[llvm] 37c3fbf - [X86] Add test showing failure to merge concatenatable VPERMV3 nodes

[Simon Pilgrim via llvm-commits]

Author: Simon Pilgrim
Date: 2025-03-17T15:10:41Z
New Revision: 37c3fbfa5ae9c124fffa4d53df708c688d8584c6

URL: https://github.com/llvm/llvm-project/commit/37c3fbfa5ae9c124fffa4d53df708c688d8584c6
DIFF: https://github.com/llvm/llvm-project/commit/37c3fbfa5ae9c124fffa4d53df708c688d8584c6.diff

LOG: [X86] Add test showing failure to merge concatenatable VPERMV3 nodes

Added: 
    

Modified: 
    llvm/test/CodeGen/X86/vector-shuffle-combining-avx512vbmi.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512vbmi.ll b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512vbmi.ll
index ccc78823623df..1301fcb0cd0b0 100644
--- a/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512vbmi.ll
+++ b/llvm/test/CodeGen/X86/vector-shuffle-combining-avx512vbmi.ll
@@ -158,3 +158,22 @@ define <64 x i8> @combine_vpermi2var_v64i8_with_mask(<64 x i8> %a0, <64 x i8> %a
   %sel = select <64 x i1> %cmp, <64 x i8> zeroinitializer, <64 x i8> %perm
   ret <64 x i8> %sel
 }
+
+define <32 x i8> @concat_vpermt2var_v16i8(<32 x i8> %x0, <32 x i8> %x1) {
+; CHECK-LABEL: concat_vpermt2var_v16i8:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,17,2,18,4,19,6,21,8,23,10,25,12,27,14,29]
+; CHECK-NEXT:    vpermi2b %xmm1, %xmm0, %xmm2
+; CHECK-NEXT:    vmovdqa {{.*#+}} xmm3 = [16,49,18,50,20,51,22,53,24,55,26,57,28,59,30,61]
+; CHECK-NEXT:    vpermi2b %ymm1, %ymm0, %ymm3
+; CHECK-NEXT:    vinserti128 $1, %xmm3, %ymm2, %ymm0
+; CHECK-NEXT:    ret{{[l|q]}}
+  %l0 = shufflevector <32 x i8> %x0, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %l1 = shufflevector <32 x i8> %x1, <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+  %h0 = shufflevector <32 x i8> %x0, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+  %h1 = shufflevector <32 x i8> %x1, <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+  %v0 = call <16 x i8> @llvm.x86.avx512.maskz.vpermt2var.qi.128(<16 x i8> <i8 0, i8 17, i8 2, i8 18, i8 4, i8 19, i8 6, i8 21, i8 8, i8 23, i8 10, i8 25, i8 12, i8 27, i8 14, i8 29>, <16 x i8> %l0, <16 x i8> %l1, i16 -1)
+  %v1 = call <16 x i8> @llvm.x86.avx512.maskz.vpermt2var.qi.128(<16 x i8> <i8 0, i8 17, i8 2, i8 18, i8 4, i8 19, i8 6, i8 21, i8 8, i8 23, i8 10, i8 25, i8 12, i8 27, i8 14, i8 29>, <16 x i8> %h0, <16 x i8> %h1, i16 -1)
+  %res = shufflevector <16 x i8> %v0, <16 x i8> %v1, <32 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
+  ret <32 x i8> %res
+}