[PATCH] target/i386: Fix exception classes for SSE/AVX instructions.

qemu-devel
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[PATCH] target/i386: Fix exception classes for SSE/AVX instructions.

From:	Ricky Zhou
Subject:	[PATCH] target/i386: Fix exception classes for SSE/AVX instructions.
Date:	Sun, 12 Feb 2023 00:28:12 -0800
Fix the exception classes for some SSE/AVX instructions to match what is
documented in the Intel manual.

Most of these changes have no functional effect on the behavior that
qemu implements (primarily >= 16-byte memory alignment checks). For
example, since qemu does not implement the AC flag, there is no
difference in behavior between Exception Classes 4 and 5 for
instructions where the SSE version only takes <16 byte memory operands.

There is one functional change:

Before this change, MOVNTPS and MOVNTPD were labeled as Exception Class
4 (only requiring alignment for legacy SSE instructions). This changes
them to Exception Class 1 (always requiring memory alignment), as
documented in the Intel manual.

Signed-off-by: Ricky Zhou <ricky@rzhou.org>
---
 target/i386/tcg/decode-new.c.inc | 79 ++++++++++++++++----------------
 1 file changed, 40 insertions(+), 39 deletions(-)

diff --git a/target/i386/tcg/decode-new.c.inc b/target/i386/tcg/decode-new.c.inc
index d5fd8d965c..a9060c3268 100644
--- a/target/i386/tcg/decode-new.c.inc
+++ b/target/i386/tcg/decode-new.c.inc
@@ -237,7 +237,7 @@ static void decode_group14(DisasContext *s, CPUX86State 
*env, X86OpEntry *entry,
 static void decode_0F6F(DisasContext *s, CPUX86State *env, X86OpEntry *entry, 
uint8_t *b)
 {
     static const X86OpEntry opcodes_0F6F[4] = {
-        X86_OP_ENTRY3(MOVDQ,       P,q, None,None, Q,q, vex1 mmx),  /* movq */
+        X86_OP_ENTRY3(MOVDQ,       P,q, None,None, Q,q, vex5 mmx),  /* movq */
         X86_OP_ENTRY3(MOVDQ,       V,x, None,None, W,x, vex1),      /* movdqa 
*/
         X86_OP_ENTRY3(MOVDQ,       V,x, None,None, W,x, vex4_unal), /* movdqu 
*/
         {},
@@ -274,9 +274,9 @@ static void decode_0F78(DisasContext *s, CPUX86State *env, 
X86OpEntry *entry, ui
 {
     static const X86OpEntry opcodes_0F78[4] = {
         {},
-        X86_OP_ENTRY3(EXTRQ_i,       V,x, None,None, I,w,  cpuid(SSE4A)),
+        X86_OP_ENTRY3(EXTRQ_i,       V,x, None,None, I,w,  cpuid(SSE4A)), /* 
AMD extension */
         {},
-        X86_OP_ENTRY3(INSERTQ_i,     V,x, U,x, I,w,        cpuid(SSE4A)),
+        X86_OP_ENTRY3(INSERTQ_i,     V,x, U,x, I,w,        cpuid(SSE4A)), /* 
AMD extension */
     };
     *entry = *decode_by_prefix(s, opcodes_0F78);
 }
@@ -284,9 +284,9 @@ static void decode_0F78(DisasContext *s, CPUX86State *env, 
X86OpEntry *entry, ui
 static void decode_0F79(DisasContext *s, CPUX86State *env, X86OpEntry *entry, 
uint8_t *b)
 {
     if (s->prefix & PREFIX_REPNZ) {
-        entry->gen = gen_INSERTQ_r;
+        entry->gen = gen_INSERTQ_r; /* AMD extension */
     } else if (s->prefix & PREFIX_DATA) {
-        entry->gen = gen_EXTRQ_r;
+        entry->gen = gen_EXTRQ_r; /* AMD extension */
     } else {
         entry->gen = NULL;
     };
@@ -306,7 +306,7 @@ static void decode_0F7E(DisasContext *s, CPUX86State *env, 
X86OpEntry *entry, ui
 static void decode_0F7F(DisasContext *s, CPUX86State *env, X86OpEntry *entry, 
uint8_t *b)
 {
     static const X86OpEntry opcodes_0F7F[4] = {
-        X86_OP_ENTRY3(MOVDQ,       W,x, None,None, V,x, vex1 mmx), /* movq */
+        X86_OP_ENTRY3(MOVDQ,       W,x, None,None, V,x, vex5 mmx), /* movq */
         X86_OP_ENTRY3(MOVDQ,       W,x, None,None, V,x, vex1), /* movdqa */
         X86_OP_ENTRY3(MOVDQ,       W,x, None,None, V,x, vex4_unal), /* movdqu 
*/
         {},
@@ -639,15 +639,15 @@ static void decode_0F10(DisasContext *s, CPUX86State 
*env, X86OpEntry *entry, ui
     static const X86OpEntry opcodes_0F10_reg[4] = {
         X86_OP_ENTRY3(MOVDQ,   V,x,  None,None, W,x, vex4_unal), /* MOVUPS */
         X86_OP_ENTRY3(MOVDQ,   V,x,  None,None, W,x, vex4_unal), /* MOVUPD */
-        X86_OP_ENTRY3(VMOVSS,  V,x,  H,x,       W,x, vex4),
-        X86_OP_ENTRY3(VMOVLPx, V,x,  H,x,       W,x, vex4), /* MOVSD */
+        X86_OP_ENTRY3(VMOVSS,  V,x,  H,x,       W,x, vex5),
+        X86_OP_ENTRY3(VMOVLPx, V,x,  H,x,       W,x, vex5), /* MOVSD */
     };
 
     static const X86OpEntry opcodes_0F10_mem[4] = {
         X86_OP_ENTRY3(MOVDQ,      V,x,  None,None, W,x,  vex4_unal), /* MOVUPS 
*/
         X86_OP_ENTRY3(MOVDQ,      V,x,  None,None, W,x,  vex4_unal), /* MOVUPD 
*/
-        X86_OP_ENTRY3(VMOVSS_ld,  V,x,  H,x,       M,ss, vex4),
-        X86_OP_ENTRY3(VMOVSD_ld,  V,x,  H,x,       M,sd, vex4),
+        X86_OP_ENTRY3(VMOVSS_ld,  V,x,  H,x,       M,ss, vex5),
+        X86_OP_ENTRY3(VMOVSD_ld,  V,x,  H,x,       M,sd, vex5),
     };
 
     if ((get_modrm(s, env) >> 6) == 3) {
@@ -660,17 +660,17 @@ static void decode_0F10(DisasContext *s, CPUX86State 
*env, X86OpEntry *entry, ui
 static void decode_0F11(DisasContext *s, CPUX86State *env, X86OpEntry *entry, 
uint8_t *b)
 {
     static const X86OpEntry opcodes_0F11_reg[4] = {
-        X86_OP_ENTRY3(MOVDQ,   W,x,  None,None, V,x, vex4), /* MOVPS */
-        X86_OP_ENTRY3(MOVDQ,   W,x,  None,None, V,x, vex4), /* MOVPD */
-        X86_OP_ENTRY3(VMOVSS,  W,x,  H,x,       V,x, vex4),
-        X86_OP_ENTRY3(VMOVLPx, W,x,  H,x,       V,q, vex4), /* MOVSD */
+        X86_OP_ENTRY3(MOVDQ,   W,x,  None,None, V,x, vex4), /* MOVUPS */
+        X86_OP_ENTRY3(MOVDQ,   W,x,  None,None, V,x, vex4), /* MOVUPD */
+        X86_OP_ENTRY3(VMOVSS,  W,x,  H,x,       V,x, vex5),
+        X86_OP_ENTRY3(VMOVLPx, W,x,  H,x,       V,q, vex5), /* MOVSD */
     };
 
     static const X86OpEntry opcodes_0F11_mem[4] = {
-        X86_OP_ENTRY3(MOVDQ,      W,x,  None,None, V,x, vex4), /* MOVPS */
-        X86_OP_ENTRY3(MOVDQ,      W,x,  None,None, V,x, vex4), /* MOVPD */
-        X86_OP_ENTRY3(VMOVSS_st,  M,ss, None,None, V,x, vex4),
-        X86_OP_ENTRY3(VMOVLPx_st, M,sd, None,None, V,x, vex4), /* MOVSD */
+        X86_OP_ENTRY3(MOVDQ,      W,x,  None,None, V,x, vex4), /* MOVUPS */
+        X86_OP_ENTRY3(MOVDQ,      W,x,  None,None, V,x, vex4), /* MOVUPD */
+        X86_OP_ENTRY3(VMOVSS_st,  M,ss, None,None, V,x, vex5),
+        X86_OP_ENTRY3(VMOVLPx_st, M,sd, None,None, V,x, vex5), /* MOVSD */
     };
 
     if ((get_modrm(s, env) >> 6) == 3) {
@@ -687,16 +687,16 @@ static void decode_0F12(DisasContext *s, CPUX86State 
*env, X86OpEntry *entry, ui
          * Use dq for operand for compatibility with gen_MOVSD and
          * to allow VEX128 only.
          */
-        X86_OP_ENTRY3(VMOVLPx_ld, V,dq, H,dq,      M,q, vex4), /* MOVLPS */
-        X86_OP_ENTRY3(VMOVLPx_ld, V,dq, H,dq,      M,q, vex4), /* MOVLPD */
+        X86_OP_ENTRY3(VMOVLPx_ld, V,dq, H,dq,      M,q, vex5), /* MOVLPS */
+        X86_OP_ENTRY3(VMOVLPx_ld, V,dq, H,dq,      M,q, vex5), /* MOVLPD */
         X86_OP_ENTRY3(VMOVSLDUP,  V,x,  None,None, W,x, vex4 cpuid(SSE3)),
-        X86_OP_ENTRY3(VMOVDDUP,   V,x,  None,None, WM,q, vex4 cpuid(SSE3)), /* 
qq if VEX.256 */
+        X86_OP_ENTRY3(VMOVDDUP,   V,x,  None,None, WM,q, vex5 cpuid(SSE3)), /* 
qq if VEX.256 */
     };
     static const X86OpEntry opcodes_0F12_reg[4] = {
-        X86_OP_ENTRY3(VMOVHLPS,  V,dq, H,dq,       U,dq, vex4),
-        X86_OP_ENTRY3(VMOVLPx,   W,x,  H,x,        U,q,  vex4), /* MOVLPD */
+        X86_OP_ENTRY3(VMOVHLPS,  V,dq, H,dq,       U,dq, vex7),
+        X86_OP_ENTRY3(VMOVLPx,   W,x,  H,x,        U,q,  vex5), /* MOVLPD */
         X86_OP_ENTRY3(VMOVSLDUP, V,x,  None,None,  U,x,  vex4 cpuid(SSE3)),
-        X86_OP_ENTRY3(VMOVDDUP,  V,x,  None,None,  U,x,  vex4 cpuid(SSE3)),
+        X86_OP_ENTRY3(VMOVDDUP,  V,x,  None,None,  U,x,  vex5 cpuid(SSE3)),
     };
 
     if ((get_modrm(s, env) >> 6) == 3) {
@@ -716,15 +716,15 @@ static void decode_0F16(DisasContext *s, CPUX86State 
*env, X86OpEntry *entry, ui
          * Operand 1 technically only reads the low 64 bits, but uses dq so 
that
          * it is easier to check for op0 == op1 in an endianness-neutral 
manner.
          */
-        X86_OP_ENTRY3(VMOVHPx_ld, V,dq, H,dq,      M,q, vex4), /* MOVHPS */
-        X86_OP_ENTRY3(VMOVHPx_ld, V,dq, H,dq,      M,q, vex4), /* MOVHPD */
+        X86_OP_ENTRY3(VMOVHPx_ld, V,dq, H,dq,      M,q, vex5), /* MOVHPS */
+        X86_OP_ENTRY3(VMOVHPx_ld, V,dq, H,dq,      M,q, vex5), /* MOVHPD */
         X86_OP_ENTRY3(VMOVSHDUP,  V,x,  None,None, W,x, vex4 cpuid(SSE3)),
         {},
     };
     static const X86OpEntry opcodes_0F16_reg[4] = {
         /* Same as above, operand 1 could be Hq if it wasn't for big-endian.  
*/
-        X86_OP_ENTRY3(VMOVLHPS,  V,dq, H,dq,      U,q, vex4),
-        X86_OP_ENTRY3(VMOVHPx,   V,x,  H,x,       U,x, vex4), /* MOVHPD */
+        X86_OP_ENTRY3(VMOVLHPS,  V,dq, H,dq,      U,q, vex7),
+        X86_OP_ENTRY3(VMOVHPx,   V,x,  H,x,       U,x, vex5), /* MOVHPD */
         X86_OP_ENTRY3(VMOVSHDUP, V,x,  None,None, U,x, vex4 cpuid(SSE3)),
         {},
     };
@@ -750,8 +750,9 @@ static void decode_0F2A(DisasContext *s, CPUX86State *env, 
X86OpEntry *entry, ui
 static void decode_0F2B(DisasContext *s, CPUX86State *env, X86OpEntry *entry, 
uint8_t *b)
 {
     static const X86OpEntry opcodes_0F2B[4] = {
-        X86_OP_ENTRY3(MOVDQ,      M,x,  None,None, V,x, vex4), /* MOVNTPS */
-        X86_OP_ENTRY3(MOVDQ,      M,x,  None,None, V,x, vex4), /* MOVNTPD */
+        X86_OP_ENTRY3(MOVDQ,      M,x,  None,None, V,x, vex1), /* MOVNTPS */
+        X86_OP_ENTRY3(MOVDQ,      M,x,  None,None, V,x, vex1), /* MOVNTPD */
+        /* AMD extensions */
         X86_OP_ENTRY3(VMOVSS_st,  M,ss, None,None, V,x, vex4 cpuid(SSE4A)), /* 
MOVNTSS */
         X86_OP_ENTRY3(VMOVLPx_st, M,sd, None,None, V,x, vex4 cpuid(SSE4A)), /* 
MOVNTSD */
     };
@@ -813,7 +814,7 @@ static void decode_0FE6(DisasContext *s, CPUX86State *env, 
X86OpEntry *entry, ui
     static const X86OpEntry opcodes_0FE6[4] = {
         {},
         X86_OP_ENTRY2(VCVTTPD2DQ,  V,x, W,x,      vex2),
-        X86_OP_ENTRY2(VCVTDQ2PD,   V,x, W,x,      vex2),
+        X86_OP_ENTRY2(VCVTDQ2PD,   V,x, W,x,      vex5),
         X86_OP_ENTRY2(VCVTPD2DQ,   V,x, W,x,      vex2),
     };
     *entry = *decode_by_prefix(s, opcodes_0FE6);
@@ -831,17 +832,17 @@ static const X86OpEntry opcodes_0F[256] = {
     [0x10] = X86_OP_GROUP0(0F10),
     [0x11] = X86_OP_GROUP0(0F11),
     [0x12] = X86_OP_GROUP0(0F12),
-    [0x13] = X86_OP_ENTRY3(VMOVLPx_st,  M,q, None,None, V,q,  vex4 p_00_66),
+    [0x13] = X86_OP_ENTRY3(VMOVLPx_st,  M,q, None,None, V,q,  vex5 p_00_66),
     [0x14] = X86_OP_ENTRY3(VUNPCKLPx,   V,x, H,x, W,x,        vex4 p_00_66),
     [0x15] = X86_OP_ENTRY3(VUNPCKHPx,   V,x, H,x, W,x,        vex4 p_00_66),
     [0x16] = X86_OP_GROUP0(0F16),
     /* Incorrectly listed as Mq,Vq in the manual */
-    [0x17] = X86_OP_ENTRY3(VMOVHPx_st,  M,q, None,None, V,dq, vex4 p_00_66),
+    [0x17] = X86_OP_ENTRY3(VMOVHPx_st,  M,q, None,None, V,dq, vex5 p_00_66),
 
     [0x50] = X86_OP_ENTRY3(MOVMSK,     G,y, None,None, U,x, vex7 p_00_66),
-    [0x51] = X86_OP_GROUP3(sse_unary,  V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
-    [0x52] = X86_OP_GROUP3(sse_unary,  V,x, H,x, W,x, vex4_rep5 p_00_f3),
-    [0x53] = X86_OP_GROUP3(sse_unary,  V,x, H,x, W,x, vex4_rep5 p_00_f3),
+    [0x51] = X86_OP_GROUP3(sse_unary,  V,x, H,x, W,x, vex2_rep3 
p_00_66_f3_f2), /* sqrtps */
+    [0x52] = X86_OP_GROUP3(sse_unary,  V,x, H,x, W,x, vex4_rep5 p_00_f3), /* 
rsqrtps */
+    [0x53] = X86_OP_GROUP3(sse_unary,  V,x, H,x, W,x, vex4_rep5 p_00_f3), /* 
rcpps */
     [0x54] = X86_OP_ENTRY3(PAND,       V,x, H,x, W,x,  vex4 p_00_66), /* vand 
*/
     [0x55] = X86_OP_ENTRY3(PANDN,      V,x, H,x, W,x,  vex4 p_00_66), /* vandn 
*/
     [0x56] = X86_OP_ENTRY3(POR,        V,x, H,x, W,x,  vex4 p_00_66), /* vor */
@@ -871,15 +872,15 @@ static const X86OpEntry opcodes_0F[256] = {
     [0x2B] = X86_OP_GROUP0(0F2B),
     [0x2C] = X86_OP_GROUP0(0F2C),
     [0x2D] = X86_OP_GROUP0(0F2D),
-    [0x2E] = X86_OP_ENTRY3(VUCOMI,     None,None, V,x, W,x,  vex4 p_00_66),
-    [0x2F] = X86_OP_ENTRY3(VCOMI,      None,None, V,x, W,x,  vex4 p_00_66),
+    [0x2E] = X86_OP_ENTRY3(VUCOMI,     None,None, V,x, W,x,  vex3 p_00_66),
+    [0x2F] = X86_OP_ENTRY3(VCOMI,      None,None, V,x, W,x,  vex3 p_00_66),
 
     [0x38] = X86_OP_GROUP0(0F38),
     [0x3a] = X86_OP_GROUP0(0F3A),
 
     [0x58] = X86_OP_ENTRY3(VADD,       V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
     [0x59] = X86_OP_ENTRY3(VMUL,       V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
-    [0x5a] = X86_OP_GROUP3(sse_unary,  V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
+    [0x5a] = X86_OP_GROUP3(sse_unary,  V,x, H,x, W,x, vex2_rep3 
p_00_66_f3_f2), /* CVTPS2PD */
     [0x5b] = X86_OP_GROUP0(0F5B),
     [0x5c] = X86_OP_ENTRY3(VSUB,       V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
     [0x5d] = X86_OP_ENTRY3(VMIN,       V,x, H,x, W,x, vex2_rep3 p_00_66_f3_f2),
-- 
2.39.1
[Prev in Thread]
Current Thread
[Next in Thread]
[PATCH] target/i386: Fix exception classes for SSE/AVX instructions., Ricky Zhou <=
Prev by Date: Re: [PATCH 8/9] bsd-user: implement sysctlbyname(2)
Next by Date: Re: [PATCH] vhost: avoid a potential use of an uninitialized variable in vhost_svq_poll()
Previous by thread: [PATCH] linux-user/mips: Low down switchable NaN2008 requirement
Next by thread: Ping [PATCH v5] Emulate dip switch language layout settings on SUN keyboard
Index(es):
- Date
- Thread