[llvm] 498c723 - [X86][SLM] +1uop for PSHUFBrm xmm

[Simon Pilgrim via llvm-commits]

Author: Simon Pilgrim
Date: 2021-10-17T18:13:14+01:00
New Revision: 498c7236bc763cdee7e7a26e4e411b05b32735c5

URL: https://github.com/llvm/llvm-project/commit/498c7236bc763cdee7e7a26e4e411b05b32735c5
DIFF: https://github.com/llvm/llvm-project/commit/498c7236bc763cdee7e7a26e4e411b05b32735c5.diff

LOG: [X86][SLM] +1uop for PSHUFBrm xmm

Extra 1uop for folded pshufb ops, based off a recent llvm-exegesis capture and what Intel AoM / Agner reports as well.

Added: 
    

Modified: 
    llvm/lib/Target/X86/X86ScheduleSLM.td
    llvm/test/tools/llvm-mca/X86/SLM/resources-ssse3.s

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/X86/X86ScheduleSLM.td b/llvm/lib/Target/X86/X86ScheduleSLM.td
index 51882bcab73d3..d2abe4099eec7 100644
--- a/llvm/lib/Target/X86/X86ScheduleSLM.td
+++ b/llvm/lib/Target/X86/X86ScheduleSLM.td
@@ -385,8 +385,8 @@ defm : SLMWriteResPair<WriteShuffleY, [SLM_FPC_RSV0],  1>;
 defm : X86WriteResPairUnsupported<WriteShuffleZ>;
 defm : SLMWriteResPair<WriteShuffleX, [SLM_FPC_RSV0],  1>;
 defm : SLMWriteResPair<WriteVarShuffle,  [SLM_FPC_RSV0],  1>;
-defm : SLMWriteResPair<WriteVarShuffleX, [SLM_FPC_RSV0],  5, [5], 4>;
-defm : SLMWriteResPair<WriteVarShuffleY, [SLM_FPC_RSV0],  5, [5], 4>;
+defm : SLMWriteResPair<WriteVarShuffleX, [SLM_FPC_RSV0],  5, [5], 4, 1>;
+defm : X86WriteResPairUnsupported<WriteVarShuffleY>;
 defm : X86WriteResPairUnsupported<WriteVarShuffleZ>;
 defm : SLMWriteResPair<WriteBlend,  [SLM_FPC_RSV0],  1>;
 defm : SLMWriteResPair<WriteBlendY, [SLM_FPC_RSV0],  1>;

diff  --git a/llvm/test/tools/llvm-mca/X86/SLM/resources-ssse3.s b/llvm/test/tools/llvm-mca/X86/SLM/resources-ssse3.s
index e74a73f5bb3d5..7fe497d9af271 100644
--- a/llvm/test/tools/llvm-mca/X86/SLM/resources-ssse3.s
+++ b/llvm/test/tools/llvm-mca/X86/SLM/resources-ssse3.s
@@ -157,7 +157,7 @@ psignw      (%rax), %xmm2
 # CHECK-NEXT:  1      1     1.00                        pshufb	%mm0, %mm2
 # CHECK-NEXT:  1      4     1.00    *                   pshufb	(%rax), %mm2
 # CHECK-NEXT:  4      5     5.00                        pshufb	%xmm0, %xmm2
-# CHECK-NEXT:  4      8     5.00    *                   pshufb	(%rax), %xmm2
+# CHECK-NEXT:  5      8     5.00    *                   pshufb	(%rax), %xmm2
 # CHECK-NEXT:  1      1     0.50                        psignb	%mm0, %mm2
 # CHECK-NEXT:  1      4     1.00    *                   psignb	(%rax), %mm2
 # CHECK-NEXT:  1      1     0.50                        psignb	%xmm0, %xmm2