|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [Xen-devel] [PATCH v5 17/47] x86emul: support AVX512{F, BW} shift/rotate insns
Note that simd_packed_fp for the opcode space 0f38 major opcodes 14 and
15 is not really correct, but sufficient for the purposes here. Further
adjustments may later be needed for the down conversion unsigned
saturating VPMOV* insns, first and foremost for the different Disp8
scaling those ones use.
Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
---
v3: New.
--- a/tools/tests/x86_emulator/evex-disp8.c
+++ b/tools/tests/x86_emulator/evex-disp8.c
@@ -178,6 +178,24 @@ static const struct test avx512f_all[] =
INSN(pmulld, 66, 0f38, 40, vl, d, vl),
INSN(pmuludq, 66, 0f, f4, vl, q, vl),
INSN(por, 66, 0f, eb, vl, dq, vl),
+ INSNX(prol, 66, 0f, 72, 1, vl, dq, vl),
+ INSN(prolv, 66, 0f38, 15, vl, dq, vl),
+ INSNX(pror, 66, 0f, 72, 0, vl, dq, vl),
+ INSN(prorv, 66, 0f38, 14, vl, dq, vl),
+ INSN(pslld, 66, 0f, f2, el_4, d, vl),
+ INSNX(pslld, 66, 0f, 72, 6, vl, d, vl),
+ INSN(psllq, 66, 0f, f3, el_2, q, vl),
+ INSNX(psllq, 66, 0f, 73, 6, vl, q, vl),
+ INSN(psllv, 66, 0f38, 47, vl, dq, vl),
+ INSNX(psra, 66, 0f, 72, 4, vl, dq, vl),
+ INSN(psrad, 66, 0f, e2, el_4, d, vl),
+ INSN(psraq, 66, 0f, e2, el_2, q, vl),
+ INSN(psrav, 66, 0f38, 46, vl, dq, vl),
+ INSN(psrld, 66, 0f, d2, el_4, d, vl),
+ INSNX(psrld, 66, 0f, 72, 2, vl, d, vl),
+ INSN(psrlq, 66, 0f, d3, el_2, q, vl),
+ INSNX(psrlq, 66, 0f, 73, 2, vl, q, vl),
+ INSN(psrlv, 66, 0f38, 45, vl, dq, vl),
INSN(psubd, 66, 0f, fa, vl, d, vl),
INSN(psubq, 66, 0f, fb, vl, q, vl),
INSN(pternlog, 66, 0f3a, 25, vl, dq, vl),
@@ -241,6 +259,17 @@ static const struct test avx512bw_all[]
INSN(pmulhw, 66, 0f, e5, vl, w, vl),
INSN(pmullw, 66, 0f, d5, vl, w, vl),
INSN(psadbw, 66, 0f, f6, vl, b, vl),
+ INSNX(pslldq, 66, 0f, 73, 7, vl, b, vl),
+ INSN(psllvw, 66, 0f38, 12, vl, w, vl),
+ INSN(psllw, 66, 0f, f1, el_8, w, vl),
+ INSNX(psllw, 66, 0f, 71, 6, vl, w, vl),
+ INSN(psravw, 66, 0f38, 11, vl, w, vl),
+ INSN(psraw, 66, 0f, e1, el_8, w, vl),
+ INSNX(psraw, 66, 0f, 71, 4, vl, w, vl),
+ INSNX(psrldq, 66, 0f, 73, 3, vl, b, vl),
+ INSN(psrlvw, 66, 0f38, 10, vl, w, vl),
+ INSN(psrlw, 66, 0f, d1, el_8, w, vl),
+ INSNX(psrlw, 66, 0f, 71, 2, vl, w, vl),
INSN(psubb, 66, 0f, f8, vl, b, vl),
INSN(psubsb, 66, 0f, e8, vl, b, vl),
INSN(psubsw, 66, 0f, e9, vl, w, vl),
--- a/xen/arch/x86/x86_emulate/x86_emulate.c
+++ b/xen/arch/x86/x86_emulate/x86_emulate.c
@@ -319,7 +319,7 @@ static const struct twobyte_table {
[0x6e] = { DstImplicit|SrcMem|ModRM|Mov, simd_none, d8s_dq64 },
[0x6f] = { DstImplicit|SrcMem|ModRM|Mov, simd_packed_int, d8s_vl },
[0x70] = { SrcImmByte|ModRM|TwoOp, simd_other },
- [0x71 ... 0x73] = { DstImplicit|SrcImmByte|ModRM },
+ [0x71 ... 0x73] = { DstImplicit|SrcImmByte|ModRM, simd_none, d8s_vl },
[0x74 ... 0x76] = { DstImplicit|SrcMem|ModRM, simd_packed_int, d8s_vl },
[0x77] = { DstImplicit|SrcNone },
[0x78] = { ImplicitOps|ModRM },
@@ -366,19 +366,19 @@ static const struct twobyte_table {
[0xc7] = { ImplicitOps|ModRM },
[0xc8 ... 0xcf] = { ImplicitOps },
[0xd0] = { DstImplicit|SrcMem|ModRM, simd_other },
- [0xd1 ... 0xd3] = { DstImplicit|SrcMem|ModRM, simd_128 },
+ [0xd1 ... 0xd3] = { DstImplicit|SrcMem|ModRM, simd_128, 4 },
[0xd4 ... 0xd5] = { DstImplicit|SrcMem|ModRM, simd_packed_int, d8s_vl },
[0xd6] = { DstMem|SrcImplicit|ModRM|Mov, simd_other, 3 },
[0xd7] = { DstReg|SrcImplicit|ModRM|Mov },
[0xd8 ... 0xdf] = { DstImplicit|SrcMem|ModRM, simd_packed_int, d8s_vl },
[0xe0] = { DstImplicit|SrcMem|ModRM, simd_packed_int, d8s_vl },
- [0xe1 ... 0xe2] = { DstImplicit|SrcMem|ModRM, simd_128 },
+ [0xe1 ... 0xe2] = { DstImplicit|SrcMem|ModRM, simd_128, 4 },
[0xe3 ... 0xe5] = { DstImplicit|SrcMem|ModRM, simd_packed_int, d8s_vl },
[0xe6] = { DstImplicit|SrcMem|ModRM|Mov, simd_other },
[0xe7] = { DstMem|SrcImplicit|ModRM|Mov, simd_packed_int, d8s_vl },
[0xe8 ... 0xef] = { DstImplicit|SrcMem|ModRM, simd_packed_int, d8s_vl },
[0xf0] = { DstImplicit|SrcMem|ModRM|Mov, simd_other },
- [0xf1 ... 0xf3] = { DstImplicit|SrcMem|ModRM, simd_128 },
+ [0xf1 ... 0xf3] = { DstImplicit|SrcMem|ModRM, simd_128, 4 },
[0xf4 ... 0xf6] = { DstImplicit|SrcMem|ModRM, simd_packed_int, d8s_vl },
[0xf7] = { DstMem|SrcMem|ModRM|Mov, simd_packed_int },
[0xf8 ... 0xfe] = { DstImplicit|SrcMem|ModRM, simd_packed_int, d8s_vl },
@@ -434,9 +434,9 @@ static const struct ext0f38_table {
} ext0f38_table[256] = {
[0x00 ... 0x0b] = { .simd_size = simd_packed_int },
[0x0c ... 0x0f] = { .simd_size = simd_packed_fp },
- [0x10] = { .simd_size = simd_packed_int },
+ [0x10 ... 0x12] = { .simd_size = simd_packed_int, .d8s = d8s_vl },
[0x13] = { .simd_size = simd_other, .two_op = 1 },
- [0x14 ... 0x16] = { .simd_size = simd_packed_fp },
+ [0x14 ... 0x16] = { .simd_size = simd_packed_fp, .d8s = d8s_vl },
[0x17] = { .simd_size = simd_packed_int, .two_op = 1 },
[0x18] = { .simd_size = simd_scalar_opc, .two_op = 1, .d8s = 2 },
[0x19] = { .simd_size = simd_scalar_opc, .two_op = 1, .d8s = 3 },
@@ -453,7 +453,7 @@ static const struct ext0f38_table {
[0x36 ... 0x3f] = { .simd_size = simd_packed_int, .d8s = d8s_vl },
[0x40] = { .simd_size = simd_packed_int, .d8s = d8s_vl },
[0x41] = { .simd_size = simd_packed_int, .two_op = 1 },
- [0x45 ... 0x47] = { .simd_size = simd_packed_int },
+ [0x45 ... 0x47] = { .simd_size = simd_packed_int, .d8s = d8s_vl },
[0x58 ... 0x59] = { .simd_size = simd_other, .two_op = 1 },
[0x5a] = { .simd_size = simd_128, .two_op = 1 },
[0x78 ... 0x79] = { .simd_size = simd_other, .two_op = 1 },
@@ -5979,10 +5979,15 @@ x86_emulate(
case X86EMUL_OPC_EVEX_66(0x0f, 0xdf): /* vpandn{d,q}
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
case X86EMUL_OPC_EVEX_66(0x0f, 0xeb): /* vpor{d,q}
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
case X86EMUL_OPC_EVEX_66(0x0f, 0xef): /* vpxor{d,q}
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+ case X86EMUL_OPC_EVEX_66(0x0f38, 0x14): /* vprorv{d,q}
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+ case X86EMUL_OPC_EVEX_66(0x0f38, 0x15): /* vprolv{d,q}
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
case X86EMUL_OPC_EVEX_66(0x0f38, 0x39): /* vpmins{d,q}
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
case X86EMUL_OPC_EVEX_66(0x0f38, 0x3b): /* vpminu{d,q}
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
case X86EMUL_OPC_EVEX_66(0x0f38, 0x3d): /* vpmaxs{d,q}
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
case X86EMUL_OPC_EVEX_66(0x0f38, 0x3f): /* vpmaxu{d,q}
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+ case X86EMUL_OPC_EVEX_66(0x0f38, 0x45): /* vpsrlv{d,q}
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+ case X86EMUL_OPC_EVEX_66(0x0f38, 0x46): /* vpsrav{d,q}
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
+ case X86EMUL_OPC_EVEX_66(0x0f38, 0x47): /* vpsllv{d,q}
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
avx512f_no_sae:
host_and_vcpu_must_have(avx512f);
generate_exception_if(ea.type != OP_MEM && evex.br, EXC_UD);
@@ -6599,6 +6604,9 @@ x86_emulate(
get_fpu(X86EMUL_FPU_mmx);
goto simd_0f_common;
+ case X86EMUL_OPC_EVEX_66(0x0f, 0xd1): /* vpsrlw
xmm/m128,[xyz]mm,[xyz]mm{k} */
+ case X86EMUL_OPC_EVEX_66(0x0f, 0xe1): /* vpsraw
xmm/m128,[xyz]mm,[xyz]mm{k} */
+ case X86EMUL_OPC_EVEX_66(0x0f, 0xf1): /* vpsllw
xmm/m128,[xyz]mm,[xyz]mm{k} */
case X86EMUL_OPC_EVEX_66(0x0f, 0xf5): /* vpmaddwd
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
case X86EMUL_OPC_EVEX_66(0x0f, 0xf6): /* vpsadbw
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
fault_suppression = false;
@@ -6898,6 +6906,37 @@ x86_emulate(
ASSERT(!state->simd_size);
break;
+ case X86EMUL_OPC_EVEX_66(0x0f, 0x71): /* Grp12 */
+ switch ( modrm_reg & 7 )
+ {
+ case 2: /* vpsrlw $imm8,[xyz]mm/mem,[xyz]mm{k} */
+ case 4: /* vpsraw $imm8,[xyz]mm/mem,[xyz]mm{k} */
+ case 6: /* vpsllw $imm8,[xyz]mm/mem,[xyz]mm{k} */
+ avx512bw_shift_imm:
+ fault_suppression = false;
+ op_bytes = 16 << evex.lr;
+ state->simd_size = simd_packed_int;
+ goto avx512bw_imm;
+ }
+ goto unrecognized_insn;
+
+ case X86EMUL_OPC_EVEX_66(0x0f, 0x72): /* Grp13 */
+ switch ( modrm_reg & 7 )
+ {
+ case 2: /* vpsrld $imm8,[xyz]mm/mem,[xyz]mm{k} */
+ case 6: /* vpslld $imm8,[xyz]mm/mem,[xyz]mm{k} */
+ generate_exception_if(evex.w, EXC_UD);
+ /* fall through */
+ case 0: /* vpror{d,q} $imm8,[xyz]mm/mem,[xyz]mm{k} */
+ case 1: /* vprol{d,q} $imm8,[xyz]mm/mem,[xyz]mm{k} */
+ case 4: /* vpsra{d,q} $imm8,[xyz]mm/mem,[xyz]mm{k} */
+ avx512f_shift_imm:
+ op_bytes = 16 << evex.lr;
+ state->simd_size = simd_packed_int;
+ goto avx512f_imm8_no_sae;
+ }
+ goto unrecognized_insn;
+
case X86EMUL_OPC(0x0f, 0x73): /* Grp14 */
switch ( modrm_reg & 7 )
{
@@ -6923,6 +6962,19 @@ x86_emulate(
}
goto unrecognized_insn;
+ case X86EMUL_OPC_EVEX_66(0x0f, 0x73): /* Grp14 */
+ switch ( modrm_reg & 7 )
+ {
+ case 2: /* vpsrlq $imm8,[xyz]mm/mem,[xyz]mm{k} */
+ case 6: /* vpsllq $imm8,[xyz]mm/mem,[xyz]mm{k} */
+ generate_exception_if(!evex.w, EXC_UD);
+ goto avx512f_shift_imm;
+ case 3: /* vpsrldq $imm8,{x,y}mm,{x,y}mm */
+ case 7: /* vpslldq $imm8,{x,y}mm,{x,y}mm */
+ goto avx512bw_shift_imm;
+ }
+ goto unrecognized_insn;
+
case X86EMUL_OPC(0x0f, 0x77): /* emms */
case X86EMUL_OPC_VEX(0x0f, 0x77): /* vzero{all,upper} */
if ( vex.opcx != vex_none )
@@ -7858,6 +7910,16 @@ x86_emulate(
}
break;
+ case X86EMUL_OPC_EVEX_66(0x0f, 0xd2): /* vpsrld
xmm/m128,[xyz]mm,[xyz]mm{k} */
+ case X86EMUL_OPC_EVEX_66(0x0f, 0xd3): /* vpsrlq
xmm/m128,[xyz]mm,[xyz]mm{k} */
+ case X86EMUL_OPC_EVEX_66(0x0f, 0xe2): /* vpsra{d,q}
xmm/m128,[xyz]mm,[xyz]mm{k} */
+ case X86EMUL_OPC_EVEX_66(0x0f, 0xf2): /* vpslld
xmm/m128,[xyz]mm,[xyz]mm{k} */
+ case X86EMUL_OPC_EVEX_66(0x0f, 0xf3): /* vpsllq
xmm/m128,[xyz]mm,[xyz]mm{k} */
+ generate_exception_if(evex.br, EXC_UD);
+ fault_suppression = false;
+ if ( b == 0xe2 )
+ goto avx512f_no_sae;
+ /* fall through */
case X86EMUL_OPC_EVEX_66(0x0f, 0xfa): /* vpsubd
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
case X86EMUL_OPC_EVEX_66(0x0f, 0xfb): /* vpsubq
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
case X86EMUL_OPC_EVEX_66(0x0f, 0xfe): /* vpaddd
[xyz]mm/mem,[xyz]mm,[xyz]mm{k} */
@@ -8098,6 +8160,14 @@ 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.br, EXC_UD);
+ elem_bytes = 1 << evex.w;
+ goto avx512f_no_sae;
+
case X86EMUL_OPC_EVEX_66(0x0f38, 0x18): /* vbroadcastss xmm/m32,[xyz]mm{k}
*/
generate_exception_if(evex.w || evex.br, EXC_UD);
avx512_broadcast:
@@ -8855,6 +8925,7 @@ x86_emulate(
generate_exception_if(!evex.r || !evex.R || evex.z, EXC_UD);
if ( !(b & 0x20) )
goto avx512f_imm8_no_sae;
+ avx512bw_imm:
host_and_vcpu_must_have(avx512bw);
generate_exception_if(evex.br, EXC_UD);
elem_bytes = 1 << evex.w;
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel
|
 |
Lists.xenproject.org is hosted with RackSpace, monitoring our |