mirrors/qemu
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

target/i386: reimplement 0x0f 0x3a, add AVX

Browse Source 
The more complicated operations here are insertions and extractions.
Otherwise, there are just more entries than usual because the PS/PD/SS/SD
variations are encoded in the opcode rater than in the prefixes.

These three-byte opcodes also include AVX new instructions, whose
implementation in the helpers was originally done by Paul Brook
<paul@nowt.org>.

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This commit is contained in:
Paolo Bonzini 2022-09-06 10:34:11 +02:00
parent a64eee3ab4
commit 7906847768
5 changed files with 491 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

95
target/i386/ops_sse.h
View File

@ -2381,6 +2381,101 @@ void glue(helper_aeskeygenassist, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
#endif
#endif
#if SHIFT >= 1
void glue(helper_vpermilpd_imm, SUFFIX)(Reg *d, Reg *s, uint32_t order)
{
uint64_t r0, r1;
int i;
for (i = 0; i < 1 << SHIFT; i += 2) {
r0 = s->Q(i + ((order >> 0) & 1));
r1 = s->Q(i + ((order >> 1) & 1));
d->Q(i) = r0;
d->Q(i+1) = r1;
order >>= 2;
}
}
void glue(helper_vpermilps_imm, SUFFIX)(Reg *d, Reg *s, uint32_t order)
{
uint32_t r0, r1, r2, r3;
int i;
for (i = 0; i < 2 << SHIFT; i += 4) {
r0 = s->L(i + ((order >> 0) & 3));
r1 = s->L(i + ((order >> 2) & 3));
r2 = s->L(i + ((order >> 4) & 3));
r3 = s->L(i + ((order >> 6) & 3));
d->L(i) = r0;
d->L(i+1) = r1;
d->L(i+2) = r2;
d->L(i+3) = r3;
}
}
#if SHIFT >= 2
void helper_vpermdq_ymm(Reg *d, Reg *v, Reg *s, uint32_t order)
{
uint64_t r0, r1, r2, r3;
switch (order & 3) {
case 0:
r0 = v->Q(0);
r1 = v->Q(1);
break;
case 1:
r0 = v->Q(2);
r1 = v->Q(3);
break;
case 2:
r0 = s->Q(0);
r1 = s->Q(1);
break;
case 3:
r0 = s->Q(2);
r1 = s->Q(3);
break;
}
switch ((order >> 4) & 3) {
case 0:
r2 = v->Q(0);
r3 = v->Q(1);
break;
case 1:
r2 = v->Q(2);
r3 = v->Q(3);
break;
case 2:
r2 = s->Q(0);
r3 = s->Q(1);
break;
case 3:
r2 = s->Q(2);
r3 = s->Q(3);
break;
}
d->Q(0) = r0;
d->Q(1) = r1;
d->Q(2) = r2;
d->Q(3) = r3;
}
void helper_vpermq_ymm(Reg *d, Reg *s, uint32_t order)
{
uint64_t r0, r1, r2, r3;
r0 = s->Q(order & 3);
r1 = s->Q((order >> 2) & 3);
r2 = s->Q((order >> 4) & 3);
r3 = s->Q((order >> 6) & 3);
d->Q(0) = r0;
d->Q(1) = r1;
d->Q(2) = r2;
d->Q(3) = r3;
}
#endif
#endif
#undef SSE_HELPER_S
#undef LANE_WIDTH

10
target/i386/ops_sse_header.h
View File

@ -411,6 +411,16 @@ DEF_HELPER_4(glue(aeskeygenassist, SUFFIX), void, env, Reg, Reg, i32)
DEF_HELPER_5(glue(pclmulqdq, SUFFIX), void, env, Reg, Reg, Reg, i32)
#endif
/* AVX helpers */
#if SHIFT >= 1
DEF_HELPER_3(glue(vpermilpd_imm, SUFFIX), void, Reg, Reg, i32)
DEF_HELPER_3(glue(vpermilps_imm, SUFFIX), void, Reg, Reg, i32)
#if SHIFT == 2
DEF_HELPER_4(vpermdq_ymm, void, Reg, Reg, Reg, i32)
DEF_HELPER_3(vpermq_ymm, void, Reg, Reg, i32)
#endif
#endif
#undef SHIFT
#undef Reg
#undef SUFFIX

75
target/i386/tcg/decode-new.c.inc
View File

@ -375,7 +375,78 @@ static void decode_0F38(DisasContext *s, CPUX86State *env, X86OpEntry *entry, ui
}
}
static void decode_VINSERTPS(DisasContext *s, CPUX86State *env, X86OpEntry *entry, uint8_t *b)
{
static const X86OpEntry
vinsertps_reg = X86_OP_ENTRY4(VINSERTPS_r, V,dq, H,dq, U,dq, vex5 cpuid(SSE41) p_66),
vinsertps_mem = X86_OP_ENTRY4(VINSERTPS_m, V,dq, H,dq, M,d, vex5 cpuid(SSE41) p_66);
int modrm = get_modrm(s, env);
*entry = (modrm >> 6) == 3 ? vinsertps_reg : vinsertps_mem;
}
static const X86OpEntry opcodes_0F3A[256] = {
/*
* These are VEX-only, but incorrectly listed in the manual as exception type 4.
* Also the "qq" instructions are sometimes omitted by Table 2-17, but are VEX256
* only.
*/
[0x00] = X86_OP_ENTRY3(VPERMQ, V,qq, W,qq, I,b, vex6 cpuid(AVX2) p_66),
[0x01] = X86_OP_ENTRY3(VPERMQ, V,qq, W,qq, I,b, vex6 cpuid(AVX2) p_66), /* VPERMPD */
[0x02] = X86_OP_ENTRY4(VBLENDPS, V,x, H,x, W,x, vex6 cpuid(AVX2) p_66), /* VPBLENDD */
[0x04] = X86_OP_ENTRY3(VPERMILPS_i, V,x, W,x, I,b, vex6 cpuid(AVX) p_66),
[0x05] = X86_OP_ENTRY3(VPERMILPD_i, V,x, W,x, I,b, vex6 cpuid(AVX) p_66),
[0x06] = X86_OP_ENTRY4(VPERM2x128, V,qq, H,qq, W,qq, vex6 cpuid(AVX) p_66),
[0x14] = X86_OP_ENTRY3(PEXTRB, E,b, V,dq, I,b, vex5 cpuid(SSE41) zext0 p_66),
[0x15] = X86_OP_ENTRY3(PEXTRW, E,w, V,dq, I,b, vex5 cpuid(SSE41) zext0 p_66),
[0x16] = X86_OP_ENTRY3(PEXTR, E,y, V,dq, I,b, vex5 cpuid(SSE41) p_66),
[0x17] = X86_OP_ENTRY3(VEXTRACTPS, E,d, V,dq, I,b, vex5 cpuid(SSE41) p_66),
[0x20] = X86_OP_ENTRY4(PINSRB, V,dq, H,dq, E,b, vex5 cpuid(SSE41) zext2 p_66),
[0x21] = X86_OP_GROUP0(VINSERTPS),
[0x22] = X86_OP_ENTRY4(PINSR, V,dq, H,dq, E,y, vex5 cpuid(SSE41) p_66),
[0x40] = X86_OP_ENTRY4(VDDPS, V,x, H,x, W,x, vex2 cpuid(SSE41) p_66),
[0x41] = X86_OP_ENTRY4(VDDPD, V,dq, H,dq, W,dq, vex2 cpuid(SSE41) p_66),
[0x42] = X86_OP_ENTRY4(VMPSADBW, V,x, H,x, W,x, vex2 cpuid(SSE41) avx2_256 p_66),
[0x44] = X86_OP_ENTRY4(PCLMULQDQ, V,dq, H,dq, W,dq, vex4 cpuid(PCLMULQDQ) p_66),
[0x46] = X86_OP_ENTRY4(VPERM2x128, V,qq, H,qq, W,qq, vex6 cpuid(AVX2) p_66),
[0x60] = X86_OP_ENTRY4(PCMPESTRM, None,None, V,dq, W,dq, vex4_unal cpuid(SSE42) p_66),
[0x61] = X86_OP_ENTRY4(PCMPESTRI, None,None, V,dq, W,dq, vex4_unal cpuid(SSE42) p_66),
[0x62] = X86_OP_ENTRY4(PCMPISTRM, None,None, V,dq, W,dq, vex4_unal cpuid(SSE42) p_66),
[0x63] = X86_OP_ENTRY4(PCMPISTRI, None,None, V,dq, W,dq, vex4_unal cpuid(SSE42) p_66),
[0x08] = X86_OP_ENTRY3(VROUNDPS, V,x, W,x, I,b, vex2 cpuid(SSE41) p_66),
[0x09] = X86_OP_ENTRY3(VROUNDPD, V,x, W,x, I,b, vex2 cpuid(SSE41) p_66),
/*
* Not listed as four operand in the manual. Also writes and reads 128-bits
* from the first two operands due to the V operand picking higher entries of
* the H operand; the "Vss,Hss,Wss" description from the manual is incorrect.
* For other unary operations such as VSQRTSx this is hidden by the "REPScalar"
* value of vex_special, because the table lists the operand types of VSQRTPx.
*/
[0x0a] = X86_OP_ENTRY4(VROUNDSS, V,x, H,x, W,ss, vex3 cpuid(SSE41) p_66),
[0x0b] = X86_OP_ENTRY4(VROUNDSD, V,x, H,x, W,sd, vex3 cpuid(SSE41) p_66),
[0x0c] = X86_OP_ENTRY4(VBLENDPS, V,x, H,x, W,x, vex4 cpuid(SSE41) p_66),
[0x0d] = X86_OP_ENTRY4(VBLENDPD, V,x, H,x, W,x, vex4 cpuid(SSE41) p_66),
[0x0e] = X86_OP_ENTRY4(VPBLENDW, V,x, H,x, W,x, vex4 cpuid(SSE41) p_66),
[0x0f] = X86_OP_ENTRY4(PALIGNR, V,x, H,x, W,x, vex4 cpuid(SSSE3) mmx p_00_66),
[0x18] = X86_OP_ENTRY4(VINSERTx128, V,qq, H,qq, W,qq, vex6 cpuid(AVX) p_66),
[0x19] = X86_OP_ENTRY3(VEXTRACTx128, W,dq, V,qq, I,b, vex6 cpuid(AVX) p_66),
[0x38] = X86_OP_ENTRY4(VINSERTx128, V,qq, H,qq, W,qq, vex6 cpuid(AVX2) p_66),
[0x39] = X86_OP_ENTRY3(VEXTRACTx128, W,dq, V,qq, I,b, vex6 cpuid(AVX2) p_66),
/* Listed incorrectly as type 4 */
[0x4a] = X86_OP_ENTRY4(VBLENDVPS, V,x, H,x, W,x, vex6 cpuid(AVX) p_66),
[0x4b] = X86_OP_ENTRY4(VBLENDVPD, V,x, H,x, W,x, vex6 cpuid(AVX) p_66),
[0x4c] = X86_OP_ENTRY4(VPBLENDVB, V,x, H,x, W,x, vex6 cpuid(AVX) p_66 avx2_256),
[0xdf] = X86_OP_ENTRY3(VAESKEYGEN, V,dq, W,dq, I,b, vex4 cpuid(AES) p_66),
[0xF0] = X86_OP_ENTRY3(RORX, G,y, E,y, I,b, vex13 cpuid(BMI2) p_f2),
};
@ -916,6 +987,10 @@ static bool decode_insn(DisasContext *s, CPUX86State *env, X86DecodeFunc decode_
}
}
if (e->op3 != X86_TYPE_None) {
/*
* A couple instructions actually use the extra immediate byte for an Lx
* register operand; those are handled in the gen_* functions as one off.
*/
assert(e->op3 == X86_TYPE_I && e->s3 == X86_SIZE_b);
s->rip_offset += 1;
}

309
target/i386/tcg/emit.c.inc
View File

@ -473,6 +473,55 @@ HORIZONTAL_FP_SSE(VHADD, hadd)
HORIZONTAL_FP_SSE(VHSUB, hsub)
HORIZONTAL_FP_SSE(VADDSUB, addsub)
static inline void gen_ternary_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
int op3, SSEFunc_0_epppp xmm, SSEFunc_0_epppp ymm)
{
SSEFunc_0_epppp fn = s->vex_l ? ymm : xmm;
TCGv_ptr ptr3 = tcg_temp_new_ptr();
/* The format of the fourth input is Lx */
tcg_gen_addi_ptr(ptr3, cpu_env, ZMM_OFFSET(op3));
fn(cpu_env, OP_PTR0, OP_PTR1, OP_PTR2, ptr3);
tcg_temp_free_ptr(ptr3);
}
#define TERNARY_SSE(uvname, lname) \
static void gen_##uvname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
{ \
gen_ternary_sse(s, env, decode, (uint8_t)decode->immediate >> 4, \
gen_helper_##lname##_xmm, gen_helper_##lname##_ymm); \
}
TERNARY_SSE(VBLENDVPS, blendvps)
TERNARY_SSE(VBLENDVPD, blendvpd)
TERNARY_SSE(VPBLENDVB, pblendvb)
static inline void gen_binary_imm_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
SSEFunc_0_epppi xmm, SSEFunc_0_epppi ymm)
{
TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
if (!s->vex_l) {
xmm(cpu_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
} else {
ymm(cpu_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
}
}
#define BINARY_IMM_SSE(uname, lname) \
static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
{ \
gen_binary_imm_sse(s, env, decode, \
gen_helper_##lname##_xmm, \
gen_helper_##lname##_ymm); \
}
BINARY_IMM_SSE(VBLENDPD, blendpd)
BINARY_IMM_SSE(VBLENDPS, blendps)
BINARY_IMM_SSE(VPBLENDW, pblendw)
BINARY_IMM_SSE(VDDPS, dpps)
#define gen_helper_dppd_ymm NULL
BINARY_IMM_SSE(VDDPD, dppd)
BINARY_IMM_SSE(VMPSADBW, mpsadbw)
BINARY_IMM_SSE(PCLMULQDQ, pclmulqdq)
#define BINARY_INT_GVEC(uname, func, ...) \
static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
{ \
@ -638,6 +687,32 @@ static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decod
UNARY_IMM_SSE(PSHUFD, pshufd)
UNARY_IMM_SSE(PSHUFHW, pshufhw)
UNARY_IMM_SSE(PSHUFLW, pshuflw)
#define gen_helper_vpermq_xmm NULL
UNARY_IMM_SSE(VPERMQ, vpermq)
UNARY_IMM_SSE(VPERMILPS_i, vpermilps_imm)
UNARY_IMM_SSE(VPERMILPD_i, vpermilpd_imm)
static inline void gen_unary_imm_fp_sse(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode,
SSEFunc_0_eppi xmm, SSEFunc_0_eppi ymm)
{
TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
if (!s->vex_l) {
xmm(cpu_env, OP_PTR0, OP_PTR1, imm);
} else {
ymm(cpu_env, OP_PTR0, OP_PTR1, imm);
}
}
#define UNARY_IMM_FP_SSE(uname, lname) \
static void gen_##uname(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode) \
{ \
gen_unary_imm_fp_sse(s, env, decode, \
gen_helper_##lname##_xmm, \
gen_helper_##lname##_ymm); \
}
UNARY_IMM_FP_SSE(VROUNDPS, roundps)
UNARY_IMM_FP_SSE(VROUNDPD, roundpd)
static void gen_ADCOX(DisasContext *s, CPUX86State *env, MemOp ot, int cc_op)
{
@ -957,6 +1032,18 @@ static void gen_MULX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
}
static void gen_PALIGNR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
if (!(s->prefix & PREFIX_DATA)) {
gen_helper_palignr_mmx(cpu_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
} else if (!s->vex_l) {
gen_helper_palignr_xmm(cpu_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
} else {
gen_helper_palignr_ymm(cpu_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
}
}
static void gen_PANDN(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
int vec_len = vector_len(s, decode);
@ -967,6 +1054,42 @@ static void gen_PANDN(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
decode->op[1].offset, vec_len, vec_len);
}
static void gen_PCMPESTRI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
gen_helper_pcmpestri_xmm(cpu_env, OP_PTR1, OP_PTR2, imm);
set_cc_op(s, CC_OP_EFLAGS);
}
static void gen_PCMPESTRM(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
gen_helper_pcmpestrm_xmm(cpu_env, OP_PTR1, OP_PTR2, imm);
set_cc_op(s, CC_OP_EFLAGS);
if ((s->prefix & PREFIX_VEX) && !s->vex_l) {
tcg_gen_gvec_dup_imm(MO_64, offsetof(CPUX86State, xmm_regs[0].ZMM_X(1)),
16, 16, 0);
}
}
static void gen_PCMPISTRI(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
gen_helper_pcmpistri_xmm(cpu_env, OP_PTR1, OP_PTR2, imm);
set_cc_op(s, CC_OP_EFLAGS);
}
static void gen_PCMPISTRM(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
gen_helper_pcmpistrm_xmm(cpu_env, OP_PTR1, OP_PTR2, imm);
set_cc_op(s, CC_OP_EFLAGS);
if ((s->prefix & PREFIX_VEX) && !s->vex_l) {
tcg_gen_gvec_dup_imm(MO_64, offsetof(CPUX86State, xmm_regs[0].ZMM_X(1)),
16, 16, 0);
}
}
static void gen_PDEP(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
MemOp ot = decode->op[1].ot;
@ -985,6 +1108,89 @@ static void gen_PEXT(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
gen_helper_pext(s->T0, s->T0, s->T1);
}
static inline void gen_pextr(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode, MemOp ot)
{
int vec_len = vector_len(s, decode);
int mask = (vec_len >> ot) - 1;
int val = decode->immediate & mask;
switch (ot) {
case MO_8:
tcg_gen_ld8u_tl(s->T0, cpu_env, vector_elem_offset(&decode->op[1], ot, val));
break;
case MO_16:
tcg_gen_ld16u_tl(s->T0, cpu_env, vector_elem_offset(&decode->op[1], ot, val));
break;
case MO_32:
#ifdef TARGET_X86_64
tcg_gen_ld32u_tl(s->T0, cpu_env, vector_elem_offset(&decode->op[1], ot, val));
break;
case MO_64:
#endif
tcg_gen_ld_tl(s->T0, cpu_env, vector_elem_offset(&decode->op[1], ot, val));
break;
default:
abort();
}
}
static void gen_PEXTRB(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
gen_pextr(s, env, decode, MO_8);
}
static void gen_PEXTRW(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
gen_pextr(s, env, decode, MO_16);
}
static void gen_PEXTR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
MemOp ot = decode->op[0].ot;
gen_pextr(s, env, decode, ot);
}
static inline void gen_pinsr(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode, MemOp ot)
{
int vec_len = vector_len(s, decode);
int mask = (vec_len >> ot) - 1;
int val = decode->immediate & mask;
if (decode->op[1].offset != decode->op[0].offset) {
assert(vec_len == 16);
gen_store_sse(s, decode, decode->op[1].offset);
}
switch (ot) {
case MO_8:
tcg_gen_st8_tl(s->T1, cpu_env, vector_elem_offset(&decode->op[0], ot, val));
break;
case MO_16:
tcg_gen_st16_tl(s->T1, cpu_env, vector_elem_offset(&decode->op[0], ot, val));
break;
case MO_32:
#ifdef TARGET_X86_64
tcg_gen_st32_tl(s->T1, cpu_env, vector_elem_offset(&decode->op[0], ot, val));
break;
case MO_64:
#endif
tcg_gen_st_tl(s->T1, cpu_env, vector_elem_offset(&decode->op[0], ot, val));
break;
default:
abort();
}
}
static void gen_PINSRB(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
gen_pinsr(s, env, decode, MO_8);
}
static void gen_PINSR(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
gen_pinsr(s, env, decode, decode->op[2].ot);
}
static void gen_PMOVMSKB(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
if (s->prefix & PREFIX_DATA) {
@ -1193,6 +1399,13 @@ static void gen_SHRX(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
tcg_gen_shr_tl(s->T0, s->T0, s->T1);
}
static void gen_VAESKEYGEN(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
assert(!s->vex_l);
gen_helper_aeskeygenassist_xmm(cpu_env, OP_PTR0, OP_PTR1, imm);
}
static void gen_VCVTfp2fp(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
gen_unary_fp_sse(s, env, decode,
@ -1201,6 +1414,102 @@ static void gen_VCVTfp2fp(DisasContext *s, CPUX86State *env, X86DecodedInsn *dec
gen_helper_cvtsd2ss, gen_helper_cvtss2sd);
}
static void gen_VEXTRACTx128(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
int mask = decode->immediate & 1;
int src_ofs = vector_elem_offset(&decode->op[1], MO_128, mask);
if (decode->op[0].has_ea) {
/* VEX-only instruction, no alignment requirements. */
gen_sto_env_A0(s, src_ofs, false);
} else {
tcg_gen_gvec_mov(MO_64, decode->op[0].offset, src_ofs, 16, 16);
}
}
static void gen_VEXTRACTPS(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
gen_pextr(s, env, decode, MO_32);
}
static void gen_vinsertps(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
int val = decode->immediate;
int dest_word = (val >> 4) & 3;
int new_mask = (val & 15) | (1 << dest_word);
int vec_len = 16;
assert(!s->vex_l);
if (new_mask == 15) {
/* All zeroes except possibly for the inserted element */
tcg_gen_gvec_dup_imm(MO_64, decode->op[0].offset, vec_len, vec_len, 0);
} else if (decode->op[1].offset != decode->op[0].offset) {
gen_store_sse(s, decode, decode->op[1].offset);
}
if (new_mask != (val & 15)) {
tcg_gen_st_i32(s->tmp2_i32, cpu_env,
vector_elem_offset(&decode->op[0], MO_32, dest_word));
}
if (new_mask != 15) {
TCGv_i32 zero = tcg_constant_i32(0); /* float32_zero */
int i;
for (i = 0; i < 4; i++) {
if ((val >> i) & 1) {
tcg_gen_st_i32(zero, cpu_env,
vector_elem_offset(&decode->op[0], MO_32, i));
}
}
}
}
static void gen_VINSERTPS_r(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
int val = decode->immediate;
tcg_gen_ld_i32(s->tmp2_i32, cpu_env,
vector_elem_offset(&decode->op[2], MO_32, (val >> 6) & 3));
gen_vinsertps(s, env, decode);
}
static void gen_VINSERTPS_m(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
tcg_gen_qemu_ld_i32(s->tmp2_i32, s->A0, s->mem_index, MO_LEUL);
gen_vinsertps(s, env, decode);
}
static void gen_VINSERTx128(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
int mask = decode->immediate & 1;
tcg_gen_gvec_mov(MO_64,
decode->op[0].offset + offsetof(YMMReg, YMM_X(mask)),
decode->op[2].offset + offsetof(YMMReg, YMM_X(0)), 16, 16);
tcg_gen_gvec_mov(MO_64,
decode->op[0].offset + offsetof(YMMReg, YMM_X(!mask)),
decode->op[1].offset + offsetof(YMMReg, YMM_X(!mask)), 16, 16);
}
static void gen_VPERM2x128(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
assert(s->vex_l);
gen_helper_vpermdq_ymm(OP_PTR0, OP_PTR1, OP_PTR2, imm);
}
static void gen_VROUNDSD(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
assert(!s->vex_l);
gen_helper_roundsd_xmm(cpu_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
}
static void gen_VROUNDSS(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
TCGv_i32 imm = tcg_constant8u_i32(decode->immediate);
assert(!s->vex_l);
gen_helper_roundss_xmm(cpu_env, OP_PTR0, OP_PTR1, OP_PTR2, imm);
}
static void gen_VZEROALL(DisasContext *s, CPUX86State *env, X86DecodedInsn *decode)
{
TCGv_ptr ptr = tcg_temp_new_ptr();

3
target/i386/tcg/translate.c
View File

@ -4782,7 +4782,8 @@ static bool disas_insn(DisasContext *s, CPUState *cpu)
use_new &= b <= limit;
#endif
if (use_new &&
((b >= 0x150 && b <= 0x17f) ||
(b == 0x13a ||
(b >= 0x150 && b <= 0x17f) ||
(b >= 0x1d0 && b <= 0x1ff))) {
disas_insn_new(s, cpu, b + 0x100);
return s->pc;