/* Purpose: test example, verify the soundness of the bitsplit operation */ + + .text + .globl _start + +_start: + { + call init + } + { + r1=#187 + } + { + r3:2=bitsplit(r1, #3) + } + { + p0 = cmp.eq(r2, #3); if (p0.new) jump:t test2 + jump fail + } + +test2: + { + p0 = cmp.eq(r3, #23); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_call.S b/tests/tcg/hexagon/test_call.S new file mode 100644 index 0000000000..fe59a8f006 --- /dev/null +++ b/tests/tcg/hexagon/test_call.S @@ -0,0 +1,60 @@ +/* Purpose: test function calls and duplex instructions. + * The string "Hello there, I'm a test string!" with the first letter replaced + * with a capital L should be printed out. + */ + + .text + .globl test +test: + { + jumpr r31 + memb(r0+#0)=#76 + } +.Lfunc_end0: +.Ltmp0: + .size test, .Ltmp0-test + + .globl _start +_start: + { + call init + } + { + call test + r0=##dummy_buffer + allocframe(#0) + } + { + call write + } + { + jump pass + } + { + r31:30=deallocframe(r30):raw + } +.Lfunc_end1: +.Ltmp1: + .size _start, .Ltmp1-_start +write: + { + r0=##dummy_buffer + } + { + r1=#256 + } + { + jumpr r31 + } +.Lfunc_end2: +.Ltmp2: + .size write, .Ltmp2-write + + .type dummy_buffer,@object + .data + .globl dummy_buffer + .p2align 3 +dummy_buffer: + .string "Hello there, I'm a test string!\n" + .space 223 + .size dummy_buffer, 256 diff --git a/tests/tcg/hexagon/test_clobber.S b/tests/tcg/hexagon/test_clobber.S new file mode 100644 index 0000000000..198817ebd5 --- /dev/null +++ b/tests/tcg/hexagon/test_clobber.S @@ -0,0 +1,35 @@ +/* Purpose: demonstrate the succesful operation of the register save mechanism, + * in which the caller saves the registers that will be clobbered, and restores + * them after the call. + */ + + .text + .globl _start + +_start: + { + call init + } + { + r16=#47 + r17=#155 + } + { + memd(r29+#-16)=r17:16; allocframe(#8) + } + { + r16=#255 + r17=#42 + } + { + r17:16=memd(r29+#0); deallocframe + } + { + r3=add(r16,r17) + } + { + p0 = cmp.eq(r3, #202); if (p0.new) jump:t pass + } + { + jump fail + } diff --git a/tests/tcg/hexagon/test_cmp.S b/tests/tcg/hexagon/test_cmp.S new file mode 100644 index 0000000000..31ee9565fe --- /dev/null +++ b/tests/tcg/hexagon/test_cmp.S @@ -0,0 +1,34 @@ +/* Purpose: test a signed and unsigned comparison */ + + .text + .globl _start + +_start: + { + call init + } + { + jump signed + } + + .globl signed +signed: + { + r0=#-2 + r1=#0 + } + { + p0 = cmp.lt(r0, r1); if (p0.new) jump:t unsigned + jump fail + } + + .globl unsigned +unsigned: + { + r0=#-2 + r1=#0 + } + { + p0 = cmp.gtu(r0, r1); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_cmpy.S b/tests/tcg/hexagon/test_cmpy.S new file mode 100644 index 0000000000..0b3dfb95de --- /dev/null +++ b/tests/tcg/hexagon/test_cmpy.S @@ -0,0 +1,31 @@ +/* Purpose: test example, verify the soundness of the cmpy operation + * + * 3j+5 * 2j+4 = 22j+14 + * + * the complex multiply between 0x00030005 and 0x00020004 is 0x000000160000000e + */ + + .text + .globl _start + +_start: + { + call init + } + { + r0=#196613 + r1=#131076 + } + { + r3:2=cmpy(r0, r1):sat + } + { + p0 = cmp.eq(r2, #14); if (p0.new) jump:t test2 + jump fail + } + +test2: + { + p0 = cmp.eq(r3, #22); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_dotnew.S b/tests/tcg/hexagon/test_dotnew.S new file mode 100644 index 0000000000..70886d9483 --- /dev/null +++ b/tests/tcg/hexagon/test_dotnew.S @@ -0,0 +1,36 @@ +/* Purpose: test the .new operator while performing memory stores. */ + .text + .globl _start + +_start: + { + call init + } + { + r0=#1 + memw(sp+#0)=r0.new + } + { + r1=#2 + memw(sp+#4)=r1.new + } + { + r2=#3 + memw(sp+#8)=r2.new + } + { + r0=memw(sp+#8) + } + { + r1=memw(sp+#4) + } + { + r2=memw(sp+#0) + } + { + r3=mpyi(r1,r2) + } + { + p0 = cmp.eq(r3, #2); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_ext.S b/tests/tcg/hexagon/test_ext.S new file mode 100644 index 0000000000..0b855bc8e8 --- /dev/null +++ b/tests/tcg/hexagon/test_ext.S @@ -0,0 +1,16 @@ +/* Purpose: test immediate extender instructions. */ + + .text + .globl _start + +_start: + { + call init + } + { + r2=##-559038737 + } + { + p0 = cmp.eq(r2, ##-559038737); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_fibonacci.S b/tests/tcg/hexagon/test_fibonacci.S new file mode 100644 index 0000000000..41cb1517cb --- /dev/null +++ b/tests/tcg/hexagon/test_fibonacci.S @@ -0,0 +1,33 @@ +/* Purpose: computes the Fibonacci series up to a constant number. */ + + .text + .globl _start + +_start: + { + call init + } + { + r2=#100 + } + { + p0=cmp.gt(r2,#0); if (!p0.new) jump:nt .LBB0_3 + } + { + r3=#0 + r4=#1 + } +.LBB0_2: + { + r5=r4 + } + { + p0=cmp.gt(r2,r5); if (p0.new) jump:nt .LBB0_2 + r4=add(r3,r4) + r3=r5 + } +.LBB0_3: + { + p0 = cmp.eq(r3, #144); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_hl.S b/tests/tcg/hexagon/test_hl.S new file mode 100644 index 0000000000..217b3143e2 --- /dev/null +++ b/tests/tcg/hexagon/test_hl.S @@ -0,0 +1,19 @@ +/* Purpose: test example, verify the soundness of the high/low assignment */ + + .text + .globl _start + +_start: + { + call init + } + { + r0.H=#42 + } + { + r0.L=#69 + } + { + p0 = cmp.eq(r0, #2752581); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_hwloops.S b/tests/tcg/hexagon/test_hwloops.S new file mode 100644 index 0000000000..b6e858bcb7 --- /dev/null +++ b/tests/tcg/hexagon/test_hwloops.S @@ -0,0 +1,22 @@ +/* Purpose: simple C Program to test hardware loops. */ + + .text + .globl _start + +_start: + { + call init + } + { + loop0(.LBB0_1,#10) + r2=#0 + } +.LBB0_1: + { + r2=add(r2,#1) + nop + }:endloop0 + { + p0 = cmp.eq(r2, #10); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_jmp.S b/tests/tcg/hexagon/test_jmp.S new file mode 100644 index 0000000000..9bf6ea34e5 --- /dev/null +++ b/tests/tcg/hexagon/test_jmp.S @@ -0,0 +1,25 @@ +/* Purpose: test example, verify the soundness of the add operation */ + + .text + .globl _start + +_start: + { + call init + } + { + r1=#0 + r2=#0 + } + { + r3=add(r1,r2) + } + { + p0 = cmp.eq(r3, #0) + } + { + if (p0) jump:t pass + } + { + jump fail + } diff --git a/tests/tcg/hexagon/test_lsr.S b/tests/tcg/hexagon/test_lsr.S new file mode 100644 index 0000000000..202eb4aec7 --- /dev/null +++ b/tests/tcg/hexagon/test_lsr.S @@ -0,0 +1,39 @@ +/* Purpose: test the soundness of the lsr operation */ + + .text + .globl _start + +_start: + { + call init + } + { + r0=#-56984 + r1=#2147483647 + } + { + r2=#0x19 + } + { + r0&=lsr(r1, r2) + } + { + p0 = cmp.eq(r0, #0x28); if (p0.new) jump:t test2 + jump fail + } + +test2: + { + r0=#0x0000000a + r1=#0x00000000 + } + { + r2=#-1 + } + { + r1:0=lsl(r1:0, r2) + } + { + p0 = cmp.eq(r0, #0x5); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_mpyi.S b/tests/tcg/hexagon/test_mpyi.S new file mode 100644 index 0000000000..b8e0d50b68 --- /dev/null +++ b/tests/tcg/hexagon/test_mpyi.S @@ -0,0 +1,20 @@ +/* Purpose: test a simple multiplication operation */ + + .text + .globl _start + +_start: + { + call init + } + { + r1=#4 + r2=#6 + } + { + r3=mpyi(r1,r2) + } + { + p0 = cmp.eq(r3, #24); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_overflow.S b/tests/tcg/hexagon/test_overflow.S new file mode 100644 index 0000000000..9e2a235616 --- /dev/null +++ b/tests/tcg/hexagon/test_overflow.S @@ -0,0 +1,63 @@ +// Purpose: test example, verify the soundness of the overflow bit +// +// a right shift with negative amount should make r0 saturate, setting the +// overflow bit + + .text + .globl _start + +_start: + { + call init + } + { + r0=#0x10000000 + r1=#-2 + } + { + r2=asr(r0, r1):sat + } + { + r4=USR + } + { + r5=and(r4,#1) + } + { + p0 = cmp.eq(r5, #0); if (p0.new) jump:t check_ovfl + jump fail + } +check_ovfl: + { + r1=#-3 + } + { + r2=asr(r0, r1):sat + } + { + r4=USR + } + { + r5=and(r4,#1) + } + { + p0 = cmp.eq(r5, #1); if (p0.new) jump:t check_sticky + jump fail + } +check_sticky: + { + r1=#-2 + } + { + r2=asr(r0, r1):sat + } + { + r4=USR + } + { + r5=and(r4,#1) + } + { + p0 = cmp.eq(r5, #1); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_packet.S b/tests/tcg/hexagon/test_packet.S new file mode 100644 index 0000000000..d26e284be9 --- /dev/null +++ b/tests/tcg/hexagon/test_packet.S @@ -0,0 +1,26 @@ +/* Purpose: test that writes of a register in a packet are performed only after + * that packet has finished its execution. + */ + + .text + .globl _start + +_start: + { + call init + } + { + r2=#4 + r3=#6 + } + { + memw(sp+#0)=r2 + } + { + r3=memw(sp+#0) + r0=add(r2,r3) + } + { + p0 = cmp.eq(r0, #10); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_reorder.S b/tests/tcg/hexagon/test_reorder.S new file mode 100644 index 0000000000..f341a0142b --- /dev/null +++ b/tests/tcg/hexagon/test_reorder.S @@ -0,0 +1,29 @@ +/* Purpose: demonstrate handling of .new uses appearing before the associated + * definition. + * Here we perform a jump that skips the code resetting R2 from 0xDEADBEEF to 0, + * only if P0.new is true, but P0 is assigned to 1 (R4) in the next instruction + * in the packet. + */ + + .text + .globl _start + +_start: + { + call init + } + { r2=#-559038737 } + { r4=#1 } + { + if (p0.new) jump:nt skip + p0=r4; + } + +fallthrough: + { r2=#0 } + +skip: + { + p0 = cmp.eq(r2, #-559038737); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_round.S b/tests/tcg/hexagon/test_round.S new file mode 100644 index 0000000000..2becd62c4c --- /dev/null +++ b/tests/tcg/hexagon/test_round.S @@ -0,0 +1,31 @@ +/* Purpose: test example, verify the soundness of the cround operation + * 106 = 0b1101010 with the comma at third digit is 12.5 which is crounded to 12 + * but rounded to 13 + */ + + .text + .globl _start + +_start: + { + call init + } + { + r1=#200 + } + { + r2=round(r1, #4) + } + { + p0 = cmp.eq(r2, #13); if (p0.new) jump:t test2 + jump fail + } + +test2: + { + r2=cround(r1, #4) + } + { + p0 = cmp.eq(r2, #12); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_vavgw.S b/tests/tcg/hexagon/test_vavgw.S new file mode 100644 index 0000000000..8f67238900 --- /dev/null +++ b/tests/tcg/hexagon/test_vavgw.S @@ -0,0 +1,33 @@ +/* Purpose: test example, verify the soundness of the vavgw operation + * + * 0x00030001 averaged with 0x00010003 results 0x00020002 + */ + + .text + .globl _start + +_start: + { + call init + } + { + r0=#3 + r1=#1 + } + { + r2=#1 + r3=#3 + } + { + r1:0=vavgw(r1:0, r3:2):crnd + } + { + p0 = cmp.eq(r0, #2); if (p0.new) jump:t test2 + jump fail + } + +test2: + { + p0 = cmp.eq(r1, #2); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_vcmpb.S b/tests/tcg/hexagon/test_vcmpb.S new file mode 100644 index 0000000000..3c6700a63a --- /dev/null +++ b/tests/tcg/hexagon/test_vcmpb.S @@ -0,0 +1,32 @@ +/* Purpose: test example, verify the soundness of the vector compare bytes + * operation + * + * Vector word comparison between 0x1234567887654321 and 0x1234567800000000 + * should result in 0x11110000 + */ + + .text + .globl _start + +_start: + { + call init + } + { + r0=#305419896 + r1=#-2023406815 + } + { + r2=#305419896 + r3=#0 + } + { + p2=vcmpb.eq(r1:0, r3:2) + } + { + r4=p2 + } + { + p0 = cmp.eq(r4, #15); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_vcmpw.S b/tests/tcg/hexagon/test_vcmpw.S new file mode 100644 index 0000000000..112f08c92f --- /dev/null +++ b/tests/tcg/hexagon/test_vcmpw.S @@ -0,0 +1,29 @@ +/* Purpose: test example, verify the soundness of the vector compare words + * operation + * + * Vector word comparison between 0x1234567887654321 and 0x1234567800000000 + * should result in 0x11110000 + */ + + .text + .globl _start + +_start: + { + call init + } + { + r0=#305419896 + r1=#-2023406815 + } + { + r2=#305419896 + r3=#0 + } + { + p2=vcmpw.eq(r1:0, r3:2) + } + { + if (p2) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_vcmpy.S b/tests/tcg/hexagon/test_vcmpy.S new file mode 100644 index 0000000000..df379f9186 --- /dev/null +++ b/tests/tcg/hexagon/test_vcmpy.S @@ -0,0 +1,50 @@ +/* Purpose: test example, verify the soundness of the vcmpy operation + * this operation is a complex multiply and accumulate on vectors of two values + * + * (3j+5 * 2j+4) + (4j+6 * 5j+2) = 22j+14 + * + * the complex multiply between 0x00030005 and 0x00020004 is 0x000000160000000e + * the complex multiply between 0x00040006 and 0x00050002 is 0x000000160000000e + */ + + .text + .globl _start + +_start: + { + call init + } + { + r0=#196613 + r1=#131076 + } + { + r2=#262150 + r3=#327682 + } + { + r5:4=vcmpyr(r1:0, r3:2):sat + r7:6=vcmpyi(r1:0, r3:2):sat + } + { + p0 = cmp.eq(r4, #18); if (p0.new) jump:t test2 + jump fail + } + +test2: + { + p0 = cmp.eq(r5, #-2); if (p0.new) jump:t test3 + jump fail + } + +test3: + { + p0 = cmp.eq(r6, #38); if (p0.new) jump:t test4 + jump fail + } + +test4: + { + p0 = cmp.eq(r7, #24); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_vlsrw.S b/tests/tcg/hexagon/test_vlsrw.S new file mode 100644 index 0000000000..962ec99543 --- /dev/null +++ b/tests/tcg/hexagon/test_vlsrw.S @@ -0,0 +1,23 @@ +/* Purpose: test the soundness of the vlsrw operation */ + + .text + .globl _start + +_start: + { + call init + } + { + r0=#0x00000001 + r1=#0x00000001 + } + { + r1:0=vlsrw(r1:0, #1) + } + { + r0 = add(r0, r1) + } + { + p0 = cmp.eq(r0, #0); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_vmaxh.S b/tests/tcg/hexagon/test_vmaxh.S new file mode 100644 index 0000000000..1fce935e35 --- /dev/null +++ b/tests/tcg/hexagon/test_vmaxh.S @@ -0,0 +1,37 @@ +/* Purpose: test example, verify the soundness of the vrmaxh operation + * + * the maximum between 0x0002000300010005 and 0x0003000200020007 is + * 0x0003000300020007. + * + * r1=0x00010003 r0=0x00010005 r3=0x00030002 r2=0x00020007 + * result: r1=0x00030003 r0=0x00020007 + */ + + .text + .globl _start + +_start: + { + call init + } + { + r0=#65541 + r1=#65539 + } + { + r2=#131079 + r3=#196610 + } + { + r1:0=vmaxh(r1:0, r3:2) + } + { + p0 = cmp.eq(r0, #131079); if (p0.new) jump:t test2 + jump fail + } + +test2: + { + p0 = cmp.eq(r1, #196611); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_vminh.S b/tests/tcg/hexagon/test_vminh.S new file mode 100644 index 0000000000..6c6d18c673 --- /dev/null +++ b/tests/tcg/hexagon/test_vminh.S @@ -0,0 +1,37 @@ +/* Purpose: test example, verify the soundness of the vrmaxh operation + * + * the minimum between 0x0002000300010005 and 0x0003000200020007 is + * 0x0003000300020007 + * + * r1=0x00010003 r0=0x00010005 r3=0x00030002 r2=0x00020007 + * result: r1=0x00010002 r0=0x00010005 + */ + + .text + .globl _start + +_start: + { + call init + } + { + r0=#65541 + r1=#65539 + } + { + r2=#131079 + r3=#196610 + } + { + r1:0=vminh(r1:0, r3:2) + } + { + p0 = cmp.eq(r0, #65541); if (p0.new) jump:t test2 + jump fail + } + +test2: + { + p0 = cmp.eq(r1, #65538); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_vpmpyh.S b/tests/tcg/hexagon/test_vpmpyh.S new file mode 100644 index 0000000000..942d691da4 --- /dev/null +++ b/tests/tcg/hexagon/test_vpmpyh.S @@ -0,0 +1,30 @@ +/* Purpose: test example, verify the soundness of the vpmpyh operator + * + * 0x01020304 vector polynomial multiplied with 0x04030201 results + * 0x000400060b060b04 + */ + + .text + .globl _start + +_start: + { + call init + } + { + r0=#16909060 + r1=#67305985 + } + { + r1:0=vpmpyh(r0, r1) + } + { + p0 = cmp.eq(r0, #184945412); if (p0.new) jump:t test2 + jump fail + } + +test2: + { + p0 = cmp.eq(r1, #262150); if (p0.new) jump:t pass + jump fail + } diff --git a/tests/tcg/hexagon/test_vspliceb.S b/tests/tcg/hexagon/test_vspliceb.S new file mode 100644 index 0000000000..bae2a9c163 --- /dev/null +++ b/tests/tcg/hexagon/test_vspliceb.S @@ -0,0 +1,33 @@ +/* Purpose: test example, verify the soundness of the vspliceb operation + * the operation is a binary splice of two 64bit operators + * + * vspliceb(0xffffffffffffffff,0x0000000000000000,5) = 0x000000000000001f + */ + .text + .globl _start + +_start: + { + call init + } + { + r0=#-1 + r1=#-1 + } + { + r2=#0 + r3=#0 + } + { + r5:4=vspliceb(r1:0, r3:2, #5) + } + { + p0 = cmp.eq(r4, #-1); if (p0.new) jump:t test2 + jump fail + } + +test2: + { + p0 = cmp.eq(r5, #255); if (p0.new) jump:t pass + jump fail + } -- 2.32.0 ^ permalink raw reply related [flat|nested] 27+ messages in thread
* [PATCH v6 12/12] gitlab-ci: do not use qemu-project Docker registry 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via ` (10 preceding siblings ...) 2021-07-20 12:30 ` [PATCH v6 11/12] target/hexagon: import additional tests Alessandro Di Federico via @ 2021-07-20 12:30 ` Alessandro Di Federico via 11 siblings, 0 replies; 27+ messages in thread From: Alessandro Di Federico via @ 2021-07-20 12:30 UTC (permalink / raw) To: qemu-devel Cc: tsimpson, bcain, babush, nizzo, richard.henderson, Alessandro Di Federico From: Alessandro Di Federico <ale@rev.ng> This commit is necessary in order to use container built by the current run of the CI. If we don't do this, we use official containers which are not affected by the additional dependencies we're introducing. This should be considered as a temporary solution in order to test this patchset. Signed-off-by: Alessandro Di Federico <ale@rev.ng> --- .gitlab-ci.d/container-cross.yml | 2 +- .gitlab-ci.d/container-template.yml | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) diff --git a/.gitlab-ci.d/container-cross.yml b/.gitlab-ci.d/container-cross.yml index 0fcebe363a..eb134e927d 100644 --- a/.gitlab-ci.d/container-cross.yml +++ b/.gitlab-ci.d/container-cross.yml @@ -63,7 +63,7 @@ hexagon-cross-container: - docker:dind before_script: - export TAG="$CI_REGISTRY_IMAGE/qemu/$NAME:latest" - - export COMMON_TAG="$CI_REGISTRY/qemu-project/qemu/qemu/$NAME:latest" + - export COMMON_TAG="$CI_REGISTRY_IMAGE/qemu/$NAME:latest" - docker info - docker login $CI_REGISTRY -u "$CI_REGISTRY_USER" -p "$CI_REGISTRY_PASSWORD" script: diff --git a/.gitlab-ci.d/container-template.yml b/.gitlab-ci.d/container-template.yml index 1baecd9460..c85ae377b8 100644 --- a/.gitlab-ci.d/container-template.yml +++ b/.gitlab-ci.d/container-template.yml @@ -5,7 +5,7 @@ - docker:dind before_script: - export TAG="$CI_REGISTRY_IMAGE/qemu/$NAME:latest" - - export COMMON_TAG="$CI_REGISTRY/qemu-project/qemu/$NAME:latest" + - export COMMON_TAG="$CI_REGISTRY_IMAGE/qemu/$NAME:latest" - apk add python3 - docker info - docker login $CI_REGISTRY -u "$CI_REGISTRY_USER" -p "$CI_REGISTRY_PASSWORD" @@ -14,7 +14,7 @@ - echo "COMMON_TAG:$COMMON_TAG" - ./tests/docker/docker.py --engine docker build -t "qemu/$NAME" -f "tests/docker/dockerfiles/$NAME.docker" - -r $CI_REGISTRY/qemu-project/qemu + -r $CI_REGISTRY_IMAGE - docker tag "qemu/$NAME" "$TAG" - docker push "$TAG" after_script: -- 2.32.0 ^ permalink raw reply related [flat|nested] 27+ messages in thread
end of thread, other threads:[~2021-10-28 16:15 UTC | newest] Thread overview: 27+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2021-07-20 12:30 [PATCH v6 00/12] target/hexagon: introduce idef-parser Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 01/12] target/hexagon: update MAINTAINERS for idef-parser Alessandro Di Federico via 2021-09-07 18:09 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 02/12] target/hexagon: import README " Alessandro Di Federico via 2021-09-07 18:09 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 03/12] target/hexagon: make slot number an unsigned Alessandro Di Federico via 2021-09-07 18:09 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 04/12] target/hexagon: make helper functions non-static Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 05/12] target/hexagon: introduce new helper functions Alessandro Di Federico via 2021-09-07 18:09 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 06/12] target/hexagon: expose next PC in DisasContext Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 07/12] target/hexagon: prepare input for the idef-parser Alessandro Di Federico via 2021-09-07 18:09 ` Taylor Simpson 2021-10-18 11:23 ` Anton Johansson via 2021-10-28 15:44 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 08/12] target/hexagon: import lexer for idef-parser Alessandro Di Federico via 2021-09-07 18:08 ` Taylor Simpson 2021-10-18 11:37 ` Anton Johansson via 2021-10-28 15:49 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 09/12] target/hexagon: import parser " Alessandro Di Federico via 2021-09-07 18:08 ` Taylor Simpson 2021-09-08 6:35 ` Richard Henderson 2021-10-18 11:42 ` Anton Johansson via 2021-10-28 15:56 ` Taylor Simpson 2021-07-20 12:30 ` [PATCH v6 10/12] target/hexagon: call idef-parser functions Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 11/12] target/hexagon: import additional tests Alessandro Di Federico via 2021-07-20 12:30 ` [PATCH v6 12/12] gitlab-ci: do not use qemu-project Docker registry Alessandro Di Federico via
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).