From 56055822b4d3a656ec0676de193c1065dc332188 Mon Sep 17 00:00:00 2001 From: "Wang, Xin10" Date: Wed, 17 May 2023 23:36:15 -0400 Subject: [X86]Fix wrong asm match for VMASKMOVDQU VMASKMOVDQU supports 32bit/64bit version in 64bitmode, previously we prefer to use VMASKMOVDQU64 in 64bitmode because the 32bit one need 0x67 prefix. After D150436, asm match table changed a little, which makes in 64bit mode "vmaskmovdqu %xmm0, %xmm1" will match VMASKMOVDQU other than VMASKMOVDQU64, this patch correct the asm match order for this instruction. Reviewed By: skan Differential Revision: https://reviews.llvm.org/D150835 --- llvm/lib/Target/X86/X86InstrSSE.td | 12 ++++++------ llvm/test/MC/X86/x86_64-asm-match.s | 6 ++++++ 2 files changed, 12 insertions(+), 6 deletions(-) diff --git a/llvm/lib/Target/X86/X86InstrSSE.td b/llvm/lib/Target/X86/X86InstrSSE.td index 4296528bfc9b..abb36e09db73 100644 --- a/llvm/lib/Target/X86/X86InstrSSE.td +++ b/llvm/lib/Target/X86/X86InstrSSE.td @@ -4070,18 +4070,18 @@ let ExeDomain = SSEPackedInt, SchedRW = [SchedWriteVecMoveLS.XMM.MR] in { // As VEX does not have separate instruction contexts for address size // overrides, VMASKMOVDQU and VMASKMOVDQU64 would have a decode conflict. // Prefer VMASKMODDQU64. -let Uses = [EDI], Predicates = [HasAVX], isAsmParserOnly = 1 in -def VMASKMOVDQU : VPDI<0xF7, MRMSrcReg, (outs), - (ins VR128:$src, VR128:$mask), - "maskmovdqu\t{$mask, $src|$src, $mask}", - [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, EDI)]>, - VEX, WIG; let Uses = [RDI], Predicates = [HasAVX,In64BitMode] in def VMASKMOVDQU64 : VPDI<0xF7, MRMSrcReg, (outs), (ins VR128:$src, VR128:$mask), "maskmovdqu\t{$mask, $src|$src, $mask}", [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, RDI)]>, VEX, WIG; +let Uses = [EDI], Predicates = [HasAVX], isAsmParserOnly = 1 in +def VMASKMOVDQU : VPDI<0xF7, MRMSrcReg, (outs), + (ins VR128:$src, VR128:$mask), + "maskmovdqu\t{$mask, $src|$src, $mask}", + [(int_x86_sse2_maskmov_dqu VR128:$src, VR128:$mask, EDI)]>, + VEX, WIG; let Uses = [EDI], Predicates = [UseSSE2] in def MASKMOVDQU : PDI<0xF7, MRMSrcReg, (outs), (ins VR128:$src, VR128:$mask), diff --git a/llvm/test/MC/X86/x86_64-asm-match.s b/llvm/test/MC/X86/x86_64-asm-match.s index f436fd2ded33..d4ec34790106 100644 --- a/llvm/test/MC/X86/x86_64-asm-match.s +++ b/llvm/test/MC/X86/x86_64-asm-match.s @@ -37,6 +37,11 @@ // CHECK: Matching formal operand class MCK_GR32 against actual operand at index 2 (Reg:ecx): match success using generic matcher // CHECK: Matching formal operand class InvalidMatchClass against actual operand at index 3: actual operand index out of range // CHECK: Opcode result: complete match, selecting this opcode +// CHECK: AsmMatcher: found 2 encodings with mnemonic 'vmaskmovdqu' +// CHECK: Trying to match opcode VMASKMOVDQU64 +// CHECK: Matching formal operand class MCK_FR16 against actual operand at index 1 (Reg:xmm0): match success using generic matcher +// CHECK: Matching formal operand class MCK_FR16 against actual operand at index 2 (Reg:xmm1): match success using generic matcher +// CHECK: Matching formal operand class InvalidMatchClass against actual operand at index 3: actual operand index out of range // CHECK: AsmMatcher: found 4 encodings with mnemonic 'punpcklbw' // CHECK: Trying to match opcode MMX_PUNPCKLBWrr // CHECK: Matching formal operand class MCK_VR64 against actual operand at index 1 (Reg:mm0): match success using generic matcher @@ -51,6 +56,7 @@ pshufb CPI1_0(%rip), %xmm1 sha1rnds4 $1, %xmm1, %xmm2 pinsrw $3, %ecx, %xmm5 crc32l %gs:0xdeadbeef(%rbx,%rcx,8),%ecx +vmaskmovdqu %xmm0, %xmm1 .intel_syntax punpcklbw mm0, dword ptr [rsp] -- cgit v1.2.1