From: Paolo Bonzini <pbonzini@redhat.com>
To: qemu-devel@nongnu.org
Cc: richard.henderson@linaro.org, paul@nowt.org
Subject: [PATCH v2 21/37] target/i386: reimplement 0x0f 0x50-0x5f, add AVX
Date: Tue, 20 Sep 2022 19:24:51 +0200 [thread overview]
Message-ID: <20220920172507.95568-22-pbonzini@redhat.com> (raw)
In-Reply-To: <20220920172507.95568-1-pbonzini@redhat.com>
These are mostly floating-point SSE operations. The odd ones out
are MOVMSK and CVTxx2yy, the others are straightforward.
Unary operations are a bit special in AVX because they have 2 operands
for PD/PS operands (VEX.vvvv must be 1111b), and 3 operands for SD/SS.
They are handled using X86_OP_GROUP3 for compactness.
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
target/i386/tcg/decode-new.c.inc | 43 ++++++++
target/i386/tcg/emit.c.inc | 166 +++++++++++++++++++++++++++++++
target/i386/tcg/translate.c | 2 +-
3 files changed, 210 insertions(+), 1 deletion(-)
diff --git a/target/i386/tcg/decode-new.c.inc b/target/i386/tcg/decode-new.c.inc
index d207a1f0c1..5b753ea329 100644
--- a/target/i386/tcg/decode-new.c.inc
+++ b/target/i386/tcg/decode-new.c.inc
@@ -252,7 +252,41 @@ static void decode_0F3A(DisasContext *s, CPUX86State *env, X86OpEntry *entry, ui
*entry = opcodes_0F3A[*b];
}
+static void decode_sse_unary(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b)
+{
+ if (!(s->prefix & (PREFIX_REPZ | PREFIX_REPNZ))) {
+ entry->op1 = X86_TYPE_None;
+ entry->s1 = X86_SIZE_None;
+ }
+ switch (*b) {
+ case 0x51: entry->gen = gen_VSQRT; break;
+ case 0x52: entry->gen = gen_VRSQRT; break;
+ case 0x53: entry->gen = gen_VRCP; break;
+ case 0x5A: entry->gen = gen_VCVTfp2fp; break;
+ }
+}
+
+static void decode_0F5B(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b)
+{
+ static const X86OpEntry opcodes_0F5B[4] = {
+ X86_OP_ENTRY2(VCVTDQ2PS, V,x, W,x, vex2),
+ X86_OP_ENTRY2(VCVTPS2DQ, V,x, W,x, vex2),
+ X86_OP_ENTRY2(VCVTTPS2DQ, V,x, W,x, vex2),
+ {},
+ };
+ *entry = *decode_by_prefix(s, opcodes_0F5B);
+}
+
static const X86OpEntry opcodes_0F[256] = {
+ [0x50] = X86_OP_ENTRY3(MOVMSK, G,y, None,None, U,x, vex7 p_00_66),
+ [0x51] = X86_OP_GROUP3(sse_unary, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
+ [0x52] = X86_OP_GROUP3(sse_unary, V,x, H,x, W,x, vex5 p_00_f3),
+ [0x53] = X86_OP_GROUP3(sse_unary, V,x, H,x, W,x, vex5 p_00_f3),
+ [0x54] = X86_OP_ENTRY3(PAND, V,x, H,x, W,x, vex4 p_00_66), /* vand */
+ [0x55] = X86_OP_ENTRY3(PANDN, V,x, H,x, W,x, vex4 p_00_66), /* vandn */
+ [0x56] = X86_OP_ENTRY3(POR, V,x, H,x, W,x, vex4 p_00_66), /* vor */
+ [0x57] = X86_OP_ENTRY3(PXOR, V,x, H,x, W,x, vex4 p_00_66), /* vxor */
+
[0x60] = X86_OP_ENTRY3(PUNPCKLBW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66),
[0x61] = X86_OP_ENTRY3(PUNPCKLWD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66),
[0x62] = X86_OP_ENTRY3(PUNPCKLDQ, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66),
@@ -265,6 +299,15 @@ static const X86OpEntry opcodes_0F[256] = {
[0x38] = X86_OP_GROUP0(0F38),
[0x3a] = X86_OP_GROUP0(0F3A),
+ [0x58] = X86_OP_ENTRY3(VADD, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
+ [0x59] = X86_OP_ENTRY3(VMUL, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
+ [0x5a] = X86_OP_GROUP3(sse_unary, V,x, H,x, W,x, vex3 p_00_66_f3_f2),
+ [0x5b] = X86_OP_GROUP0(0F5B),
+ [0x5c] = X86_OP_ENTRY3(VSUB, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
+ [0x5d] = X86_OP_ENTRY3(VMIN, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
+ [0x5e] = X86_OP_ENTRY3(VDIV, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
+ [0x5f] = X86_OP_ENTRY3(VMAX, V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
+
[0x68] = X86_OP_ENTRY3(PUNPCKHBW, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66),
[0x69] = X86_OP_ENTRY3(PUNPCKHWD, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66),
[0x6a] = X86_OP_ENTRY3(PUNPCKHDQ, V,x, H,x, W,x, vex4 mmx avx2_256 p_00_66),
diff --git a/target/i386/tcg/emit.c.inc b/target/i386/tcg/emit.c.inc
index f1f7397869..58b2fd7a2a 100644
--- a/target/i386/tcg/emit.c.inc
+++ b/target/i386/tcg/emit.c.inc
@@ -318,6 +318,131 @@ static void gen_store_sse(DisasContext *s, X86DecodedInsn *decode, int src_ofs)
}
}
+/*
+ * 00 = v*ps Vps, Hps, Wpd
+ * 66 = v*pd Vpd, Hpd, Wps
+ * f3 = v*ss Vss, Hss, Wps
+ * f2 = v*sd Vsd, Hsd, Wps
+ */
+static inline void gen_unary_fp_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
+ SSEFunc_0_epp pd_xmm, SSEFunc_0_epp ps_xmm,
+ SSEFunc_0_epp pd_ymm, SSEFunc_0_epp ps_ymm,
+ SSEFunc_0_eppp sd, SSEFunc_0_eppp ss)
+{
+ if ((s->prefix & (PREFIX_REPZ | PREFIX_REPNZ)) != 0) {
+ SSEFunc_0_eppp fn = s->prefix & PREFIX_REPZ ? ss : sd;
+ if (!fn) {
+ gen_illegal_opcode(s);
+ return;
+ }
+ fn(cpu_env, OP_PTR0, OP_PTR1, OP_PTR2);
+ } else {
+ SSEFunc_0_epp ps, pd, fn;
+ ps = s->vex_l ? ps_ymm : ps_xmm;
+ pd = s->vex_l ? pd_ymm : pd_xmm;
+ fn = s->prefix & PREFIX_DATA ? pd : ps;
+ if (!fn) {
+ gen_illegal_opcode(s);
+ return;
+ }
+ fn(cpu_env, OP_PTR0, OP_PTR2);
+ }
+}
+#define UNARY_FP_SSE(uname, lname) \
+static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
+{ \
+ gen_unary_fp_sse(s, env, decode, \
+ gen_helper_##lname##pd_xmm, \
+ gen_helper_##lname##ps_xmm, \
+ gen_helper_##lname##pd_ymm, \
+ gen_helper_##lname##ps_ymm, \
+ gen_helper_##lname##sd, \
+ gen_helper_##lname##ss); \
+}
+UNARY_FP_SSE(VSQRT, sqrt)
+
+/*
+ * 00 = v*ps Vps, Hps, Wpd
+ * 66 = v*pd Vpd, Hpd, Wps
+ * f3 = v*ss Vss, Hss, Wps
+ * f2 = v*sd Vsd, Hsd, Wps
+ */
+static inline void gen_fp_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
+ SSEFunc_0_eppp pd_xmm, SSEFunc_0_eppp ps_xmm,
+ SSEFunc_0_eppp pd_ymm, SSEFunc_0_eppp ps_ymm,
+ SSEFunc_0_eppp sd, SSEFunc_0_eppp ss)
+{
+ SSEFunc_0_eppp ps, pd, fn;
+ if ((s->prefix & (PREFIX_REPZ | PREFIX_REPNZ)) != 0) {
+ fn = s->prefix & PREFIX_REPZ ? ss : sd;
+ } else {
+ ps = s->vex_l ? ps_ymm : ps_xmm;
+ pd = s->vex_l ? pd_ymm : pd_xmm;
+ fn = s->prefix & PREFIX_DATA ? pd : ps;
+ }
+ if (fn) {
+ fn(cpu_env, OP_PTR0, OP_PTR1, OP_PTR2);
+ } else {
+ gen_illegal_opcode(s);
+ }
+}
+#define FP_SSE(uname, lname) \
+static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
+{ \
+ gen_fp_sse(s, env, decode, \
+ gen_helper_##lname##pd_xmm, \
+ gen_helper_##lname##ps_xmm, \
+ gen_helper_##lname##pd_ymm, \
+ gen_helper_##lname##ps_ymm, \
+ gen_helper_##lname##sd, \
+ gen_helper_##lname##ss); \
+}
+FP_SSE(VADD, add)
+FP_SSE(VMUL, mul)
+FP_SSE(VSUB, sub)
+FP_SSE(VMIN, min)
+FP_SSE(VDIV, div)
+FP_SSE(VMAX, max)
+
+/*
+ * 00 = v*ps Vps, Wpd
+ * f3 = v*ss Vss, Wps
+ */
+static inline void gen_unary_fp32_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
+ SSEFunc_0_epp ps_xmm,
+ SSEFunc_0_epp ps_ymm,
+ SSEFunc_0_eppp ss)
+{
+ if ((s->prefix & (PREFIX_DATA | PREFIX_REPNZ)) != 0) {
+ goto illegal_op;
+ } else if (s->prefix & PREFIX_REPZ) {
+ if (!ss) {
+ goto illegal_op;
+ }
+ ss(cpu_env, OP_PTR0, OP_PTR1, OP_PTR2);
+ } else {
+ SSEFunc_0_epp fn = s->vex_l ? ps_ymm : ps_xmm;
+ if (!fn) {
+ goto illegal_op;
+ }
+ fn(cpu_env, OP_PTR0, OP_PTR2);
+ }
+ return;
+
+illegal_op:
+ gen_illegal_opcode(s);
+}
+#define UNARY_FP32_SSE(uname, lname) \
+static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
+{ \
+ gen_unary_fp32_sse(s, env, decode, \
+ gen_helper_##lname##ps_xmm, \
+ gen_helper_##lname##ps_ymm, \
+ gen_helper_##lname##ss); \
+}
+UNARY_FP32_SSE(VRSQRT, rsqrt)
+UNARY_FP32_SSE(VRCP, rcp)
+
#define BINARY_INT_GVEC(uname, func, ...) \
static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
{ \
@@ -413,6 +538,29 @@ static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decod
BINARY_INT_SSE(PUNPCKLQDQ, punpcklqdq)
BINARY_INT_SSE(PUNPCKHQDQ, punpckhqdq)
+static inline void gen_unary_int_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
+ SSEFunc_0_epp xmm, SSEFunc_0_epp ymm)
+{
+ if (!s->vex_l) {
+ xmm(cpu_env, OP_PTR0, OP_PTR2);
+ } else {
+ ymm(cpu_env, OP_PTR0, OP_PTR2);
+ }
+}
+
+#define UNARY_INT_SSE(uname, lname) \
+static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
+{ \
+ gen_unary_int_sse(s, env, decode, \
+ gen_helper_##lname##_xmm, \
+ gen_helper_##lname##_ymm); \
+}
+
+UNARY_INT_SSE(VCVTDQ2PS, cvtdq2ps)
+UNARY_INT_SSE(VCVTPS2DQ, cvtps2dq)
+UNARY_INT_SSE(VCVTTPS2DQ, cvttps2dq)
+
+
static void gen_ADCOX(DisasContext *s, CPUX86State *env, MemOp ot, int cc_op)
{
TCGv carry_in = NULL;
@@ -608,6 +756,16 @@ static void gen_MOVDQ(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
gen_store_sse(s, decode, decode->op[2].offset);
}
+static void gen_MOVMSK(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
+{
+ typeof(gen_helper_movmskps_ymm) *ps, *pd, *fn;
+ ps = s->vex_l ? gen_helper_movmskps_ymm : gen_helper_movmskps_xmm;
+ pd = s->vex_l ? gen_helper_movmskpd_ymm : gen_helper_movmskpd_xmm;
+ fn = s->prefix & PREFIX_DATA ? pd : ps;
+ fn(s->tmp2_i32, cpu_env, OP_PTR2);
+ tcg_gen_extu_i32_tl(s->T0, s->tmp2_i32);
+}
+
static void gen_MULX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
MemOp ot = decode->op[0].ot;
@@ -708,3 +866,11 @@ static void gen_SHRX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
}
tcg_gen_shr_tl(s->T0, s->T0, s->T1);
}
+
+static void gen_VCVTfp2fp(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
+{
+ gen_unary_fp_sse(s, env, decode,
+ gen_helper_cvtpd2ps_xmm, gen_helper_cvtps2pd_xmm,
+ gen_helper_cvtpd2ps_ymm, gen_helper_cvtps2pd_ymm,
+ gen_helper_cvtsd2ss, gen_helper_cvtss2sd);
+}
diff --git a/target/i386/tcg/translate.c b/target/i386/tcg/translate.c
index fd565650a9..95172c30d0 100644
--- a/target/i386/tcg/translate.c
+++ b/target/i386/tcg/translate.c
@@ -4695,7 +4695,7 @@ static target_ulong disas_insn(DisasContext *s, CPUState *cpu)
use_new &= b <= limit;
#endif
if (use_new &&
- ((b >= 0x160 && b <= 0x16f) ||
+ ((b >= 0x150 && b <= 0x16f) ||
(b >= 0x1d8 && b <= 0x1ff && (b & 8)))) {
disas_insn_new(s, cpu, b + 0x100);
return s->pc;
--
2.37.2
next prev parent reply other threads:[~2022-09-20 21:48 UTC|newest]
Thread overview: 61+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-09-20 17:24 [PATCH v2 00/37] target/i386: new decoder + AVX implementation Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 01/37] target/i386: Define XMMReg and access macros, align ZMM registers Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 02/37] target/i386: make ldo/sto operations consistent with ldq Paolo Bonzini
2022-09-24 15:00 ` Richard Henderson
2022-09-20 17:24 ` [PATCH v2 03/37] target/i386: REPZ and REPNZ are mutually exclusive Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 04/37] target/i386: introduce insn_get_addr Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 05/37] target/i386: add core of new i386 decoder Paolo Bonzini
2022-09-24 15:09 ` Richard Henderson
2022-09-20 17:24 ` [PATCH v2 06/37] target/i386: add ALU load/writeback core Paolo Bonzini
2022-09-24 15:13 ` Richard Henderson
2022-09-20 17:24 ` [PATCH v2 07/37] target/i386: add CPUID[EAX=7, ECX=0].ECX to DisasContext Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 08/37] target/i386: add CPUID feature checks to new decoder Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 09/37] target/i386: add AVX_EN hflag Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 10/37] target/i386: validate VEX prefixes via the instructions' exception classes Paolo Bonzini
2022-09-24 20:19 ` Richard Henderson
2022-09-20 17:24 ` [PATCH v2 11/37] target/i386: validate SSE prefixes directly in the decoding table Paolo Bonzini
2022-09-24 20:23 ` Richard Henderson
2022-09-20 17:24 ` [PATCH v2 12/37] target/i386: move scalar 0F 38 and 0F 3A instruction to new decoder Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 13/37] target/i386: Prepare ops_sse_header.h for 256 bit AVX Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 14/37] target/i386: extend helpers to support VEX.V 3- and 4- operand encodings Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 15/37] target/i386: support operand merging in binary scalar helpers Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 16/37] target/i386: provide 3-operand versions of unary " Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 17/37] target/i386: implement additional AVX comparison operators Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 18/37] target/i386: Introduce 256-bit vector helpers Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 19/37] target/i386: reimplement 0x0f 0x60-0x6f, add AVX Paolo Bonzini
2022-09-24 20:31 ` Richard Henderson
2022-09-20 17:24 ` [PATCH v2 20/37] target/i386: reimplement 0x0f 0xd8-0xdf, 0xe8-0xef, 0xf8-0xff, " Paolo Bonzini
2022-09-24 20:32 ` Richard Henderson
2022-09-20 17:24 ` Paolo Bonzini [this message]
2022-09-24 20:37 ` [PATCH v2 21/37] target/i386: reimplement 0x0f 0x50-0x5f, " Richard Henderson
2022-09-20 17:24 ` [PATCH v2 22/37] target/i386: reimplement 0x0f 0x78-0x7f, " Paolo Bonzini
2022-09-24 20:43 ` Richard Henderson
2022-09-26 7:24 ` Paolo Bonzini
2022-09-26 7:46 ` Richard Henderson
2022-09-20 17:24 ` [PATCH v2 23/37] target/i386: reimplement 0x0f 0x70-0x77, " Paolo Bonzini
2022-09-24 20:53 ` Richard Henderson
2022-09-20 17:24 ` [PATCH v2 24/37] target/i386: reimplement 0x0f 0xd0-0xd7, 0xe0-0xe7, 0xf0-0xf7, " Paolo Bonzini
2022-09-24 20:56 ` Richard Henderson
2022-09-26 7:56 ` Paolo Bonzini
2022-09-20 17:24 ` [PATCH v2 25/37] target/i386: clarify (un)signedness of immediates from 0F3Ah opcodes Paolo Bonzini
2022-09-24 20:57 ` Richard Henderson
2022-09-20 17:24 ` [PATCH v2 26/37] target/i386: reimplement 0x0f 0x3a, add AVX Paolo Bonzini
2022-09-24 21:02 ` Richard Henderson
2022-09-20 17:24 ` [PATCH v2 27/37] target/i386: Use tcg gvec ops for pmovmskb Paolo Bonzini
2022-09-24 21:08 ` Richard Henderson
2022-09-20 17:24 ` [PATCH v2 28/37] target/i386: reimplement 0x0f 0x38, add AVX Paolo Bonzini
2022-09-24 21:14 ` Richard Henderson
2022-09-20 17:24 ` [PATCH v2 29/37] target/i386: reimplement 0x0f 0xc2, 0xc4-0xc6, " Paolo Bonzini
2022-09-20 17:25 ` [PATCH v2 30/37] target/i386: reimplement 0x0f 0x10-0x17, " Paolo Bonzini
2022-09-24 21:16 ` Richard Henderson
2022-09-20 17:25 ` [PATCH v2 31/37] target/i386: reimplement 0x0f 0x28-0x2f, " Paolo Bonzini
2022-09-24 21:18 ` Richard Henderson
2022-09-20 17:25 ` [PATCH v2 32/37] target/i386: implement XSAVE and XRSTOR of AVX registers Paolo Bonzini
2022-09-24 21:19 ` Richard Henderson
2022-09-20 17:25 ` [PATCH v2 33/37] target/i386: implement VLDMXCSR/VSTMXCSR Paolo Bonzini
2022-09-24 21:20 ` Richard Henderson
2022-09-20 17:25 ` [PATCH v2 34/37] target/i386: Enable AVX cpuid bits when using TCG Paolo Bonzini
2022-09-20 17:25 ` [PATCH v2 35/37] tests/tcg: extend SSE tests to AVX Paolo Bonzini
2022-09-20 17:25 ` [PATCH v2 36/37] target/i386: move 3DNow to the new decoder Paolo Bonzini
2022-09-24 21:24 ` Richard Henderson
2022-09-20 17:25 ` [PATCH v2 37/37] target/i386: remove old SSE decoder Paolo Bonzini
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=20220920172507.95568-22-pbonzini@redhat.com \
--to=pbonzini@redhat.com \
--cc=paul@nowt.org \
--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).