From: Paolo Bonzini <pbonzini@redhat.com>
To: qemu-devel@nongnu.org
Cc: Richard Henderson <richard.henderson@linaro.org>
Subject: [PULL 11/46] target/i386: add X86_SPECIALs for MOVSX and MOVZX
Date: Sun, 31 Dec 2023 09:44:27 +0100 [thread overview]
Message-ID: <20231231084502.235366-12-pbonzini@redhat.com> (raw)
In-Reply-To: <20231231084502.235366-1-pbonzini@redhat.com>
Usually the registers are just moved into s->T0 without much care for
their operand size. However, in some cases we can get more efficient
code if the operand fetching logic syncs with the emission function
on what is nicer.
All the current uses are mostly demonstrative and only reduce the code
in the emission functions, because the instructions do not support
memory operands. However the logic is generic and applies to several
more instructions such as MOVSXD (aka movslq), one-byte shift
instructions, multiplications, XLAT, and indirect calls/jumps.
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
target/i386/tcg/decode-new.c.inc | 18 ++++++++++----
target/i386/tcg/decode-new.h | 4 +++
target/i386/tcg/emit.c.inc | 42 +++++++++++++++++---------------
3 files changed, 40 insertions(+), 24 deletions(-)
diff --git a/target/i386/tcg/decode-new.c.inc b/target/i386/tcg/decode-new.c.inc
index 00fdb243857..d7a86d96c0c 100644
--- a/target/i386/tcg/decode-new.c.inc
+++ b/target/i386/tcg/decode-new.c.inc
@@ -156,6 +156,8 @@
#define op0_Rd .special = X86_SPECIAL_Op0_Rd,
#define op2_Ry .special = X86_SPECIAL_Op2_Ry,
#define avx_movx .special = X86_SPECIAL_AVXExtMov,
+#define sextT0 .special = X86_SPECIAL_SExtT0,
+#define zextT0 .special = X86_SPECIAL_ZExtT0,
#define vex1 .vex_class = 1,
#define vex1_rep3 .vex_class = 1, .vex_special = X86_VEX_REPScalar,
@@ -571,8 +573,8 @@ static const X86OpEntry opcodes_0F38_F0toFF[16][5] = {
[5] = {
X86_OP_ENTRY3(BZHI, G,y, E,y, B,y, vex13 cpuid(BMI1)),
{},
- X86_OP_ENTRY3(PEXT, G,y, B,y, E,y, vex13 cpuid(BMI2)),
- X86_OP_ENTRY3(PDEP, G,y, B,y, E,y, vex13 cpuid(BMI2)),
+ X86_OP_ENTRY3(PEXT, G,y, B,y, E,y, vex13 zextT0 cpuid(BMI2)),
+ X86_OP_ENTRY3(PDEP, G,y, B,y, E,y, vex13 zextT0 cpuid(BMI2)),
{},
},
[6] = {
@@ -583,10 +585,10 @@ static const X86OpEntry opcodes_0F38_F0toFF[16][5] = {
{},
},
[7] = {
- X86_OP_ENTRY3(BEXTR, G,y, E,y, B,y, vex13 cpuid(BMI1)),
+ X86_OP_ENTRY3(BEXTR, G,y, E,y, B,y, vex13 zextT0 cpuid(BMI1)),
X86_OP_ENTRY3(SHLX, G,y, E,y, B,y, vex13 cpuid(BMI1)),
- X86_OP_ENTRY3(SARX, G,y, E,y, B,y, vex13 cpuid(BMI1)),
- X86_OP_ENTRY3(SHRX, G,y, E,y, B,y, vex13 cpuid(BMI1)),
+ X86_OP_ENTRY3(SARX, G,y, E,y, B,y, vex13 sextT0 cpuid(BMI1)),
+ X86_OP_ENTRY3(SHRX, G,y, E,y, B,y, vex13 zextT0 cpuid(BMI1)),
{},
},
};
@@ -1905,6 +1907,12 @@ static void disas_insn_new(DisasContext *s, CPUState *cpu, int b)
}
break;
+ case X86_SPECIAL_SExtT0:
+ case X86_SPECIAL_ZExtT0:
+ /* Handled in gen_load. */
+ assert(decode.op[1].unit == X86_OP_INT);
+ break;
+
default:
break;
}
diff --git a/target/i386/tcg/decode-new.h b/target/i386/tcg/decode-new.h
index b253f7457ae..70b6717227f 100644
--- a/target/i386/tcg/decode-new.h
+++ b/target/i386/tcg/decode-new.h
@@ -191,6 +191,10 @@ typedef enum X86InsnSpecial {
* become P/P/Q/N, and size "x" becomes "q".
*/
X86_SPECIAL_MMX,
+
+ /* When loaded into s->T0, register operand 1 is zero/sign extended. */
+ X86_SPECIAL_SExtT0,
+ X86_SPECIAL_ZExtT0,
} X86InsnSpecial;
/*
diff --git a/target/i386/tcg/emit.c.inc b/target/i386/tcg/emit.c.inc
index f5e44117eab..4c2006fdd09 100644
--- a/target/i386/tcg/emit.c.inc
+++ b/target/i386/tcg/emit.c.inc
@@ -232,9 +232,30 @@ static void gen_load(DisasContext *s, X86DecodedInsn *decode, int opn, TCGv v)
break;
case X86_OP_INT:
if (op->has_ea) {
- gen_op_ld_v(s, op->ot, v, s->A0);
+ if (v == s->T0 && decode->e.special == X86_SPECIAL_SExtT0) {
+ gen_op_ld_v(s, op->ot | MO_SIGN, v, s->A0);
+ } else {
+ gen_op_ld_v(s, op->ot, v, s->A0);
+ }
+
+ } else if (op->ot == MO_8 && byte_reg_is_xH(s, op->n)) {
+ if (v == s->T0 && decode->e.special == X86_SPECIAL_SExtT0) {
+ tcg_gen_sextract_tl(v, cpu_regs[op->n - 4], 8, 8);
+ } else {
+ tcg_gen_extract_tl(v, cpu_regs[op->n - 4], 8, 8);
+ }
+
+ } else if (op->ot < MO_TL && v == s->T0 &&
+ (decode->e.special == X86_SPECIAL_SExtT0 ||
+ decode->e.special == X86_SPECIAL_ZExtT0)) {
+ if (decode->e.special == X86_SPECIAL_SExtT0) {
+ tcg_gen_ext_tl(v, cpu_regs[op->n], op->ot | MO_SIGN);
+ } else {
+ tcg_gen_ext_tl(v, cpu_regs[op->n], op->ot);
+ }
+
} else {
- gen_op_mov_v_reg(s, op->ot, v, op->n);
+ tcg_gen_mov_tl(v, cpu_regs[op->n]);
}
break;
case X86_OP_IMM:
@@ -1084,9 +1105,6 @@ static void gen_BEXTR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
* Shifts larger than operand size get zeros.
*/
tcg_gen_ext8u_tl(s->A0, s->T1);
- if (TARGET_LONG_BITS == 64 && ot == MO_32) {
- tcg_gen_ext32u_tl(s->T0, s->T0);
- }
tcg_gen_shr_tl(s->T0, s->T0, s->A0);
tcg_gen_movcond_tl(TCG_COND_LEU, s->T0, s->A0, bound, s->T0, zero);
@@ -1428,19 +1446,11 @@ static void gen_PCMPISTRM(DisasContext *s, CPUX86State *env, X86DecodedInsn *dec
static void gen_PDEP(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
- MemOp ot = decode->op[1].ot;
- if (ot < MO_64) {
- tcg_gen_ext32u_tl(s->T0, s->T0);
- }
gen_helper_pdep(s->T0, s->T0, s->T1);
}
static void gen_PEXT(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
- MemOp ot = decode->op[1].ot;
- if (ot < MO_64) {
- tcg_gen_ext32u_tl(s->T0, s->T0);
- }
gen_helper_pext(s->T0, s->T0, s->T1);
}
@@ -1796,9 +1806,6 @@ static void gen_SARX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
mask = ot == MO_64 ? 63 : 31;
tcg_gen_andi_tl(s->T1, s->T1, mask);
- if (ot != MO_64) {
- tcg_gen_ext32s_tl(s->T0, s->T0);
- }
tcg_gen_sar_tl(s->T0, s->T0, s->T1);
}
@@ -1873,9 +1880,6 @@ static void gen_SHRX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
mask = ot == MO_64 ? 63 : 31;
tcg_gen_andi_tl(s->T1, s->T1, mask);
- if (ot != MO_64) {
- tcg_gen_ext32u_tl(s->T0, s->T0);
- }
tcg_gen_shr_tl(s->T0, s->T0, s->T1);
}
--
2.43.0
next prev parent reply other threads:[~2023-12-31 8:49 UTC|newest]
Thread overview: 48+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-12-31 8:44 [PULL 00/46] (mostly) target/i386 and meson changes for 2023-12-31 Paolo Bonzini
2023-12-31 8:44 ` [PULL 01/46] configure: use a native non-cross compiler for linux-user Paolo Bonzini
2023-12-31 8:44 ` [PULL 02/46] target/i386: optimize computation of JL and JLE from flags Paolo Bonzini
2023-12-31 8:44 ` [PULL 03/46] target/i386: speedup JO/SETO after MUL or IMUL Paolo Bonzini
2023-12-31 8:44 ` [PULL 04/46] target/i386: remove unnecessary arguments from raise_interrupt Paolo Bonzini
2023-12-31 8:44 ` [PULL 05/46] target/i386: remove unnecessary truncations Paolo Bonzini
2023-12-31 8:44 ` [PULL 06/46] target/i386: clean up cpu_cc_compute_all Paolo Bonzini
2023-12-31 8:44 ` [PULL 07/46] target/i386: document more deviations from the manual Paolo Bonzini
2023-12-31 8:44 ` [PULL 08/46] target/i386: reimplement check for validity of LOCK prefix Paolo Bonzini
2023-12-31 8:44 ` [PULL 09/46] target/i386: avoid trunc and ext for MULX and RORX Paolo Bonzini
2023-12-31 8:44 ` [PULL 10/46] target/i386: rename zext0/zext2 and make them closer to the manual Paolo Bonzini
2023-12-31 8:44 ` Paolo Bonzini [this message]
2023-12-31 8:44 ` [PULL 12/46] target/i386: do not decode string source/destination into decode->mem Paolo Bonzini
2023-12-31 8:44 ` [PULL 13/46] target/i386: do not clobber A0 in POP translation Paolo Bonzini
2023-12-31 8:44 ` [PULL 14/46] target/i386: do not clobber T0 on string operations Paolo Bonzini
2023-12-31 8:44 ` [PULL 15/46] target/i386: split eflags computation out of gen_compute_eflags Paolo Bonzini
2023-12-31 8:44 ` [PULL 16/46] target/i386: do not use s->tmp4 for push Paolo Bonzini
2023-12-31 8:44 ` [PULL 17/46] target/i386: do not use s->tmp0 for jumps on ECX ==/!= 0 Paolo Bonzini
2023-12-31 8:44 ` [PULL 18/46] target/i386: prepare for implementation of STOS/SCAS in new decoder Paolo Bonzini
2023-12-31 8:44 ` [PULL 19/46] target/i386: move operand load and writeback out of gen_cmovcc1 Paolo Bonzini
2023-12-31 8:44 ` [PULL 20/46] target/i386: adjust decoding of J operand Paolo Bonzini
2023-12-31 8:44 ` [PULL 21/46] target/i386: introduce flags writeback mechanism Paolo Bonzini
2023-12-31 8:44 ` [PULL 22/46] target/i386: implement CMPccXADD Paolo Bonzini
2023-12-31 8:44 ` [PULL 23/46] target/i386: the sgx_epc_get_section stub is reachable Paolo Bonzini
2023-12-31 8:44 ` [PULL 24/46] esp: check for NULL result from scsi_device_find() Paolo Bonzini
2023-12-31 8:44 ` [PULL 25/46] meson: fix type of "relocatable" option Paolo Bonzini
2023-12-31 8:44 ` [PULL 26/46] meson: remove unused variable Paolo Bonzini
2023-12-31 8:44 ` [PULL 27/46] meson: use version_compare() to compare version Paolo Bonzini
2023-12-31 8:44 ` [PULL 28/46] Makefile: clean qemu-iotests output Paolo Bonzini
2023-12-31 8:44 ` [PULL 29/46] configure: remove unnecessary subshell Paolo Bonzini
2023-12-31 8:44 ` [PULL 30/46] configure: unify again the case arms in probe_target_compiler Paolo Bonzini
2023-12-31 8:44 ` [PULL 31/46] meson: add more sections to main meson.build Paolo Bonzini
2023-12-31 8:44 ` [PULL 32/46] meson: move program checks together Paolo Bonzini
2023-12-31 8:44 ` [PULL 33/46] meson: move option validation together Paolo Bonzini
2023-12-31 8:44 ` [PULL 34/46] meson: move accelerator dependency checks together Paolo Bonzini
2023-12-31 8:44 ` [PULL 35/46] meson: keep subprojects together Paolo Bonzini
2023-12-31 8:44 ` [PULL 36/46] meson: move CFI detection code with other compiler flags Paolo Bonzini
2023-12-31 8:44 ` [PULL 37/46] meson: move config-host.h definitions together Paolo Bonzini
2023-12-31 8:44 ` [PULL 38/46] meson: move subdirs to "Collect sources" section Paolo Bonzini
2023-12-31 8:44 ` [PULL 39/46] meson: always probe u2f and canokey if the option is enabled Paolo Bonzini
2023-12-31 8:44 ` [PULL 40/46] meson: remove OS definitions from config_targetos Paolo Bonzini
2023-12-31 8:44 ` [PULL 41/46] meson: remove CONFIG_POSIX and CONFIG_WIN32 " Paolo Bonzini
2023-12-31 8:44 ` [PULL 42/46] meson: remove config_targetos Paolo Bonzini
2023-12-31 8:44 ` [PULL 43/46] meson: remove CONFIG_ALL Paolo Bonzini
2023-12-31 8:45 ` [PULL 44/46] meson: rename config_all Paolo Bonzini
2023-12-31 8:45 ` [PULL 45/46] configure, meson: rename targetos to host_os Paolo Bonzini
2023-12-31 8:45 ` [PULL 46/46] meson.build: report graphics backends separately Paolo Bonzini
2024-01-05 12:53 ` [PULL 00/46] (mostly) target/i386 and meson changes for 2023-12-31 Peter Maydell
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20231231084502.235366-12-pbonzini@redhat.com \
--to=pbonzini@redhat.com \
--cc=qemu-devel@nongnu.org \
--cc=richard.henderson@linaro.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).