qemu/target/i386/ops_sse.h

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

2567 lines
79 KiB
C
Raw Normal View History

/*
* MMX/3DNow!/SSE/SSE2/SSE3/SSSE3/SSE4/PNI support
*
* Copyright (c) 2005 Fabrice Bellard
* Copyright (c) 2008 Intel Corporation <andrew.zaborowski@intel.com>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
*/
#include "crypto/aes.h"
#if SHIFT == 0
#define Reg MMXReg
#define XMM_ONLY(...)
#define B(n) MMX_B(n)
#define W(n) MMX_W(n)
#define L(n) MMX_L(n)
#define Q(n) MMX_Q(n)
#define SUFFIX _mmx
#else
#define Reg ZMMReg
#define XMM_ONLY(...) __VA_ARGS__
#define B(n) ZMM_B(n)
#define W(n) ZMM_W(n)
#define L(n) ZMM_L(n)
#define Q(n) ZMM_Q(n)
#if SHIFT == 1
#define SUFFIX _xmm
#else
#define SUFFIX _ymm
#endif
#endif
#define LANE_WIDTH (SHIFT ? 16 : 8)
#define PACK_WIDTH (LANE_WIDTH / 2)
#if SHIFT == 0
#define FPSRL(x, c) ((x) >> shift)
#define FPSRAW(x, c) ((int16_t)(x) >> shift)
#define FPSRAL(x, c) ((int32_t)(x) >> shift)
#define FPSLL(x, c) ((x) << shift)
#endif
void glue(helper_psrlw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, Reg *c)
{
int shift;
if (c->Q(0) > 15) {
for (int i = 0; i < 1 << SHIFT; i++) {
d->Q(i) = 0;
}
} else {
shift = c->B(0);
for (int i = 0; i < 4 << SHIFT; i++) {
d->W(i) = FPSRL(s->W(i), shift);
}
}
}
void glue(helper_psllw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, Reg *c)
{
int shift;
if (c->Q(0) > 15) {
for (int i = 0; i < 1 << SHIFT; i++) {
d->Q(i) = 0;
}
} else {
shift = c->B(0);
for (int i = 0; i < 4 << SHIFT; i++) {
d->W(i) = FPSLL(s->W(i), shift);
}
}
}
void glue(helper_psraw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, Reg *c)
{
int shift;
if (c->Q(0) > 15) {
shift = 15;
} else {
shift = c->B(0);
}
for (int i = 0; i < 4 << SHIFT; i++) {
d->W(i) = FPSRAW(s->W(i), shift);
}
}
void glue(helper_psrld, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, Reg *c)
{
int shift;
if (c->Q(0) > 31) {
for (int i = 0; i < 1 << SHIFT; i++) {
d->Q(i) = 0;
}
} else {
shift = c->B(0);
for (int i = 0; i < 2 << SHIFT; i++) {
d->L(i) = FPSRL(s->L(i), shift);
}
}
}
void glue(helper_pslld, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, Reg *c)
{
int shift;
if (c->Q(0) > 31) {
for (int i = 0; i < 1 << SHIFT; i++) {
d->Q(i) = 0;
}
} else {
shift = c->B(0);
for (int i = 0; i < 2 << SHIFT; i++) {
d->L(i) = FPSLL(s->L(i), shift);
}
}
}
void glue(helper_psrad, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, Reg *c)
{
int shift;
if (c->Q(0) > 31) {
shift = 31;
} else {
shift = c->B(0);
}
for (int i = 0; i < 2 << SHIFT; i++) {
d->L(i) = FPSRAL(s->L(i), shift);
}
}
void glue(helper_psrlq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, Reg *c)
{
int shift;
if (c->Q(0) > 63) {
for (int i = 0; i < 1 << SHIFT; i++) {
d->Q(i) = 0;
}
} else {
shift = c->B(0);
for (int i = 0; i < 1 << SHIFT; i++) {
d->Q(i) = FPSRL(s->Q(i), shift);
}
}
}
void glue(helper_psllq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, Reg *c)
{
int shift;
if (c->Q(0) > 63) {
for (int i = 0; i < 1 << SHIFT; i++) {
d->Q(i) = 0;
}
} else {
shift = c->B(0);
for (int i = 0; i < 1 << SHIFT; i++) {
d->Q(i) = FPSLL(s->Q(i), shift);
}
}
}
#if SHIFT >= 1
void glue(helper_psrldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, Reg *c)
{
int shift, i, j;
shift = c->L(0);
if (shift > 16) {
shift = 16;
}
for (j = 0; j < 8 << SHIFT; j += LANE_WIDTH) {
for (i = 0; i < 16 - shift; i++) {
d->B(j + i) = s->B(j + i + shift);
}
for (i = 16 - shift; i < 16; i++) {
d->B(j + i) = 0;
}
}
}
void glue(helper_pslldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, Reg *c)
{
int shift, i, j;
shift = c->L(0);
if (shift > 16) {
shift = 16;
}
for (j = 0; j < 8 << SHIFT; j += LANE_WIDTH) {
for (i = 15; i >= shift; i--) {
d->B(j + i) = s->B(j + i - shift);
}
for (i = 0; i < shift; i++) {
d->B(j + i) = 0;
}
}
}
#endif
#define SSE_HELPER_1(name, elem, num, F) \
void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s) \
{ \
int n = num; \
for (int i = 0; i < n; i++) { \
d->elem(i) = F(s->elem(i)); \
} \
}
#define SSE_HELPER_2(name, elem, num, F) \
void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s) \
{ \
int n = num; \
for (int i = 0; i < n; i++) { \
d->elem(i) = F(v->elem(i), s->elem(i)); \
} \
}
#define SSE_HELPER_B(name, F) \
SSE_HELPER_2(name, B, 8 << SHIFT, F)
#define SSE_HELPER_W(name, F) \
SSE_HELPER_2(name, W, 4 << SHIFT, F)
#define SSE_HELPER_L(name, F) \
SSE_HELPER_2(name, L, 2 << SHIFT, F)
#define SSE_HELPER_Q(name, F) \
SSE_HELPER_2(name, Q, 1 << SHIFT, F)
#if SHIFT == 0
static inline int satub(int x)
{
if (x < 0) {
return 0;
} else if (x > 255) {
return 255;
} else {
return x;
}
}
static inline int satuw(int x)
{
if (x < 0) {
return 0;
} else if (x > 65535) {
return 65535;
} else {
return x;
}
}
static inline int satsb(int x)
{
if (x < -128) {
return -128;
} else if (x > 127) {
return 127;
} else {
return x;
}
}
static inline int satsw(int x)
{
if (x < -32768) {
return -32768;
} else if (x > 32767) {
return 32767;
} else {
return x;
}
}
#define FADD(a, b) ((a) + (b))
#define FADDUB(a, b) satub((a) + (b))
#define FADDUW(a, b) satuw((a) + (b))
#define FADDSB(a, b) satsb((int8_t)(a) + (int8_t)(b))
#define FADDSW(a, b) satsw((int16_t)(a) + (int16_t)(b))
#define FSUB(a, b) ((a) - (b))
#define FSUBUB(a, b) satub((a) - (b))
#define FSUBUW(a, b) satuw((a) - (b))
#define FSUBSB(a, b) satsb((int8_t)(a) - (int8_t)(b))
#define FSUBSW(a, b) satsw((int16_t)(a) - (int16_t)(b))
#define FMINUB(a, b) ((a) < (b)) ? (a) : (b)
#define FMINSW(a, b) ((int16_t)(a) < (int16_t)(b)) ? (a) : (b)
#define FMAXUB(a, b) ((a) > (b)) ? (a) : (b)
#define FMAXSW(a, b) ((int16_t)(a) > (int16_t)(b)) ? (a) : (b)
#define FMULHRW(a, b) (((int16_t)(a) * (int16_t)(b) + 0x8000) >> 16)
#define FMULHUW(a, b) ((a) * (b) >> 16)
#define FMULHW(a, b) ((int16_t)(a) * (int16_t)(b) >> 16)
#define FAVG(a, b) (((a) + (b) + 1) >> 1)
#endif
SSE_HELPER_W(helper_pmulhuw, FMULHUW)
SSE_HELPER_W(helper_pmulhw, FMULHW)
#if SHIFT == 0
void glue(helper_pmulhrw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
d->W(0) = FMULHRW(d->W(0), s->W(0));
d->W(1) = FMULHRW(d->W(1), s->W(1));
d->W(2) = FMULHRW(d->W(2), s->W(2));
d->W(3) = FMULHRW(d->W(3), s->W(3));
}
#endif
SSE_HELPER_B(helper_pavgb, FAVG)
SSE_HELPER_W(helper_pavgw, FAVG)
void glue(helper_pmuludq, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
for (i = 0; i < (1 << SHIFT); i++) {
d->Q(i) = (uint64_t)s->L(i * 2) * (uint64_t)v->L(i * 2);
}
}
void glue(helper_pmaddwd, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
for (i = 0; i < (2 << SHIFT); i++) {
d->L(i) = (int16_t)s->W(2 * i) * (int16_t)v->W(2 * i) +
(int16_t)s->W(2 * i + 1) * (int16_t)v->W(2 * i + 1);
}
}
#if SHIFT == 0
static inline int abs1(int a)
{
if (a < 0) {
return -a;
} else {
return a;
}
}
#endif
void glue(helper_psadbw, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
for (i = 0; i < (1 << SHIFT); i++) {
unsigned int val = 0;
val += abs1(v->B(8 * i + 0) - s->B(8 * i + 0));
val += abs1(v->B(8 * i + 1) - s->B(8 * i + 1));
val += abs1(v->B(8 * i + 2) - s->B(8 * i + 2));
val += abs1(v->B(8 * i + 3) - s->B(8 * i + 3));
val += abs1(v->B(8 * i + 4) - s->B(8 * i + 4));
val += abs1(v->B(8 * i + 5) - s->B(8 * i + 5));
val += abs1(v->B(8 * i + 6) - s->B(8 * i + 6));
val += abs1(v->B(8 * i + 7) - s->B(8 * i + 7));
d->Q(i) = val;
}
}
#if SHIFT < 2
void glue(helper_maskmov, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
target_ulong a0)
{
int i;
for (i = 0; i < (8 << SHIFT); i++) {
if (s->B(i) & 0x80) {
cpu_stb_data_ra(env, a0 + i, d->B(i), GETPC());
}
}
}
#endif
#define SHUFFLE4(F, a, b, offset) do { \
r0 = a->F((order & 3) + offset); \
r1 = a->F(((order >> 2) & 3) + offset); \
r2 = b->F(((order >> 4) & 3) + offset); \
r3 = b->F(((order >> 6) & 3) + offset); \
d->F(offset) = r0; \
d->F(offset + 1) = r1; \
d->F(offset + 2) = r2; \
d->F(offset + 3) = r3; \
} while (0)
#if SHIFT == 0
void glue(helper_pshufw, SUFFIX)(Reg *d, Reg *s, int order)
{
uint16_t r0, r1, r2, r3;
SHUFFLE4(W, s, s, 0);
}
#else
void glue(helper_shufps, SUFFIX)(Reg *d, Reg *v, Reg *s, int order)
{
uint32_t r0, r1, r2, r3;
int i;
for (i = 0; i < 2 << SHIFT; i += 4) {
SHUFFLE4(L, v, s, i);
}
}
void glue(helper_shufpd, SUFFIX)(Reg *d, Reg *v, Reg *s, int order)
{
uint64_t r0, r1;
int i;
for (i = 0; i < 1 << SHIFT; i += 2) {
r0 = v->Q(((order & 1) & 1) + i);
r1 = s->Q(((order >> 1) & 1) + i);
d->Q(i) = r0;
d->Q(i + 1) = r1;
order >>= 2;
}
}
void glue(helper_pshufd, SUFFIX)(Reg *d, Reg *s, int order)
{
uint32_t r0, r1, r2, r3;
int i;
for (i = 0; i < 2 << SHIFT; i += 4) {
SHUFFLE4(L, s, s, i);
}
}
void glue(helper_pshuflw, SUFFIX)(Reg *d, Reg *s, int order)
{
uint16_t r0, r1, r2, r3;
int i, j;
for (i = 0, j = 1; j < 1 << SHIFT; i += 8, j += 2) {
SHUFFLE4(W, s, s, i);
d->Q(j) = s->Q(j);
}
}
void glue(helper_pshufhw, SUFFIX)(Reg *d, Reg *s, int order)
{
uint16_t r0, r1, r2, r3;
int i, j;
for (i = 4, j = 0; j < 1 << SHIFT; i += 8, j += 2) {
d->Q(j) = s->Q(j);
SHUFFLE4(W, s, s, i);
}
}
#endif
#if SHIFT >= 1
/* FPU ops */
/* XXX: not accurate */
#define SSE_HELPER_P(name, F) \
void glue(helper_ ## name ## ps, SUFFIX)(CPUX86State *env, \
Reg *d, Reg *v, Reg *s) \
{ \
int i; \
for (i = 0; i < 2 << SHIFT; i++) { \
d->ZMM_S(i) = F(32, v->ZMM_S(i), s->ZMM_S(i)); \
} \
} \
\
void glue(helper_ ## name ## pd, SUFFIX)(CPUX86State *env, \
Reg *d, Reg *v, Reg *s) \
{ \
int i; \
for (i = 0; i < 1 << SHIFT; i++) { \
d->ZMM_D(i) = F(64, v->ZMM_D(i), s->ZMM_D(i)); \
} \
}
#if SHIFT == 1
#define SSE_HELPER_S(name, F) \
SSE_HELPER_P(name, F) \
\
void helper_ ## name ## ss(CPUX86State *env, Reg *d, Reg *v, Reg *s)\
{ \
int i; \
d->ZMM_S(0) = F(32, v->ZMM_S(0), s->ZMM_S(0)); \
for (i = 1; i < 2 << SHIFT; i++) { \
d->ZMM_L(i) = v->ZMM_L(i); \
} \
} \
\
void helper_ ## name ## sd(CPUX86State *env, Reg *d, Reg *v, Reg *s)\
{ \
int i; \
d->ZMM_D(0) = F(64, v->ZMM_D(0), s->ZMM_D(0)); \
for (i = 1; i < 1 << SHIFT; i++) { \
d->ZMM_Q(i) = v->ZMM_Q(i); \
} \
}
#else
#define SSE_HELPER_S(name, F) SSE_HELPER_P(name, F)
#endif
#define FPU_ADD(size, a, b) float ## size ## _add(a, b, &env->sse_status)
#define FPU_SUB(size, a, b) float ## size ## _sub(a, b, &env->sse_status)
#define FPU_MUL(size, a, b) float ## size ## _mul(a, b, &env->sse_status)
#define FPU_DIV(size, a, b) float ## size ## _div(a, b, &env->sse_status)
/* Note that the choice of comparison op here is important to get the
* special cases right: for min and max Intel specifies that (-0,0),
* (NaN, anything) and (anything, NaN) return the second argument.
*/
#define FPU_MIN(size, a, b) \
(float ## size ## _lt(a, b, &env->sse_status) ? (a) : (b))
#define FPU_MAX(size, a, b) \
(float ## size ## _lt(b, a, &env->sse_status) ? (a) : (b))
SSE_HELPER_S(add, FPU_ADD)
SSE_HELPER_S(sub, FPU_SUB)
SSE_HELPER_S(mul, FPU_MUL)
SSE_HELPER_S(div, FPU_DIV)
SSE_HELPER_S(min, FPU_MIN)
SSE_HELPER_S(max, FPU_MAX)
void glue(helper_sqrtps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
int i;
for (i = 0; i < 2 << SHIFT; i++) {
d->ZMM_S(i) = float32_sqrt(s->ZMM_S(i), &env->sse_status);
}
}
void glue(helper_sqrtpd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
int i;
for (i = 0; i < 1 << SHIFT; i++) {
d->ZMM_D(i) = float64_sqrt(s->ZMM_D(i), &env->sse_status);
}
}
#if SHIFT == 1
void helper_sqrtss(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
d->ZMM_S(0) = float32_sqrt(s->ZMM_S(0), &env->sse_status);
for (i = 1; i < 2 << SHIFT; i++) {
d->ZMM_L(i) = v->ZMM_L(i);
}
}
void helper_sqrtsd(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
d->ZMM_D(0) = float64_sqrt(s->ZMM_D(0), &env->sse_status);
for (i = 1; i < 1 << SHIFT; i++) {
d->ZMM_Q(i) = v->ZMM_Q(i);
}
}
#endif
/* float to float conversions */
void glue(helper_cvtps2pd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
int i;
for (i = 1 << SHIFT; --i >= 0; ) {
d->ZMM_D(i) = float32_to_float64(s->ZMM_S(i), &env->sse_status);
}
}
void glue(helper_cvtpd2ps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
int i;
for (i = 0; i < 1 << SHIFT; i++) {
d->ZMM_S(i) = float64_to_float32(s->ZMM_D(i), &env->sse_status);
}
for (i >>= 1; i < 1 << SHIFT; i++) {
d->Q(i) = 0;
}
}
#if SHIFT >= 1
void glue(helper_cvtph2ps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
int i;
for (i = 2 << SHIFT; --i >= 0; ) {
d->ZMM_S(i) = float16_to_float32(s->ZMM_H(i), true, &env->sse_status);
}
}
void glue(helper_cvtps2ph, SUFFIX)(CPUX86State *env, Reg *d, Reg *s, int mode)
{
int i;
FloatRoundMode prev_rounding_mode = env->sse_status.float_rounding_mode;
if (!(mode & (1 << 2))) {
set_x86_rounding_mode(mode & 3, &env->sse_status);
}
for (i = 0; i < 2 << SHIFT; i++) {
d->ZMM_H(i) = float32_to_float16(s->ZMM_S(i), true, &env->sse_status);
}
for (i >>= 2; i < 1 << SHIFT; i++) {
d->Q(i) = 0;
}
env->sse_status.float_rounding_mode = prev_rounding_mode;
}
#endif
#if SHIFT == 1
void helper_cvtss2sd(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
d->ZMM_D(0) = float32_to_float64(s->ZMM_S(0), &env->sse_status);
for (i = 1; i < 1 << SHIFT; i++) {
d->ZMM_Q(i) = v->ZMM_Q(i);
}
}
void helper_cvtsd2ss(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
d->ZMM_S(0) = float64_to_float32(s->ZMM_D(0), &env->sse_status);
for (i = 1; i < 2 << SHIFT; i++) {
d->ZMM_L(i) = v->ZMM_L(i);
}
}
#endif
/* integer to float */
void glue(helper_cvtdq2ps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
int i;
for (i = 0; i < 2 << SHIFT; i++) {
d->ZMM_S(i) = int32_to_float32(s->ZMM_L(i), &env->sse_status);
}
}
void glue(helper_cvtdq2pd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
int i;
for (i = 1 << SHIFT; --i >= 0; ) {
int32_t l = s->ZMM_L(i);
d->ZMM_D(i) = int32_to_float64(l, &env->sse_status);
}
}
#if SHIFT == 1
void helper_cvtpi2ps(CPUX86State *env, ZMMReg *d, MMXReg *s)
{
d->ZMM_S(0) = int32_to_float32(s->MMX_L(0), &env->sse_status);
d->ZMM_S(1) = int32_to_float32(s->MMX_L(1), &env->sse_status);
}
void helper_cvtpi2pd(CPUX86State *env, ZMMReg *d, MMXReg *s)
{
d->ZMM_D(0) = int32_to_float64(s->MMX_L(0), &env->sse_status);
d->ZMM_D(1) = int32_to_float64(s->MMX_L(1), &env->sse_status);
}
void helper_cvtsi2ss(CPUX86State *env, ZMMReg *d, uint32_t val)
{
d->ZMM_S(0) = int32_to_float32(val, &env->sse_status);
}
void helper_cvtsi2sd(CPUX86State *env, ZMMReg *d, uint32_t val)
{
d->ZMM_D(0) = int32_to_float64(val, &env->sse_status);
}
#ifdef TARGET_X86_64
void helper_cvtsq2ss(CPUX86State *env, ZMMReg *d, uint64_t val)
{
d->ZMM_S(0) = int64_to_float32(val, &env->sse_status);
}
void helper_cvtsq2sd(CPUX86State *env, ZMMReg *d, uint64_t val)
{
d->ZMM_D(0) = int64_to_float64(val, &env->sse_status);
}
#endif
#endif
/* float to integer */
#if SHIFT == 1
/*
* x86 mandates that we return the indefinite integer value for the result
* of any float-to-integer conversion that raises the 'invalid' exception.
* Wrap the softfloat functions to get this behaviour.
*/
#define WRAP_FLOATCONV(RETTYPE, FN, FLOATTYPE, INDEFVALUE) \
static inline RETTYPE x86_##FN(FLOATTYPE a, float_status *s) \
{ \
int oldflags, newflags; \
RETTYPE r; \
\
oldflags = get_float_exception_flags(s); \
set_float_exception_flags(0, s); \
r = FN(a, s); \
newflags = get_float_exception_flags(s); \
if (newflags & float_flag_invalid) { \
r = INDEFVALUE; \
} \
set_float_exception_flags(newflags | oldflags, s); \
return r; \
}
WRAP_FLOATCONV(int32_t, float32_to_int32, float32, INT32_MIN)
WRAP_FLOATCONV(int32_t, float32_to_int32_round_to_zero, float32, INT32_MIN)
WRAP_FLOATCONV(int32_t, float64_to_int32, float64, INT32_MIN)
WRAP_FLOATCONV(int32_t, float64_to_int32_round_to_zero, float64, INT32_MIN)
WRAP_FLOATCONV(int64_t, float32_to_int64, float32, INT64_MIN)
WRAP_FLOATCONV(int64_t, float32_to_int64_round_to_zero, float32, INT64_MIN)
WRAP_FLOATCONV(int64_t, float64_to_int64, float64, INT64_MIN)
WRAP_FLOATCONV(int64_t, float64_to_int64_round_to_zero, float64, INT64_MIN)
#endif
void glue(helper_cvtps2dq, SUFFIX)(CPUX86State *env, ZMMReg *d, ZMMReg *s)
{
int i;
for (i = 0; i < 2 << SHIFT; i++) {
d->ZMM_L(i) = x86_float32_to_int32(s->ZMM_S(i), &env->sse_status);
}
}
void glue(helper_cvtpd2dq, SUFFIX)(CPUX86State *env, ZMMReg *d, ZMMReg *s)
{
int i;
for (i = 0; i < 1 << SHIFT; i++) {
d->ZMM_L(i) = x86_float64_to_int32(s->ZMM_D(i), &env->sse_status);
}
for (i >>= 1; i < 1 << SHIFT; i++) {
d->Q(i) = 0;
}
}
#if SHIFT == 1
void helper_cvtps2pi(CPUX86State *env, MMXReg *d, ZMMReg *s)
{
d->MMX_L(0) = x86_float32_to_int32(s->ZMM_S(0), &env->sse_status);
d->MMX_L(1) = x86_float32_to_int32(s->ZMM_S(1), &env->sse_status);
}
void helper_cvtpd2pi(CPUX86State *env, MMXReg *d, ZMMReg *s)
{
d->MMX_L(0) = x86_float64_to_int32(s->ZMM_D(0), &env->sse_status);
d->MMX_L(1) = x86_float64_to_int32(s->ZMM_D(1), &env->sse_status);
}
int32_t helper_cvtss2si(CPUX86State *env, ZMMReg *s)
{
return x86_float32_to_int32(s->ZMM_S(0), &env->sse_status);
}
int32_t helper_cvtsd2si(CPUX86State *env, ZMMReg *s)
{
return x86_float64_to_int32(s->ZMM_D(0), &env->sse_status);
}
#ifdef TARGET_X86_64
int64_t helper_cvtss2sq(CPUX86State *env, ZMMReg *s)
{
return x86_float32_to_int64(s->ZMM_S(0), &env->sse_status);
}
int64_t helper_cvtsd2sq(CPUX86State *env, ZMMReg *s)
{
return x86_float64_to_int64(s->ZMM_D(0), &env->sse_status);
}
#endif
#endif
/* float to integer truncated */
void glue(helper_cvttps2dq, SUFFIX)(CPUX86State *env, ZMMReg *d, ZMMReg *s)
{
int i;
for (i = 0; i < 2 << SHIFT; i++) {
d->ZMM_L(i) = x86_float32_to_int32_round_to_zero(s->ZMM_S(i),
&env->sse_status);
}
}
void glue(helper_cvttpd2dq, SUFFIX)(CPUX86State *env, ZMMReg *d, ZMMReg *s)
{
int i;
for (i = 0; i < 1 << SHIFT; i++) {
d->ZMM_L(i) = x86_float64_to_int32_round_to_zero(s->ZMM_D(i),
&env->sse_status);
}
for (i >>= 1; i < 1 << SHIFT; i++) {
d->Q(i) = 0;
}
}
#if SHIFT == 1
void helper_cvttps2pi(CPUX86State *env, MMXReg *d, ZMMReg *s)
{
d->MMX_L(0) = x86_float32_to_int32_round_to_zero(s->ZMM_S(0), &env->sse_status);
d->MMX_L(1) = x86_float32_to_int32_round_to_zero(s->ZMM_S(1), &env->sse_status);
}
void helper_cvttpd2pi(CPUX86State *env, MMXReg *d, ZMMReg *s)
{
d->MMX_L(0) = x86_float64_to_int32_round_to_zero(s->ZMM_D(0), &env->sse_status);
d->MMX_L(1) = x86_float64_to_int32_round_to_zero(s->ZMM_D(1), &env->sse_status);
}
int32_t helper_cvttss2si(CPUX86State *env, ZMMReg *s)
{
return x86_float32_to_int32_round_to_zero(s->ZMM_S(0), &env->sse_status);
}
int32_t helper_cvttsd2si(CPUX86State *env, ZMMReg *s)
{
return x86_float64_to_int32_round_to_zero(s->ZMM_D(0), &env->sse_status);
}
#ifdef TARGET_X86_64
int64_t helper_cvttss2sq(CPUX86State *env, ZMMReg *s)
{
return x86_float32_to_int64_round_to_zero(s->ZMM_S(0), &env->sse_status);
}
int64_t helper_cvttsd2sq(CPUX86State *env, ZMMReg *s)
{
return x86_float64_to_int64_round_to_zero(s->ZMM_D(0), &env->sse_status);
}
#endif
#endif
void glue(helper_rsqrtps, SUFFIX)(CPUX86State *env, ZMMReg *d, ZMMReg *s)
{
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
uint8_t old_flags = get_float_exception_flags(&env->sse_status);
int i;
for (i = 0; i < 2 << SHIFT; i++) {
d->ZMM_S(i) = float32_div(float32_one,
float32_sqrt(s->ZMM_S(i), &env->sse_status),
&env->sse_status);
}
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
set_float_exception_flags(old_flags, &env->sse_status);
}
#if SHIFT == 1
void helper_rsqrtss(CPUX86State *env, ZMMReg *d, ZMMReg *v, ZMMReg *s)
{
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
uint8_t old_flags = get_float_exception_flags(&env->sse_status);
int i;
d->ZMM_S(0) = float32_div(float32_one,
float32_sqrt(s->ZMM_S(0), &env->sse_status),
&env->sse_status);
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
set_float_exception_flags(old_flags, &env->sse_status);
for (i = 1; i < 2 << SHIFT; i++) {
d->ZMM_L(i) = v->ZMM_L(i);
}
}
#endif
void glue(helper_rcpps, SUFFIX)(CPUX86State *env, ZMMReg *d, ZMMReg *s)
{
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
uint8_t old_flags = get_float_exception_flags(&env->sse_status);
int i;
for (i = 0; i < 2 << SHIFT; i++) {
d->ZMM_S(i) = float32_div(float32_one, s->ZMM_S(i), &env->sse_status);
}
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
set_float_exception_flags(old_flags, &env->sse_status);
}
#if SHIFT == 1
void helper_rcpss(CPUX86State *env, ZMMReg *d, ZMMReg *v, ZMMReg *s)
{
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
uint8_t old_flags = get_float_exception_flags(&env->sse_status);
int i;
d->ZMM_S(0) = float32_div(float32_one, s->ZMM_S(0), &env->sse_status);
for (i = 1; i < 2 << SHIFT; i++) {
d->ZMM_L(i) = v->ZMM_L(i);
}
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
set_float_exception_flags(old_flags, &env->sse_status);
}
#endif
#if SHIFT == 1
static inline uint64_t helper_extrq(uint64_t src, int shift, int len)
{
uint64_t mask;
if (len == 0) {
mask = ~0LL;
} else {
mask = (1ULL << len) - 1;
}
return (src >> shift) & mask;
}
void helper_extrq_r(CPUX86State *env, ZMMReg *d, ZMMReg *s)
{
d->ZMM_Q(0) = helper_extrq(d->ZMM_Q(0), s->ZMM_B(1) & 63, s->ZMM_B(0) & 63);
}
void helper_extrq_i(CPUX86State *env, ZMMReg *d, int index, int length)
{
d->ZMM_Q(0) = helper_extrq(d->ZMM_Q(0), index, length);
}
static inline uint64_t helper_insertq(uint64_t dest, uint64_t src, int shift, int len)
{
uint64_t mask;
if (len == 0) {
mask = ~0ULL;
} else {
mask = (1ULL << len) - 1;
}
return (dest & ~(mask << shift)) | ((src & mask) << shift);
}
void helper_insertq_r(CPUX86State *env, ZMMReg *d, ZMMReg *s)
{
d->ZMM_Q(0) = helper_insertq(d->ZMM_Q(0), s->ZMM_Q(0), s->ZMM_B(9) & 63, s->ZMM_B(8) & 63);
}
void helper_insertq_i(CPUX86State *env, ZMMReg *d, ZMMReg *s, int index, int length)
{
d->ZMM_Q(0) = helper_insertq(d->ZMM_Q(0), s->ZMM_Q(0), index, length);
}
#endif
#define SSE_HELPER_HPS(name, F) \
void glue(helper_ ## name, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s) \
{ \
float32 r[2 << SHIFT]; \
int i, j, k; \
for (k = 0; k < 2 << SHIFT; k += LANE_WIDTH / 4) { \
for (i = j = 0; j < 4; i++, j += 2) { \
r[i + k] = F(v->ZMM_S(j + k), v->ZMM_S(j + k + 1), &env->sse_status); \
} \
for (j = 0; j < 4; i++, j += 2) { \
r[i + k] = F(s->ZMM_S(j + k), s->ZMM_S(j + k + 1), &env->sse_status); \
} \
} \
for (i = 0; i < 2 << SHIFT; i++) { \
d->ZMM_S(i) = r[i]; \
} \
}
SSE_HELPER_HPS(haddps, float32_add)
SSE_HELPER_HPS(hsubps, float32_sub)
#define SSE_HELPER_HPD(name, F) \
void glue(helper_ ## name, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s) \
{ \
float64 r[1 << SHIFT]; \
int i, j, k; \
for (k = 0; k < 1 << SHIFT; k += LANE_WIDTH / 8) { \
for (i = j = 0; j < 2; i++, j += 2) { \
r[i + k] = F(v->ZMM_D(j + k), v->ZMM_D(j + k + 1), &env->sse_status); \
} \
for (j = 0; j < 2; i++, j += 2) { \
r[i + k] = F(s->ZMM_D(j + k), s->ZMM_D(j + k + 1), &env->sse_status); \
} \
} \
for (i = 0; i < 1 << SHIFT; i++) { \
d->ZMM_D(i) = r[i]; \
} \
}
SSE_HELPER_HPD(haddpd, float64_add)
SSE_HELPER_HPD(hsubpd, float64_sub)
void glue(helper_addsubps, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
for (i = 0; i < 2 << SHIFT; i += 2) {
d->ZMM_S(i) = float32_sub(v->ZMM_S(i), s->ZMM_S(i), &env->sse_status);
d->ZMM_S(i+1) = float32_add(v->ZMM_S(i+1), s->ZMM_S(i+1), &env->sse_status);
}
}
void glue(helper_addsubpd, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
for (i = 0; i < 1 << SHIFT; i += 2) {
d->ZMM_D(i) = float64_sub(v->ZMM_D(i), s->ZMM_D(i), &env->sse_status);
d->ZMM_D(i+1) = float64_add(v->ZMM_D(i+1), s->ZMM_D(i+1), &env->sse_status);
}
}
#define SSE_HELPER_CMP_P(name, F, C) \
void glue(helper_ ## name ## ps, SUFFIX)(CPUX86State *env, \
Reg *d, Reg *v, Reg *s) \
{ \
int i; \
for (i = 0; i < 2 << SHIFT; i++) { \
d->ZMM_L(i) = C(F(32, v->ZMM_S(i), s->ZMM_S(i))) ? -1 : 0; \
} \
} \
\
void glue(helper_ ## name ## pd, SUFFIX)(CPUX86State *env, \
Reg *d, Reg *v, Reg *s) \
{ \
int i; \
for (i = 0; i < 1 << SHIFT; i++) { \
d->ZMM_Q(i) = C(F(64, v->ZMM_D(i), s->ZMM_D(i))) ? -1 : 0; \
} \
}
#if SHIFT == 1
#define SSE_HELPER_CMP(name, F, C) \
SSE_HELPER_CMP_P(name, F, C) \
void helper_ ## name ## ss(CPUX86State *env, Reg *d, Reg *v, Reg *s) \
{ \
int i; \
d->ZMM_L(0) = C(F(32, v->ZMM_S(0), s->ZMM_S(0))) ? -1 : 0; \
for (i = 1; i < 2 << SHIFT; i++) { \
d->ZMM_L(i) = v->ZMM_L(i); \
} \
} \
\
void helper_ ## name ## sd(CPUX86State *env, Reg *d, Reg *v, Reg *s) \
{ \
int i; \
d->ZMM_Q(0) = C(F(64, v->ZMM_D(0), s->ZMM_D(0))) ? -1 : 0; \
for (i = 1; i < 1 << SHIFT; i++) { \
d->ZMM_Q(i) = v->ZMM_Q(i); \
} \
}
static inline bool FPU_EQU(FloatRelation x)
{
return (x == float_relation_equal || x == float_relation_unordered);
}
static inline bool FPU_GE(FloatRelation x)
{
return (x == float_relation_equal || x == float_relation_greater);
}
#define FPU_EQ(x) (x == float_relation_equal)
#define FPU_LT(x) (x == float_relation_less)
#define FPU_LE(x) (x <= float_relation_equal)
#define FPU_GT(x) (x == float_relation_greater)
#define FPU_UNORD(x) (x == float_relation_unordered)
/* We must make sure we evaluate the argument in case it is a signalling NAN */
#define FPU_FALSE(x) (x == float_relation_equal && 0)
#define FPU_CMPQ(size, a, b) \
float ## size ## _compare_quiet(a, b, &env->sse_status)
#define FPU_CMPS(size, a, b) \
float ## size ## _compare(a, b, &env->sse_status)
#else
#define SSE_HELPER_CMP(name, F, C) SSE_HELPER_CMP_P(name, F, C)
#endif
SSE_HELPER_CMP(cmpeq, FPU_CMPQ, FPU_EQ)
SSE_HELPER_CMP(cmplt, FPU_CMPS, FPU_LT)
SSE_HELPER_CMP(cmple, FPU_CMPS, FPU_LE)
SSE_HELPER_CMP(cmpunord, FPU_CMPQ, FPU_UNORD)
SSE_HELPER_CMP(cmpneq, FPU_CMPQ, !FPU_EQ)
SSE_HELPER_CMP(cmpnlt, FPU_CMPS, !FPU_LT)
SSE_HELPER_CMP(cmpnle, FPU_CMPS, !FPU_LE)
SSE_HELPER_CMP(cmpord, FPU_CMPQ, !FPU_UNORD)
SSE_HELPER_CMP(cmpequ, FPU_CMPQ, FPU_EQU)
SSE_HELPER_CMP(cmpnge, FPU_CMPS, !FPU_GE)
SSE_HELPER_CMP(cmpngt, FPU_CMPS, !FPU_GT)
SSE_HELPER_CMP(cmpfalse, FPU_CMPQ, FPU_FALSE)
SSE_HELPER_CMP(cmpnequ, FPU_CMPQ, !FPU_EQU)
SSE_HELPER_CMP(cmpge, FPU_CMPS, FPU_GE)
SSE_HELPER_CMP(cmpgt, FPU_CMPS, FPU_GT)
SSE_HELPER_CMP(cmptrue, FPU_CMPQ, !FPU_FALSE)
SSE_HELPER_CMP(cmpeqs, FPU_CMPS, FPU_EQ)
SSE_HELPER_CMP(cmpltq, FPU_CMPQ, FPU_LT)
SSE_HELPER_CMP(cmpleq, FPU_CMPQ, FPU_LE)
SSE_HELPER_CMP(cmpunords, FPU_CMPS, FPU_UNORD)
SSE_HELPER_CMP(cmpneqq, FPU_CMPS, !FPU_EQ)
SSE_HELPER_CMP(cmpnltq, FPU_CMPQ, !FPU_LT)
SSE_HELPER_CMP(cmpnleq, FPU_CMPQ, !FPU_LE)
SSE_HELPER_CMP(cmpords, FPU_CMPS, !FPU_UNORD)
SSE_HELPER_CMP(cmpequs, FPU_CMPS, FPU_EQU)
SSE_HELPER_CMP(cmpngeq, FPU_CMPQ, !FPU_GE)
SSE_HELPER_CMP(cmpngtq, FPU_CMPQ, !FPU_GT)
SSE_HELPER_CMP(cmpfalses, FPU_CMPS, FPU_FALSE)
SSE_HELPER_CMP(cmpnequs, FPU_CMPS, !FPU_EQU)
SSE_HELPER_CMP(cmpgeq, FPU_CMPQ, FPU_GE)
SSE_HELPER_CMP(cmpgtq, FPU_CMPQ, FPU_GT)
SSE_HELPER_CMP(cmptrues, FPU_CMPS, !FPU_FALSE)
#undef SSE_HELPER_CMP
#if SHIFT == 1
static const int comis_eflags[4] = {CC_C, CC_Z, 0, CC_Z | CC_P | CC_C};
void helper_ucomiss(CPUX86State *env, Reg *d, Reg *s)
{
FloatRelation ret;
float32 s0, s1;
s0 = d->ZMM_S(0);
s1 = s->ZMM_S(0);
ret = float32_compare_quiet(s0, s1, &env->sse_status);
CC_SRC = comis_eflags[ret + 1];
}
void helper_comiss(CPUX86State *env, Reg *d, Reg *s)
{
FloatRelation ret;
float32 s0, s1;
s0 = d->ZMM_S(0);
s1 = s->ZMM_S(0);
ret = float32_compare(s0, s1, &env->sse_status);
CC_SRC = comis_eflags[ret + 1];
}
void helper_ucomisd(CPUX86State *env, Reg *d, Reg *s)
{
FloatRelation ret;
float64 d0, d1;
d0 = d->ZMM_D(0);
d1 = s->ZMM_D(0);
ret = float64_compare_quiet(d0, d1, &env->sse_status);
CC_SRC = comis_eflags[ret + 1];
}
void helper_comisd(CPUX86State *env, Reg *d, Reg *s)
{
FloatRelation ret;
float64 d0, d1;
d0 = d->ZMM_D(0);
d1 = s->ZMM_D(0);
ret = float64_compare(d0, d1, &env->sse_status);
CC_SRC = comis_eflags[ret + 1];
}
#endif
uint32_t glue(helper_movmskps, SUFFIX)(CPUX86State *env, Reg *s)
{
uint32_t mask;
int i;
mask = 0;
for (i = 0; i < 2 << SHIFT; i++) {
mask |= (s->ZMM_L(i) >> (31 - i)) & (1 << i);
}
return mask;
}
uint32_t glue(helper_movmskpd, SUFFIX)(CPUX86State *env, Reg *s)
{
uint32_t mask;
int i;
mask = 0;
for (i = 0; i < 1 << SHIFT; i++) {
mask |= (s->ZMM_Q(i) >> (63 - i)) & (1 << i);
}
return mask;
}
#endif
#define PACK_HELPER_B(name, F) \
void glue(helper_pack ## name, SUFFIX)(CPUX86State *env, \
Reg *d, Reg *v, Reg *s) \
{ \
uint8_t r[PACK_WIDTH * 2]; \
int j, k; \
for (j = 0; j < 4 << SHIFT; j += PACK_WIDTH) { \
for (k = 0; k < PACK_WIDTH; k++) { \
r[k] = F((int16_t)v->W(j + k)); \
} \
for (k = 0; k < PACK_WIDTH; k++) { \
r[PACK_WIDTH + k] = F((int16_t)s->W(j + k)); \
} \
for (k = 0; k < PACK_WIDTH * 2; k++) { \
d->B(2 * j + k) = r[k]; \
} \
} \
}
PACK_HELPER_B(sswb, satsb)
PACK_HELPER_B(uswb, satub)
void glue(helper_packssdw, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
uint16_t r[PACK_WIDTH];
int j, k;
for (j = 0; j < 2 << SHIFT; j += PACK_WIDTH / 2) {
for (k = 0; k < PACK_WIDTH / 2; k++) {
r[k] = satsw(v->L(j + k));
}
for (k = 0; k < PACK_WIDTH / 2; k++) {
r[PACK_WIDTH / 2 + k] = satsw(s->L(j + k));
}
for (k = 0; k < PACK_WIDTH; k++) {
d->W(2 * j + k) = r[k];
}
}
}
#define UNPCK_OP(base_name, base) \
\
void glue(helper_punpck ## base_name ## bw, SUFFIX)(CPUX86State *env,\
Reg *d, Reg *v, Reg *s) \
{ \
uint8_t r[PACK_WIDTH * 2]; \
int j, i; \
\
for (j = 0; j < 8 << SHIFT; ) { \
int k = j + base * PACK_WIDTH; \
for (i = 0; i < PACK_WIDTH; i++) { \
r[2 * i] = v->B(k + i); \
r[2 * i + 1] = s->B(k + i); \
} \
for (i = 0; i < PACK_WIDTH * 2; i++, j++) { \
d->B(j) = r[i]; \
} \
} \
} \
\
void glue(helper_punpck ## base_name ## wd, SUFFIX)(CPUX86State *env,\
Reg *d, Reg *v, Reg *s) \
{ \
uint16_t r[PACK_WIDTH]; \
int j, i; \
\
for (j = 0; j < 4 << SHIFT; ) { \
int k = j + base * PACK_WIDTH / 2; \
for (i = 0; i < PACK_WIDTH / 2; i++) { \
r[2 * i] = v->W(k + i); \
r[2 * i + 1] = s->W(k + i); \
} \
for (i = 0; i < PACK_WIDTH; i++, j++) { \
d->W(j) = r[i]; \
} \
} \
} \
\
void glue(helper_punpck ## base_name ## dq, SUFFIX)(CPUX86State *env,\
Reg *d, Reg *v, Reg *s) \
{ \
uint32_t r[PACK_WIDTH / 2]; \
int j, i; \
\
for (j = 0; j < 2 << SHIFT; ) { \
int k = j + base * PACK_WIDTH / 4; \
for (i = 0; i < PACK_WIDTH / 4; i++) { \
r[2 * i] = v->L(k + i); \
r[2 * i + 1] = s->L(k + i); \
} \
for (i = 0; i < PACK_WIDTH / 2; i++, j++) { \
d->L(j) = r[i]; \
} \
} \
} \
\
XMM_ONLY( \
void glue(helper_punpck ## base_name ## qdq, SUFFIX)( \
CPUX86State *env, Reg *d, Reg *v, Reg *s) \
{ \
uint64_t r[2]; \
int i; \
\
for (i = 0; i < 1 << SHIFT; i += 2) { \
r[0] = v->Q(base + i); \
r[1] = s->Q(base + i); \
d->Q(i) = r[0]; \
d->Q(i + 1) = r[1]; \
} \
} \
)
UNPCK_OP(l, 0)
UNPCK_OP(h, 1)
#undef PACK_WIDTH
#undef PACK_HELPER_B
#undef UNPCK_OP
/* 3DNow! float ops */
#if SHIFT == 0
void helper_pi2fd(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_S(0) = int32_to_float32(s->MMX_L(0), &env->mmx_status);
d->MMX_S(1) = int32_to_float32(s->MMX_L(1), &env->mmx_status);
}
void helper_pi2fw(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_S(0) = int32_to_float32((int16_t)s->MMX_W(0), &env->mmx_status);
d->MMX_S(1) = int32_to_float32((int16_t)s->MMX_W(2), &env->mmx_status);
}
void helper_pf2id(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_L(0) = float32_to_int32_round_to_zero(s->MMX_S(0), &env->mmx_status);
d->MMX_L(1) = float32_to_int32_round_to_zero(s->MMX_S(1), &env->mmx_status);
}
void helper_pf2iw(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_L(0) = satsw(float32_to_int32_round_to_zero(s->MMX_S(0),
&env->mmx_status));
d->MMX_L(1) = satsw(float32_to_int32_round_to_zero(s->MMX_S(1),
&env->mmx_status));
}
void helper_pfacc(CPUX86State *env, MMXReg *d, MMXReg *s)
{
float32 r;
r = float32_add(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
d->MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
d->MMX_S(0) = r;
}
void helper_pfadd(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_S(0) = float32_add(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
d->MMX_S(1) = float32_add(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
}
void helper_pfcmpeq(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_L(0) = float32_eq_quiet(d->MMX_S(0), s->MMX_S(0),
&env->mmx_status) ? -1 : 0;
d->MMX_L(1) = float32_eq_quiet(d->MMX_S(1), s->MMX_S(1),
&env->mmx_status) ? -1 : 0;
}
void helper_pfcmpge(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_L(0) = float32_le(s->MMX_S(0), d->MMX_S(0),
&env->mmx_status) ? -1 : 0;
d->MMX_L(1) = float32_le(s->MMX_S(1), d->MMX_S(1),
&env->mmx_status) ? -1 : 0;
}
void helper_pfcmpgt(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_L(0) = float32_lt(s->MMX_S(0), d->MMX_S(0),
&env->mmx_status) ? -1 : 0;
d->MMX_L(1) = float32_lt(s->MMX_S(1), d->MMX_S(1),
&env->mmx_status) ? -1 : 0;
}
void helper_pfmax(CPUX86State *env, MMXReg *d, MMXReg *s)
{
if (float32_lt(d->MMX_S(0), s->MMX_S(0), &env->mmx_status)) {
d->MMX_S(0) = s->MMX_S(0);
}
if (float32_lt(d->MMX_S(1), s->MMX_S(1), &env->mmx_status)) {
d->MMX_S(1) = s->MMX_S(1);
}
}
void helper_pfmin(CPUX86State *env, MMXReg *d, MMXReg *s)
{
if (float32_lt(s->MMX_S(0), d->MMX_S(0), &env->mmx_status)) {
d->MMX_S(0) = s->MMX_S(0);
}
if (float32_lt(s->MMX_S(1), d->MMX_S(1), &env->mmx_status)) {
d->MMX_S(1) = s->MMX_S(1);
}
}
void helper_pfmul(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_S(0) = float32_mul(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
d->MMX_S(1) = float32_mul(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
}
void helper_pfnacc(CPUX86State *env, MMXReg *d, MMXReg *s)
{
float32 r;
r = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
d->MMX_S(1) = float32_sub(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
d->MMX_S(0) = r;
}
void helper_pfpnacc(CPUX86State *env, MMXReg *d, MMXReg *s)
{
float32 r;
r = float32_sub(d->MMX_S(0), d->MMX_S(1), &env->mmx_status);
d->MMX_S(1) = float32_add(s->MMX_S(0), s->MMX_S(1), &env->mmx_status);
d->MMX_S(0) = r;
}
void helper_pfrcp(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_S(0) = float32_div(float32_one, s->MMX_S(0), &env->mmx_status);
d->MMX_S(1) = d->MMX_S(0);
}
void helper_pfrsqrt(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_L(1) = s->MMX_L(0) & 0x7fffffff;
d->MMX_S(1) = float32_div(float32_one,
float32_sqrt(d->MMX_S(1), &env->mmx_status),
&env->mmx_status);
d->MMX_L(1) |= s->MMX_L(0) & 0x80000000;
d->MMX_L(0) = d->MMX_L(1);
}
void helper_pfsub(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_S(0) = float32_sub(d->MMX_S(0), s->MMX_S(0), &env->mmx_status);
d->MMX_S(1) = float32_sub(d->MMX_S(1), s->MMX_S(1), &env->mmx_status);
}
void helper_pfsubr(CPUX86State *env, MMXReg *d, MMXReg *s)
{
d->MMX_S(0) = float32_sub(s->MMX_S(0), d->MMX_S(0), &env->mmx_status);
d->MMX_S(1) = float32_sub(s->MMX_S(1), d->MMX_S(1), &env->mmx_status);
}
void helper_pswapd(CPUX86State *env, MMXReg *d, MMXReg *s)
{
uint32_t r;
r = s->MMX_L(0);
d->MMX_L(0) = s->MMX_L(1);
d->MMX_L(1) = r;
}
#endif
/* SSSE3 op helpers */
void glue(helper_pshufb, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
#if SHIFT == 0
uint8_t r[8];
for (i = 0; i < 8; i++) {
r[i] = (s->B(i) & 0x80) ? 0 : (v->B(s->B(i) & 7));
}
for (i = 0; i < 8; i++) {
d->B(i) = r[i];
}
#else
uint8_t r[8 << SHIFT];
for (i = 0; i < 8 << SHIFT; i++) {
int j = i & ~0xf;
r[i] = (s->B(i) & 0x80) ? 0 : v->B(j | (s->B(i) & 0xf));
}
for (i = 0; i < 8 << SHIFT; i++) {
d->B(i) = r[i];
}
#endif
}
#define SSE_HELPER_HW(name, F) \
void glue(helper_ ## name, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s) \
{ \
uint16_t r[4 << SHIFT]; \
int i, j, k; \
for (k = 0; k < 4 << SHIFT; k += LANE_WIDTH / 2) { \
for (i = j = 0; j < LANE_WIDTH / 2; i++, j += 2) { \
r[i + k] = F(v->W(j + k), v->W(j + k + 1)); \
} \
for (j = 0; j < LANE_WIDTH / 2; i++, j += 2) { \
r[i + k] = F(s->W(j + k), s->W(j + k + 1)); \
} \
} \
for (i = 0; i < 4 << SHIFT; i++) { \
d->W(i) = r[i]; \
} \
}
#define SSE_HELPER_HL(name, F) \
void glue(helper_ ## name, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s) \
{ \
uint32_t r[2 << SHIFT]; \
int i, j, k; \
for (k = 0; k < 2 << SHIFT; k += LANE_WIDTH / 4) { \
for (i = j = 0; j < LANE_WIDTH / 4; i++, j += 2) { \
r[i + k] = F(v->L(j + k), v->L(j + k + 1)); \
} \
for (j = 0; j < LANE_WIDTH / 4; i++, j += 2) { \
r[i + k] = F(s->L(j + k), s->L(j + k + 1)); \
} \
} \
for (i = 0; i < 2 << SHIFT; i++) { \
d->L(i) = r[i]; \
} \
}
SSE_HELPER_HW(phaddw, FADD)
SSE_HELPER_HW(phsubw, FSUB)
SSE_HELPER_HW(phaddsw, FADDSW)
SSE_HELPER_HW(phsubsw, FSUBSW)
SSE_HELPER_HL(phaddd, FADD)
SSE_HELPER_HL(phsubd, FSUB)
#undef SSE_HELPER_HW
#undef SSE_HELPER_HL
void glue(helper_pmaddubsw, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
for (i = 0; i < 4 << SHIFT; i++) {
d->W(i) = satsw((int8_t)s->B(i * 2) * (uint8_t)v->B(i * 2) +
(int8_t)s->B(i * 2 + 1) * (uint8_t)v->B(i * 2 + 1));
}
}
#define FMULHRSW(d, s) (((int16_t) d * (int16_t)s + 0x4000) >> 15)
SSE_HELPER_W(helper_pmulhrsw, FMULHRSW)
#define FSIGNB(d, s) (s <= INT8_MAX ? s ? d : 0 : -(int8_t)d)
#define FSIGNW(d, s) (s <= INT16_MAX ? s ? d : 0 : -(int16_t)d)
#define FSIGNL(d, s) (s <= INT32_MAX ? s ? d : 0 : -(int32_t)d)
SSE_HELPER_B(helper_psignb, FSIGNB)
SSE_HELPER_W(helper_psignw, FSIGNW)
SSE_HELPER_L(helper_psignd, FSIGNL)
void glue(helper_palignr, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s,
uint32_t imm)
{
int i;
/* XXX could be checked during translation */
if (imm >= (SHIFT ? 32 : 16)) {
for (i = 0; i < (1 << SHIFT); i++) {
d->Q(i) = 0;
}
} else {
int shift = imm * 8;
#define SHR(v, i) (i < 64 && i > -64 ? i > 0 ? v >> (i) : (v << -(i)) : 0)
#if SHIFT == 0
d->Q(0) = SHR(s->Q(0), shift - 0) |
SHR(v->Q(0), shift - 64);
#else
for (i = 0; i < (1 << SHIFT); i += 2) {
uint64_t r0, r1;
r0 = SHR(s->Q(i), shift - 0) |
SHR(s->Q(i + 1), shift - 64) |
SHR(v->Q(i), shift - 128) |
SHR(v->Q(i + 1), shift - 192);
r1 = SHR(s->Q(i), shift + 64) |
SHR(s->Q(i + 1), shift - 0) |
SHR(v->Q(i), shift - 64) |
SHR(v->Q(i + 1), shift - 128);
d->Q(i) = r0;
d->Q(i + 1) = r1;
}
#endif
#undef SHR
}
}
#if SHIFT >= 1
#define SSE_HELPER_V(name, elem, num, F) \
void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s, \
Reg *m) \
{ \
int i; \
for (i = 0; i < num; i++) { \
d->elem(i) = F(v->elem(i), s->elem(i), m->elem(i)); \
} \
}
#define SSE_HELPER_I(name, elem, num, F) \
void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s, \
uint32_t imm) \
{ \
int i; \
for (i = 0; i < num; i++) { \
int j = i & 7; \
d->elem(i) = F(v->elem(i), s->elem(i), (imm >> j) & 1); \
} \
}
/* SSE4.1 op helpers */
#define FBLENDVB(v, s, m) ((m & 0x80) ? s : v)
#define FBLENDVPS(v, s, m) ((m & 0x80000000) ? s : v)
#define FBLENDVPD(v, s, m) ((m & 0x8000000000000000LL) ? s : v)
SSE_HELPER_V(helper_pblendvb, B, 8 << SHIFT, FBLENDVB)
SSE_HELPER_V(helper_blendvps, L, 2 << SHIFT, FBLENDVPS)
SSE_HELPER_V(helper_blendvpd, Q, 1 << SHIFT, FBLENDVPD)
void glue(helper_ptest, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
uint64_t zf = 0, cf = 0;
int i;
for (i = 0; i < 1 << SHIFT; i++) {
zf |= (s->Q(i) & d->Q(i));
cf |= (s->Q(i) & ~d->Q(i));
}
CC_SRC = (zf ? 0 : CC_Z) | (cf ? 0 : CC_C);
}
#define FMOVSLDUP(i) s->L((i) & ~1)
#define FMOVSHDUP(i) s->L((i) | 1)
#define FMOVDLDUP(i) s->Q((i) & ~1)
#define SSE_HELPER_F(name, elem, num, F) \
void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *s) \
{ \
int n = num; \
for (int i = n; --i >= 0; ) { \
d->elem(i) = F(i); \
} \
}
#if SHIFT > 0
SSE_HELPER_F(helper_pmovsxbw, W, 4 << SHIFT, (int8_t) s->B)
SSE_HELPER_F(helper_pmovsxbd, L, 2 << SHIFT, (int8_t) s->B)
SSE_HELPER_F(helper_pmovsxbq, Q, 1 << SHIFT, (int8_t) s->B)
SSE_HELPER_F(helper_pmovsxwd, L, 2 << SHIFT, (int16_t) s->W)
SSE_HELPER_F(helper_pmovsxwq, Q, 1 << SHIFT, (int16_t) s->W)
SSE_HELPER_F(helper_pmovsxdq, Q, 1 << SHIFT, (int32_t) s->L)
SSE_HELPER_F(helper_pmovzxbw, W, 4 << SHIFT, s->B)
SSE_HELPER_F(helper_pmovzxbd, L, 2 << SHIFT, s->B)
SSE_HELPER_F(helper_pmovzxbq, Q, 1 << SHIFT, s->B)
SSE_HELPER_F(helper_pmovzxwd, L, 2 << SHIFT, s->W)
SSE_HELPER_F(helper_pmovzxwq, Q, 1 << SHIFT, s->W)
SSE_HELPER_F(helper_pmovzxdq, Q, 1 << SHIFT, s->L)
SSE_HELPER_F(helper_pmovsldup, L, 2 << SHIFT, FMOVSLDUP)
SSE_HELPER_F(helper_pmovshdup, L, 2 << SHIFT, FMOVSHDUP)
SSE_HELPER_F(helper_pmovdldup, Q, 1 << SHIFT, FMOVDLDUP)
#endif
void glue(helper_pmuldq, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
for (i = 0; i < 1 << SHIFT; i++) {
d->Q(i) = (int64_t)(int32_t) v->L(2 * i) * (int32_t) s->L(2 * i);
}
}
void glue(helper_packusdw, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
uint16_t r[8];
int i, j, k;
for (i = 0, j = 0; i <= 2 << SHIFT; i += 8, j += 4) {
r[0] = satuw(v->L(j));
r[1] = satuw(v->L(j + 1));
r[2] = satuw(v->L(j + 2));
r[3] = satuw(v->L(j + 3));
r[4] = satuw(s->L(j));
r[5] = satuw(s->L(j + 1));
r[6] = satuw(s->L(j + 2));
r[7] = satuw(s->L(j + 3));
for (k = 0; k < 8; k++) {
d->W(i + k) = r[k];
}
}
}
#if SHIFT == 1
void glue(helper_phminposuw, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
int idx = 0;
if (s->W(1) < s->W(idx)) {
idx = 1;
}
if (s->W(2) < s->W(idx)) {
idx = 2;
}
if (s->W(3) < s->W(idx)) {
idx = 3;
}
if (s->W(4) < s->W(idx)) {
idx = 4;
}
if (s->W(5) < s->W(idx)) {
idx = 5;
}
if (s->W(6) < s->W(idx)) {
idx = 6;
}
if (s->W(7) < s->W(idx)) {
idx = 7;
}
d->W(0) = s->W(idx);
d->W(1) = idx;
d->L(1) = 0;
d->Q(1) = 0;
}
#endif
void glue(helper_roundps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
uint32_t mode)
{
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
uint8_t old_flags = get_float_exception_flags(&env->sse_status);
signed char prev_rounding_mode;
int i;
prev_rounding_mode = env->sse_status.float_rounding_mode;
if (!(mode & (1 << 2))) {
set_x86_rounding_mode(mode & 3, &env->sse_status);
}
for (i = 0; i < 2 << SHIFT; i++) {
d->ZMM_S(i) = float32_round_to_int(s->ZMM_S(i), &env->sse_status);
}
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
if (mode & (1 << 3) && !(old_flags & float_flag_inexact)) {
set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
~float_flag_inexact,
&env->sse_status);
}
env->sse_status.float_rounding_mode = prev_rounding_mode;
}
void glue(helper_roundpd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
uint32_t mode)
{
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
uint8_t old_flags = get_float_exception_flags(&env->sse_status);
signed char prev_rounding_mode;
int i;
prev_rounding_mode = env->sse_status.float_rounding_mode;
if (!(mode & (1 << 2))) {
set_x86_rounding_mode(mode & 3, &env->sse_status);
}
for (i = 0; i < 1 << SHIFT; i++) {
d->ZMM_D(i) = float64_round_to_int(s->ZMM_D(i), &env->sse_status);
}
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
if (mode & (1 << 3) && !(old_flags & float_flag_inexact)) {
set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
~float_flag_inexact,
&env->sse_status);
}
env->sse_status.float_rounding_mode = prev_rounding_mode;
}
#if SHIFT == 1
void glue(helper_roundss, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s,
uint32_t mode)
{
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
uint8_t old_flags = get_float_exception_flags(&env->sse_status);
signed char prev_rounding_mode;
int i;
prev_rounding_mode = env->sse_status.float_rounding_mode;
if (!(mode & (1 << 2))) {
set_x86_rounding_mode(mode & 3, &env->sse_status);
}
d->ZMM_S(0) = float32_round_to_int(s->ZMM_S(0), &env->sse_status);
for (i = 1; i < 2 << SHIFT; i++) {
d->ZMM_L(i) = v->ZMM_L(i);
}
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
if (mode & (1 << 3) && !(old_flags & float_flag_inexact)) {
set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
~float_flag_inexact,
&env->sse_status);
}
env->sse_status.float_rounding_mode = prev_rounding_mode;
}
void glue(helper_roundsd, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s,
uint32_t mode)
{
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
uint8_t old_flags = get_float_exception_flags(&env->sse_status);
signed char prev_rounding_mode;
int i;
prev_rounding_mode = env->sse_status.float_rounding_mode;
if (!(mode & (1 << 2))) {
set_x86_rounding_mode(mode & 3, &env->sse_status);
}
d->ZMM_D(0) = float64_round_to_int(s->ZMM_D(0), &env->sse_status);
for (i = 1; i < 1 << SHIFT; i++) {
d->ZMM_Q(i) = v->ZMM_Q(i);
}
target/i386: fix IEEE SSE floating-point exception raising The SSE instruction implementations all fail to raise the expected IEEE floating-point exceptions because they do nothing to convert the exception state from the softfloat machinery into the exception flags in MXCSR. Fix this by adding such conversions. Unlike for x87, emulated SSE floating-point operations might be optimized using hardware floating point on the host, and so a different approach is taken that is compatible with such optimizations. The required invariant is that all exceptions set in env->sse_status (other than "denormal operand", for which the SSE semantics are different from those in the softfloat code) are ones that are set in the MXCSR; the emulated MXCSR is updated lazily when code reads MXCSR, while when code sets MXCSR, the exceptions in env->sse_status are set accordingly. A few instructions do not raise all the exceptions that would be raised by the softfloat code, and those instructions are made to save and restore the softfloat exception state accordingly. Nothing is done about "denormal operand"; setting that (only for the case when input denormals are *not* flushed to zero, the opposite of the logic in the softfloat code for such an exception) will require custom code for relevant instructions, or else architecture-specific conditionals in the softfloat code for when to set such an exception together with custom code for various SSE conversion and rounding instructions that do not set that exception. Nothing is done about trapping exceptions (for which there is minimal and largely broken support in QEMU's emulation in the x87 case and no support at all in the SSE case). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006252358000.3832@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-26 02:58:31 +03:00
if (mode & (1 << 3) && !(old_flags & float_flag_inexact)) {
set_float_exception_flags(get_float_exception_flags(&env->sse_status) &
~float_flag_inexact,
&env->sse_status);
}
env->sse_status.float_rounding_mode = prev_rounding_mode;
}
#endif
#define FBLENDP(v, s, m) (m ? s : v)
SSE_HELPER_I(helper_blendps, L, 2 << SHIFT, FBLENDP)
SSE_HELPER_I(helper_blendpd, Q, 1 << SHIFT, FBLENDP)
SSE_HELPER_I(helper_pblendw, W, 4 << SHIFT, FBLENDP)
void glue(helper_dpps, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s,
uint32_t mask)
{
float32 prod1, prod2, temp2, temp3, temp4;
int i;
for (i = 0; i < 2 << SHIFT; i += 4) {
/*
* We must evaluate (A+B)+(C+D), not ((A+B)+C)+D
* to correctly round the intermediate results
*/
if (mask & (1 << 4)) {
prod1 = float32_mul(v->ZMM_S(i), s->ZMM_S(i), &env->sse_status);
} else {
prod1 = float32_zero;
}
if (mask & (1 << 5)) {
prod2 = float32_mul(v->ZMM_S(i+1), s->ZMM_S(i+1), &env->sse_status);
} else {
prod2 = float32_zero;
}
temp2 = float32_add(prod1, prod2, &env->sse_status);
if (mask & (1 << 6)) {
prod1 = float32_mul(v->ZMM_S(i+2), s->ZMM_S(i+2), &env->sse_status);
} else {
prod1 = float32_zero;
}
if (mask & (1 << 7)) {
prod2 = float32_mul(v->ZMM_S(i+3), s->ZMM_S(i+3), &env->sse_status);
} else {
prod2 = float32_zero;
}
temp3 = float32_add(prod1, prod2, &env->sse_status);
temp4 = float32_add(temp2, temp3, &env->sse_status);
d->ZMM_S(i) = (mask & (1 << 0)) ? temp4 : float32_zero;
d->ZMM_S(i+1) = (mask & (1 << 1)) ? temp4 : float32_zero;
d->ZMM_S(i+2) = (mask & (1 << 2)) ? temp4 : float32_zero;
d->ZMM_S(i+3) = (mask & (1 << 3)) ? temp4 : float32_zero;
}
}
#if SHIFT == 1
/* Oddly, there is no ymm version of dppd */
void glue(helper_dppd, SUFFIX)(CPUX86State *env,
Reg *d, Reg *v, Reg *s, uint32_t mask)
{
float64 prod1, prod2, temp2;
if (mask & (1 << 4)) {
prod1 = float64_mul(v->ZMM_D(0), s->ZMM_D(0), &env->sse_status);
} else {
prod1 = float64_zero;
}
if (mask & (1 << 5)) {
prod2 = float64_mul(v->ZMM_D(1), s->ZMM_D(1), &env->sse_status);
} else {
prod2 = float64_zero;
}
temp2 = float64_add(prod1, prod2, &env->sse_status);
d->ZMM_D(0) = (mask & (1 << 0)) ? temp2 : float64_zero;
d->ZMM_D(1) = (mask & (1 << 1)) ? temp2 : float64_zero;
}
#endif
void glue(helper_mpsadbw, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s,
uint32_t offset)
{
int i, j;
uint16_t r[8];
for (j = 0; j < 4 << SHIFT; ) {
int s0 = (j * 2) + ((offset & 3) << 2);
int d0 = (j * 2) + ((offset & 4) << 0);
for (i = 0; i < LANE_WIDTH / 2; i++, d0++) {
r[i] = 0;
r[i] += abs1(v->B(d0 + 0) - s->B(s0 + 0));
r[i] += abs1(v->B(d0 + 1) - s->B(s0 + 1));
r[i] += abs1(v->B(d0 + 2) - s->B(s0 + 2));
r[i] += abs1(v->B(d0 + 3) - s->B(s0 + 3));
}
for (i = 0; i < LANE_WIDTH / 2; i++, j++) {
d->W(j) = r[i];
}
offset >>= 3;
}
}
/* SSE4.2 op helpers */
#if SHIFT == 1
static inline int pcmp_elen(CPUX86State *env, int reg, uint32_t ctrl)
{
target_long val, limit;
/* Presence of REX.W is indicated by a bit higher than 7 set */
if (ctrl >> 8) {
val = (target_long)env->regs[reg];
} else {
val = (int32_t)env->regs[reg];
}
if (ctrl & 1) {
limit = 8;
} else {
limit = 16;
}
if ((val > limit) || (val < -limit)) {
return limit;
}
return abs1(val);
}
static inline int pcmp_ilen(Reg *r, uint8_t ctrl)
{
int val = 0;
if (ctrl & 1) {
while (val < 8 && r->W(val)) {
val++;
}
} else {
while (val < 16 && r->B(val)) {
val++;
}
}
return val;
}
static inline int pcmp_val(Reg *r, uint8_t ctrl, int i)
{
switch ((ctrl >> 0) & 3) {
case 0:
return r->B(i);
case 1:
return r->W(i);
case 2:
return (int8_t)r->B(i);
case 3:
default:
return (int16_t)r->W(i);
}
}
static inline unsigned pcmpxstrx(CPUX86State *env, Reg *d, Reg *s,
uint8_t ctrl, int valids, int validd)
{
unsigned int res = 0;
int v;
int j, i;
int upper = (ctrl & 1) ? 7 : 15;
valids--;
validd--;
CC_SRC = (valids < upper ? CC_Z : 0) | (validd < upper ? CC_S : 0);
switch ((ctrl >> 2) & 3) {
case 0:
for (j = valids; j >= 0; j--) {
res <<= 1;
v = pcmp_val(s, ctrl, j);
for (i = validd; i >= 0; i--) {
res |= (v == pcmp_val(d, ctrl, i));
}
}
break;
case 1:
for (j = valids; j >= 0; j--) {
res <<= 1;
v = pcmp_val(s, ctrl, j);
for (i = ((validd - 1) | 1); i >= 0; i -= 2) {
res |= (pcmp_val(d, ctrl, i - 0) >= v &&
pcmp_val(d, ctrl, i - 1) <= v);
}
}
break;
case 2:
res = (1 << (upper - MAX(valids, validd))) - 1;
res <<= MAX(valids, validd) - MIN(valids, validd);
for (i = MIN(valids, validd); i >= 0; i--) {
res <<= 1;
v = pcmp_val(s, ctrl, i);
res |= (v == pcmp_val(d, ctrl, i));
}
break;
case 3:
if (validd == -1) {
res = (2 << upper) - 1;
break;
}
target/i386: correct fix for pcmpxstrx substring search This corrects a bug introduced in my previous fix for SSE4.2 pcmpestri / pcmpestrm / pcmpistri / pcmpistrm substring search, commit ae35eea7e4a9f21dd147406dfbcd0c4c6aaf2a60. That commit fixed a bug that showed up in four GCC tests with one libc implementation. The tests in question generate random inputs to the intrinsics and compare results to a C implementation, but they only test 1024 possible random inputs, and when the tests use the cases of those instructions that work with word rather than byte inputs, it's easy to have problematic cases that show up much less frequently than that. Thus, testing with a different libc implementation, and so a different random number generator, showed up a problem with the previous patch. When investigating the previous test failures, I found the description of these instructions in the Intel manuals (starting from computing a 16x16 or 8x8 set of comparison results) confusing and hard to match up with the more optimized implementation in QEMU, and referred to AMD manuals which described the instructions in a different way. Those AMD descriptions are very explicit that the whole of the string being searched for must be found in the other operand, not running off the end of that operand; they say "If the prototype and the SUT are equal in length, the two strings must be identical for the comparison to be TRUE.". However, that statement is incorrect. In my previous commit message, I noted: The operation in this case is a search for a string (argument d to the helper) in another string (argument s to the helper); if a copy of d at a particular position would run off the end of s, the resulting output bit should be 0 whether or not the strings match in the region where they overlap, but the QEMU implementation was wrongly comparing only up to the point where s ends and counting it as a match if an initial segment of d matched a terminal segment of s. Here, "run off the end of s" means that some byte of d would overlap some byte outside of s; thus, if d has zero length, it is considered to match everywhere, including after the end of s. The description "some byte of d would overlap some byte outside of s" is accurate only when understood to refer to overlapping some byte *within the 16-byte operand* but at or after the zero terminator; it is valid to run over the end of s if the end of s is the end of the 16-byte operand. So the fix in the previous patch for the case of d being empty was correct, but the other part of that patch was not correct (as it never allowed partial matches even at the end of the 16-byte operand). Nor was the code before the previous patch correct for the case of d nonempty, as it would always have allowed partial matches at the end of s. Fix with a partial revert of my previous change, combined with inserting a check for the special case of s having maximum length to determine where it is necessary to check for matches. In the added test, test 1 is for the case of empty strings, which failed before my 2017 patch, test 2 is for the bug introduced by my 2017 patch and test 3 deals with the case where a match of an initial segment at the end of the string is not valid when the string ends before the end of the 16-byte operand (that is, the case that would be broken by a simple revert of the non-empty-string part of my 2017 patch). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006121344290.9881@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-12 16:45:23 +03:00
for (j = valids == upper ? valids : valids - validd; j >= 0; j--) {
res <<= 1;
v = 1;
target/i386: correct fix for pcmpxstrx substring search This corrects a bug introduced in my previous fix for SSE4.2 pcmpestri / pcmpestrm / pcmpistri / pcmpistrm substring search, commit ae35eea7e4a9f21dd147406dfbcd0c4c6aaf2a60. That commit fixed a bug that showed up in four GCC tests with one libc implementation. The tests in question generate random inputs to the intrinsics and compare results to a C implementation, but they only test 1024 possible random inputs, and when the tests use the cases of those instructions that work with word rather than byte inputs, it's easy to have problematic cases that show up much less frequently than that. Thus, testing with a different libc implementation, and so a different random number generator, showed up a problem with the previous patch. When investigating the previous test failures, I found the description of these instructions in the Intel manuals (starting from computing a 16x16 or 8x8 set of comparison results) confusing and hard to match up with the more optimized implementation in QEMU, and referred to AMD manuals which described the instructions in a different way. Those AMD descriptions are very explicit that the whole of the string being searched for must be found in the other operand, not running off the end of that operand; they say "If the prototype and the SUT are equal in length, the two strings must be identical for the comparison to be TRUE.". However, that statement is incorrect. In my previous commit message, I noted: The operation in this case is a search for a string (argument d to the helper) in another string (argument s to the helper); if a copy of d at a particular position would run off the end of s, the resulting output bit should be 0 whether or not the strings match in the region where they overlap, but the QEMU implementation was wrongly comparing only up to the point where s ends and counting it as a match if an initial segment of d matched a terminal segment of s. Here, "run off the end of s" means that some byte of d would overlap some byte outside of s; thus, if d has zero length, it is considered to match everywhere, including after the end of s. The description "some byte of d would overlap some byte outside of s" is accurate only when understood to refer to overlapping some byte *within the 16-byte operand* but at or after the zero terminator; it is valid to run over the end of s if the end of s is the end of the 16-byte operand. So the fix in the previous patch for the case of d being empty was correct, but the other part of that patch was not correct (as it never allowed partial matches even at the end of the 16-byte operand). Nor was the code before the previous patch correct for the case of d nonempty, as it would always have allowed partial matches at the end of s. Fix with a partial revert of my previous change, combined with inserting a check for the special case of s having maximum length to determine where it is necessary to check for matches. In the added test, test 1 is for the case of empty strings, which failed before my 2017 patch, test 2 is for the bug introduced by my 2017 patch and test 3 deals with the case where a match of an initial segment at the end of the string is not valid when the string ends before the end of the 16-byte operand (that is, the case that would be broken by a simple revert of the non-empty-string part of my 2017 patch). Signed-off-by: Joseph Myers <joseph@codesourcery.com> Message-Id: <alpine.DEB.2.21.2006121344290.9881@digraph.polyomino.org.uk> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2020-06-12 16:45:23 +03:00
for (i = MIN(valids - j, validd); i >= 0; i--) {
v &= (pcmp_val(s, ctrl, i + j) == pcmp_val(d, ctrl, i));
}
res |= v;
}
break;
}
switch ((ctrl >> 4) & 3) {
case 1:
res ^= (2 << upper) - 1;
break;
case 3:
res ^= (1 << (valids + 1)) - 1;
break;
}
if (res) {
CC_SRC |= CC_C;
}
if (res & 1) {
CC_SRC |= CC_O;
}
return res;
}
void glue(helper_pcmpestri, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
uint32_t ctrl)
{
unsigned int res = pcmpxstrx(env, d, s, ctrl,
pcmp_elen(env, R_EDX, ctrl),
pcmp_elen(env, R_EAX, ctrl));
if (res) {
env->regs[R_ECX] = (ctrl & (1 << 6)) ? 31 - clz32(res) : ctz32(res);
} else {
env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
}
}
void glue(helper_pcmpestrm, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
uint32_t ctrl)
{
int i;
unsigned int res = pcmpxstrx(env, d, s, ctrl,
pcmp_elen(env, R_EDX, ctrl),
pcmp_elen(env, R_EAX, ctrl));
if ((ctrl >> 6) & 1) {
if (ctrl & 1) {
for (i = 0; i < 8; i++, res >>= 1) {
env->xmm_regs[0].W(i) = (res & 1) ? ~0 : 0;
}
} else {
for (i = 0; i < 16; i++, res >>= 1) {
env->xmm_regs[0].B(i) = (res & 1) ? ~0 : 0;
}
}
} else {
env->xmm_regs[0].Q(1) = 0;
env->xmm_regs[0].Q(0) = res;
}
}
void glue(helper_pcmpistri, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
uint32_t ctrl)
{
unsigned int res = pcmpxstrx(env, d, s, ctrl,
pcmp_ilen(s, ctrl),
pcmp_ilen(d, ctrl));
if (res) {
env->regs[R_ECX] = (ctrl & (1 << 6)) ? 31 - clz32(res) : ctz32(res);
} else {
env->regs[R_ECX] = 16 >> (ctrl & (1 << 0));
}
}
void glue(helper_pcmpistrm, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
uint32_t ctrl)
{
int i;
unsigned int res = pcmpxstrx(env, d, s, ctrl,
pcmp_ilen(s, ctrl),
pcmp_ilen(d, ctrl));
if ((ctrl >> 6) & 1) {
if (ctrl & 1) {
for (i = 0; i < 8; i++, res >>= 1) {
env->xmm_regs[0].W(i) = (res & 1) ? ~0 : 0;
}
} else {
for (i = 0; i < 16; i++, res >>= 1) {
env->xmm_regs[0].B(i) = (res & 1) ? ~0 : 0;
}
}
} else {
env->xmm_regs[0].Q(1) = 0;
env->xmm_regs[0].Q(0) = res;
}
}
#define CRCPOLY 0x1edc6f41
#define CRCPOLY_BITREV 0x82f63b78
target_ulong helper_crc32(uint32_t crc1, target_ulong msg, uint32_t len)
{
target_ulong crc = (msg & ((target_ulong) -1 >>
(TARGET_LONG_BITS - len))) ^ crc1;
while (len--) {
crc = (crc >> 1) ^ ((crc & 1) ? CRCPOLY_BITREV : 0);
}
return crc;
}
#endif
#if SHIFT == 1
static void clmulq(uint64_t *dest_l, uint64_t *dest_h,
uint64_t a, uint64_t b)
{
uint64_t al, ah, resh, resl;
ah = 0;
al = a;
resh = resl = 0;
while (b) {
if (b & 1) {
resl ^= al;
resh ^= ah;
}
ah = (ah << 1) | (al >> 63);
al <<= 1;
b >>= 1;
}
*dest_l = resl;
*dest_h = resh;
}
#endif
void glue(helper_pclmulqdq, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s,
uint32_t ctrl)
{
uint64_t a, b;
int i;
for (i = 0; i < 1 << SHIFT; i += 2) {
a = v->Q(((ctrl & 1) != 0) + i);
b = s->Q(((ctrl & 16) != 0) + i);
clmulq(&d->Q(i), &d->Q(i + 1), a, b);
}
}
void glue(helper_aesdec, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
Reg st = *v;
Reg rk = *s;
for (i = 0 ; i < 2 << SHIFT ; i++) {
int j = i & 3;
d->L(i) = rk.L(i) ^ bswap32(AES_Td0[st.B(AES_ishifts[4 * j + 0])] ^
AES_Td1[st.B(AES_ishifts[4 * j + 1])] ^
AES_Td2[st.B(AES_ishifts[4 * j + 2])] ^
AES_Td3[st.B(AES_ishifts[4 * j + 3])]);
}
}
void glue(helper_aesdeclast, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
Reg st = *v;
Reg rk = *s;
for (i = 0; i < 8 << SHIFT; i++) {
d->B(i) = rk.B(i) ^ (AES_isbox[st.B(AES_ishifts[i & 15] + (i & ~15))]);
}
}
void glue(helper_aesenc, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
Reg st = *v;
Reg rk = *s;
for (i = 0 ; i < 2 << SHIFT ; i++) {
int j = i & 3;
d->L(i) = rk.L(i) ^ bswap32(AES_Te0[st.B(AES_shifts[4 * j + 0])] ^
AES_Te1[st.B(AES_shifts[4 * j + 1])] ^
AES_Te2[st.B(AES_shifts[4 * j + 2])] ^
AES_Te3[st.B(AES_shifts[4 * j + 3])]);
}
}
void glue(helper_aesenclast, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
Reg st = *v;
Reg rk = *s;
for (i = 0; i < 8 << SHIFT; i++) {
d->B(i) = rk.B(i) ^ (AES_sbox[st.B(AES_shifts[i & 15] + (i & ~15))]);
}
}
#if SHIFT == 1
void glue(helper_aesimc, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
int i;
Reg tmp = *s;
for (i = 0 ; i < 4 ; i++) {
d->L(i) = bswap32(AES_imc[tmp.B(4 * i + 0)][0] ^
AES_imc[tmp.B(4 * i + 1)][1] ^
AES_imc[tmp.B(4 * i + 2)][2] ^
AES_imc[tmp.B(4 * i + 3)][3]);
}
}
void glue(helper_aeskeygenassist, SUFFIX)(CPUX86State *env, Reg *d, Reg *s,
uint32_t ctrl)
{
int i;
Reg tmp = *s;
for (i = 0 ; i < 4 ; i++) {
d->B(i) = AES_sbox[tmp.B(i + 4)];
d->B(i + 8) = AES_sbox[tmp.B(i + 12)];
}
d->L(1) = (d->L(0) << 24 | d->L(0) >> 8) ^ ctrl;
d->L(3) = (d->L(2) << 24 | d->L(2) >> 8) ^ ctrl;
}
#endif
#endif
#if SHIFT >= 1
void glue(helper_vpermilpd, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
uint64_t r0, r1;
int i;
for (i = 0; i < 1 << SHIFT; i += 2) {
r0 = v->Q(i + ((s->Q(i) >> 1) & 1));
r1 = v->Q(i + ((s->Q(i+1) >> 1) & 1));
d->Q(i) = r0;
d->Q(i+1) = r1;
}
}
void glue(helper_vpermilps, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
uint32_t r0, r1, r2, r3;
int i;
for (i = 0; i < 2 << SHIFT; i += 4) {
r0 = v->L(i + (s->L(i) & 3));
r1 = v->L(i + (s->L(i+1) & 3));
r2 = v->L(i + (s->L(i+2) & 3));
r3 = v->L(i + (s->L(i+3) & 3));
d->L(i) = r0;
d->L(i+1) = r1;
d->L(i+2) = r2;
d->L(i+3) = r3;
}
}
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 == 1
#define FPSRLVD(x, c) (c < 32 ? ((x) >> c) : 0)
#define FPSRLVQ(x, c) (c < 64 ? ((x) >> c) : 0)
#define FPSRAVD(x, c) ((int32_t)(x) >> (c < 32 ? c : 31))
#define FPSRAVQ(x, c) ((int64_t)(x) >> (c < 64 ? c : 63))
#define FPSLLVD(x, c) (c < 32 ? ((x) << c) : 0)
#define FPSLLVQ(x, c) (c < 64 ? ((x) << c) : 0)
#endif
SSE_HELPER_L(helper_vpsrlvd, FPSRLVD)
SSE_HELPER_L(helper_vpsravd, FPSRAVD)
SSE_HELPER_L(helper_vpsllvd, FPSLLVD)
SSE_HELPER_Q(helper_vpsrlvq, FPSRLVQ)
SSE_HELPER_Q(helper_vpsravq, FPSRAVQ)
SSE_HELPER_Q(helper_vpsllvq, FPSLLVQ)
void glue(helper_vtestps, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
uint32_t zf = 0, cf = 0;
int i;
for (i = 0; i < 2 << SHIFT; i++) {
zf |= (s->L(i) & d->L(i));
cf |= (s->L(i) & ~d->L(i));
}
CC_SRC = ((zf >> 31) ? 0 : CC_Z) | ((cf >> 31) ? 0 : CC_C);
}
void glue(helper_vtestpd, SUFFIX)(CPUX86State *env, Reg *d, Reg *s)
{
uint64_t zf = 0, cf = 0;
int i;
for (i = 0; i < 1 << SHIFT; i++) {
zf |= (s->Q(i) & d->Q(i));
cf |= (s->Q(i) & ~d->Q(i));
}
CC_SRC = ((zf >> 63) ? 0 : CC_Z) | ((cf >> 63) ? 0 : CC_C);
}
void glue(helper_vpmaskmovd_st, SUFFIX)(CPUX86State *env,
Reg *v, Reg *s, target_ulong a0)
{
int i;
for (i = 0; i < (2 << SHIFT); i++) {
if (v->L(i) >> 31) {
cpu_stl_data_ra(env, a0 + i * 4, s->L(i), GETPC());
}
}
}
void glue(helper_vpmaskmovq_st, SUFFIX)(CPUX86State *env,
Reg *v, Reg *s, target_ulong a0)
{
int i;
for (i = 0; i < (1 << SHIFT); i++) {
if (v->Q(i) >> 63) {
cpu_stq_data_ra(env, a0 + i * 8, s->Q(i), GETPC());
}
}
}
void glue(helper_vpmaskmovd, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
for (i = 0; i < (2 << SHIFT); i++) {
d->L(i) = (v->L(i) >> 31) ? s->L(i) : 0;
}
}
void glue(helper_vpmaskmovq, SUFFIX)(CPUX86State *env, Reg *d, Reg *v, Reg *s)
{
int i;
for (i = 0; i < (1 << SHIFT); i++) {
d->Q(i) = (v->Q(i) >> 63) ? s->Q(i) : 0;
}
}
void glue(helper_vpgatherdd, SUFFIX)(CPUX86State *env,
Reg *d, Reg *v, Reg *s, target_ulong a0, unsigned scale)
{
int i;
for (i = 0; i < (2 << SHIFT); i++) {
if (v->L(i) >> 31) {
target_ulong addr = a0
+ ((target_ulong)(int32_t)s->L(i) << scale);
d->L(i) = cpu_ldl_data_ra(env, addr, GETPC());
}
v->L(i) = 0;
}
}
void glue(helper_vpgatherdq, SUFFIX)(CPUX86State *env,
Reg *d, Reg *v, Reg *s, target_ulong a0, unsigned scale)
{
int i;
for (i = 0; i < (1 << SHIFT); i++) {
if (v->Q(i) >> 63) {
target_ulong addr = a0
+ ((target_ulong)(int32_t)s->L(i) << scale);
d->Q(i) = cpu_ldq_data_ra(env, addr, GETPC());
}
v->Q(i) = 0;
}
}
void glue(helper_vpgatherqd, SUFFIX)(CPUX86State *env,
Reg *d, Reg *v, Reg *s, target_ulong a0, unsigned scale)
{
int i;
for (i = 0; i < (1 << SHIFT); i++) {
if (v->L(i) >> 31) {
target_ulong addr = a0
+ ((target_ulong)(int64_t)s->Q(i) << scale);
d->L(i) = cpu_ldl_data_ra(env, addr, GETPC());
}
v->L(i) = 0;
}
for (i /= 2; i < 1 << SHIFT; i++) {
d->Q(i) = 0;
v->Q(i) = 0;
}
}
void glue(helper_vpgatherqq, SUFFIX)(CPUX86State *env,
Reg *d, Reg *v, Reg *s, target_ulong a0, unsigned scale)
{
int i;
for (i = 0; i < (1 << SHIFT); i++) {
if (v->Q(i) >> 63) {
target_ulong addr = a0
+ ((target_ulong)(int64_t)s->Q(i) << scale);
d->Q(i) = cpu_ldq_data_ra(env, addr, GETPC());
}
v->Q(i) = 0;
}
}
#endif
#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;
default: /* default case added to help the compiler to avoid warnings */
g_assert_not_reached();
}
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;
default: /* default case added to help the compiler to avoid warnings */
g_assert_not_reached();
}
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;
}
void helper_vpermd_ymm(Reg *d, Reg *v, Reg *s)
{
uint32_t r[8];
int i;
for (i = 0; i < 8; i++) {
r[i] = s->L(v->L(i) & 7);
}
for (i = 0; i < 8; i++) {
d->L(i) = r[i];
}
}
#endif
/* FMA3 op helpers */
#if SHIFT == 1
#define SSE_HELPER_FMAS(name, elem, F) \
void name(CPUX86State *env, Reg *d, Reg *a, Reg *b, Reg *c, int flags) \
{ \
d->elem(0) = F(a->elem(0), b->elem(0), c->elem(0), flags, &env->sse_status); \
}
#define SSE_HELPER_FMAP(name, elem, num, F) \
void glue(name, SUFFIX)(CPUX86State *env, Reg *d, Reg *a, Reg *b, Reg *c, \
int flags, int flip) \
{ \
int i; \
for (i = 0; i < num; i++) { \
d->elem(i) = F(a->elem(i), b->elem(i), c->elem(i), flags, &env->sse_status); \
flags ^= flip; \
} \
}
SSE_HELPER_FMAS(helper_fma4ss, ZMM_S, float32_muladd)
SSE_HELPER_FMAS(helper_fma4sd, ZMM_D, float64_muladd)
#endif
#if SHIFT >= 1
SSE_HELPER_FMAP(helper_fma4ps, ZMM_S, 2 << SHIFT, float32_muladd)
SSE_HELPER_FMAP(helper_fma4pd, ZMM_D, 1 << SHIFT, float64_muladd)
#endif
#undef SSE_HELPER_S
#undef LANE_WIDTH
#undef SHIFT
#undef XMM_ONLY
#undef Reg
#undef B
#undef W
#undef L
#undef Q
#undef SUFFIX