This is an archive of the discontinued LLVM Phabricator instance.

Differential D28024

[X86][AVX2] Passing the appropriate memory operand class to VPMADDWD instruction.
ClosedPublic

Authored by aymanmus on Dec 21 2016, 7:46 AM.

Details

Reviewers
zvi
delena
igorb
Commits
rG9ff608cdc648: [X86][AVX2] Passing the appropriate memory operand class to VPMADDWD…
rL290333: [X86][AVX2] Passing the appropriate memory operand class to VPMADDWD…
Summary

Replacing the memory operand in the ymm version of VPMADDWD from i128mem to i256mem.

Diff Detail

Event Timeline

aymanmus updated this revision to Diff 82236.Dec 21 2016, 7:46 AM
aymanmus retitled this revision from to [X86][AVX2] Passing the appropriate memory operand class to VPMADDWD instruction..
aymanmus updated this object.
aymanmus added reviewers: zvi, igorb, delena.
aymanmus added a subscriber: llvm-commits.
igorb accepted this revision.Dec 22 2016, 12:13 AM
igorb edited edge metadata.

LGTM

This revision is now accepted and ready to land.Dec 22 2016, 12:13 AM
Closed by commit rL290333: [X86][AVX2] Passing the appropriate memory operand class to VPMADDWD… (authored by aymanmus). · Explain WhyDec 22 2016, 12:53 AM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
lib/
Target/
X86/
2 lines
test/
MC/
X86/
56 lines
58 lines

Diff 82236

lib/Target/X86/X86InstrSSE.td

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 4,037 Lines • ▼ Show 20 Linesdefm PAVGW : PDI_binop_all<0xE3, "pavgw", X86avg, v8i16, v16i16,
SSE_INTALU_ITINS_P, 1, NoVLX_Or_NoBWI>; SSE_INTALU_ITINS_P, 1, NoVLX_Or_NoBWI>;
let Predicates = [HasAVX, NoVLX_Or_NoBWI] in let Predicates = [HasAVX, NoVLX_Or_NoBWI] in
defm VPMADDWD : PDI_binop_rm2<0xF5, "vpmaddwd", X86vpmaddwd, v4i32, v8i16, VR128, defm VPMADDWD : PDI_binop_rm2<0xF5, "vpmaddwd", X86vpmaddwd, v4i32, v8i16, VR128,
loadv2i64, i128mem, SSE_PMADD, 0>, VEX_4V; loadv2i64, i128mem, SSE_PMADD, 0>, VEX_4V;
let Predicates = [HasAVX2, NoVLX_Or_NoBWI] in let Predicates = [HasAVX2, NoVLX_Or_NoBWI] in
defm VPMADDWDY : PDI_binop_rm2<0xF5, "vpmaddwd", X86vpmaddwd, v8i32, v16i16, defm VPMADDWDY : PDI_binop_rm2<0xF5, "vpmaddwd", X86vpmaddwd, v8i32, v16i16,
VR256, loadv4i64, i128mem, SSE_PMADD, VR256, loadv4i64, i256mem, SSE_PMADD,
0>, VEX_4V, VEX_L; 0>, VEX_4V, VEX_L;
let Constraints = "$src1 = $dst" in let Constraints = "$src1 = $dst" in
defm PMADDWD : PDI_binop_rm2<0xF5, "pmaddwd", X86vpmaddwd, v4i32, v8i16, VR128, defm PMADDWD : PDI_binop_rm2<0xF5, "pmaddwd", X86vpmaddwd, v4i32, v8i16, VR128,
memopv2i64, i128mem, SSE_PMADD>; memopv2i64, i128mem, SSE_PMADD>;
let Predicates = [HasAVX, NoVLX_Or_NoBWI] in let Predicates = [HasAVX, NoVLX_Or_NoBWI] in
defm VPSADBW : PDI_binop_rm2<0xF6, "vpsadbw", X86psadbw, v2i64, v16i8, VR128, defm VPSADBW : PDI_binop_rm2<0xF6, "vpsadbw", X86psadbw, v2i64, v16i8, VR128,
loadv2i64, i128mem, SSE_INTMUL_ITINS_P, 0>, loadv2i64, i128mem, SSE_INTMUL_ITINS_P, 0>,
▲ Show 20 Lines • Show All 4,748 Lines • Show Last 20 Lines

test/MC/X86/intel-syntax-x86-64-avx.s

Show First 20 Lines • Show All 153 Lines • ▼ Show 20 Lines
// CHECK: vcvttpd2dq xmm0, ymmword ptr [rax] // CHECK: vcvttpd2dq xmm0, ymmword ptr [rax]
// CHECK: encoding: [0xc5,0xfd,0xe6,0x00] // CHECK: encoding: [0xc5,0xfd,0xe6,0x00]
vcvttpd2dq xmm0, ymmword ptr [rax] vcvttpd2dq xmm0, ymmword ptr [rax]
// CHECK: vcvttpd2dq xmm0, ymmword ptr [rax] // CHECK: vcvttpd2dq xmm0, ymmword ptr [rax]
// CHECK: encoding: [0xc5,0xfd,0xe6,0x00] // CHECK: encoding: [0xc5,0xfd,0xe6,0x00]
vcvttpd2dqy xmm0, ymmword ptr [rax] vcvttpd2dqy xmm0, ymmword ptr [rax]
// CHECK: vpmaddwd xmm1, xmm2, xmm3
// CHECK: encoding: [0xc5,0xe9,0xf5,0xcb]
vpmaddwd xmm1, xmm2, xmm3
// CHECK: vpmaddwd xmm1, xmm2, xmmword ptr [rcx]
// CHECK: encoding: [0xc5,0xe9,0xf5,0x09]
vpmaddwd xmm1, xmm2, xmmword ptr [rcx]
// CHECK: vpmaddwd xmm1, xmm2, xmmword ptr [rsp - 4]
// CHECK: encoding: [0xc5,0xe9,0xf5,0x4c,0x24,0xfc]
vpmaddwd xmm1, xmm2, xmmword ptr [rsp - 4]
// CHECK: vpmaddwd xmm1, xmm2, xmmword ptr [rsp + 4]
// CHECK: encoding: [0xc5,0xe9,0xf5,0x4c,0x24,0x04]
vpmaddwd xmm1, xmm2, xmmword ptr [rsp + 4]
// CHECK: vpmaddwd xmm1, xmm2, xmmword ptr [rcx + 8*r14 + 268435456]
// CHECK: encoding: [0xc4,0xa1,0x69,0xf5,0x8c,0xf1,0x00,0x00,0x00,0x10]
vpmaddwd xmm1, xmm2, xmmword ptr [rcx + 8*r14 + 268435456]
// CHECK: vpmaddwd xmm1, xmm2, xmmword ptr [rcx + 8*r14 - 536870912]
// CHECK: encoding: [0xc4,0xa1,0x69,0xf5,0x8c,0xf1,0x00,0x00,0x00,0xe0]
vpmaddwd xmm1, xmm2, xmmword ptr [rcx + 8*r14 - 536870912]
// CHECK: vpmaddwd xmm1, xmm2, xmmword ptr [rcx + 8*r14 - 536870910]
// CHECK: encoding: [0xc4,0xa1,0x69,0xf5,0x8c,0xf1,0x02,0x00,0x00,0xe0]
vpmaddwd xmm1, xmm2, xmmword ptr [rcx + 8*r14 - 536870910]
// CHECK: vpmaddwd ymm1, ymm2, ymm3
// CHECK: encoding: [0xc5,0xed,0xf5,0xcb]
vpmaddwd ymm1, ymm2, ymm3
// CHECK: vpmaddwd ymm1, ymm2, ymmword ptr [rcx]
// CHECK: encoding: [0xc5,0xed,0xf5,0x09]
vpmaddwd ymm1, ymm2, ymmword ptr [rcx]
// CHECK: vpmaddwd ymm1, ymm2, ymmword ptr [rsp - 4]
// CHECK: encoding: [0xc5,0xed,0xf5,0x4c,0x24,0xfc]
vpmaddwd ymm1, ymm2, ymmword ptr [rsp - 4]
// CHECK: vpmaddwd ymm1, ymm2, ymmword ptr [rsp + 4]
// CHECK: encoding: [0xc5,0xed,0xf5,0x4c,0x24,0x04]
vpmaddwd ymm1, ymm2, ymmword ptr [rsp + 4]
// CHECK: vpmaddwd ymm1, ymm2, ymmword ptr [rcx + 8*r14 + 268435456]
// CHECK: encoding: [0xc4,0xa1,0x6d,0xf5,0x8c,0xf1,0x00,0x00,0x00,0x10]
vpmaddwd ymm1, ymm2, ymmword ptr [rcx + 8*r14 + 268435456]
// CHECK: vpmaddwd ymm1, ymm2, ymmword ptr [rcx + 8*r14 - 536870912]
// CHECK: encoding: [0xc4,0xa1,0x6d,0xf5,0x8c,0xf1,0x00,0x00,0x00,0xe0]
vpmaddwd ymm1, ymm2, ymmword ptr [rcx + 8*r14 - 536870912]
// CHECK: vpmaddwd ymm1, ymm2, ymmword ptr [rcx + 8*r14 - 536870910]
// CHECK: encoding: [0xc4,0xa1,0x6d,0xf5,0x8c,0xf1,0x02,0x00,0x00,0xe0]
vpmaddwd ymm1, ymm2, ymmword ptr [rcx + 8*r14 - 536870910]

test/MC/X86/x86_64-avx-encoding.s

Show First 20 Lines • Show All 4,239 Lines • ▼ Show 20 Lines
// CHECK: vmovsd %xmm0, %xmm8, %xmm0 // CHECK: vmovsd %xmm0, %xmm8, %xmm0
// CHECK: encoding: [0xc5,0xbb,0x10,0xc0] // CHECK: encoding: [0xc5,0xbb,0x10,0xc0]
vmovsd %xmm0, %xmm8, %xmm0 vmovsd %xmm0, %xmm8, %xmm0
// CHECK: vmovsd %xmm8, %xmm0, %xmm0 // CHECK: vmovsd %xmm8, %xmm0, %xmm0
// CHECK: encoding: [0xc5,0x7b,0x11,0xc0] // CHECK: encoding: [0xc5,0x7b,0x11,0xc0]
vmovsd %xmm8, %xmm0, %xmm0 vmovsd %xmm8, %xmm0, %xmm0
// CHECK: vpmaddwd %xmm3, %xmm2, %xmm1
// CHECK: encoding: [0xc5,0xe9,0xf5,0xcb]
vpmaddwd %xmm3, %xmm2, %xmm1
// CHECK: vpmaddwd (%rcx), %xmm2, %xmm1
// CHECK: encoding: [0xc5,0xe9,0xf5,0x09]
vpmaddwd (%rcx), %xmm2, %xmm1
// CHECK: vpmaddwd -4(%rsp), %xmm2, %xmm1
// CHECK: encoding: [0xc5,0xe9,0xf5,0x4c,0x24,0xfc]
vpmaddwd -4(%rsp), %xmm2, %xmm1
// CHECK: vpmaddwd 4(%rsp), %xmm2, %xmm1
// CHECK: encoding: [0xc5,0xe9,0xf5,0x4c,0x24,0x04]
vpmaddwd 4(%rsp), %xmm2, %xmm1
// CHECK: vpmaddwd 268435456(%rcx,%r14,8), %xmm2, %xmm1
// CHECK: encoding: [0xc4,0xa1,0x69,0xf5,0x8c,0xf1,0x00,0x00,0x00,0x10]
vpmaddwd 268435456(%rcx,%r14,8), %xmm2, %xmm1
// CHECK: vpmaddwd -536870912(%rcx,%r14,8), %xmm2, %xmm1
// CHECK: encoding: [0xc4,0xa1,0x69,0xf5,0x8c,0xf1,0x00,0x00,0x00,0xe0]
vpmaddwd -536870912(%rcx,%r14,8), %xmm2, %xmm1
// CHECK: vpmaddwd -536870910(%rcx,%r14,8), %xmm2, %xmm1
// CHECK: encoding: [0xc4,0xa1,0x69,0xf5,0x8c,0xf1,0x02,0x00,0x00,0xe0]
vpmaddwd -536870910(%rcx,%r14,8), %xmm2, %xmm1
// CHECK: vpmaddwd %ymm3, %ymm2, %ymm1
// CHECK: encoding: [0xc5,0xed,0xf5,0xcb]
vpmaddwd %ymm3, %ymm2, %ymm1
// CHECK: vpmaddwd (%rcx), %ymm2, %ymm1
// CHECK: encoding: [0xc5,0xed,0xf5,0x09]
vpmaddwd (%rcx), %ymm2, %ymm1
// CHECK: vpmaddwd -4(%rsp), %ymm2, %ymm1
// CHECK: encoding: [0xc5,0xed,0xf5,0x4c,0x24,0xfc]
vpmaddwd -4(%rsp), %ymm2, %ymm1
// CHECK: vpmaddwd 4(%rsp), %ymm2, %ymm1
// CHECK: encoding: [0xc5,0xed,0xf5,0x4c,0x24,0x04]
vpmaddwd 4(%rsp), %ymm2, %ymm1
// CHECK: vpmaddwd 268435456(%rcx,%r14,8), %ymm2, %ymm1
// CHECK: encoding: [0xc4,0xa1,0x6d,0xf5,0x8c,0xf1,0x00,0x00,0x00,0x10]
vpmaddwd 268435456(%rcx,%r14,8), %ymm2, %ymm1
// CHECK: vpmaddwd -536870912(%rcx,%r14,8), %ymm2, %ymm1
// CHECK: encoding: [0xc4,0xa1,0x6d,0xf5,0x8c,0xf1,0x00,0x00,0x00,0xe0]
vpmaddwd -536870912(%rcx,%r14,8), %ymm2, %ymm1
// CHECK: vpmaddwd -536870910(%rcx,%r14,8), %ymm2, %ymm1
// CHECK: encoding: [0xc4,0xa1,0x6d,0xf5,0x8c,0xf1,0x02,0x00,0x00,0xe0]
vpmaddwd -536870910(%rcx,%r14,8), %ymm2, %ymm1