[PULL 07/68] target/i386: Fix exception classes for SSE/AVX instructions.

[Paolo Bonzini]

From: Ricky Zhou <ricky@rzhou.org>

Fix the exception classes for some SSE/AVX instructions to match what is
documented in the Intel manual.

These changes are expected to have no functional effect on the behavior
that qemu implements (primarily >= 16-byte memory alignment checks). For
instance, 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.
Message-Id: <20230501111428.95998-2-ricky@rzhou.org>

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 target/i386/tcg/decode-new.c.inc | 46 ++++++++++++++++----------------
 1 file changed, 23 insertions(+), 23 deletions(-)

diff --git a/target/i386/tcg/decode-new.c.inc b/target/i386/tcg/decode-new.c.inc
index e0f25e7fd8ed..1a789201f740 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 
*/
         {},
@@ -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) {
@@ -662,15 +662,15 @@ static void decode_0F11(DisasContext *s, CPUX86State 
*env, X86OpEntry *entry, ui
     static const X86OpEntry opcodes_0F11_reg[4] = {
         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, vex4),
-        X86_OP_ENTRY3(VMOVLPx, W,x,  H,x,       V,q, vex4), /* MOVSD */
+        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), /* MOVUPS */
         X86_OP_ENTRY3(MOVDQ,      W,x,  None,None, V,x, vex4), /* MOVUPD */
-        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(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)),
         {},
     };
@@ -824,7 +824,7 @@ 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);
@@ -842,12 +842,12 @@ 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), /* sqrtps */
-- 
2.40.1