[llvm-branch-commits] [llvm] release/18.x: [Codegen][X86] Fix /HOTPATCH with clang-cl and inline asm (#87639) (PR #88388)
https://github.com/KanRobert approved this pull request. LGTM https://github.com/llvm/llvm-project/pull/88388 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [llvm] eb64697 - [X86][Codegen] Correct the domain of VP2INTERSECT
Author: Shengchen Kan
Date: 2023-11-30T17:56:21+08:00
New Revision: eb64697a7b75d2b22041cc992fad0c8dfa7989cb
URL:
https://github.com/llvm/llvm-project/commit/eb64697a7b75d2b22041cc992fad0c8dfa7989cb
DIFF:
https://github.com/llvm/llvm-project/commit/eb64697a7b75d2b22041cc992fad0c8dfa7989cb.diff
LOG: [X86][Codegen] Correct the domain of VP2INTERSECT
GenericDomain -> SSEPackedInt
Found by #73654
Added:
Modified:
llvm/lib/Target/X86/X86InstrAVX512.td
llvm/test/CodeGen/X86/avx512vlvp2intersect-intrinsics.ll
llvm/test/CodeGen/X86/avx512vp2intersect-intrinsics.ll
llvm/test/CodeGen/X86/stack-folding-avx512vp2intersect.ll
llvm/test/CodeGen/X86/vp2intersect_multiple_pairs.ll
Removed:
diff --git a/llvm/lib/Target/X86/X86InstrAVX512.td
b/llvm/lib/Target/X86/X86InstrAVX512.td
index f325f47d46464c3..0514f0d19506707 100644
--- a/llvm/lib/Target/X86/X86InstrAVX512.td
+++ b/llvm/lib/Target/X86/X86InstrAVX512.td
@@ -12875,8 +12875,10 @@ multiclass avx512_vp2intersect;
defm VP2INTERSECTQ : avx512_vp2intersect,
REX_W;
+}
multiclass avx512_binop_all2 opc, string OpcodeStr,
X86SchedWriteWidths sched,
diff --git a/llvm/test/CodeGen/X86/avx512vlvp2intersect-intrinsics.ll
b/llvm/test/CodeGen/X86/avx512vlvp2intersect-intrinsics.ll
index ef07d30299e9dad..9741972767bcdcd 100644
--- a/llvm/test/CodeGen/X86/avx512vlvp2intersect-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512vlvp2intersect-intrinsics.ll
@@ -84,7 +84,7 @@ define void @test_mm256_2intersect_epi32_p(ptr nocapture
readonly %a, ptr nocapt
; X86-NEXT:movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x0c]
; X86-NEXT:movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x08]
; X86-NEXT:movl {{[0-9]+}}(%esp), %edx # encoding: [0x8b,0x54,0x24,0x04]
-; X86-NEXT:vmovaps (%edx), %ymm0 # EVEX TO VEX Compression encoding:
[0xc5,0xfc,0x28,0x02]
+; X86-NEXT:vmovdqa (%edx), %ymm0 # EVEX TO VEX Compression encoding:
[0xc5,0xfd,0x6f,0x02]
; X86-NEXT:vp2intersectd (%ecx), %ymm0, %k0 # encoding:
[0x62,0xf2,0x7f,0x28,0x68,0x01]
; X86-NEXT:kmovw %k1, %ecx # encoding: [0xc5,0xf8,0x93,0xc9]
; X86-NEXT:kmovw %k0, %edx # encoding: [0xc5,0xf8,0x93,0xd0]
@@ -96,7 +96,7 @@ define void @test_mm256_2intersect_epi32_p(ptr nocapture
readonly %a, ptr nocapt
;
; X64-LABEL: test_mm256_2intersect_epi32_p:
; X64: # %bb.0: # %entry
-; X64-NEXT:vmovaps (%rdi), %ymm0 # EVEX TO VEX Compression encoding:
[0xc5,0xfc,0x28,0x07]
+; X64-NEXT:vmovdqa (%rdi), %ymm0 # EVEX TO VEX Compression encoding:
[0xc5,0xfd,0x6f,0x07]
; X64-NEXT:vp2intersectd (%rsi), %ymm0, %k0 # encoding:
[0x62,0xf2,0x7f,0x28,0x68,0x06]
; X64-NEXT:kmovw %k1, %eax # encoding: [0xc5,0xf8,0x93,0xc1]
; X64-NEXT:kmovw %k0, %esi # encoding: [0xc5,0xf8,0x93,0xf0]
@@ -125,7 +125,7 @@ define void @test_mm256_2intersect_epi64_p(ptr nocapture
readonly %a, ptr nocapt
; X86-NEXT:movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x10]
; X86-NEXT:movl {{[0-9]+}}(%esp), %edx # encoding: [0x8b,0x54,0x24,0x0c]
; X86-NEXT:movl {{[0-9]+}}(%esp), %esi # encoding: [0x8b,0x74,0x24,0x08]
-; X86-NEXT:vmovaps (%esi), %ymm0 # EVEX TO VEX Compression encoding:
[0xc5,0xfc,0x28,0x06]
+; X86-NEXT:vmovdqa (%esi), %ymm0 # EVEX TO VEX Compression encoding:
[0xc5,0xfd,0x6f,0x06]
; X86-NEXT:vp2intersectq (%edx), %ymm0, %k0 # encoding:
[0x62,0xf2,0xff,0x28,0x68,0x02]
; X86-NEXT:kshiftlw $12, %k0, %k2 # encoding:
[0xc4,0xe3,0xf9,0x32,0xd0,0x0c]
; X86-NEXT:kshiftrw $12, %k2, %k2 # encoding:
[0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
@@ -142,7 +142,7 @@ define void @test_mm256_2intersect_epi64_p(ptr nocapture
readonly %a, ptr nocapt
;
; X64-LABEL: test_mm256_2intersect_epi64_p:
; X64: # %bb.0: # %entry
-; X64-NEXT:vmovaps (%rdi), %ymm0 # EVEX TO VEX Compression encoding:
[0xc5,0xfc,0x28,0x07]
+; X64-NEXT:vmovdqa (%rdi), %ymm0 # EVEX TO VEX Compression encoding:
[0xc5,0xfd,0x6f,0x07]
; X64-NEXT:vp2intersectq (%rsi), %ymm0, %k0 # encoding:
[0x62,0xf2,0xff,0x28,0x68,0x06]
; X64-NEXT:kshiftlw $12, %k0, %k2 # encoding:
[0xc4,0xe3,0xf9,0x32,0xd0,0x0c]
; X64-NEXT:kshiftrw $12, %k2, %k2 # encoding:
[0xc4,0xe3,0xf9,0x30,0xd2,0x0c]
@@ -175,7 +175,7 @@ define void @test_mm256_2intersect_epi32_b(ptr nocapture
readonly %a, ptr nocapt
; X86-NEXT:movl {{[0-9]+}}(%esp), %eax # encoding: [0x8b,0x44,0x24,0x0c]
; X86-NEXT:movl {{[0-9]+}}(%esp), %ecx # encoding: [0x8b,0x4c,0x24,0x08]
; X86-NEXT:movl {{[0-9]+}}(%esp), %edx # encoding: [0x8b,0x54,0x24,0x04]
-; X86-NEXT:vbroadcastss (%edx), %ymm0 # EVEX TO VEX Compression encoding:
[0xc4,0xe2,0x7d,0x18,0x02]
+; X86-NEXT:vpbroadcastd (%edx), %ymm0 # EVEX TO VEX Compression encoding:
[0xc4,0xe2,0x7d,0x58,0x02]
; X86-NEXT:vp2intersectd (%ecx){1to8}, %ymm0, %k0 # encoding:
[0x62,0xf2,0x7f,0x38,0x68,0x01]
;
[llvm-branch-commits] [llvm] ae3cd2d - [X86][NFC] Clang-format X86FoldTablesUtils.h
Author: Shengchen Kan
Date: 2023-11-30T17:16:40+08:00
New Revision: ae3cd2d04a88937881ac1f9c5d34f0e0cf96d5bc
URL:
https://github.com/llvm/llvm-project/commit/ae3cd2d04a88937881ac1f9c5d34f0e0cf96d5bc
DIFF:
https://github.com/llvm/llvm-project/commit/ae3cd2d04a88937881ac1f9c5d34f0e0cf96d5bc.diff
LOG: [X86][NFC] Clang-format X86FoldTablesUtils.h
Added:
Modified:
llvm/include/llvm/Support/X86FoldTablesUtils.h
Removed:
diff --git a/llvm/include/llvm/Support/X86FoldTablesUtils.h
b/llvm/include/llvm/Support/X86FoldTablesUtils.h
index bddff7068b82690..1cce9cdaf65f8d6 100644
--- a/llvm/include/llvm/Support/X86FoldTablesUtils.h
+++ b/llvm/include/llvm/Support/X86FoldTablesUtils.h
@@ -13,23 +13,23 @@ namespace llvm {
enum {
// Select which memory operand is being unfolded.
// (stored in bits 0 - 2)
- TB_INDEX_0= 0,
- TB_INDEX_1= 1,
- TB_INDEX_2= 2,
- TB_INDEX_3= 3,
- TB_INDEX_4= 4,
+ TB_INDEX_0 = 0,
+ TB_INDEX_1 = 1,
+ TB_INDEX_2 = 2,
+ TB_INDEX_3 = 3,
+ TB_INDEX_4 = 4,
TB_INDEX_MASK = 0x7,
// Do not insert the reverse map (MemOp -> RegOp) into the table.
// This may be needed because there is a many -> one mapping.
- TB_NO_REVERSE = 1 << 3,
+ TB_NO_REVERSE = 1 << 3,
// Do not insert the forward map (RegOp -> MemOp) into the table.
// This is needed for Native Client, which prohibits branch
// instructions from using a memory operand.
- TB_NO_FORWARD = 1 << 4,
+ TB_NO_FORWARD = 1 << 4,
- TB_FOLDED_LOAD = 1 << 5,
+ TB_FOLDED_LOAD = 1 << 5,
TB_FOLDED_STORE = 1 << 6,
TB_FOLDED_BCAST = 1 << 7,
@@ -37,19 +37,19 @@ enum {
// Used for RegOp->MemOp conversion. Encoded as Log2(Align)
// (stored in bits 9 - 11)
TB_ALIGN_SHIFT = 8,
- TB_ALIGN_1 = 0 << TB_ALIGN_SHIFT,
- TB_ALIGN_16= 4 << TB_ALIGN_SHIFT,
- TB_ALIGN_32= 5 << TB_ALIGN_SHIFT,
- TB_ALIGN_64= 6 << TB_ALIGN_SHIFT,
- TB_ALIGN_MASK = 0x7 << TB_ALIGN_SHIFT,
+ TB_ALIGN_1 = 0 << TB_ALIGN_SHIFT,
+ TB_ALIGN_16 = 4 << TB_ALIGN_SHIFT,
+ TB_ALIGN_32 = 5 << TB_ALIGN_SHIFT,
+ TB_ALIGN_64 = 6 << TB_ALIGN_SHIFT,
+ TB_ALIGN_MASK = 0x7 << TB_ALIGN_SHIFT,
// Broadcast type.
// (stored in bits 12 - 13)
TB_BCAST_TYPE_SHIFT = TB_ALIGN_SHIFT + 3,
- TB_BCAST_D= 0 << TB_BCAST_TYPE_SHIFT,
- TB_BCAST_Q= 1 << TB_BCAST_TYPE_SHIFT,
- TB_BCAST_SS = 2 << TB_BCAST_TYPE_SHIFT,
- TB_BCAST_SD = 3 << TB_BCAST_TYPE_SHIFT,
+ TB_BCAST_D = 0 << TB_BCAST_TYPE_SHIFT,
+ TB_BCAST_Q = 1 << TB_BCAST_TYPE_SHIFT,
+ TB_BCAST_SS = 2 << TB_BCAST_TYPE_SHIFT,
+ TB_BCAST_SD = 3 << TB_BCAST_TYPE_SHIFT,
TB_BCAST_MASK = 0x3 << TB_BCAST_TYPE_SHIFT,
// Unused bits 14-15
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [llvm] PR for llvm/llvm-project#79675 (PR #79721)
KanRobert wrote: > > @phoebewang What do you think about merging this PR to the release branch? > > This patch fixes prior mistakes, so should be merged to release branch. > @KanRobert Do I understand it right? Agree, though I haven't notice any runfail due to this mistake. https://github.com/llvm/llvm-project/pull/79721 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [llvm] release/19.x: [X86][APX] Fix wrong encoding of promoted KMOV instructions due to missing NoCD8 (#109759) (PR #109767)
KanRobert wrote: > @KanRobert @KanRobert What do you think about merging this PR to the release > branch? LGTM https://github.com/llvm/llvm-project/pull/109767 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [llvm] release/19.x: [X86][APX] Fix wrong encoding of promoted KMOV instructions due to missing NoCD8 (#109759) (PR #109767)
https://github.com/KanRobert approved this pull request. https://github.com/llvm/llvm-project/pull/109767 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
