* [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup
@ 2012-09-22 2:04 Richard Henderson
2012-09-22 2:04 ` [Qemu-devel] [PATCH 01/14] linux-user: Use memcpy in get_user/put_user Richard Henderson
` (14 more replies)
0 siblings, 15 replies; 16+ messages in thread
From: Richard Henderson @ 2012-09-22 2:04 UTC (permalink / raw)
To: qemu-devel; +Cc: Blue Swirl
Changes since v2:
* Patch 4 split out from patch 5 (afaerber feedback)
* TB chaining preserves wrt retranslation
* Last patch for branch retranslation is new
The patch set is rebased on cfb75cb9807463ebe18b127096b48b5d0db1ce03
and is available at
git://repo.or.cz/qemu/rth.git tcg-sparc
r~
Richard Henderson (14):
linux-user: Use memcpy in get_user/put_user.
tcg-sparc: Hack in qemu_ld/st64 for 32-bit.
tcg-sparc: Fix ADDX opcode.
tcg-sparc: Don't MAP_FIXED on top of the program
tcg-sparc: Assume v9 cpu always, i.e. force v8plus in 32-bit mode.
tcg-sparc: Fix qemu_ld/st to handle 32-bit host.
tcg-sparc: Support GUEST_BASE.
tcg-sparc: Change AREG0 in generated code to %i0.
tcg-sparc: Clean up cruft stemming from attempts to use global
registers.
tcg-sparc: Mask shift immediates to avoid illegal insns.
tcg-sparc: Use defines for temporaries.
tcg-sparc: Add %g/%o registers to alloc_order
tcg-sparc: Fix and enable direct TB chaining.
tcg-sparc: Preserve branch destinations during retranslation
configure | 52 +--
disas.c | 2 -
exec-all.h | 9 +-
exec.c | 12 +-
linux-user/qemu.h | 47 +--
qemu-timer.h | 8 +-
tcg/sparc/tcg-target.c | 970 +++++++++++++++++++++++--------------------------
tcg/sparc/tcg-target.h | 33 +-
tcg/tcg.c | 3 +-
9 files changed, 514 insertions(+), 622 deletions(-)
--
1.7.11.4
^ permalink raw reply [flat|nested] 16+ messages in thread* [Qemu-devel] [PATCH 01/14] linux-user: Use memcpy in get_user/put_user. 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson @ 2012-09-22 2:04 ` Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 02/14] tcg-sparc: Hack in qemu_ld/st64 for 32-bit Richard Henderson ` (13 subsequent siblings) 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:04 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl When host and target have differing alignment rules, using a cast and direct memory operation can result in SIGBUS. Use memcpy instead, which the compiler will happily optimize when alignment is satisfied. Signed-off-by: Richard Henderson <rth@twiddle.net> Reviewed-by: Peter Maydell <peter.maydell@linaro.org> --- linux-user/qemu.h | 47 +++++++++++++++-------------------------------- 1 file changed, 15 insertions(+), 32 deletions(-) diff --git a/linux-user/qemu.h b/linux-user/qemu.h index 69b27d7..fc4cc00 100644 --- a/linux-user/qemu.h +++ b/linux-user/qemu.h @@ -289,46 +289,29 @@ static inline int access_ok(int type, abi_ulong addr, abi_ulong size) * struct has been locked - usually with lock_user_struct(). */ #define __put_user(x, hptr)\ -({\ +({ __typeof(*hptr) pu_ = (x);\ switch(sizeof(*hptr)) {\ - case 1:\ - *(uint8_t *)(hptr) = (uint8_t)(typeof(*hptr))(x);\ - break;\ - case 2:\ - *(uint16_t *)(hptr) = tswap16((uint16_t)(typeof(*hptr))(x));\ - break;\ - case 4:\ - *(uint32_t *)(hptr) = tswap32((uint32_t)(typeof(*hptr))(x));\ - break;\ - case 8:\ - *(uint64_t *)(hptr) = tswap64((typeof(*hptr))(x));\ - break;\ - default:\ - abort();\ + case 1: break;\ + case 2: pu_ = tswap16(pu_); break; \ + case 4: pu_ = tswap32(pu_); break; \ + case 8: pu_ = tswap64(pu_); break; \ + default: abort();\ }\ + memcpy(hptr, &pu_, sizeof(pu_)); \ 0;\ }) #define __get_user(x, hptr) \ -({\ +({ __typeof(*hptr) gu_; \ + memcpy(&gu_, hptr, sizeof(gu_)); \ switch(sizeof(*hptr)) {\ - case 1:\ - x = (typeof(*hptr))*(uint8_t *)(hptr);\ - break;\ - case 2:\ - x = (typeof(*hptr))tswap16(*(uint16_t *)(hptr));\ - break;\ - case 4:\ - x = (typeof(*hptr))tswap32(*(uint32_t *)(hptr));\ - break;\ - case 8:\ - x = (typeof(*hptr))tswap64(*(uint64_t *)(hptr));\ - break;\ - default:\ - /* avoid warning */\ - x = 0;\ - abort();\ + case 1: break; \ + case 2: gu_ = tswap16(gu_); break; \ + case 4: gu_ = tswap32(gu_); break; \ + case 8: gu_ = tswap64(gu_); break; \ + default: abort();\ }\ + (x) = gu_; \ 0;\ }) -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 02/14] tcg-sparc: Hack in qemu_ld/st64 for 32-bit. 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 01/14] linux-user: Use memcpy in get_user/put_user Richard Henderson @ 2012-09-22 2:04 ` Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 03/14] tcg-sparc: Fix ADDX opcode Richard Henderson ` (12 subsequent siblings) 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:04 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl Not actually implemented, but at least we avoid the tcg assert at startup. Signed-off-by: Richard Henderson <rth@twiddle.net> --- tcg/sparc/tcg-target.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index baed3b4..608fc46 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -1556,6 +1556,9 @@ static const TCGTargetOpDef sparc_op_defs[] = { { INDEX_op_brcond_i64, { "r", "rJ" } }, { INDEX_op_setcond_i64, { "r", "r", "rJ" } }, +#else + { INDEX_op_qemu_ld64, { "L", "L", "L" } }, + { INDEX_op_qemu_st64, { "L", "L", "L" } }, #endif { -1 }, }; -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 03/14] tcg-sparc: Fix ADDX opcode. 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 01/14] linux-user: Use memcpy in get_user/put_user Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 02/14] tcg-sparc: Hack in qemu_ld/st64 for 32-bit Richard Henderson @ 2012-09-22 2:04 ` Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 04/14] tcg-sparc: Don't MAP_FIXED on top of the program Richard Henderson ` (11 subsequent siblings) 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:04 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl Signed-off-by: Richard Henderson <rth@twiddle.net> --- tcg/sparc/tcg-target.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index 608fc46..0a19313 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -236,7 +236,7 @@ static inline int tcg_target_const_match(tcg_target_long val, #define ARITH_XOR (INSN_OP(2) | INSN_OP3(0x03)) #define ARITH_SUB (INSN_OP(2) | INSN_OP3(0x04)) #define ARITH_SUBCC (INSN_OP(2) | INSN_OP3(0x14)) -#define ARITH_ADDX (INSN_OP(2) | INSN_OP3(0x10)) +#define ARITH_ADDX (INSN_OP(2) | INSN_OP3(0x08)) #define ARITH_SUBX (INSN_OP(2) | INSN_OP3(0x0c)) #define ARITH_UMUL (INSN_OP(2) | INSN_OP3(0x0a)) #define ARITH_UDIV (INSN_OP(2) | INSN_OP3(0x0e)) -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 04/14] tcg-sparc: Don't MAP_FIXED on top of the program 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson ` (2 preceding siblings ...) 2012-09-22 2:04 ` [Qemu-devel] [PATCH 03/14] tcg-sparc: Fix ADDX opcode Richard Henderson @ 2012-09-22 2:04 ` Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 05/14] tcg-sparc: Assume v9 cpu always, i.e. force v8plus in 32-bit mode Richard Henderson ` (10 subsequent siblings) 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:04 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl The address we pick in sparc64.ld is also 0x60000000, so doing a fixed map on top of that is guaranteed to blow up. Choosing 0x40000000 is exactly right for the max of code_gen_buffer_size set below. No need to ever use MAP_FIXED. While getting our desired address helps optimize the generated code, we won't fail if we don't get it. Signed-off-by: Richard Henderson <rth@twiddle.net> --- exec.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/exec.c b/exec.c index f22e9e6..4510985 100644 --- a/exec.c +++ b/exec.c @@ -543,8 +543,7 @@ static void code_gen_alloc(unsigned long tb_size) code_gen_buffer_size = (800 * 1024 * 1024); #elif defined(__sparc_v9__) // Map the buffer below 2G, so we can use direct calls and branches - flags |= MAP_FIXED; - start = (void *) 0x60000000UL; + start = (void *) 0x40000000UL; if (code_gen_buffer_size > (512 * 1024 * 1024)) code_gen_buffer_size = (512 * 1024 * 1024); #elif defined(__arm__) @@ -584,8 +583,7 @@ static void code_gen_alloc(unsigned long tb_size) code_gen_buffer_size = (800 * 1024 * 1024); #elif defined(__sparc_v9__) // Map the buffer below 2G, so we can use direct calls and branches - flags |= MAP_FIXED; - addr = (void *) 0x60000000UL; + addr = (void *) 0x40000000UL; if (code_gen_buffer_size > (512 * 1024 * 1024)) { code_gen_buffer_size = (512 * 1024 * 1024); } -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 05/14] tcg-sparc: Assume v9 cpu always, i.e. force v8plus in 32-bit mode. 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson ` (3 preceding siblings ...) 2012-09-22 2:04 ` [Qemu-devel] [PATCH 04/14] tcg-sparc: Don't MAP_FIXED on top of the program Richard Henderson @ 2012-09-22 2:04 ` Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 06/14] tcg-sparc: Fix qemu_ld/st to handle 32-bit host Richard Henderson ` (9 subsequent siblings) 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:04 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl Current code doesn't actually work in 32-bit mode at all. Since no one really noticed, drop the complication of v7 and v8 cpus. Eliminate the --sparc_cpu configure option and standardize macro testing on TCG_TARGET_REG_BITS / HOST_LONG_BITS Signed-off-by: Richard Henderson <rth@twiddle.net> --- configure | 40 ++++------------------------------------ disas.c | 2 -- exec.c | 6 +++--- qemu-timer.h | 8 +++++--- tcg/sparc/tcg-target.c | 20 +++++--------------- tcg/sparc/tcg-target.h | 7 ++++--- tcg/tcg.c | 3 ++- 7 files changed, 23 insertions(+), 63 deletions(-) diff --git a/configure b/configure index 9caa545..df888f2 100755 --- a/configure +++ b/configure @@ -111,7 +111,6 @@ source_path=`dirname "$0"` cpu="" interp_prefix="/usr/gnemul/qemu-%M" static="no" -sparc_cpu="" cross_prefix="" audio_drv_list="" audio_card_list="ac97 es1370 sb16 hda" @@ -241,21 +240,6 @@ for opt do ;; --disable-debug-info) debug_info="no" ;; - --sparc_cpu=*) - sparc_cpu="$optarg" - case $sparc_cpu in - v7|v8|v8plus|v8plusa) - cpu="sparc" - ;; - v9) - cpu="sparc64" - ;; - *) - echo "undefined SPARC architecture. Exiting"; - exit 1 - ;; - esac - ;; esac done # OS specific @@ -343,8 +327,6 @@ elif check_define __i386__ ; then elif check_define __x86_64__ ; then cpu="x86_64" elif check_define __sparc__ ; then - # We can't check for 64 bit (when gcc is biarch) or V8PLUSA - # They must be specified using --sparc_cpu if check_define __arch64__ ; then cpu="sparc64" else @@ -792,8 +774,6 @@ for opt do ;; --enable-uname-release=*) uname_release="$optarg" ;; - --sparc_cpu=*) - ;; --enable-werror) werror="yes" ;; --disable-werror) werror="no" @@ -881,31 +861,19 @@ for opt do esac done -# -# If cpu ~= sparc and sparc_cpu hasn't been defined, plug in the right -# QEMU_CFLAGS/LDFLAGS (assume sparc_v8plus for 32-bit and sparc_v9 for 64-bit) -# host_guest_base="no" case "$cpu" in - sparc) case $sparc_cpu in - v7|v8) - QEMU_CFLAGS="-mcpu=${sparc_cpu} -D__sparc_${sparc_cpu}__ $QEMU_CFLAGS" - ;; - v8plus|v8plusa) - QEMU_CFLAGS="-mcpu=ultrasparc -D__sparc_${sparc_cpu}__ $QEMU_CFLAGS" - ;; - *) # sparc_cpu not defined in the command line - QEMU_CFLAGS="-mcpu=ultrasparc -D__sparc_v8plus__ $QEMU_CFLAGS" - esac + sparc) LDFLAGS="-m32 $LDFLAGS" - QEMU_CFLAGS="-m32 -ffixed-g2 -ffixed-g3 $QEMU_CFLAGS" + QEMU_CFLAGS="-m32 -mcpu=ultrasparc $QEMU_CFLAGS" + QEMU_CFLAGS="-ffixed-g2 -ffixed-g3 $QEMU_CFLAGS" if test "$solaris" = "no" ; then QEMU_CFLAGS="-ffixed-g1 -ffixed-g6 $QEMU_CFLAGS" fi ;; sparc64) - QEMU_CFLAGS="-m64 -mcpu=ultrasparc -D__sparc_v9__ $QEMU_CFLAGS" LDFLAGS="-m64 $LDFLAGS" + QEMU_CFLAGS="-m64 -mcpu=ultrasparc $QEMU_CFLAGS" QEMU_CFLAGS="-ffixed-g5 -ffixed-g6 -ffixed-g7 $QEMU_CFLAGS" if test "$solaris" != "no" ; then QEMU_CFLAGS="-ffixed-g1 $QEMU_CFLAGS" diff --git a/disas.c b/disas.c index 7b2acc9..b801c8f 100644 --- a/disas.c +++ b/disas.c @@ -316,9 +316,7 @@ void disas(FILE *out, void *code, unsigned long size) print_insn = print_insn_alpha; #elif defined(__sparc__) print_insn = print_insn_sparc; -#if defined(__sparc_v8plus__) || defined(__sparc_v8plusa__) || defined(__sparc_v9__) disasm_info.mach = bfd_mach_sparc_v9b; -#endif #elif defined(__arm__) print_insn = print_insn_arm; #elif defined(__MIPSEB__) diff --git a/exec.c b/exec.c index 4510985..bb6aa4a 100644 --- a/exec.c +++ b/exec.c @@ -86,7 +86,7 @@ static int nb_tbs; /* any access to the tbs or the page table must use this lock */ spinlock_t tb_lock = SPIN_LOCK_UNLOCKED; -#if defined(__arm__) || defined(__sparc_v9__) +#if defined(__arm__) || defined(__sparc__) /* The prologue must be reachable with a direct jump. ARM and Sparc64 have limited branch ranges (possibly also PPC) so place it in a section close to code segment. */ @@ -541,7 +541,7 @@ static void code_gen_alloc(unsigned long tb_size) /* Cannot map more than that */ if (code_gen_buffer_size > (800 * 1024 * 1024)) code_gen_buffer_size = (800 * 1024 * 1024); -#elif defined(__sparc_v9__) +#elif defined(__sparc__) && HOST_LONG_BITS == 64 // Map the buffer below 2G, so we can use direct calls and branches start = (void *) 0x40000000UL; if (code_gen_buffer_size > (512 * 1024 * 1024)) @@ -581,7 +581,7 @@ static void code_gen_alloc(unsigned long tb_size) /* Cannot map more than that */ if (code_gen_buffer_size > (800 * 1024 * 1024)) code_gen_buffer_size = (800 * 1024 * 1024); -#elif defined(__sparc_v9__) +#elif defined(__sparc__) && HOST_LONG_BITS == 64 // Map the buffer below 2G, so we can use direct calls and branches addr = (void *) 0x40000000UL; if (code_gen_buffer_size > (512 * 1024 * 1024)) { diff --git a/qemu-timer.h b/qemu-timer.h index f8af595..da7e97c 100644 --- a/qemu-timer.h +++ b/qemu-timer.h @@ -218,7 +218,7 @@ static inline int64_t cpu_get_real_ticks(void) return val; } -#elif defined(__sparc_v8plus__) || defined(__sparc_v8plusa__) || defined(__sparc_v9__) +#elif defined(__sparc__) static inline int64_t cpu_get_real_ticks (void) { @@ -227,6 +227,8 @@ static inline int64_t cpu_get_real_ticks (void) asm volatile("rd %%tick,%0" : "=r"(rval)); return rval; #else + /* We need an %o or %g register for this. For recent enough gcc + there is an "h" constraint for that. Don't bother with that. */ union { uint64_t i64; struct { @@ -234,8 +236,8 @@ static inline int64_t cpu_get_real_ticks (void) uint32_t low; } i32; } rval; - asm volatile("rd %%tick,%1; srlx %1,32,%0" - : "=r"(rval.i32.high), "=r"(rval.i32.low)); + asm volatile("rd %%tick,%%g1; srlx %%g1,32,%0; mov %%g1,%1" + : "=r"(rval.i32.high), "=r"(rval.i32.low) : : "g1"); return rval.i64; #endif } diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index 0a19313..23c2fda 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -621,18 +621,10 @@ static void tcg_out_setcond_i32(TCGContext *s, TCGCond cond, TCGArg ret, default: tcg_out_cmp(s, c1, c2, c2const); -#if defined(__sparc_v9__) || defined(__sparc_v8plus__) tcg_out_movi_imm13(s, ret, 0); - tcg_out32 (s, ARITH_MOVCC | INSN_RD(ret) - | INSN_RS1(tcg_cond_to_bcond[cond]) - | MOVCC_ICC | INSN_IMM11(1)); -#else - t = gen_new_label(); - tcg_out_branch_i32(s, INSN_COND(tcg_cond_to_bcond[cond], 1), t); - tcg_out_movi_imm13(s, ret, 1); - tcg_out_movi_imm13(s, ret, 0); - tcg_out_label(s, t, s->code_ptr); -#endif + tcg_out32(s, ARITH_MOVCC | INSN_RD(ret) + | INSN_RS1(tcg_cond_to_bcond[cond]) + | MOVCC_ICC | INSN_IMM11(1)); return; } @@ -742,7 +734,7 @@ static const void * const qemu_st_helpers[4] = { #endif #endif -#ifdef __arch64__ +#if TCG_TARGET_REG_BITS == 64 #define HOST_LD_OP LDX #define HOST_ST_OP STX #define HOST_SLL_OP SHIFT_SLLX @@ -1600,11 +1592,9 @@ static void tcg_target_init(TCGContext *s) #if TCG_TARGET_REG_BITS == 64 # define ELF_HOST_MACHINE EM_SPARCV9 -#elif defined(__sparc_v8plus__) +#else # define ELF_HOST_MACHINE EM_SPARC32PLUS # define ELF_HOST_FLAGS EF_SPARC_32PLUS -#else -# define ELF_HOST_MACHINE EM_SPARC #endif typedef struct { diff --git a/tcg/sparc/tcg-target.h b/tcg/sparc/tcg-target.h index d762574..adca1d2 100644 --- a/tcg/sparc/tcg-target.h +++ b/tcg/sparc/tcg-target.h @@ -67,7 +67,8 @@ typedef enum { /* used for function call generation */ #define TCG_REG_CALL_STACK TCG_REG_I6 -#ifdef __arch64__ + +#if TCG_TARGET_REG_BITS == 64 // Reserve space for AREG0 #define TCG_TARGET_STACK_MINFRAME (176 + 4 * (int)sizeof(long) + \ TCG_STATIC_CALL_ARGS_SIZE) @@ -81,7 +82,7 @@ typedef enum { #define TCG_TARGET_STACK_ALIGN 8 #endif -#ifdef __arch64__ +#if TCG_TARGET_REG_BITS == 64 #define TCG_TARGET_EXTEND_ARGS 1 #endif @@ -129,7 +130,7 @@ typedef enum { #ifdef CONFIG_SOLARIS #define TCG_AREG0 TCG_REG_G2 -#elif defined(__sparc_v9__) +#elif HOST_LONG_BITS == 64 #define TCG_AREG0 TCG_REG_G5 #else #define TCG_AREG0 TCG_REG_G6 diff --git a/tcg/tcg.c b/tcg/tcg.c index bb9c995..45a26f2 100644 --- a/tcg/tcg.c +++ b/tcg/tcg.c @@ -1457,7 +1457,8 @@ static void temp_allocate_frame(TCGContext *s, int temp) { TCGTemp *ts; ts = &s->temps[temp]; -#ifndef __sparc_v9__ /* Sparc64 stack is accessed with offset of 2047 */ +#if !(defined(__sparc__) && TCG_TARGET_REG_BITS == 64) + /* Sparc64 stack is accessed with offset of 2047 */ s->current_frame_offset = (s->current_frame_offset + (tcg_target_long)sizeof(tcg_target_long) - 1) & ~(sizeof(tcg_target_long) - 1); -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 06/14] tcg-sparc: Fix qemu_ld/st to handle 32-bit host. 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson ` (4 preceding siblings ...) 2012-09-22 2:04 ` [Qemu-devel] [PATCH 05/14] tcg-sparc: Assume v9 cpu always, i.e. force v8plus in 32-bit mode Richard Henderson @ 2012-09-22 2:04 ` Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 07/14] tcg-sparc: Support GUEST_BASE Richard Henderson ` (8 subsequent siblings) 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:04 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl At the same time, split out the tlb load logic to a new function. Fixes the cases of two data registers and two address registers. Fixes the signature of, and adds missing, qemu_ld/st opcodes. Signed-off-by: Richard Henderson <rth@twiddle.net> --- tcg/sparc/tcg-target.c | 777 ++++++++++++++++++++++--------------------------- 1 file changed, 348 insertions(+), 429 deletions(-) diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index 23c2fda..d89c19b 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -59,8 +59,6 @@ static const char * const tcg_target_reg_names[TCG_TARGET_NB_REGS] = { }; #endif -#define ARG_OFFSET 1 - static const int tcg_target_reg_alloc_order[] = { TCG_REG_L0, TCG_REG_L1, @@ -288,6 +286,16 @@ static inline int tcg_target_const_match(tcg_target_long val, #define ASI_PRIMARY_LITTLE 0x88 #endif +#define LDUH_LE (LDUHA | INSN_ASI(ASI_PRIMARY_LITTLE)) +#define LDSH_LE (LDSHA | INSN_ASI(ASI_PRIMARY_LITTLE)) +#define LDUW_LE (LDUWA | INSN_ASI(ASI_PRIMARY_LITTLE)) +#define LDSW_LE (LDSWA | INSN_ASI(ASI_PRIMARY_LITTLE)) +#define LDX_LE (LDXA | INSN_ASI(ASI_PRIMARY_LITTLE)) + +#define STH_LE (STHA | INSN_ASI(ASI_PRIMARY_LITTLE)) +#define STW_LE (STWA | INSN_ASI(ASI_PRIMARY_LITTLE)) +#define STX_LE (STXA | INSN_ASI(ASI_PRIMARY_LITTLE)) + static inline void tcg_out_arith(TCGContext *s, int rd, int rs1, int rs2, int op) { @@ -360,64 +368,43 @@ static inline void tcg_out_movi(TCGContext *s, TCGType type, } } -static inline void tcg_out_ld_raw(TCGContext *s, int ret, - tcg_target_long arg) -{ - tcg_out_sethi(s, ret, arg); - tcg_out32(s, LDUW | INSN_RD(ret) | INSN_RS1(ret) | - INSN_IMM13(arg & 0x3ff)); -} - -static inline void tcg_out_ld_ptr(TCGContext *s, int ret, - tcg_target_long arg) +static inline void tcg_out_ldst_rr(TCGContext *s, int data, int a1, + int a2, int op) { - if (!check_fit_tl(arg, 10)) - tcg_out_movi(s, TCG_TYPE_PTR, ret, arg & ~0x3ffULL); - if (TCG_TARGET_REG_BITS == 64) { - tcg_out32(s, LDX | INSN_RD(ret) | INSN_RS1(ret) | - INSN_IMM13(arg & 0x3ff)); - } else { - tcg_out32(s, LDUW | INSN_RD(ret) | INSN_RS1(ret) | - INSN_IMM13(arg & 0x3ff)); - } + tcg_out32(s, op | INSN_RD(data) | INSN_RS1(a1) | INSN_RS2(a2)); } -static inline void tcg_out_ldst(TCGContext *s, int ret, int addr, int offset, int op) +static inline void tcg_out_ldst(TCGContext *s, int ret, int addr, + int offset, int op) { - if (check_fit_tl(offset, 13)) + if (check_fit_tl(offset, 13)) { tcg_out32(s, op | INSN_RD(ret) | INSN_RS1(addr) | INSN_IMM13(offset)); - else { + } else { tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_I5, offset); - tcg_out32(s, op | INSN_RD(ret) | INSN_RS1(TCG_REG_I5) | - INSN_RS2(addr)); + tcg_out_ldst_rr(s, ret, addr, TCG_REG_I5, op); } } -static inline void tcg_out_ldst_asi(TCGContext *s, int ret, int addr, - int offset, int op, int asi) -{ - tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_I5, offset); - tcg_out32(s, op | INSN_RD(ret) | INSN_RS1(TCG_REG_I5) | - INSN_ASI(asi) | INSN_RS2(addr)); -} - static inline void tcg_out_ld(TCGContext *s, TCGType type, TCGReg ret, TCGReg arg1, tcg_target_long arg2) { - if (type == TCG_TYPE_I32) - tcg_out_ldst(s, ret, arg1, arg2, LDUW); - else - tcg_out_ldst(s, ret, arg1, arg2, LDX); + tcg_out_ldst(s, ret, arg1, arg2, (type == TCG_TYPE_I32 ? LDUW : LDX)); } static inline void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg, TCGReg arg1, tcg_target_long arg2) { - if (type == TCG_TYPE_I32) - tcg_out_ldst(s, arg, arg1, arg2, STW); - else - tcg_out_ldst(s, arg, arg1, arg2, STX); + tcg_out_ldst(s, arg, arg1, arg2, (type == TCG_TYPE_I32 ? STW : STX)); +} + +static inline void tcg_out_ld_ptr(TCGContext *s, int ret, + tcg_target_long arg) +{ + if (!check_fit_tl(arg, 10)) { + tcg_out_movi(s, TCG_TYPE_PTR, ret, arg & ~0x3ff); + } + tcg_out_ld(s, TCG_TYPE_PTR, ret, ret, arg & 0x3ff); } static inline void tcg_out_sety(TCGContext *s, int rs) @@ -442,14 +429,15 @@ static inline void tcg_out_addi(TCGContext *s, int reg, tcg_target_long val) } } -static inline void tcg_out_andi(TCGContext *s, int reg, tcg_target_long val) +static inline void tcg_out_andi(TCGContext *s, int rd, int rs, + tcg_target_long val) { if (val != 0) { if (check_fit_tl(val, 13)) - tcg_out_arithi(s, reg, reg, val, ARITH_AND); + tcg_out_arithi(s, rd, rs, val, ARITH_AND); else { tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_I5, val); - tcg_out_arith(s, reg, reg, TCG_REG_I5, ARITH_AND); + tcg_out_arith(s, rd, rs, TCG_REG_I5, ARITH_AND); } } } @@ -718,418 +706,328 @@ static const void * const qemu_st_helpers[4] = { helper_stl_mmu, helper_stq_mmu, }; -#endif -#if TARGET_LONG_BITS == 32 -#define TARGET_LD_OP LDUW -#else -#define TARGET_LD_OP LDX -#endif +/* Perform the TLB load and compare. -#if defined(CONFIG_SOFTMMU) -#if HOST_LONG_BITS == 32 -#define TARGET_ADDEND_LD_OP LDUW + Inputs: + ADDRLO_IDX contains the index into ARGS of the low part of the + address; the high part of the address is at ADDR_LOW_IDX+1. + + MEM_INDEX and S_BITS are the memory context and log2 size of the load. + + WHICH is the offset into the CPUTLBEntry structure of the slot to read. + This should be offsetof addr_read or addr_write. + + The result of the TLB comparison is in %[ix]cc. The sanitized address + is in the returned register, maybe %o0. The TLB addend is in %o1. */ + +static int tcg_out_tlb_load(TCGContext *s, int addrlo_idx, int mem_index, + int s_bits, const TCGArg *args, int which) +{ + const int addrlo = args[addrlo_idx]; + const int r0 = TCG_REG_O0; + const int r1 = TCG_REG_O1; + const int r2 = TCG_REG_O2; + int addr = addrlo; + int tlb_ofs; + + if (TCG_TARGET_REG_BITS == 32 && TARGET_LONG_BITS == 64) { + /* Assemble the 64-bit address in R0. */ + tcg_out_arithi(s, r0, addrlo, 0, SHIFT_SRL); + tcg_out_arithi(s, r1, args[addrlo_idx + 1], 32, SHIFT_SLLX); + tcg_out_arith(s, r0, r0, r1, ARITH_OR); + } + + /* Shift the page number down to tlb-entry. */ + tcg_out_arithi(s, r1, addrlo, + TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS, SHIFT_SRL); + + /* Mask out the page offset, except for the required alignment. */ + tcg_out_andi(s, r0, addr, TARGET_PAGE_MASK | ((1 << s_bits) - 1)); + + /* Compute tlb index, modulo tlb size. */ + tcg_out_andi(s, r1, r1, (CPU_TLB_SIZE - 1) << CPU_TLB_ENTRY_BITS); + + /* Relative to the current ENV. */ + tcg_out_arith(s, r1, TCG_AREG0, r1, ARITH_ADD); + + /* Find a base address that can load both tlb comparator and addend. */ + tlb_ofs = offsetof(CPUArchState, tlb_table[mem_index][0]); + if (!check_fit_tl(tlb_ofs + sizeof(CPUTLBEntry), 13)) { + tcg_out_addi(s, r1, tlb_ofs); + tlb_ofs = 0; + } + + /* Load the tlb comparator and the addend. */ + tcg_out_ld(s, TCG_TYPE_TL, r2, r1, tlb_ofs + which); + tcg_out_ld(s, TCG_TYPE_PTR, r1, r1, tlb_ofs+offsetof(CPUTLBEntry, addend)); + + /* subcc arg0, arg2, %g0 */ + tcg_out_cmp(s, r0, r2, 0); + + /* If the guest address must be zero-extended, do so now. */ + if (TCG_TARGET_REG_BITS == 64 && TARGET_LONG_BITS == 32) { + tcg_out_arithi(s, r0, addrlo, 0, SHIFT_SRL); + return r0; + } + return addrlo; +} +#endif /* CONFIG_SOFTMMU */ + +static const int qemu_ld_opc[8] = { +#ifdef TARGET_WORDS_BIGENDIAN + LDUB, LDUH, LDUW, LDX, LDSB, LDSH, LDSW, LDX #else -#define TARGET_ADDEND_LD_OP LDX -#endif + LDUB, LDUH_LE, LDUW_LE, LDX_LE, LDSB, LDSH_LE, LDSW_LE, LDX_LE #endif +}; -#if TCG_TARGET_REG_BITS == 64 -#define HOST_LD_OP LDX -#define HOST_ST_OP STX -#define HOST_SLL_OP SHIFT_SLLX -#define HOST_SRA_OP SHIFT_SRAX +static const int qemu_st_opc[4] = { +#ifdef TARGET_WORDS_BIGENDIAN + STB, STH, STW, STX #else -#define HOST_LD_OP LDUW -#define HOST_ST_OP STW -#define HOST_SLL_OP SHIFT_SLL -#define HOST_SRA_OP SHIFT_SRA + STB, STH_LE, STW_LE, STX_LE #endif +}; -static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, - int opc) +static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int sizeop) { - int addr_reg, data_reg, arg0, arg1, arg2, mem_index, s_bits; + int addrlo_idx = 1, datalo, datahi, addr_reg; #if defined(CONFIG_SOFTMMU) - uint32_t *label1_ptr, *label2_ptr; + int memi_idx, memi, s_bits, n; + uint32_t *label_ptr[2]; #endif - data_reg = *args++; - addr_reg = *args++; - mem_index = *args; - s_bits = opc & 3; - - arg0 = TCG_REG_O0; - arg1 = TCG_REG_O1; - arg2 = TCG_REG_O2; + datahi = datalo = args[0]; + if (TCG_TARGET_REG_BITS == 32 && sizeop == 3) { + datahi = args[1]; + addrlo_idx = 2; + } #if defined(CONFIG_SOFTMMU) - /* srl addr_reg, x, arg1 */ - tcg_out_arithi(s, arg1, addr_reg, TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS, - SHIFT_SRL); - /* and addr_reg, x, arg0 */ - tcg_out_arithi(s, arg0, addr_reg, TARGET_PAGE_MASK | ((1 << s_bits) - 1), - ARITH_AND); + memi_idx = addrlo_idx + 1 + (TARGET_LONG_BITS > TCG_TARGET_REG_BITS); + memi = args[memi_idx]; + s_bits = sizeop & 3; + + addr_reg = tcg_out_tlb_load(s, addrlo_idx, memi, s_bits, args, + offsetof(CPUTLBEntry, addr_read)); + + if (TCG_TARGET_REG_BITS == 32 && sizeop == 3) { + int reg64; + + /* bne,pn %[xi]cc, label0 */ + label_ptr[0] = (uint32_t *)s->code_ptr; + tcg_out32(s, (INSN_OP(0) | INSN_COND(COND_NE, 0) | INSN_OP2(0x1) + | ((TARGET_LONG_BITS == 64) << 21))); + + /* TLB Hit. */ + /* Load all 64-bits into an O/G register. */ + reg64 = (datalo < 16 ? datalo : TCG_REG_O0); + tcg_out_ldst_rr(s, reg64, addr_reg, TCG_REG_O1, qemu_ld_opc[sizeop]); + + /* Move the two 32-bit pieces into the destination registers. */ + tcg_out_arithi(s, datahi, reg64, 32, SHIFT_SRLX); + if (reg64 != datalo) { + tcg_out_mov(s, TCG_TYPE_I32, datalo, reg64); + } - /* and arg1, x, arg1 */ - tcg_out_andi(s, arg1, (CPU_TLB_SIZE - 1) << CPU_TLB_ENTRY_BITS); + /* b,a,pt label1 */ + label_ptr[1] = (uint32_t *)s->code_ptr; + tcg_out32(s, (INSN_OP(0) | INSN_COND(COND_A, 0) | INSN_OP2(0x1) + | (1 << 29) | (1 << 19))); + } else { + /* The fast path is exactly one insn. Thus we can perform the + entire TLB Hit in the (annulled) delay slot of the branch + over the TLB Miss case. */ + + /* beq,a,pt %[xi]cc, label0 */ + label_ptr[0] = NULL; + label_ptr[1] = (uint32_t *)s->code_ptr; + tcg_out32(s, (INSN_OP(0) | INSN_COND(COND_E, 0) | INSN_OP2(0x1) + | ((TARGET_LONG_BITS == 64) << 21) + | (1 << 29) | (1 << 19))); + /* delay slot */ + tcg_out_ldst_rr(s, datalo, addr_reg, TCG_REG_O1, qemu_ld_opc[sizeop]); + } - /* add arg1, x, arg1 */ - tcg_out_addi(s, arg1, offsetof(CPUArchState, - tlb_table[mem_index][0].addr_read)); + /* TLB Miss. */ - /* add env, arg1, arg1 */ - tcg_out_arith(s, arg1, TCG_AREG0, arg1, ARITH_ADD); + if (label_ptr[0]) { + *label_ptr[0] |= INSN_OFF19((unsigned long)s->code_ptr - + (unsigned long)label_ptr[0]); + } + n = 0; + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[n++], TCG_AREG0); + if (TARGET_LONG_BITS > TCG_TARGET_REG_BITS) { + tcg_out_mov(s, TCG_TYPE_REG, tcg_target_call_iarg_regs[n++], + args[addrlo_idx + 1]); + } + tcg_out_mov(s, TCG_TYPE_REG, tcg_target_call_iarg_regs[n++], + args[addrlo_idx]); - /* ld [arg1], arg2 */ - tcg_out32(s, TARGET_LD_OP | INSN_RD(arg2) | INSN_RS1(arg1) | - INSN_RS2(TCG_REG_G0)); + /* Store AREG0 in stack to avoid ugly glibc bugs that mangle + global registers */ + tcg_out_st(s, TCG_TYPE_REG, TCG_AREG0, TCG_REG_CALL_STACK, + TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - + sizeof(long)); - /* subcc arg0, arg2, %g0 */ - tcg_out_arith(s, TCG_REG_G0, arg0, arg2, ARITH_SUBCC); - - /* will become: - be label1 - or - be,pt %xcc label1 */ - label1_ptr = (uint32_t *)s->code_ptr; - tcg_out32(s, 0); - - /* mov (delay slot) */ - tcg_out_mov(s, TCG_TYPE_PTR, arg0, addr_reg); - - /* mov */ - tcg_out_movi(s, TCG_TYPE_I32, arg1, mem_index); - /* XXX/FIXME: suboptimal */ - tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[3], - tcg_target_call_iarg_regs[2]); - tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[2], - tcg_target_call_iarg_regs[1]); - tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], - tcg_target_call_iarg_regs[0]); - tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], - TCG_AREG0); - - /* XXX: move that code at the end of the TB */ /* qemu_ld_helper[s_bits](arg0, arg1) */ tcg_out32(s, CALL | ((((tcg_target_ulong)qemu_ld_helpers[s_bits] - (tcg_target_ulong)s->code_ptr) >> 2) & 0x3fffffff)); - /* Store AREG0 in stack to avoid ugly glibc bugs that mangle - global registers */ - // delay slot - tcg_out_ldst(s, TCG_AREG0, TCG_REG_CALL_STACK, - TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - - sizeof(long), HOST_ST_OP); - tcg_out_ldst(s, TCG_AREG0, TCG_REG_CALL_STACK, - TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - - sizeof(long), HOST_LD_OP); - - /* data_reg = sign_extend(arg0) */ - switch(opc) { + /* delay slot */ + tcg_out_movi(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[n], memi); + + /* Reload AREG0. */ + tcg_out_ld(s, TCG_TYPE_REG, TCG_AREG0, TCG_REG_CALL_STACK, + TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - + sizeof(long)); + + n = tcg_target_call_oarg_regs[0]; + /* datalo = sign_extend(arg0) */ + switch (sizeop) { case 0 | 4: - /* sll arg0, 24/56, data_reg */ - tcg_out_arithi(s, data_reg, arg0, (int)sizeof(tcg_target_long) * 8 - 8, - HOST_SLL_OP); - /* sra data_reg, 24/56, data_reg */ - tcg_out_arithi(s, data_reg, data_reg, - (int)sizeof(tcg_target_long) * 8 - 8, HOST_SRA_OP); + /* Recall that SRA sign extends from bit 31 through bit 63. */ + tcg_out_arithi(s, datalo, n, 24, SHIFT_SLL); + tcg_out_arithi(s, datalo, datalo, 24, SHIFT_SRA); break; case 1 | 4: - /* sll arg0, 16/48, data_reg */ - tcg_out_arithi(s, data_reg, arg0, - (int)sizeof(tcg_target_long) * 8 - 16, HOST_SLL_OP); - /* sra data_reg, 16/48, data_reg */ - tcg_out_arithi(s, data_reg, data_reg, - (int)sizeof(tcg_target_long) * 8 - 16, HOST_SRA_OP); + tcg_out_arithi(s, datalo, n, 16, SHIFT_SLL); + tcg_out_arithi(s, datalo, datalo, 16, SHIFT_SRA); break; case 2 | 4: - /* sll arg0, 32, data_reg */ - tcg_out_arithi(s, data_reg, arg0, 32, HOST_SLL_OP); - /* sra data_reg, 32, data_reg */ - tcg_out_arithi(s, data_reg, data_reg, 32, HOST_SRA_OP); + tcg_out_arithi(s, datalo, n, 0, SHIFT_SRA); break; + case 3: + if (TCG_TARGET_REG_BITS == 32) { + tcg_out_mov(s, TCG_TYPE_REG, datahi, n); + tcg_out_mov(s, TCG_TYPE_REG, datalo, n + 1); + break; + } + /* FALLTHRU */ case 0: case 1: case 2: - case 3: default: /* mov */ - tcg_out_mov(s, TCG_TYPE_REG, data_reg, arg0); + tcg_out_mov(s, TCG_TYPE_REG, datalo, n); break; } - /* will become: - ba label2 */ - label2_ptr = (uint32_t *)s->code_ptr; - tcg_out32(s, 0); - - /* nop (delay slot */ - tcg_out_nop(s); - - /* label1: */ -#if TARGET_LONG_BITS == 32 - /* be label1 */ - *label1_ptr = (INSN_OP(0) | INSN_COND(COND_E, 0) | INSN_OP2(0x2) | - INSN_OFF22((unsigned long)s->code_ptr - - (unsigned long)label1_ptr)); -#else - /* be,pt %xcc label1 */ - *label1_ptr = (INSN_OP(0) | INSN_COND(COND_E, 0) | INSN_OP2(0x1) | - (0x5 << 19) | INSN_OFF19((unsigned long)s->code_ptr - - (unsigned long)label1_ptr)); -#endif - - /* ld [arg1 + x], arg1 */ - tcg_out_ldst(s, arg1, arg1, offsetof(CPUTLBEntry, addend) - - offsetof(CPUTLBEntry, addr_read), TARGET_ADDEND_LD_OP); - -#if TARGET_LONG_BITS == 32 - /* and addr_reg, x, arg0 */ - tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_I5, 0xffffffff); - tcg_out_arith(s, arg0, addr_reg, TCG_REG_I5, ARITH_AND); - /* add arg0, arg1, arg0 */ - tcg_out_arith(s, arg0, arg0, arg1, ARITH_ADD); + *label_ptr[1] |= INSN_OFF19((unsigned long)s->code_ptr - + (unsigned long)label_ptr[1]); #else - /* add addr_reg, arg1, arg0 */ - tcg_out_arith(s, arg0, addr_reg, arg1, ARITH_ADD); -#endif + addr_reg = args[addrlo_idx]; + if (TCG_TARGET_REG_BITS == 64 && TARGET_LONG_BITS == 32) { + tcg_out_arithi(s, TCG_REG_I5, addr_reg, 0, SHIFT_SRL); + addr_reg = TCG_REG_I5; + } + if (TCG_TARGET_REG_BITS == 32 && sizeop == 3) { + int reg64 = (datalo < 16 ? datalo : TCG_REG_O0); -#else - arg0 = addr_reg; -#endif + tcg_out_ldst_rr(s, reg64, addr_reg, TCG_REG_G0, qemu_ld_opc[sizeop]); - switch(opc) { - case 0: - /* ldub [arg0], data_reg */ - tcg_out_ldst(s, data_reg, arg0, 0, LDUB); - break; - case 0 | 4: - /* ldsb [arg0], data_reg */ - tcg_out_ldst(s, data_reg, arg0, 0, LDSB); - break; - case 1: -#ifdef TARGET_WORDS_BIGENDIAN - /* lduh [arg0], data_reg */ - tcg_out_ldst(s, data_reg, arg0, 0, LDUH); -#else - /* lduha [arg0] ASI_PRIMARY_LITTLE, data_reg */ - tcg_out_ldst_asi(s, data_reg, arg0, 0, LDUHA, ASI_PRIMARY_LITTLE); -#endif - break; - case 1 | 4: -#ifdef TARGET_WORDS_BIGENDIAN - /* ldsh [arg0], data_reg */ - tcg_out_ldst(s, data_reg, arg0, 0, LDSH); -#else - /* ldsha [arg0] ASI_PRIMARY_LITTLE, data_reg */ - tcg_out_ldst_asi(s, data_reg, arg0, 0, LDSHA, ASI_PRIMARY_LITTLE); -#endif - break; - case 2: -#ifdef TARGET_WORDS_BIGENDIAN - /* lduw [arg0], data_reg */ - tcg_out_ldst(s, data_reg, arg0, 0, LDUW); -#else - /* lduwa [arg0] ASI_PRIMARY_LITTLE, data_reg */ - tcg_out_ldst_asi(s, data_reg, arg0, 0, LDUWA, ASI_PRIMARY_LITTLE); -#endif - break; - case 2 | 4: -#ifdef TARGET_WORDS_BIGENDIAN - /* ldsw [arg0], data_reg */ - tcg_out_ldst(s, data_reg, arg0, 0, LDSW); -#else - /* ldswa [arg0] ASI_PRIMARY_LITTLE, data_reg */ - tcg_out_ldst_asi(s, data_reg, arg0, 0, LDSWA, ASI_PRIMARY_LITTLE); -#endif - break; - case 3: -#ifdef TARGET_WORDS_BIGENDIAN - /* ldx [arg0], data_reg */ - tcg_out_ldst(s, data_reg, arg0, 0, LDX); -#else - /* ldxa [arg0] ASI_PRIMARY_LITTLE, data_reg */ - tcg_out_ldst_asi(s, data_reg, arg0, 0, LDXA, ASI_PRIMARY_LITTLE); -#endif - break; - default: - tcg_abort(); + tcg_out_arithi(s, datahi, reg64, 32, SHIFT_SRLX); + if (reg64 != datalo) { + tcg_out_mov(s, TCG_TYPE_I32, datalo, reg64); + } + } else { + tcg_out_ldst_rr(s, datalo, addr_reg, TCG_REG_G0, qemu_ld_opc[sizeop]); } - -#if defined(CONFIG_SOFTMMU) - /* label2: */ - *label2_ptr = (INSN_OP(0) | INSN_COND(COND_A, 0) | INSN_OP2(0x2) | - INSN_OFF22((unsigned long)s->code_ptr - - (unsigned long)label2_ptr)); -#endif +#endif /* CONFIG_SOFTMMU */ } -static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, - int opc) +static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int sizeop) { - int addr_reg, data_reg, arg0, arg1, arg2, mem_index, s_bits; + int addrlo_idx = 1, datalo, datahi, addr_reg; #if defined(CONFIG_SOFTMMU) - uint32_t *label1_ptr, *label2_ptr; + int memi_idx, memi, n; + uint32_t *label_ptr; #endif - data_reg = *args++; - addr_reg = *args++; - mem_index = *args; - - s_bits = opc; - - arg0 = TCG_REG_O0; - arg1 = TCG_REG_O1; - arg2 = TCG_REG_O2; + datahi = datalo = args[0]; + if (TCG_TARGET_REG_BITS == 32 && sizeop == 3) { + datahi = args[1]; + addrlo_idx = 2; + } #if defined(CONFIG_SOFTMMU) - /* srl addr_reg, x, arg1 */ - tcg_out_arithi(s, arg1, addr_reg, TARGET_PAGE_BITS - CPU_TLB_ENTRY_BITS, - SHIFT_SRL); - - /* and addr_reg, x, arg0 */ - tcg_out_arithi(s, arg0, addr_reg, TARGET_PAGE_MASK | ((1 << s_bits) - 1), - ARITH_AND); - - /* and arg1, x, arg1 */ - tcg_out_andi(s, arg1, (CPU_TLB_SIZE - 1) << CPU_TLB_ENTRY_BITS); - - /* add arg1, x, arg1 */ - tcg_out_addi(s, arg1, offsetof(CPUArchState, - tlb_table[mem_index][0].addr_write)); + memi_idx = addrlo_idx + 1 + (TARGET_LONG_BITS > TCG_TARGET_REG_BITS); + memi = args[memi_idx]; + + addr_reg = tcg_out_tlb_load(s, addrlo_idx, memi, sizeop, args, + offsetof(CPUTLBEntry, addr_write)); + + if (TCG_TARGET_REG_BITS == 32 && sizeop == 3) { + /* Reconstruct the full 64-bit value in %g1, using %o2 as temp. */ + /* ??? Redefine the temps from %i4/%i5 so that we have a o/g temp. */ + tcg_out_arithi(s, TCG_REG_G1, datalo, 0, SHIFT_SRL); + tcg_out_arithi(s, TCG_REG_O2, datahi, 32, SHIFT_SLLX); + tcg_out_arith(s, TCG_REG_G1, TCG_REG_G1, TCG_REG_O2, ARITH_OR); + datalo = TCG_REG_G1; + } - /* add env, arg1, arg1 */ - tcg_out_arith(s, arg1, TCG_AREG0, arg1, ARITH_ADD); + /* The fast path is exactly one insn. Thus we can perform the entire + TLB Hit in the (annulled) delay slot of the branch over TLB Miss. */ + /* beq,a,pt %[xi]cc, label0 */ + label_ptr = (uint32_t *)s->code_ptr; + tcg_out32(s, (INSN_OP(0) | INSN_COND(COND_E, 0) | INSN_OP2(0x1) + | ((TARGET_LONG_BITS == 64) << 21) + | (1 << 29) | (1 << 19))); + /* delay slot */ + tcg_out_ldst_rr(s, datalo, addr_reg, TCG_REG_O1, qemu_st_opc[sizeop]); + + /* TLB Miss. */ + + n = 0; + tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[n++], TCG_AREG0); + if (TARGET_LONG_BITS > TCG_TARGET_REG_BITS) { + tcg_out_mov(s, TCG_TYPE_REG, tcg_target_call_iarg_regs[n++], + args[addrlo_idx + 1]); + } + tcg_out_mov(s, TCG_TYPE_REG, tcg_target_call_iarg_regs[n++], + args[addrlo_idx]); + if (TCG_TARGET_REG_BITS == 32 && sizeop == 3) { + tcg_out_mov(s, TCG_TYPE_REG, tcg_target_call_iarg_regs[n++], datahi); + } + tcg_out_mov(s, TCG_TYPE_REG, tcg_target_call_iarg_regs[n++], datalo); - /* ld [arg1], arg2 */ - tcg_out32(s, TARGET_LD_OP | INSN_RD(arg2) | INSN_RS1(arg1) | - INSN_RS2(TCG_REG_G0)); + /* Store AREG0 in stack to avoid ugly glibc bugs that mangle + global registers */ + tcg_out_st(s, TCG_TYPE_REG, TCG_AREG0, TCG_REG_CALL_STACK, + TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - + sizeof(long)); - /* subcc arg0, arg2, %g0 */ - tcg_out_arith(s, TCG_REG_G0, arg0, arg2, ARITH_SUBCC); - - /* will become: - be label1 - or - be,pt %xcc label1 */ - label1_ptr = (uint32_t *)s->code_ptr; - tcg_out32(s, 0); - - /* mov (delay slot) */ - tcg_out_mov(s, TCG_TYPE_PTR, arg0, addr_reg); - - /* mov */ - tcg_out_mov(s, TCG_TYPE_REG, arg1, data_reg); - - /* mov */ - tcg_out_movi(s, TCG_TYPE_I32, arg2, mem_index); - - /* XXX/FIXME: suboptimal */ - tcg_out_mov(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[3], - tcg_target_call_iarg_regs[2]); - tcg_out_mov(s, TCG_TYPE_I64, tcg_target_call_iarg_regs[2], - tcg_target_call_iarg_regs[1]); - tcg_out_mov(s, TCG_TYPE_TL, tcg_target_call_iarg_regs[1], - tcg_target_call_iarg_regs[0]); - tcg_out_mov(s, TCG_TYPE_PTR, tcg_target_call_iarg_regs[0], - TCG_AREG0); - /* XXX: move that code at the end of the TB */ /* qemu_st_helper[s_bits](arg0, arg1, arg2) */ - tcg_out32(s, CALL | ((((tcg_target_ulong)qemu_st_helpers[s_bits] + tcg_out32(s, CALL | ((((tcg_target_ulong)qemu_st_helpers[sizeop] - (tcg_target_ulong)s->code_ptr) >> 2) & 0x3fffffff)); - /* Store AREG0 in stack to avoid ugly glibc bugs that mangle - global registers */ - // delay slot - tcg_out_ldst(s, TCG_AREG0, TCG_REG_CALL_STACK, - TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - - sizeof(long), HOST_ST_OP); - tcg_out_ldst(s, TCG_AREG0, TCG_REG_CALL_STACK, - TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - - sizeof(long), HOST_LD_OP); - - /* will become: - ba label2 */ - label2_ptr = (uint32_t *)s->code_ptr; - tcg_out32(s, 0); - - /* nop (delay slot) */ - tcg_out_nop(s); + /* delay slot */ + tcg_out_movi(s, TCG_TYPE_REG, tcg_target_call_iarg_regs[n], memi); -#if TARGET_LONG_BITS == 32 - /* be label1 */ - *label1_ptr = (INSN_OP(0) | INSN_COND(COND_E, 0) | INSN_OP2(0x2) | - INSN_OFF22((unsigned long)s->code_ptr - - (unsigned long)label1_ptr)); -#else - /* be,pt %xcc label1 */ - *label1_ptr = (INSN_OP(0) | INSN_COND(COND_E, 0) | INSN_OP2(0x1) | - (0x5 << 19) | INSN_OFF19((unsigned long)s->code_ptr - - (unsigned long)label1_ptr)); -#endif - - /* ld [arg1 + x], arg1 */ - tcg_out_ldst(s, arg1, arg1, offsetof(CPUTLBEntry, addend) - - offsetof(CPUTLBEntry, addr_write), TARGET_ADDEND_LD_OP); + /* Reload AREG0. */ + tcg_out_ld(s, TCG_TYPE_REG, TCG_AREG0, TCG_REG_CALL_STACK, + TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - + sizeof(long)); -#if TARGET_LONG_BITS == 32 - /* and addr_reg, x, arg0 */ - tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_I5, 0xffffffff); - tcg_out_arith(s, arg0, addr_reg, TCG_REG_I5, ARITH_AND); - /* add arg0, arg1, arg0 */ - tcg_out_arith(s, arg0, arg0, arg1, ARITH_ADD); + *label_ptr |= INSN_OFF19((unsigned long)s->code_ptr - + (unsigned long)label_ptr); #else - /* add addr_reg, arg1, arg0 */ - tcg_out_arith(s, arg0, addr_reg, arg1, ARITH_ADD); -#endif - -#else - arg0 = addr_reg; -#endif - - switch(opc) { - case 0: - /* stb data_reg, [arg0] */ - tcg_out_ldst(s, data_reg, arg0, 0, STB); - break; - case 1: -#ifdef TARGET_WORDS_BIGENDIAN - /* sth data_reg, [arg0] */ - tcg_out_ldst(s, data_reg, arg0, 0, STH); -#else - /* stha data_reg, [arg0] ASI_PRIMARY_LITTLE */ - tcg_out_ldst_asi(s, data_reg, arg0, 0, STHA, ASI_PRIMARY_LITTLE); -#endif - break; - case 2: -#ifdef TARGET_WORDS_BIGENDIAN - /* stw data_reg, [arg0] */ - tcg_out_ldst(s, data_reg, arg0, 0, STW); -#else - /* stwa data_reg, [arg0] ASI_PRIMARY_LITTLE */ - tcg_out_ldst_asi(s, data_reg, arg0, 0, STWA, ASI_PRIMARY_LITTLE); -#endif - break; - case 3: -#ifdef TARGET_WORDS_BIGENDIAN - /* stx data_reg, [arg0] */ - tcg_out_ldst(s, data_reg, arg0, 0, STX); -#else - /* stxa data_reg, [arg0] ASI_PRIMARY_LITTLE */ - tcg_out_ldst_asi(s, data_reg, arg0, 0, STXA, ASI_PRIMARY_LITTLE); -#endif - break; - default: - tcg_abort(); + addr_reg = args[addrlo_idx]; + if (TCG_TARGET_REG_BITS == 64 && TARGET_LONG_BITS == 32) { + tcg_out_arithi(s, TCG_REG_I5, addr_reg, 0, SHIFT_SRL); + addr_reg = TCG_REG_I5; } - -#if defined(CONFIG_SOFTMMU) - /* label2: */ - *label2_ptr = (INSN_OP(0) | INSN_COND(COND_A, 0) | INSN_OP2(0x2) | - INSN_OFF22((unsigned long)s->code_ptr - - (unsigned long)label2_ptr)); -#endif + if (TCG_TARGET_REG_BITS == 32 && sizeop == 3) { + /* Reconstruct the full 64-bit value in %g1, using %o2 as temp. */ + /* ??? Redefine the temps from %i4/%i5 so that we have a o/g temp. */ + tcg_out_arithi(s, TCG_REG_G1, datalo, 0, SHIFT_SRL); + tcg_out_arithi(s, TCG_REG_O2, datahi, 32, SHIFT_SLLX); + tcg_out_arith(s, TCG_REG_G1, TCG_REG_G1, TCG_REG_O2, ARITH_OR); + datalo = TCG_REG_G1; + } + tcg_out_ldst_rr(s, datalo, addr_reg, TCG_REG_G0, qemu_st_opc[sizeop]); +#endif /* CONFIG_SOFTMMU */ } static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, @@ -1175,12 +1073,12 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, /* Store AREG0 in stack to avoid ugly glibc bugs that mangle global registers */ // delay slot - tcg_out_ldst(s, TCG_AREG0, TCG_REG_CALL_STACK, - TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - - sizeof(long), HOST_ST_OP); - tcg_out_ldst(s, TCG_AREG0, TCG_REG_CALL_STACK, - TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - - sizeof(long), HOST_LD_OP); + tcg_out_st(s, TCG_TYPE_REG, TCG_AREG0, TCG_REG_CALL_STACK, + TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - + sizeof(long)); + tcg_out_ld(s, TCG_TYPE_REG, TCG_AREG0, TCG_REG_CALL_STACK, + TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - + sizeof(long)); break; case INDEX_op_jmp: case INDEX_op_br: @@ -1348,6 +1246,9 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, tcg_out_qemu_ld(s, args, 2 | 4); break; #endif + case INDEX_op_qemu_ld64: + tcg_out_qemu_ld(s, args, 3); + break; case INDEX_op_qemu_st8: tcg_out_qemu_st(s, args, 0); break; @@ -1357,6 +1258,9 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, case INDEX_op_qemu_st32: tcg_out_qemu_st(s, args, 2); break; + case INDEX_op_qemu_st64: + tcg_out_qemu_st(s, args, 3); + break; #if TCG_TARGET_REG_BITS == 64 case INDEX_op_movi_i64: @@ -1421,13 +1325,6 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, args[2], const_args[2]); break; - case INDEX_op_qemu_ld64: - tcg_out_qemu_ld(s, args, 3); - break; - case INDEX_op_qemu_st64: - tcg_out_qemu_st(s, args, 3); - break; - #endif gen_arith: tcg_out_arithc(s, args[0], args[1], args[2], const_args[2], c); @@ -1492,20 +1389,6 @@ static const TCGTargetOpDef sparc_op_defs[] = { { INDEX_op_mulu2_i32, { "r", "r", "r", "rJ" } }, #endif - { INDEX_op_qemu_ld8u, { "r", "L" } }, - { INDEX_op_qemu_ld8s, { "r", "L" } }, - { INDEX_op_qemu_ld16u, { "r", "L" } }, - { INDEX_op_qemu_ld16s, { "r", "L" } }, - { INDEX_op_qemu_ld32, { "r", "L" } }, -#if TCG_TARGET_REG_BITS == 64 - { INDEX_op_qemu_ld32u, { "r", "L" } }, - { INDEX_op_qemu_ld32s, { "r", "L" } }, -#endif - - { INDEX_op_qemu_st8, { "L", "L" } }, - { INDEX_op_qemu_st16, { "L", "L" } }, - { INDEX_op_qemu_st32, { "L", "L" } }, - #if TCG_TARGET_REG_BITS == 64 { INDEX_op_mov_i64, { "r", "r" } }, { INDEX_op_movi_i64, { "r" } }, @@ -1520,8 +1403,6 @@ static const TCGTargetOpDef sparc_op_defs[] = { { INDEX_op_st16_i64, { "r", "r" } }, { INDEX_op_st32_i64, { "r", "r" } }, { INDEX_op_st_i64, { "r", "r" } }, - { INDEX_op_qemu_ld64, { "L", "L" } }, - { INDEX_op_qemu_st64, { "L", "L" } }, { INDEX_op_add_i64, { "r", "r", "rJ" } }, { INDEX_op_mul_i64, { "r", "r", "rJ" } }, @@ -1548,10 +1429,48 @@ static const TCGTargetOpDef sparc_op_defs[] = { { INDEX_op_brcond_i64, { "r", "rJ" } }, { INDEX_op_setcond_i64, { "r", "r", "rJ" } }, -#else - { INDEX_op_qemu_ld64, { "L", "L", "L" } }, +#endif + +#if TCG_TARGET_REG_BITS == 64 + { INDEX_op_qemu_ld8u, { "r", "L" } }, + { INDEX_op_qemu_ld8s, { "r", "L" } }, + { INDEX_op_qemu_ld16u, { "r", "L" } }, + { INDEX_op_qemu_ld16s, { "r", "L" } }, + { INDEX_op_qemu_ld32, { "r", "L" } }, + { INDEX_op_qemu_ld32u, { "r", "L" } }, + { INDEX_op_qemu_ld32s, { "r", "L" } }, + { INDEX_op_qemu_ld64, { "r", "L" } }, + + { INDEX_op_qemu_st8, { "L", "L" } }, + { INDEX_op_qemu_st16, { "L", "L" } }, + { INDEX_op_qemu_st32, { "L", "L" } }, + { INDEX_op_qemu_st64, { "L", "L" } }, +#elif TARGET_LONG_BITS <= TCG_TARGET_REG_BITS + { INDEX_op_qemu_ld8u, { "r", "L" } }, + { INDEX_op_qemu_ld8s, { "r", "L" } }, + { INDEX_op_qemu_ld16u, { "r", "L" } }, + { INDEX_op_qemu_ld16s, { "r", "L" } }, + { INDEX_op_qemu_ld32, { "r", "L" } }, + { INDEX_op_qemu_ld64, { "r", "r", "L" } }, + + { INDEX_op_qemu_st8, { "L", "L" } }, + { INDEX_op_qemu_st16, { "L", "L" } }, + { INDEX_op_qemu_st32, { "L", "L" } }, { INDEX_op_qemu_st64, { "L", "L", "L" } }, +#else + { INDEX_op_qemu_ld8u, { "r", "L", "L" } }, + { INDEX_op_qemu_ld8s, { "r", "L", "L" } }, + { INDEX_op_qemu_ld16u, { "r", "L", "L" } }, + { INDEX_op_qemu_ld16s, { "r", "L", "L" } }, + { INDEX_op_qemu_ld32, { "r", "L", "L" } }, + { INDEX_op_qemu_ld64, { "L", "L", "L", "L" } }, + + { INDEX_op_qemu_st8, { "L", "L", "L" } }, + { INDEX_op_qemu_st16, { "L", "L", "L" } }, + { INDEX_op_qemu_st32, { "L", "L", "L" } }, + { INDEX_op_qemu_st64, { "L", "L", "L", "L" } }, #endif + { -1 }, }; -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 07/14] tcg-sparc: Support GUEST_BASE. 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson ` (5 preceding siblings ...) 2012-09-22 2:04 ` [Qemu-devel] [PATCH 06/14] tcg-sparc: Fix qemu_ld/st to handle 32-bit host Richard Henderson @ 2012-09-22 2:05 ` Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 08/14] tcg-sparc: Change AREG0 in generated code to %i0 Richard Henderson ` (7 subsequent siblings) 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:05 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl Signed-off-by: Richard Henderson <rth@twiddle.net> --- configure | 2 ++ tcg/sparc/tcg-target.c | 26 +++++++++++++++++++++++--- tcg/sparc/tcg-target.h | 2 ++ 3 files changed, 27 insertions(+), 3 deletions(-) diff --git a/configure b/configure index df888f2..0dc4170 100755 --- a/configure +++ b/configure @@ -870,6 +870,7 @@ case "$cpu" in if test "$solaris" = "no" ; then QEMU_CFLAGS="-ffixed-g1 -ffixed-g6 $QEMU_CFLAGS" fi + host_guest_base="yes" ;; sparc64) LDFLAGS="-m64 $LDFLAGS" @@ -878,6 +879,7 @@ case "$cpu" in if test "$solaris" != "no" ; then QEMU_CFLAGS="-ffixed-g1 $QEMU_CFLAGS" fi + host_guest_base="yes" ;; s390) QEMU_CFLAGS="-m31 -march=z990 $QEMU_CFLAGS" diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index d89c19b..5acfeba 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -59,6 +59,12 @@ static const char * const tcg_target_reg_names[TCG_TARGET_NB_REGS] = { }; #endif +#ifdef CONFIG_USE_GUEST_BASE +# define TCG_GUEST_BASE_REG TCG_REG_I3 +#else +# define TCG_GUEST_BASE_REG TCG_REG_G0 +#endif + static const int tcg_target_reg_alloc_order[] = { TCG_REG_L0, TCG_REG_L1, @@ -680,6 +686,14 @@ static void tcg_target_qemu_prologue(TCGContext *s) tcg_out32(s, SAVE | INSN_RD(TCG_REG_O6) | INSN_RS1(TCG_REG_O6) | INSN_IMM13(-(TCG_TARGET_STACK_MINFRAME + CPU_TEMP_BUF_NLONGS * (int)sizeof(long)))); + +#ifdef CONFIG_USE_GUEST_BASE + if (GUEST_BASE != 0) { + tcg_out_movi(s, TCG_TYPE_PTR, TCG_GUEST_BASE_REG, GUEST_BASE); + tcg_regset_set_reg(s->reserved_regs, TCG_GUEST_BASE_REG); + } +#endif + tcg_out32(s, JMPL | INSN_RD(TCG_REG_G0) | INSN_RS1(TCG_REG_I1) | INSN_RS2(TCG_REG_G0)); tcg_out_mov(s, TCG_TYPE_PTR, TCG_AREG0, TCG_REG_I0); @@ -925,14 +939,18 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int sizeop) if (TCG_TARGET_REG_BITS == 32 && sizeop == 3) { int reg64 = (datalo < 16 ? datalo : TCG_REG_O0); - tcg_out_ldst_rr(s, reg64, addr_reg, TCG_REG_G0, qemu_ld_opc[sizeop]); + tcg_out_ldst_rr(s, reg64, addr_reg, + (GUEST_BASE ? TCG_GUEST_BASE_REG : TCG_REG_G0), + qemu_ld_opc[sizeop]); tcg_out_arithi(s, datahi, reg64, 32, SHIFT_SRLX); if (reg64 != datalo) { tcg_out_mov(s, TCG_TYPE_I32, datalo, reg64); } } else { - tcg_out_ldst_rr(s, datalo, addr_reg, TCG_REG_G0, qemu_ld_opc[sizeop]); + tcg_out_ldst_rr(s, datalo, addr_reg, + (GUEST_BASE ? TCG_GUEST_BASE_REG : TCG_REG_G0), + qemu_ld_opc[sizeop]); } #endif /* CONFIG_SOFTMMU */ } @@ -1026,7 +1044,9 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int sizeop) tcg_out_arith(s, TCG_REG_G1, TCG_REG_G1, TCG_REG_O2, ARITH_OR); datalo = TCG_REG_G1; } - tcg_out_ldst_rr(s, datalo, addr_reg, TCG_REG_G0, qemu_st_opc[sizeop]); + tcg_out_ldst_rr(s, datalo, addr_reg, + (GUEST_BASE ? TCG_GUEST_BASE_REG : TCG_REG_G0), + qemu_st_opc[sizeop]); #endif /* CONFIG_SOFTMMU */ } diff --git a/tcg/sparc/tcg-target.h b/tcg/sparc/tcg-target.h index adca1d2..99e9f57 100644 --- a/tcg/sparc/tcg-target.h +++ b/tcg/sparc/tcg-target.h @@ -128,6 +128,8 @@ typedef enum { #define TCG_TARGET_HAS_movcond_i64 0 #endif +#define TCG_TARGET_HAS_GUEST_BASE + #ifdef CONFIG_SOLARIS #define TCG_AREG0 TCG_REG_G2 #elif HOST_LONG_BITS == 64 -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 08/14] tcg-sparc: Change AREG0 in generated code to %i0. 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson ` (6 preceding siblings ...) 2012-09-22 2:05 ` [Qemu-devel] [PATCH 07/14] tcg-sparc: Support GUEST_BASE Richard Henderson @ 2012-09-22 2:05 ` Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 09/14] tcg-sparc: Clean up cruft stemming from attempts to use global registers Richard Henderson ` (6 subsequent siblings) 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:05 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl We can now move the TCG variable from %g[56] to a call-preserved windowed register. Signed-off-by: Richard Henderson <rth@twiddle.net> --- tcg/sparc/tcg-target.c | 3 ++- tcg/sparc/tcg-target.h | 8 +------- 2 files changed, 3 insertions(+), 8 deletions(-) diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index 5acfeba..9ab5746 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -696,7 +696,8 @@ static void tcg_target_qemu_prologue(TCGContext *s) tcg_out32(s, JMPL | INSN_RD(TCG_REG_G0) | INSN_RS1(TCG_REG_I1) | INSN_RS2(TCG_REG_G0)); - tcg_out_mov(s, TCG_TYPE_PTR, TCG_AREG0, TCG_REG_I0); + /* delay slot */ + tcg_out_nop(s); } #if defined(CONFIG_SOFTMMU) diff --git a/tcg/sparc/tcg-target.h b/tcg/sparc/tcg-target.h index 99e9f57..ee154d0 100644 --- a/tcg/sparc/tcg-target.h +++ b/tcg/sparc/tcg-target.h @@ -130,13 +130,7 @@ typedef enum { #define TCG_TARGET_HAS_GUEST_BASE -#ifdef CONFIG_SOLARIS -#define TCG_AREG0 TCG_REG_G2 -#elif HOST_LONG_BITS == 64 -#define TCG_AREG0 TCG_REG_G5 -#else -#define TCG_AREG0 TCG_REG_G6 -#endif +#define TCG_AREG0 TCG_REG_I0 static inline void flush_icache_range(tcg_target_ulong start, tcg_target_ulong stop) -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 09/14] tcg-sparc: Clean up cruft stemming from attempts to use global registers. 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson ` (7 preceding siblings ...) 2012-09-22 2:05 ` [Qemu-devel] [PATCH 08/14] tcg-sparc: Change AREG0 in generated code to %i0 Richard Henderson @ 2012-09-22 2:05 ` Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 10/14] tcg-sparc: Mask shift immediates to avoid illegal insns Richard Henderson ` (5 subsequent siblings) 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:05 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl Don't use -ffixed-gN. Don't link statically. Don't save/restore AREG0 around calls. Don't allocate space on the stack for AREG0 save. Signed-off-by: Richard Henderson <rth@twiddle.net> --- configure | 12 ----------- tcg/sparc/tcg-target.c | 55 +++++++++++++++++--------------------------------- tcg/sparc/tcg-target.h | 18 +++++++---------- 3 files changed, 26 insertions(+), 59 deletions(-) diff --git a/configure b/configure index 0dc4170..1b86517 100755 --- a/configure +++ b/configure @@ -866,19 +866,11 @@ case "$cpu" in sparc) LDFLAGS="-m32 $LDFLAGS" QEMU_CFLAGS="-m32 -mcpu=ultrasparc $QEMU_CFLAGS" - QEMU_CFLAGS="-ffixed-g2 -ffixed-g3 $QEMU_CFLAGS" - if test "$solaris" = "no" ; then - QEMU_CFLAGS="-ffixed-g1 -ffixed-g6 $QEMU_CFLAGS" - fi host_guest_base="yes" ;; sparc64) LDFLAGS="-m64 $LDFLAGS" QEMU_CFLAGS="-m64 -mcpu=ultrasparc $QEMU_CFLAGS" - QEMU_CFLAGS="-ffixed-g5 -ffixed-g6 -ffixed-g7 $QEMU_CFLAGS" - if test "$solaris" != "no" ; then - QEMU_CFLAGS="-ffixed-g1 $QEMU_CFLAGS" - fi host_guest_base="yes" ;; s390) @@ -4101,10 +4093,6 @@ fi if test "$target_linux_user" = "yes" -o "$target_bsd_user" = "yes" ; then case "$ARCH" in - sparc) - # -static is used to avoid g1/g3 usage by the dynamic linker - ldflags="$linker_script -static $ldflags" - ;; alpha | s390x) # The default placement of the application is fine. ;; diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index 9ab5746..e625aa3 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -161,7 +161,6 @@ static int target_parse_constraint(TCGArgConstraint *ct, const char **pct_str) tcg_regset_reset_reg(ct->u.regs, TCG_REG_O0); tcg_regset_reset_reg(ct->u.regs, TCG_REG_O1); tcg_regset_reset_reg(ct->u.regs, TCG_REG_O2); - tcg_regset_reset_reg(ct->u.regs, TCG_REG_O3); break; case 'I': ct->ct |= TCG_CT_CONST_S11; @@ -681,11 +680,22 @@ static void tcg_out_setcond2_i32(TCGContext *s, TCGCond cond, TCGArg ret, /* Generate global QEMU prologue and epilogue code */ static void tcg_target_qemu_prologue(TCGContext *s) { - tcg_set_frame(s, TCG_REG_I6, TCG_TARGET_CALL_STACK_OFFSET, - CPU_TEMP_BUF_NLONGS * (int)sizeof(long)); + int tmp_buf_size, frame_size; + + /* The TCG temp buffer is at the top of the frame, immediately + below the frame pointer. */ + tmp_buf_size = CPU_TEMP_BUF_NLONGS * (int)sizeof(long); + tcg_set_frame(s, TCG_REG_I6, TCG_TARGET_STACK_BIAS - tmp_buf_size, + tmp_buf_size); + + /* TCG_TARGET_CALL_STACK_OFFSET includes the stack bias, but is + otherwise the minimal frame usable by callees. */ + frame_size = TCG_TARGET_CALL_STACK_OFFSET - TCG_TARGET_STACK_BIAS; + frame_size += TCG_STATIC_CALL_ARGS_SIZE + tmp_buf_size; + frame_size += TCG_TARGET_STACK_ALIGN - 1; + frame_size &= -TCG_TARGET_STACK_ALIGN; tcg_out32(s, SAVE | INSN_RD(TCG_REG_O6) | INSN_RS1(TCG_REG_O6) | - INSN_IMM13(-(TCG_TARGET_STACK_MINFRAME + - CPU_TEMP_BUF_NLONGS * (int)sizeof(long)))); + INSN_IMM13(-frame_size)); #ifdef CONFIG_USE_GUEST_BASE if (GUEST_BASE != 0) { @@ -698,6 +708,8 @@ static void tcg_target_qemu_prologue(TCGContext *s) INSN_RS2(TCG_REG_G0)); /* delay slot */ tcg_out_nop(s); + + /* No epilogue required. We issue ret + restore directly in the TB. */ } #if defined(CONFIG_SOFTMMU) @@ -880,12 +892,6 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int sizeop) tcg_out_mov(s, TCG_TYPE_REG, tcg_target_call_iarg_regs[n++], args[addrlo_idx]); - /* Store AREG0 in stack to avoid ugly glibc bugs that mangle - global registers */ - tcg_out_st(s, TCG_TYPE_REG, TCG_AREG0, TCG_REG_CALL_STACK, - TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - - sizeof(long)); - /* qemu_ld_helper[s_bits](arg0, arg1) */ tcg_out32(s, CALL | ((((tcg_target_ulong)qemu_ld_helpers[s_bits] - (tcg_target_ulong)s->code_ptr) >> 2) @@ -893,11 +899,6 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int sizeop) /* delay slot */ tcg_out_movi(s, TCG_TYPE_I32, tcg_target_call_iarg_regs[n], memi); - /* Reload AREG0. */ - tcg_out_ld(s, TCG_TYPE_REG, TCG_AREG0, TCG_REG_CALL_STACK, - TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - - sizeof(long)); - n = tcg_target_call_oarg_regs[0]; /* datalo = sign_extend(arg0) */ switch (sizeop) { @@ -1011,12 +1012,6 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int sizeop) } tcg_out_mov(s, TCG_TYPE_REG, tcg_target_call_iarg_regs[n++], datalo); - /* Store AREG0 in stack to avoid ugly glibc bugs that mangle - global registers */ - tcg_out_st(s, TCG_TYPE_REG, TCG_AREG0, TCG_REG_CALL_STACK, - TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - - sizeof(long)); - /* qemu_st_helper[s_bits](arg0, arg1, arg2) */ tcg_out32(s, CALL | ((((tcg_target_ulong)qemu_st_helpers[sizeop] - (tcg_target_ulong)s->code_ptr) >> 2) @@ -1024,11 +1019,6 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int sizeop) /* delay slot */ tcg_out_movi(s, TCG_TYPE_REG, tcg_target_call_iarg_regs[n], memi); - /* Reload AREG0. */ - tcg_out_ld(s, TCG_TYPE_REG, TCG_AREG0, TCG_REG_CALL_STACK, - TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - - sizeof(long)); - *label_ptr |= INSN_OFF19((unsigned long)s->code_ptr - (unsigned long)label_ptr); #else @@ -1091,15 +1081,8 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, tcg_out32(s, JMPL | INSN_RD(TCG_REG_O7) | INSN_RS1(TCG_REG_I5) | INSN_RS2(TCG_REG_G0)); } - /* Store AREG0 in stack to avoid ugly glibc bugs that mangle - global registers */ - // delay slot - tcg_out_st(s, TCG_TYPE_REG, TCG_AREG0, TCG_REG_CALL_STACK, - TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - - sizeof(long)); - tcg_out_ld(s, TCG_TYPE_REG, TCG_AREG0, TCG_REG_CALL_STACK, - TCG_TARGET_CALL_STACK_OFFSET - TCG_STATIC_CALL_ARGS_SIZE - - sizeof(long)); + /* delay slot */ + tcg_out_nop(s); break; case INDEX_op_jmp: case INDEX_op_br: diff --git a/tcg/sparc/tcg-target.h b/tcg/sparc/tcg-target.h index ee154d0..6314ffb 100644 --- a/tcg/sparc/tcg-target.h +++ b/tcg/sparc/tcg-target.h @@ -66,20 +66,16 @@ typedef enum { #define TCG_CT_CONST_S13 0x200 /* used for function call generation */ -#define TCG_REG_CALL_STACK TCG_REG_I6 +#define TCG_REG_CALL_STACK TCG_REG_O6 #if TCG_TARGET_REG_BITS == 64 -// Reserve space for AREG0 -#define TCG_TARGET_STACK_MINFRAME (176 + 4 * (int)sizeof(long) + \ - TCG_STATIC_CALL_ARGS_SIZE) -#define TCG_TARGET_CALL_STACK_OFFSET (2047 - 16) -#define TCG_TARGET_STACK_ALIGN 16 +#define TCG_TARGET_STACK_BIAS 2047 +#define TCG_TARGET_STACK_ALIGN 16 +#define TCG_TARGET_CALL_STACK_OFFSET (128 + 6*8 + TCG_TARGET_STACK_BIAS) #else -// AREG0 + one word for alignment -#define TCG_TARGET_STACK_MINFRAME (92 + (2 + 1) * (int)sizeof(long) + \ - TCG_STATIC_CALL_ARGS_SIZE) -#define TCG_TARGET_CALL_STACK_OFFSET TCG_TARGET_STACK_MINFRAME -#define TCG_TARGET_STACK_ALIGN 8 +#define TCG_TARGET_STACK_BIAS 0 +#define TCG_TARGET_STACK_ALIGN 8 +#define TCG_TARGET_CALL_STACK_OFFSET (64 + 4 + 6*4) #endif #if TCG_TARGET_REG_BITS == 64 -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 10/14] tcg-sparc: Mask shift immediates to avoid illegal insns. 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson ` (8 preceding siblings ...) 2012-09-22 2:05 ` [Qemu-devel] [PATCH 09/14] tcg-sparc: Clean up cruft stemming from attempts to use global registers Richard Henderson @ 2012-09-22 2:05 ` Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 11/14] tcg-sparc: Use defines for temporaries Richard Henderson ` (4 subsequent siblings) 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:05 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl The xtensa-test image generates a sra_i32 with count 0x40. Whether this is accident of tcg constant propagation or originating directly from the instruction stream is immaterial. Signed-off-by: Richard Henderson <rth@twiddle.net> --- tcg/sparc/tcg-target.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index e625aa3..be5c170 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -1154,13 +1154,16 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, goto gen_arith; case INDEX_op_shl_i32: c = SHIFT_SLL; - goto gen_arith; + do_shift32: + /* Limit immediate shift count lest we create an illegal insn. */ + tcg_out_arithc(s, args[0], args[1], args[2] & 31, const_args[2], c); + break; case INDEX_op_shr_i32: c = SHIFT_SRL; - goto gen_arith; + goto do_shift32; case INDEX_op_sar_i32: c = SHIFT_SRA; - goto gen_arith; + goto do_shift32; case INDEX_op_mul_i32: c = ARITH_UMUL; goto gen_arith; @@ -1281,13 +1284,16 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, break; case INDEX_op_shl_i64: c = SHIFT_SLLX; - goto gen_arith; + do_shift64: + /* Limit immediate shift count lest we create an illegal insn. */ + tcg_out_arithc(s, args[0], args[1], args[2] & 63, const_args[2], c); + break; case INDEX_op_shr_i64: c = SHIFT_SRLX; - goto gen_arith; + goto do_shift64; case INDEX_op_sar_i64: c = SHIFT_SRAX; - goto gen_arith; + goto do_shift64; case INDEX_op_mul_i64: c = ARITH_MULX; goto gen_arith; -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 11/14] tcg-sparc: Use defines for temporaries. 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson ` (9 preceding siblings ...) 2012-09-22 2:05 ` [Qemu-devel] [PATCH 10/14] tcg-sparc: Mask shift immediates to avoid illegal insns Richard Henderson @ 2012-09-22 2:05 ` Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 12/14] tcg-sparc: Add %g/%o registers to alloc_order Richard Henderson ` (3 subsequent siblings) 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:05 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl And change from %i4/%i5 to %g1/%o7 to remove a v8plus fixme. Signed-off-by: Richard Henderson <rth@twiddle.net> --- tcg/sparc/tcg-target.c | 115 +++++++++++++++++++++++++------------------------ 1 file changed, 59 insertions(+), 56 deletions(-) diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index be5c170..d401f8e 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -59,8 +59,12 @@ static const char * const tcg_target_reg_names[TCG_TARGET_NB_REGS] = { }; #endif +/* Define some temporary registers. T2 is used for constant generation. */ +#define TCG_REG_T1 TCG_REG_G1 +#define TCG_REG_T2 TCG_REG_O7 + #ifdef CONFIG_USE_GUEST_BASE -# define TCG_GUEST_BASE_REG TCG_REG_I3 +# define TCG_GUEST_BASE_REG TCG_REG_I5 #else # define TCG_GUEST_BASE_REG TCG_REG_G0 #endif @@ -79,6 +83,7 @@ static const int tcg_target_reg_alloc_order[] = { TCG_REG_I2, TCG_REG_I3, TCG_REG_I4, + TCG_REG_I5, }; static const int tcg_target_call_iarg_regs[6] = { @@ -366,10 +371,10 @@ static inline void tcg_out_movi(TCGContext *s, TCGType type, tcg_out_sethi(s, ret, ~arg); tcg_out_arithi(s, ret, ret, (arg & 0x3ff) | -0x400, ARITH_XOR); } else { - tcg_out_movi_imm32(s, TCG_REG_I4, arg >> (TCG_TARGET_REG_BITS / 2)); - tcg_out_arithi(s, TCG_REG_I4, TCG_REG_I4, 32, SHIFT_SLLX); - tcg_out_movi_imm32(s, ret, arg); - tcg_out_arith(s, ret, ret, TCG_REG_I4, ARITH_OR); + tcg_out_movi_imm32(s, ret, arg >> (TCG_TARGET_REG_BITS / 2)); + tcg_out_arithi(s, ret, ret, 32, SHIFT_SLLX); + tcg_out_movi_imm32(s, TCG_REG_T2, arg); + tcg_out_arith(s, ret, ret, TCG_REG_T2, ARITH_OR); } } @@ -386,8 +391,8 @@ static inline void tcg_out_ldst(TCGContext *s, int ret, int addr, tcg_out32(s, op | INSN_RD(ret) | INSN_RS1(addr) | INSN_IMM13(offset)); } else { - tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_I5, offset); - tcg_out_ldst_rr(s, ret, addr, TCG_REG_I5, op); + tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_T1, offset); + tcg_out_ldst_rr(s, ret, addr, TCG_REG_T1, op); } } @@ -428,8 +433,8 @@ static inline void tcg_out_addi(TCGContext *s, int reg, tcg_target_long val) if (check_fit_tl(val, 13)) tcg_out_arithi(s, reg, reg, val, ARITH_ADD); else { - tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_I5, val); - tcg_out_arith(s, reg, reg, TCG_REG_I5, ARITH_ADD); + tcg_out_movi(s, TCG_TYPE_PTR, TCG_REG_T1, val); + tcg_out_arith(s, reg, reg, TCG_REG_T1, ARITH_ADD); } } } @@ -441,8 +446,8 @@ static inline void tcg_out_andi(TCGContext *s, int rd, int rs, if (check_fit_tl(val, 13)) tcg_out_arithi(s, rd, rs, val, ARITH_AND); else { - tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_I5, val); - tcg_out_arith(s, rd, rs, TCG_REG_I5, ARITH_AND); + tcg_out_movi(s, TCG_TYPE_I32, TCG_REG_T1, val); + tcg_out_arith(s, rd, rs, TCG_REG_T1, ARITH_AND); } } } @@ -454,8 +459,8 @@ static void tcg_out_div32(TCGContext *s, int rd, int rs1, if (uns) { tcg_out_sety(s, TCG_REG_G0); } else { - tcg_out_arithi(s, TCG_REG_I5, rs1, 31, SHIFT_SRA); - tcg_out_sety(s, TCG_REG_I5); + tcg_out_arithi(s, TCG_REG_T1, rs1, 31, SHIFT_SRA); + tcg_out_sety(s, TCG_REG_T1); } tcg_out_arithc(s, rd, rs1, val2, val2const, @@ -601,8 +606,8 @@ static void tcg_out_setcond_i32(TCGContext *s, TCGCond cond, TCGArg ret, case TCG_COND_GTU: case TCG_COND_GEU: if (c2const && c2 != 0) { - tcg_out_movi_imm13(s, TCG_REG_I5, c2); - c2 = TCG_REG_I5; + tcg_out_movi_imm13(s, TCG_REG_T1, c2); + c2 = TCG_REG_T1; } t = c1, c1 = c2, c2 = t, c2const = 0; cond = tcg_swap_cond(cond); @@ -649,15 +654,15 @@ static void tcg_out_setcond2_i32(TCGContext *s, TCGCond cond, TCGArg ret, switch (cond) { case TCG_COND_EQ: - tcg_out_setcond_i32(s, TCG_COND_EQ, TCG_REG_I5, al, bl, blconst); + tcg_out_setcond_i32(s, TCG_COND_EQ, TCG_REG_T1, al, bl, blconst); tcg_out_setcond_i32(s, TCG_COND_EQ, ret, ah, bh, bhconst); - tcg_out_arith(s, ret, ret, TCG_REG_I5, ARITH_AND); + tcg_out_arith(s, ret, ret, TCG_REG_T1, ARITH_AND); break; case TCG_COND_NE: - tcg_out_setcond_i32(s, TCG_COND_NE, TCG_REG_I5, al, al, blconst); + tcg_out_setcond_i32(s, TCG_COND_NE, TCG_REG_T1, al, al, blconst); tcg_out_setcond_i32(s, TCG_COND_NE, ret, ah, bh, bhconst); - tcg_out_arith(s, ret, ret, TCG_REG_I5, ARITH_OR); + tcg_out_arith(s, ret, ret, TCG_REG_T1, ARITH_OR); break; default: @@ -935,8 +940,8 @@ static void tcg_out_qemu_ld(TCGContext *s, const TCGArg *args, int sizeop) #else addr_reg = args[addrlo_idx]; if (TCG_TARGET_REG_BITS == 64 && TARGET_LONG_BITS == 32) { - tcg_out_arithi(s, TCG_REG_I5, addr_reg, 0, SHIFT_SRL); - addr_reg = TCG_REG_I5; + tcg_out_arithi(s, TCG_REG_T1, addr_reg, 0, SHIFT_SRL); + addr_reg = TCG_REG_T1; } if (TCG_TARGET_REG_BITS == 32 && sizeop == 3) { int reg64 = (datalo < 16 ? datalo : TCG_REG_O0); @@ -979,12 +984,11 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int sizeop) offsetof(CPUTLBEntry, addr_write)); if (TCG_TARGET_REG_BITS == 32 && sizeop == 3) { - /* Reconstruct the full 64-bit value in %g1, using %o2 as temp. */ - /* ??? Redefine the temps from %i4/%i5 so that we have a o/g temp. */ - tcg_out_arithi(s, TCG_REG_G1, datalo, 0, SHIFT_SRL); + /* Reconstruct the full 64-bit value. */ + tcg_out_arithi(s, TCG_REG_T1, datalo, 0, SHIFT_SRL); tcg_out_arithi(s, TCG_REG_O2, datahi, 32, SHIFT_SLLX); - tcg_out_arith(s, TCG_REG_G1, TCG_REG_G1, TCG_REG_O2, ARITH_OR); - datalo = TCG_REG_G1; + tcg_out_arith(s, TCG_REG_O2, TCG_REG_T1, TCG_REG_O2, ARITH_OR); + datalo = TCG_REG_O2; } /* The fast path is exactly one insn. Thus we can perform the entire @@ -1024,16 +1028,14 @@ static void tcg_out_qemu_st(TCGContext *s, const TCGArg *args, int sizeop) #else addr_reg = args[addrlo_idx]; if (TCG_TARGET_REG_BITS == 64 && TARGET_LONG_BITS == 32) { - tcg_out_arithi(s, TCG_REG_I5, addr_reg, 0, SHIFT_SRL); - addr_reg = TCG_REG_I5; + tcg_out_arithi(s, TCG_REG_T1, addr_reg, 0, SHIFT_SRL); + addr_reg = TCG_REG_T1; } if (TCG_TARGET_REG_BITS == 32 && sizeop == 3) { - /* Reconstruct the full 64-bit value in %g1, using %o2 as temp. */ - /* ??? Redefine the temps from %i4/%i5 so that we have a o/g temp. */ - tcg_out_arithi(s, TCG_REG_G1, datalo, 0, SHIFT_SRL); + tcg_out_arithi(s, TCG_REG_T1, datalo, 0, SHIFT_SRL); tcg_out_arithi(s, TCG_REG_O2, datahi, 32, SHIFT_SLLX); - tcg_out_arith(s, TCG_REG_G1, TCG_REG_G1, TCG_REG_O2, ARITH_OR); - datalo = TCG_REG_G1; + tcg_out_arith(s, TCG_REG_O2, TCG_REG_T1, TCG_REG_O2, ARITH_OR); + datalo = TCG_REG_O2; } tcg_out_ldst_rr(s, datalo, addr_reg, (GUEST_BASE ? TCG_GUEST_BASE_REG : TCG_REG_G0), @@ -1057,28 +1059,29 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, case INDEX_op_goto_tb: if (s->tb_jmp_offset) { /* direct jump method */ - tcg_out_sethi(s, TCG_REG_I5, args[0] & 0xffffe000); - tcg_out32(s, JMPL | INSN_RD(TCG_REG_G0) | INSN_RS1(TCG_REG_I5) | + tcg_out_sethi(s, TCG_REG_T1, args[0] & 0xffffe000); + tcg_out32(s, JMPL | INSN_RD(TCG_REG_G0) | INSN_RS1(TCG_REG_T1) | INSN_IMM13((args[0] & 0x1fff))); s->tb_jmp_offset[args[0]] = s->code_ptr - s->code_buf; } else { /* indirect jump method */ - tcg_out_ld_ptr(s, TCG_REG_I5, (tcg_target_long)(s->tb_next + args[0])); - tcg_out32(s, JMPL | INSN_RD(TCG_REG_G0) | INSN_RS1(TCG_REG_I5) | + tcg_out_ld_ptr(s, TCG_REG_T1, + (tcg_target_long)(s->tb_next + args[0])); + tcg_out32(s, JMPL | INSN_RD(TCG_REG_G0) | INSN_RS1(TCG_REG_T1) | INSN_RS2(TCG_REG_G0)); } tcg_out_nop(s); s->tb_next_offset[args[0]] = s->code_ptr - s->code_buf; break; case INDEX_op_call: - if (const_args[0]) + if (const_args[0]) { tcg_out32(s, CALL | ((((tcg_target_ulong)args[0] - (tcg_target_ulong)s->code_ptr) >> 2) & 0x3fffffff)); - else { - tcg_out_ld_ptr(s, TCG_REG_I5, + } else { + tcg_out_ld_ptr(s, TCG_REG_T1, (tcg_target_long)(s->tb_next + args[0])); - tcg_out32(s, JMPL | INSN_RD(TCG_REG_O7) | INSN_RS1(TCG_REG_I5) | + tcg_out32(s, JMPL | INSN_RD(TCG_REG_O7) | INSN_RS1(TCG_REG_T1) | INSN_RS2(TCG_REG_G0)); } /* delay slot */ @@ -1184,11 +1187,11 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, case INDEX_op_rem_i32: case INDEX_op_remu_i32: - tcg_out_div32(s, TCG_REG_I5, args[1], args[2], const_args[2], + tcg_out_div32(s, TCG_REG_T1, args[1], args[2], const_args[2], opc == INDEX_op_remu_i32); - tcg_out_arithc(s, TCG_REG_I5, TCG_REG_I5, args[2], const_args[2], + tcg_out_arithc(s, TCG_REG_T1, TCG_REG_T1, args[2], const_args[2], ARITH_UMUL); - tcg_out_arith(s, args[0], args[1], TCG_REG_I5, ARITH_SUB); + tcg_out_arith(s, args[0], args[1], TCG_REG_T1, ARITH_SUB); break; case INDEX_op_brcond_i32: @@ -1305,11 +1308,11 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, goto gen_arith; case INDEX_op_rem_i64: case INDEX_op_remu_i64: - tcg_out_arithc(s, TCG_REG_I5, args[1], args[2], const_args[2], + tcg_out_arithc(s, TCG_REG_T1, args[1], args[2], const_args[2], opc == INDEX_op_rem_i64 ? ARITH_SDIVX : ARITH_UDIVX); - tcg_out_arithc(s, TCG_REG_I5, TCG_REG_I5, args[2], const_args[2], + tcg_out_arithc(s, TCG_REG_T1, TCG_REG_T1, args[2], const_args[2], ARITH_MULX); - tcg_out_arith(s, args[0], args[1], TCG_REG_I5, ARITH_SUB); + tcg_out_arith(s, args[0], args[1], TCG_REG_T1, ARITH_SUB); break; case INDEX_op_ext32s_i64: if (const_args[1]) { @@ -1507,15 +1510,15 @@ static void tcg_target_init(TCGContext *s) (1 << TCG_REG_O7)); tcg_regset_clear(s->reserved_regs); - tcg_regset_set_reg(s->reserved_regs, TCG_REG_G0); -#if TCG_TARGET_REG_BITS == 64 - tcg_regset_set_reg(s->reserved_regs, TCG_REG_I4); // for internal use -#endif - tcg_regset_set_reg(s->reserved_regs, TCG_REG_I5); // for internal use - tcg_regset_set_reg(s->reserved_regs, TCG_REG_I6); - tcg_regset_set_reg(s->reserved_regs, TCG_REG_I7); - tcg_regset_set_reg(s->reserved_regs, TCG_REG_O6); - tcg_regset_set_reg(s->reserved_regs, TCG_REG_O7); + tcg_regset_set_reg(s->reserved_regs, TCG_REG_G0); /* zero */ + tcg_regset_set_reg(s->reserved_regs, TCG_REG_G6); /* reserved for os */ + tcg_regset_set_reg(s->reserved_regs, TCG_REG_G7); /* thread pointer */ + tcg_regset_set_reg(s->reserved_regs, TCG_REG_I6); /* frame pointer */ + tcg_regset_set_reg(s->reserved_regs, TCG_REG_I7); /* return address */ + tcg_regset_set_reg(s->reserved_regs, TCG_REG_O6); /* stack pointer */ + tcg_regset_set_reg(s->reserved_regs, TCG_REG_T1); /* for internal use */ + tcg_regset_set_reg(s->reserved_regs, TCG_REG_T2); /* for internal use */ + tcg_add_target_add_op_defs(sparc_op_defs); } -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 12/14] tcg-sparc: Add %g/%o registers to alloc_order 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson ` (10 preceding siblings ...) 2012-09-22 2:05 ` [Qemu-devel] [PATCH 11/14] tcg-sparc: Use defines for temporaries Richard Henderson @ 2012-09-22 2:05 ` Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 13/14] tcg-sparc: Fix and enable direct TB chaining Richard Henderson ` (2 subsequent siblings) 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:05 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl Signed-off-by: Richard Henderson <rth@twiddle.net> --- tcg/sparc/tcg-target.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index d401f8e..03c385a 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -78,12 +78,25 @@ static const int tcg_target_reg_alloc_order[] = { TCG_REG_L5, TCG_REG_L6, TCG_REG_L7, + TCG_REG_I0, TCG_REG_I1, TCG_REG_I2, TCG_REG_I3, TCG_REG_I4, TCG_REG_I5, + + TCG_REG_G2, + TCG_REG_G3, + TCG_REG_G4, + TCG_REG_G5, + + TCG_REG_O0, + TCG_REG_O1, + TCG_REG_O2, + TCG_REG_O3, + TCG_REG_O4, + TCG_REG_O5, }; static const int tcg_target_call_iarg_regs[6] = { -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 13/14] tcg-sparc: Fix and enable direct TB chaining. 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson ` (11 preceding siblings ...) 2012-09-22 2:05 ` [Qemu-devel] [PATCH 12/14] tcg-sparc: Add %g/%o registers to alloc_order Richard Henderson @ 2012-09-22 2:05 ` Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 14/14] tcg-sparc: Preserve branch destinations during retranslation Richard Henderson 2012-09-22 18:01 ` [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Blue Swirl 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:05 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl Signed-off-by: Richard Henderson <rth@twiddle.net> --- exec-all.h | 9 ++++++--- tcg/sparc/tcg-target.c | 21 ++++++++++++++++++--- 2 files changed, 24 insertions(+), 6 deletions(-) diff --git a/exec-all.h b/exec-all.h index dba9609..6516da0 100644 --- a/exec-all.h +++ b/exec-all.h @@ -132,9 +132,10 @@ static inline void tlb_flush(CPUArchState *env, int flush_global) #define CODE_GEN_AVG_BLOCK_SIZE 64 #endif -#if defined(_ARCH_PPC) || defined(__x86_64__) || defined(__arm__) || defined(__i386__) -#define USE_DIRECT_JUMP -#elif defined(CONFIG_TCG_INTERPRETER) +#if defined(__arm__) || defined(_ARCH_PPC) \ + || defined(__x86_64__) || defined(__i386__) \ + || defined(__sparc__) \ + || defined(CONFIG_TCG_INTERPRETER) #define USE_DIRECT_JUMP #endif @@ -244,6 +245,8 @@ static inline void tb_set_jmp_target1(uintptr_t jmp_addr, uintptr_t addr) __asm __volatile__ ("swi 0x9f0002" : : "r" (_beg), "r" (_end), "r" (_flg)); #endif } +#elif defined(__sparc__) +void tb_set_jmp_target1(uintptr_t jmp_addr, uintptr_t addr); #else #error tb_set_jmp_target1 is missing #endif diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index 03c385a..1db0c9d 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -1072,10 +1072,10 @@ static inline void tcg_out_op(TCGContext *s, TCGOpcode opc, const TCGArg *args, case INDEX_op_goto_tb: if (s->tb_jmp_offset) { /* direct jump method */ - tcg_out_sethi(s, TCG_REG_T1, args[0] & 0xffffe000); - tcg_out32(s, JMPL | INSN_RD(TCG_REG_G0) | INSN_RS1(TCG_REG_T1) | - INSN_IMM13((args[0] & 0x1fff))); + uint32_t old_insn = *(uint32_t *)s->code_ptr; s->tb_jmp_offset[args[0]] = s->code_ptr - s->code_buf; + /* Make sure to preserve links during retranslation. */ + tcg_out32(s, CALL | (old_insn & ~INSN_OP(-1))); } else { /* indirect jump method */ tcg_out_ld_ptr(s, TCG_REG_T1, @@ -1595,3 +1595,18 @@ void tcg_register_jit(void *buf, size_t buf_size) tcg_register_jit_int(buf, buf_size, &debug_frame, sizeof(debug_frame)); } + +void tb_set_jmp_target1(uintptr_t jmp_addr, uintptr_t addr) +{ + uint32_t *ptr = (uint32_t *)jmp_addr; + tcg_target_long disp = (tcg_target_long)(addr - jmp_addr) >> 2; + + /* We can reach the entire address space for 32-bit. For 64-bit + the code_gen_buffer can't be larger than 2GB. */ + if (TCG_TARGET_REG_BITS == 64 && !check_fit_tl(disp, 30)) { + tcg_abort(); + } + + *ptr = CALL | (disp & 0x3fffffff); + flush_icache_range(jmp_addr, jmp_addr + 4); +} -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* [Qemu-devel] [PATCH 14/14] tcg-sparc: Preserve branch destinations during retranslation 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson ` (12 preceding siblings ...) 2012-09-22 2:05 ` [Qemu-devel] [PATCH 13/14] tcg-sparc: Fix and enable direct TB chaining Richard Henderson @ 2012-09-22 2:05 ` Richard Henderson 2012-09-22 18:01 ` [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Blue Swirl 14 siblings, 0 replies; 16+ messages in thread From: Richard Henderson @ 2012-09-22 2:05 UTC (permalink / raw) To: qemu-devel; +Cc: Blue Swirl Signed-off-by: Richard Henderson <rth@twiddle.net> --- tcg/sparc/tcg-target.c | 19 +++++++++++-------- 1 file changed, 11 insertions(+), 8 deletions(-) diff --git a/tcg/sparc/tcg-target.c b/tcg/sparc/tcg-target.c index 1db0c9d..876da4f 100644 --- a/tcg/sparc/tcg-target.c +++ b/tcg/sparc/tcg-target.c @@ -488,30 +488,33 @@ static inline void tcg_out_nop(TCGContext *s) static void tcg_out_branch_i32(TCGContext *s, int opc, int label_index) { TCGLabel *l = &s->labels[label_index]; + uint32_t off22; if (l->has_value) { - tcg_out32(s, (INSN_OP(0) | INSN_COND(opc, 0) | INSN_OP2(0x2) - | INSN_OFF22(l->u.value - (unsigned long)s->code_ptr))); + off22 = INSN_OFF22(l->u.value - (unsigned long)s->code_ptr); } else { + /* Make sure to preserve destinations during retranslation. */ + off22 = *(uint32_t *)s->code_ptr & INSN_OFF22(-1); tcg_out_reloc(s, s->code_ptr, R_SPARC_WDISP22, label_index, 0); - tcg_out32(s, (INSN_OP(0) | INSN_COND(opc, 0) | INSN_OP2(0x2) | 0)); } + tcg_out32(s, INSN_OP(0) | INSN_COND(opc, 0) | INSN_OP2(0x2) | off22); } #if TCG_TARGET_REG_BITS == 64 static void tcg_out_branch_i64(TCGContext *s, int opc, int label_index) { TCGLabel *l = &s->labels[label_index]; + uint32_t off19; if (l->has_value) { - tcg_out32(s, (INSN_OP(0) | INSN_COND(opc, 0) | INSN_OP2(0x1) | - (0x5 << 19) | - INSN_OFF19(l->u.value - (unsigned long)s->code_ptr))); + off19 = INSN_OFF19(l->u.value - (unsigned long)s->code_ptr); } else { + /* Make sure to preserve destinations during retranslation. */ + off19 = *(uint32_t *)s->code_ptr & INSN_OFF19(-1); tcg_out_reloc(s, s->code_ptr, R_SPARC_WDISP19, label_index, 0); - tcg_out32(s, (INSN_OP(0) | INSN_COND(opc, 0) | INSN_OP2(0x1) | - (0x5 << 19) | 0)); } + tcg_out32(s, (INSN_OP(0) | INSN_COND(opc, 0) | INSN_OP2(0x1) | + (0x5 << 19) | off19)); } #endif -- 1.7.11.4 ^ permalink raw reply related [flat|nested] 16+ messages in thread
* Re: [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson ` (13 preceding siblings ...) 2012-09-22 2:05 ` [Qemu-devel] [PATCH 14/14] tcg-sparc: Preserve branch destinations during retranslation Richard Henderson @ 2012-09-22 18:01 ` Blue Swirl 14 siblings, 0 replies; 16+ messages in thread From: Blue Swirl @ 2012-09-22 18:01 UTC (permalink / raw) To: Richard Henderson; +Cc: qemu-devel On Sat, Sep 22, 2012 at 2:04 AM, Richard Henderson <rth@twiddle.net> wrote: > Changes since v2: > * Patch 4 split out from patch 5 (afaerber feedback) > * TB chaining preserves wrt retranslation > * Last patch for branch retranslation is new > > The patch set is rebased on cfb75cb9807463ebe18b127096b48b5d0db1ce03 > and is available at > > git://repo.or.cz/qemu/rth.git tcg-sparc Thanks, pulled. > > > r~ > > > Richard Henderson (14): > linux-user: Use memcpy in get_user/put_user. > tcg-sparc: Hack in qemu_ld/st64 for 32-bit. > tcg-sparc: Fix ADDX opcode. > tcg-sparc: Don't MAP_FIXED on top of the program > tcg-sparc: Assume v9 cpu always, i.e. force v8plus in 32-bit mode. > tcg-sparc: Fix qemu_ld/st to handle 32-bit host. > tcg-sparc: Support GUEST_BASE. > tcg-sparc: Change AREG0 in generated code to %i0. > tcg-sparc: Clean up cruft stemming from attempts to use global > registers. > tcg-sparc: Mask shift immediates to avoid illegal insns. > tcg-sparc: Use defines for temporaries. > tcg-sparc: Add %g/%o registers to alloc_order > tcg-sparc: Fix and enable direct TB chaining. > tcg-sparc: Preserve branch destinations during retranslation > > configure | 52 +-- > disas.c | 2 - > exec-all.h | 9 +- > exec.c | 12 +- > linux-user/qemu.h | 47 +-- > qemu-timer.h | 8 +- > tcg/sparc/tcg-target.c | 970 +++++++++++++++++++++++-------------------------- > tcg/sparc/tcg-target.h | 33 +- > tcg/tcg.c | 3 +- > 9 files changed, 514 insertions(+), 622 deletions(-) > > -- > 1.7.11.4 > ^ permalink raw reply [flat|nested] 16+ messages in thread
end of thread, other threads:[~2012-09-22 18:01 UTC | newest] Thread overview: 16+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2012-09-22 2:04 [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 01/14] linux-user: Use memcpy in get_user/put_user Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 02/14] tcg-sparc: Hack in qemu_ld/st64 for 32-bit Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 03/14] tcg-sparc: Fix ADDX opcode Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 04/14] tcg-sparc: Don't MAP_FIXED on top of the program Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 05/14] tcg-sparc: Assume v9 cpu always, i.e. force v8plus in 32-bit mode Richard Henderson 2012-09-22 2:04 ` [Qemu-devel] [PATCH 06/14] tcg-sparc: Fix qemu_ld/st to handle 32-bit host Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 07/14] tcg-sparc: Support GUEST_BASE Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 08/14] tcg-sparc: Change AREG0 in generated code to %i0 Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 09/14] tcg-sparc: Clean up cruft stemming from attempts to use global registers Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 10/14] tcg-sparc: Mask shift immediates to avoid illegal insns Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 11/14] tcg-sparc: Use defines for temporaries Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 12/14] tcg-sparc: Add %g/%o registers to alloc_order Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 13/14] tcg-sparc: Fix and enable direct TB chaining Richard Henderson 2012-09-22 2:05 ` [Qemu-devel] [PATCH 14/14] tcg-sparc: Preserve branch destinations during retranslation Richard Henderson 2012-09-22 18:01 ` [Qemu-devel] [PATCH v3 00/14] Sparc TCG cleanup Blue Swirl
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).