diff mbox series

[v8,47/50] x86emul: restore ordering within main switch statement

Message ID 5C8B875C020000780021F326@prv1-mh.provo.novell.com (mailing list archive)
State New, archived
Headers show
Series x86emul: remaining AVX512 support | expand

Commit Message

Jan Beulich March 15, 2019, 11:07 a.m. UTC 
Incremental additions and/or mistakes have lead to some code blocks
sitting in "unexpected" places. Re-sort the case blocks (opcode space;
major opcode; 66/F3/F2 prefix; legacy/VEX/EVEX encoding).

As an exception the opcode space 0x0f EVEX-encoded VPEXTRW is left at
its current place, to keep it close to the "pextr" label.

Pure code movement.

Signed-off-by: Jan Beulich <jbeulich@suse.com>
---
v7: New.

Comments

Andrew Cooper June 21, 2019, 1:20 p.m. UTC | #1 
On 15/03/2019 11:07, Jan Beulich wrote:
> Incremental additions and/or mistakes have lead to some code blocks
> sitting in "unexpected" places. Re-sort the case blocks (opcode space;
> major opcode; 66/F3/F2 prefix; legacy/VEX/EVEX encoding).
>
> As an exception the opcode space 0x0f EVEX-encoded VPEXTRW is left at
> its current place, to keep it close to the "pextr" label.
>
> Pure code movement.
>
> Signed-off-by: Jan Beulich <jbeulich@suse.com>

Acked-by: Andrew Cooper <andrew.cooper3@citirx.com>
diff mbox series

Patch

--- a/xen/arch/x86/x86_emulate/x86_emulate.c
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c
@@ -7129,15 +7129,6 @@  x86_emulate(
         ASSERT(!state->simd_size);
         break;
 
-    case X86EMUL_OPC_EVEX_F3(0x0f, 0x7e): /* vmovq xmm/m64,xmm */
-    case X86EMUL_OPC_EVEX_66(0x0f, 0xd6): /* vmovq xmm,xmm/m64 */
-        generate_exception_if(evex.lr || !evex.w || evex.opmsk || evex.brs,
-                              EXC_UD);
-        host_and_vcpu_must_have(avx512f);
-        d |= TwoOp;
-        op_bytes = 8;
-        goto simd_zmm;
-
     case X86EMUL_OPC_66(0x0f, 0xe7):     /* movntdq xmm,m128 */
     case X86EMUL_OPC_VEX_66(0x0f, 0xe7): /* vmovntdq {x,y}mm,mem */
         generate_exception_if(ea.type != OP_MEM, EXC_UD);
@@ -7535,6 +7526,15 @@  x86_emulate(
         op_bytes = 8;
         goto simd_0f_int;
 
+    case X86EMUL_OPC_EVEX_F3(0x0f, 0x7e): /* vmovq xmm/m64,xmm */
+    case X86EMUL_OPC_EVEX_66(0x0f, 0xd6): /* vmovq xmm,xmm/m64 */
+        generate_exception_if(evex.lr || !evex.w || evex.opmsk || evex.brs,
+                              EXC_UD);
+        host_and_vcpu_must_have(avx512f);
+        d |= TwoOp;
+        op_bytes = 8;
+        goto simd_zmm;
+
     case X86EMUL_OPC(0x0f, 0x80) ... X86EMUL_OPC(0x0f, 0x8f): /* jcc (near) */
         if ( test_cc(b, _regs.eflags) )
             jmp_rel((int32_t)src.val);
@@ -8635,63 +8635,6 @@  x86_emulate(
         dst.type = OP_NONE;
         break;
 
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x10): /* vpsrlvw [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x11): /* vpsravw [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x12): /* vpsllvw [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-        host_and_vcpu_must_have(avx512bw);
-        generate_exception_if(!evex.w || evex.brs, EXC_UD);
-        elem_bytes = 2;
-        goto avx512f_no_sae;
-
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x18): /* vbroadcastss xmm/m32,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x58): /* vpbroadcastd xmm/m32,[xyz]mm{k} */
-        op_bytes = elem_bytes;
-        generate_exception_if(evex.w || evex.brs, EXC_UD);
-    avx512_broadcast:
-        /*
-         * For the respective code below the main switch() to work we need to
-         * fold op_mask here: A source element gets read whenever any of its
-         * respective destination elements' mask bits is set.
-         */
-        if ( fault_suppression )
-        {
-            n = 1 << ((b & 3) - evex.w);
-            EXPECT(elem_bytes > 0);
-            ASSERT(op_bytes == n * elem_bytes);
-            for ( i = n; i < (16 << evex.lr) / elem_bytes; i += n )
-                op_mask |= (op_mask >> i) & ((1 << n) - 1);
-        }
-        goto avx512f_no_sae;
-
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x1b): /* vbroadcastf32x8 m256,zmm{k} */
-                                            /* vbroadcastf64x4 m256,zmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x5b): /* vbroadcasti32x8 m256,zmm{k} */
-                                            /* vbroadcasti64x4 m256,zmm{k} */
-        generate_exception_if(ea.type != OP_MEM || evex.lr != 2, EXC_UD);
-        /* fall through */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x19): /* vbroadcastsd xmm/m64,{y,z}mm{k} */
-                                            /* vbroadcastf32x2 xmm/m64,{y,z}mm{k} */
-        generate_exception_if(!evex.lr, EXC_UD);
-        /* fall through */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x59): /* vpbroadcastq xmm/m64,[xyz]mm{k} */
-                                            /* vbroadcasti32x2 xmm/m64,[xyz]mm{k} */
-        if ( b == 0x59 )
-            op_bytes = 8;
-        generate_exception_if(evex.brs, EXC_UD);
-        if ( !evex.w )
-            host_and_vcpu_must_have(avx512dq);
-        goto avx512_broadcast;
-
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x1a): /* vbroadcastf32x4 m128,{y,z}mm{k} */
-                                            /* vbroadcastf64x2 m128,{y,z}mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x5a): /* vbroadcasti32x4 m128,{y,z}mm{k} */
-                                            /* vbroadcasti64x2 m128,{y,z}mm{k} */
-        generate_exception_if(ea.type != OP_MEM || !evex.lr || evex.brs,
-                              EXC_UD);
-        if ( evex.w )
-            host_and_vcpu_must_have(avx512dq);
-        goto avx512_broadcast;
-
     case X86EMUL_OPC_66(0x0f38, 0x20): /* pmovsxbw xmm/m64,xmm */
     case X86EMUL_OPC_66(0x0f38, 0x21): /* pmovsxbd xmm/m32,xmm */
     case X86EMUL_OPC_66(0x0f38, 0x22): /* pmovsxbq xmm/m16,xmm */
@@ -8725,47 +8668,14 @@  x86_emulate(
         host_and_vcpu_must_have(sse4_1);
         goto simd_0f38_common;
 
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x13): /* vcvtph2ps xmm/mem,{x,y}mm */
-        generate_exception_if(vex.w, EXC_UD);
-        host_and_vcpu_must_have(f16c);
-        op_bytes = 8 << vex.l;
-        goto simd_0f_ymm;
-
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x13): /* vcvtph2ps {x,y}mm/mem,[xyz]mm{k} */
-        generate_exception_if(evex.w || (ea.type != OP_REG && evex.brs), EXC_UD);
-        host_and_vcpu_must_have(avx512f);
-        if ( !evex.brs )
-            avx512_vlen_check(false);
-        op_bytes = 8 << evex.lr;
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x10): /* vpsrlvw [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x11): /* vpsravw [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x12): /* vpsllvw [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+        host_and_vcpu_must_have(avx512bw);
+        generate_exception_if(!evex.w || evex.brs, EXC_UD);
         elem_bytes = 2;
-        goto simd_zmm;
-
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x16): /* vpermps ymm/m256,ymm,ymm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x36): /* vpermd ymm/m256,ymm,ymm */
-        generate_exception_if(!vex.l || vex.w, EXC_UD);
-        goto simd_0f_avx2;
-
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x16): /* vpermp{s,d} {y,z}mm/mem,{y,z}mm,{y,z}mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x36): /* vperm{d,q} {y,z}mm/mem,{y,z}mm,{y,z}mm{k} */
-        generate_exception_if(!evex.lr, EXC_UD);
-        fault_suppression = false;
         goto avx512f_no_sae;
 
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x20): /* vpmovsxbw xmm/mem,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x21): /* vpmovsxbd xmm/mem,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x22): /* vpmovsxbq xmm/mem,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x23): /* vpmovsxwd xmm/mem,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x24): /* vpmovsxwq xmm/mem,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x25): /* vpmovsxdq xmm/mem,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x30): /* vpmovzxbw xmm/mem,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x31): /* vpmovzxbd xmm/mem,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x32): /* vpmovzxbq xmm/mem,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x33): /* vpmovzxwd xmm/mem,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x34): /* vpmovzxwq xmm/mem,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x35): /* vpmovzxdq xmm/mem,{x,y}mm */
-        op_bytes = 16 >> (pmov_convert_delta[b & 7] - vex.l);
-        goto simd_0f_int;
-
     case X86EMUL_OPC_EVEX_F3(0x0f38, 0x10): /* vpmovuswb [xyz]mm,{x,y}mm/mem{k} */
     case X86EMUL_OPC_EVEX_66(0x0f38, 0x20): /* vpmovsxbw {x,y}mm/mem,[xyz]mm{k} */
     case X86EMUL_OPC_EVEX_F3(0x0f38, 0x20): /* vpmovswb [xyz]mm,{x,y}mm/mem{k} */
@@ -8811,6 +8721,96 @@  x86_emulate(
         elem_bytes = (b & 7) < 3 ? 1 : (b & 7) != 5 ? 2 : 4;
         goto avx512f_no_sae;
 
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x13): /* vcvtph2ps xmm/mem,{x,y}mm */
+        generate_exception_if(vex.w, EXC_UD);
+        host_and_vcpu_must_have(f16c);
+        op_bytes = 8 << vex.l;
+        goto simd_0f_ymm;
+
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x13): /* vcvtph2ps {x,y}mm/mem,[xyz]mm{k} */
+        generate_exception_if(evex.w || (ea.type != OP_REG && evex.brs), EXC_UD);
+        host_and_vcpu_must_have(avx512f);
+        if ( !evex.brs )
+            avx512_vlen_check(false);
+        op_bytes = 8 << evex.lr;
+        elem_bytes = 2;
+        goto simd_zmm;
+
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x16): /* vpermps ymm/m256,ymm,ymm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x36): /* vpermd ymm/m256,ymm,ymm */
+        generate_exception_if(!vex.l || vex.w, EXC_UD);
+        goto simd_0f_avx2;
+
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x16): /* vpermp{s,d} {y,z}mm/mem,{y,z}mm,{y,z}mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x36): /* vperm{d,q} {y,z}mm/mem,{y,z}mm,{y,z}mm{k} */
+        generate_exception_if(!evex.lr, EXC_UD);
+        fault_suppression = false;
+        goto avx512f_no_sae;
+
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x18): /* vbroadcastss xmm/m32,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x58): /* vpbroadcastd xmm/m32,[xyz]mm{k} */
+        op_bytes = elem_bytes;
+        generate_exception_if(evex.w || evex.brs, EXC_UD);
+    avx512_broadcast:
+        /*
+         * For the respective code below the main switch() to work we need to
+         * fold op_mask here: A source element gets read whenever any of its
+         * respective destination elements' mask bits is set.
+         */
+        if ( fault_suppression )
+        {
+            n = 1 << ((b & 3) - evex.w);
+            EXPECT(elem_bytes > 0);
+            ASSERT(op_bytes == n * elem_bytes);
+            for ( i = n; i < (16 << evex.lr) / elem_bytes; i += n )
+                op_mask |= (op_mask >> i) & ((1 << n) - 1);
+        }
+        goto avx512f_no_sae;
+
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x1b): /* vbroadcastf32x8 m256,zmm{k} */
+                                            /* vbroadcastf64x4 m256,zmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x5b): /* vbroadcasti32x8 m256,zmm{k} */
+                                            /* vbroadcasti64x4 m256,zmm{k} */
+        generate_exception_if(ea.type != OP_MEM || evex.lr != 2, EXC_UD);
+        /* fall through */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x19): /* vbroadcastsd xmm/m64,{y,z}mm{k} */
+                                            /* vbroadcastf32x2 xmm/m64,{y,z}mm{k} */
+        generate_exception_if(!evex.lr, EXC_UD);
+        /* fall through */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x59): /* vpbroadcastq xmm/m64,[xyz]mm{k} */
+                                            /* vbroadcasti32x2 xmm/m64,[xyz]mm{k} */
+        if ( b == 0x59 )
+            op_bytes = 8;
+        generate_exception_if(evex.brs, EXC_UD);
+        if ( !evex.w )
+            host_and_vcpu_must_have(avx512dq);
+        goto avx512_broadcast;
+
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x1a): /* vbroadcastf32x4 m128,{y,z}mm{k} */
+                                            /* vbroadcastf64x2 m128,{y,z}mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x5a): /* vbroadcasti32x4 m128,{y,z}mm{k} */
+                                            /* vbroadcasti64x2 m128,{y,z}mm{k} */
+        generate_exception_if(ea.type != OP_MEM || !evex.lr || evex.brs,
+                              EXC_UD);
+        if ( evex.w )
+            host_and_vcpu_must_have(avx512dq);
+        goto avx512_broadcast;
+
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x20): /* vpmovsxbw xmm/mem,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x21): /* vpmovsxbd xmm/mem,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x22): /* vpmovsxbq xmm/mem,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x23): /* vpmovsxwd xmm/mem,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x24): /* vpmovsxwq xmm/mem,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x25): /* vpmovsxdq xmm/mem,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x30): /* vpmovzxbw xmm/mem,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x31): /* vpmovzxbd xmm/mem,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x32): /* vpmovzxbq xmm/mem,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x33): /* vpmovzxwd xmm/mem,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x34): /* vpmovzxwq xmm/mem,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x35): /* vpmovzxdq xmm/mem,{x,y}mm */
+        op_bytes = 16 >> (pmov_convert_delta[b & 7] - vex.l);
+        goto simd_0f_int;
+
     case X86EMUL_OPC_EVEX_F3(0x0f38, 0x29): /* vpmov{b,w}2m [xyz]mm,k */
     case X86EMUL_OPC_EVEX_F3(0x0f38, 0x39): /* vpmov{d,q}2m [xyz]mm,k */
         generate_exception_if(!evex.r || !evex.R, EXC_UD);
@@ -8918,6 +8918,52 @@  x86_emulate(
         break;
     }
 
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x2c): /* vscalefp{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x42): /* vgetexpp{s,d} [xyz]mm/mem,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x96): /* vfmaddsub132p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x97): /* vfmsubadd132p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x98): /* vfmadd132p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x9a): /* vfmsub132p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x9c): /* vfnmadd132p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x9e): /* vfnmsub132p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xa6): /* vfmaddsub213p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xa7): /* vfmsubadd213p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xa8): /* vfmadd213p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xaa): /* vfmsub213p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xac): /* vfnmadd213p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xae): /* vfnmsub213p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xb6): /* vfmaddsub231p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xb7): /* vfmsubadd231p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xb8): /* vfmadd231p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xba): /* vfmsub231p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xbc): /* vfnmadd231p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xbe): /* vfnmsub231p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+        host_and_vcpu_must_have(avx512f);
+        if ( ea.type != OP_REG || !evex.brs )
+            avx512_vlen_check(false);
+        goto simd_zmm;
+
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x2d): /* vscalefs{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x43): /* vgetexps{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x99): /* vfmadd132s{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x9b): /* vfmsub132s{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x9d): /* vfnmadd132s{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x9f): /* vfnmsub132s{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xa9): /* vfmadd213s{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xab): /* vfmsub213s{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xad): /* vfnmadd213s{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xaf): /* vfnmsub213s{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xb9): /* vfmadd231s{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xbb): /* vfmsub231s{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xbd): /* vfnmadd231s{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xbf): /* vfnmsub231s{s,d} xmm/mem,xmm,xmm{k} */
+        host_and_vcpu_must_have(avx512f);
+    simd_zmm_scalar_sae:
+        generate_exception_if(ea.type != OP_REG && evex.brs, EXC_UD);
+        if ( !evex.brs )
+            avx512_vlen_check(true);
+        goto simd_zmm;
+
     case X86EMUL_OPC_66(0x0f38, 0x37): /* pcmpgtq xmm/m128,xmm */
         host_and_vcpu_must_have(sse4_2);
         goto simd_0f38_common;
@@ -8950,6 +8996,31 @@  x86_emulate(
         generate_exception_if(vex.l, EXC_UD);
         goto simd_0f_avx;
 
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x50): /* vpdpbusd [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x51): /* vpdpbusds [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x52): /* vpdpwssd [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x53): /* vpdpwssds [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+        host_and_vcpu_must_have(avx512_vnni);
+        generate_exception_if(evex.w, EXC_UD);
+        goto avx512f_no_sae;
+
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x58): /* vpbroadcastd xmm/m32,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x59): /* vpbroadcastq xmm/m64,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x78): /* vpbroadcastb xmm/m8,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x79): /* vpbroadcastw xmm/m16,{x,y}mm */
+        op_bytes = 1 << ((!(b & 0x20) * 2) + (b & 1));
+        /* fall through */
+    case X86EMUL_OPC_VEX_66(0x0f38, 0x46): /* vpsravd {x,y}mm/mem,{x,y}mm,{x,y}mm */
+        generate_exception_if(vex.w, EXC_UD);
+        goto simd_0f_avx2;
+
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x4d): /* vrcp14s{s,d} xmm/mem,xmm,xmm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0x4f): /* vrsqrt14s{s,d} xmm/mem,xmm,xmm{k} */
+        host_and_vcpu_must_have(avx512f);
+        generate_exception_if(evex.brs, EXC_UD);
+        avx512_vlen_check(true);
+        goto simd_zmm;
+
     case X86EMUL_OPC_EVEX_F2(0x0f38, 0x52): /* vp4dpwssd m128,zmm+3,zmm{k} */
     case X86EMUL_OPC_EVEX_F2(0x0f38, 0x53): /* vp4dpwssds m128,zmm+3,zmm{k} */
         host_and_vcpu_must_have(avx512_4vnniw);
@@ -8972,23 +9043,6 @@  x86_emulate(
         host_and_vcpu_must_have(avx512_vpopcntdq);
         goto avx512f_no_sae;
 
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x58): /* vpbroadcastd xmm/m32,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x59): /* vpbroadcastq xmm/m64,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x78): /* vpbroadcastb xmm/m8,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x79): /* vpbroadcastw xmm/m16,{x,y}mm */
-        op_bytes = 1 << ((!(b & 0x20) * 2) + (b & 1));
-        /* fall through */
-    case X86EMUL_OPC_VEX_66(0x0f38, 0x46): /* vpsravd {x,y}mm/mem,{x,y}mm,{x,y}mm */
-        generate_exception_if(vex.w, EXC_UD);
-        goto simd_0f_avx2;
-
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x4d): /* vrcp14s{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x4f): /* vrsqrt14s{s,d} xmm/mem,xmm,xmm{k} */
-        host_and_vcpu_must_have(avx512f);
-        generate_exception_if(evex.brs, EXC_UD);
-        avx512_vlen_check(true);
-        goto simd_zmm;
-
     case X86EMUL_OPC_VEX_66(0x0f38, 0x5a): /* vbroadcasti128 m128,ymm */
         generate_exception_if(ea.type != OP_MEM || !vex.l || vex.w, EXC_UD);
         goto simd_0f_avx2;
@@ -9370,60 +9424,6 @@  x86_emulate(
         host_and_vcpu_must_have(fma);
         goto simd_0f_ymm;
 
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x2c): /* vscalefp{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x42): /* vgetexpp{s,d} [xyz]mm/mem,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x96): /* vfmaddsub132p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x97): /* vfmsubadd132p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x98): /* vfmadd132p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x9a): /* vfmsub132p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x9c): /* vfnmadd132p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x9e): /* vfnmsub132p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xa6): /* vfmaddsub213p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xa7): /* vfmsubadd213p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xa8): /* vfmadd213p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xaa): /* vfmsub213p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xac): /* vfnmadd213p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xae): /* vfnmsub213p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xb6): /* vfmaddsub231p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xb7): /* vfmsubadd231p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xb8): /* vfmadd231p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xba): /* vfmsub231p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xbc): /* vfnmadd231p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xbe): /* vfnmsub231p{s,d} [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-        host_and_vcpu_must_have(avx512f);
-        if ( ea.type != OP_REG || !evex.brs )
-            avx512_vlen_check(false);
-        goto simd_zmm;
-
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x2d): /* vscalefs{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x43): /* vgetexps{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x99): /* vfmadd132s{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x9b): /* vfmsub132s{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x9d): /* vfnmadd132s{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x9f): /* vfnmsub132s{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xa9): /* vfmadd213s{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xab): /* vfmsub213s{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xad): /* vfnmadd213s{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xaf): /* vfnmsub213s{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xb9): /* vfmadd231s{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xbb): /* vfmsub231s{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xbd): /* vfnmadd231s{s,d} xmm/mem,xmm,xmm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0xbf): /* vfnmsub231s{s,d} xmm/mem,xmm,xmm{k} */
-        host_and_vcpu_must_have(avx512f);
-    simd_zmm_scalar_sae:
-        generate_exception_if(ea.type != OP_REG && evex.brs, EXC_UD);
-        if ( !evex.brs )
-            avx512_vlen_check(true);
-        goto simd_zmm;
-
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x50): /* vpdpbusd [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x51): /* vpdpbusds [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x52): /* vpdpwssd [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-    case X86EMUL_OPC_EVEX_66(0x0f38, 0x53): /* vpdpwssds [xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
-        host_and_vcpu_must_have(avx512_vnni);
-        generate_exception_if(evex.w, EXC_UD);
-        goto avx512f_no_sae;
-
     case X86EMUL_OPC_EVEX_F2(0x0f38, 0x9a): /* v4fmaddps m128,zmm+3,zmm{k} */
     case X86EMUL_OPC_EVEX_F2(0x0f38, 0xaa): /* v4fnmaddps m128,zmm+3,zmm{k} */
         host_and_vcpu_must_have(avx512_4fmaps);
@@ -10266,11 +10266,6 @@  x86_emulate(
         generate_exception_if(evex.brs || evex.opmsk, EXC_UD);
         goto avx512f_imm8_no_sae;
 
-    case X86EMUL_OPC_VEX_66(0x0f3a, 0x4a): /* vblendvps {x,y}mm,{x,y}mm/mem,{x,y}mm,{x,y}mm */
-    case X86EMUL_OPC_VEX_66(0x0f3a, 0x4b): /* vblendvpd {x,y}mm,{x,y}mm/mem,{x,y}mm,{x,y}mm */
-        generate_exception_if(vex.w, EXC_UD);
-        goto simd_0f_imm8_avx;
-
     case X86EMUL_OPC_VEX_66(0x0f3a, 0x48): /* vpermil2ps $imm,{x,y}mm/mem,{x,y}mm,{x,y}mm,{x,y}mm */
                                            /* vpermil2ps $imm,{x,y}mm,{x,y}mm/mem,{x,y}mm,{x,y}mm */
     case X86EMUL_OPC_VEX_66(0x0f3a, 0x49): /* vpermil2pd $imm,{x,y}mm/mem,{x,y}mm,{x,y}mm,{x,y}mm */
@@ -10278,6 +10273,11 @@  x86_emulate(
         host_and_vcpu_must_have(xop);
         goto simd_0f_imm8_ymm;
 
+    case X86EMUL_OPC_VEX_66(0x0f3a, 0x4a): /* vblendvps {x,y}mm,{x,y}mm/mem,{x,y}mm,{x,y}mm */
+    case X86EMUL_OPC_VEX_66(0x0f3a, 0x4b): /* vblendvpd {x,y}mm,{x,y}mm/mem,{x,y}mm,{x,y}mm */
+        generate_exception_if(vex.w, EXC_UD);
+        goto simd_0f_imm8_avx;
+
     case X86EMUL_OPC_VEX_66(0x0f3a, 0x4c): /* vpblendvb {x,y}mm,{x,y}mm/mem,{x,y}mm,{x,y}mm */
         generate_exception_if(vex.w, EXC_UD);
         goto simd_0f_int_imm8;