[PATCH v4 11/16] x86emul: support AVX10.2 media insns

[Jan Beulich <jbeulich@xxxxxxxx>]

These are all very similar to various existing insns.

Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
---
SDE: -dmr / -future
---
In evex-disp8.c, should we separately have avx10_v1_aux_all[]? If so, all
the convert insns would also need moving/copying there.
---
v4: Series re-ordering adjustments. Update to spec version 6.
v3: New.

--- a/tools/tests/x86_emulator/evex-disp8.c
+++ b/tools/tests/x86_emulator/evex-disp8.c
@@ -725,6 +725,20 @@ static const struct test avx10_2_all[] =
     INSN(comisbf16,        66, map5, 2f,    el, bf16, el),
     INSN_SFP(comx,               0f, 2f),
     INSN(comxsh,           f3, map5, 2f,    el, fp16, el),
+    INSN(dpphps,             , 0f38, 52,    vl,    d, vl),
+    INSN(mpsadbw,          f3, 0f3a, 42,    vl, d_nb, vl),
+    INSN(pdpbssd,          f2, 0f38, 50,    vl,    d, vl),
+    INSN(pdpbssds,         f2, 0f38, 51,    vl,    d, vl),
+    INSN(pdpbsud,          f3, 0f38, 50,    vl,    d, vl),
+    INSN(pdpbsuds,         f3, 0f38, 51,    vl,    d, vl),
+    INSN(pdpbuud,            , 0f38, 50,    vl,    d, vl),
+    INSN(pdpbuuds,           , 0f38, 51,    vl,    d, vl),
+    INSN(pdpwsud,          f3, 0f38, d2,    vl,    d, vl),
+    INSN(pdpwsuds,         f3, 0f38, d3,    vl,    d, vl),
+    INSN(pdpwusd,          66, 0f38, d2,    vl,    d, vl),
+    INSN(pdpwusds,         66, 0f38, d3,    vl,    d, vl),
+    INSN(pdpwuud,            , 0f38, d2,    vl,    d, vl),
+    INSN(pdpwuuds,           , 0f38, d3,    vl,    d, vl),
     INSN_SFP(ucomx,              0f, 2e),
     INSN(ucomxsh,          f3, map5, 2e,    el, fp16, el),
 };
--- a/tools/tests/x86_emulator/predicates.c
+++ b/tools/tests/x86_emulator/predicates.c
@@ -1927,8 +1927,15 @@ static const struct evex {
     { { 0x4d }, 2, T, R, pfx_66, Wn, LIG }, /* vrcp14s{s,d} */
     { { 0x4e }, 2, T, R, pfx_66, Wn, Ln }, /* vrsqrt14p{s,d} */
     { { 0x4f }, 2, T, R, pfx_66, Wn, LIG }, /* vrsqrt14s{s,d} */
+    { { 0x50 }, 2, T, R, pfx_no, W0, Ln }, /* vpdpbuud */
     { { 0x50 }, 2, T, R, pfx_66, W0, Ln }, /* vpdpbusd */
+    { { 0x50 }, 2, T, R, pfx_f3, W0, Ln }, /* vpdpbsud */
+    { { 0x50 }, 2, T, R, pfx_f2, W0, Ln }, /* vpdpbssd */
+    { { 0x51 }, 2, T, R, pfx_no, W0, Ln }, /* vpdpbuuds */
     { { 0x51 }, 2, T, R, pfx_66, W0, Ln }, /* vpdpbusds */
+    { { 0x51 }, 2, T, R, pfx_f3, W0, Ln }, /* vpdpbsuds */
+    { { 0x51 }, 2, T, R, pfx_f2, W0, Ln }, /* vpdpbssds */
+    { { 0x52 }, 2, T, R, pfx_no, W0, Ln }, /* vdpphps */
     { { 0x52 }, 2, T, R, pfx_66, W0, Ln }, /* vpdpwssd */
     { { 0x52 }, 2, T, R, pfx_f3, W0, Ln }, /* vdpbf16ps */
     { { 0x52 }, 2, T, R, pfx_f2, W0, L2 }, /* vp4dpwssd */
@@ -2029,6 +2036,12 @@ static const struct evex {
     { { 0xcc }, 2, T, R, pfx_66, Wn, L2 }, /* vrsqrt28p{s,d} */
     { { 0xcd }, 2, T, R, pfx_66, Wn, LIG }, /* vrsqrt28s{s,d} */
     { { 0xcf }, 2, T, R, pfx_66, W0, Ln }, /* vgf2p8mulb */
+    { { 0xd2 }, 2, T, R, pfx_no, W0, Ln }, /* vpdpwuud */
+    { { 0xd2 }, 2, T, R, pfx_66, W0, Ln }, /* vpdpwusd */
+    { { 0xd2 }, 2, T, R, pfx_f3, W0, Ln }, /* vpdpwsud */
+    { { 0xd3 }, 2, T, R, pfx_no, W0, Ln }, /* vpdpwuuds */
+    { { 0xd3 }, 2, T, R, pfx_66, W0, Ln }, /* vpdpwusds */
+    { { 0xd3 }, 2, T, R, pfx_f3, W0, Ln }, /* vpdpwsuds */
     { { 0xda }, 2, T, R, pfx_f3, W0, Ln }, /* vsm4key4 */
     { { 0xda }, 2, T, R, pfx_f2, W0, Ln }, /* vsm4rnds4 */
     { { 0xdc }, 2, T, R, pfx_66, WIG, Ln }, /* vaesenc */
@@ -2075,6 +2088,7 @@ static const struct evex {
     { { 0x3e }, 3, T, R, pfx_66, Wn, Ln }, /* vpcmpu{b,w} */
     { { 0x3f }, 3, T, R, pfx_66, Wn, Ln }, /* vpcmp{b,w} */
     { { 0x42 }, 3, T, R, pfx_66, W0, Ln }, /* vdbpsadbw */
+    { { 0x42 }, 3, T, R, pfx_f3, W0, Ln }, /* vmpsadbw */
     { { 0x43 }, 3, T, R, pfx_66, Wn, L1|L2 }, /* vshufi{32x4,64x2} */
     { { 0x44 }, 3, T, R, pfx_66, WIG, Ln }, /* vpclmulqdq */
     { { 0x50 }, 3, T, R, pfx_66, Wn, Ln }, /* vrangep{s,d} */
--- a/xen/arch/x86/x86_emulate/decode.c
+++ b/xen/arch/x86/x86_emulate/decode.c
@@ -434,8 +434,8 @@ static const struct ext0f38_table {
     [0xcb] = { .simd_size = simd_other, .d8s = d8s_vl },
     [0xcc ... 0xcd] = { .simd_size = simd_other, .two_op = 1, .d8s = d8s_vl },
     [0xcf] = { .simd_size = simd_packed_int, .d8s = d8s_vl },
-    [0xd2] = { .simd_size = simd_other },
-    [0xd3] = { .simd_size = simd_other },
+    [0xd2] = { .simd_size = simd_other, .d8s = d8s_vl },
+    [0xd3] = { .simd_size = simd_other, .d8s = d8s_vl },
     [0xd6] = { .simd_size = simd_other, .d8s = d8s_vl },
     [0xd7] = { .simd_size = simd_scalar_vexw, .d8s = d8s_dq },
     [0xda] = { .simd_size = simd_other, .d8s = d8s_vl },
--- a/xen/arch/x86/x86_emulate/x86_emulate.c
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c
@@ -6169,6 +6169,30 @@ x86_emulate(
         avx512_vlen_check(true);
         goto simd_zmm;
 
+    case X86EMUL_OPC_EVEX   (0x0f38, 0x50): /* vpdpbuud 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_F3(0x0f38, 0x50): /* vpdpbsud 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_F2(0x0f38, 0x50): /* vpdpbssd 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX   (0x0f38, 0x51): /* vpdpbuuds 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_F3(0x0f38, 0x51): /* vpdpbsuds 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_F2(0x0f38, 0x51): /* vpdpbssds 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX   (0x0f38, 0xd2): /* vpdpwuud 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xd2): /* vpdpwusd 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_F3(0x0f38, 0xd2): /* vpdpwsud 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX   (0x0f38, 0xd3): /* vpdpwuuds 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_66(0x0f38, 0xd3): /* vpdpwusds 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+    case X86EMUL_OPC_EVEX_F3(0x0f38, 0xd3): /* vpdpwsuds 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+        generate_exception_if(evex.w, X86_EXC_UD);
+        if ( !cp->avx10.avx10_v1_aux )
+            vcpu_must_have(avx10, 2);
+        op_bytes = 16 << evex.lr;
+        goto avx512f_no_sae;
+
+    case X86EMUL_OPC_EVEX   (0x0f38, 0x52): /* vdpphps 
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+        generate_exception_if(evex.w, X86_EXC_UD);
+        vcpu_must_have(avx10, 2);
+        op_bytes = 16 << evex.lr;
+        goto avx512f_no_sae;
+
     case X86EMUL_OPC_EVEX_66(0x0f38, 0x8f): /* vpshufbitqmb 
[xyz]mm/mem,[xyz]mm,k{k} */
         generate_exception_if(evex.w || !evex.r || !evex.R || evex.z, 
X86_EXC_UD);
         /* fall through */
@@ -7635,6 +7659,14 @@ x86_emulate(
         visa_check(bw);
         goto opmask_shift_imm;
 
+    case X86EMUL_OPC_EVEX_F3(0x0f3a, 0x42): /* vmpsadbw 
$imm8,[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+        generate_exception_if(evex.w || evex.brs, X86_EXC_UD);
+        vcpu_must_have(avx10, 2);
+        avx512_vlen_check(false);
+        op_bytes = 16 << evex.lr;
+        fault_suppression = false;
+        goto simd_imm8_zmm;
+
     case X86EMUL_OPC_66(0x0f3a, 0x44):     /* pclmulqdq $imm8,xmm/m128,xmm */
     case X86EMUL_OPC_VEX_66(0x0f3a, 0x44): /* vpclmulqdq 
$imm8,{x,y}mm/mem,{x,y}mm,{x,y}mm */
         host_and_vcpu_must_have(pclmulqdq);