Automerge: [X86] min-legal-vector-width.ll - fix AVX512 skylake check prefix clashes (#163783)

Fix check prefix clash warnings in update_llc_test_checks.py by adding an additional prefix for skylake targets with/without vbmi support

Author: RKSimon

llvm/test/CodeGen/X86/min-legal-vector-width.ll

@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=skylake-avx512 -mattr=prefer-256-bit | FileCheck %s --check-prefixes=CHECK,CHECK-SKX
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=skylake-avx512 -mattr=prefer-256-bit | FileCheck %s --check-prefixes=CHECK,CHECK-SKX,CHECK-SKX-NOVBMI
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=skylake-avx512 -mattr=prefer-256-bit,avx512vbmi | FileCheck %s --check-prefixes=CHECK,CHECK-SKX,CHECK-SKX-VBMI
 ; Make sure CPUs default to prefer-256-bit. avx512vnni isn't interesting as it just adds an isel peephole for vpmaddwd+vpaddd
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=skylake-avx512 | FileCheck %s --check-prefixes=CHECK,CHECK-AVX512
@@ -883,6 +883,30 @@ define <16 x i16> @test_16f32tosb_512(ptr %ptr, <16 x i16> %passthru) "min-legal
 }
 
 define dso_local void @mul256(ptr %a, ptr %b, ptr %c) "min-legal-vector-width"="256" {
+; CHECK-SKX-NOVBMI-LABEL: mul256:
+; CHECK-SKX-NOVBMI:       # %bb.0:
+; CHECK-SKX-NOVBMI-NEXT:    vmovdqa (%rdi), %ymm0
+; CHECK-SKX-NOVBMI-NEXT:    vmovdqa 32(%rdi), %ymm1
+; CHECK-SKX-NOVBMI-NEXT:    vmovdqa (%rsi), %ymm2
+; CHECK-SKX-NOVBMI-NEXT:    vmovdqa 32(%rsi), %ymm3
+; CHECK-SKX-NOVBMI-NEXT:    vpbroadcastd {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; CHECK-SKX-NOVBMI-NEXT:    vpand %ymm4, %ymm3, %ymm5
+; CHECK-SKX-NOVBMI-NEXT:    vpmaddubsw %ymm5, %ymm1, %ymm5
+; CHECK-SKX-NOVBMI-NEXT:    vpandn %ymm3, %ymm4, %ymm3
+; CHECK-SKX-NOVBMI-NEXT:    vpmaddubsw %ymm3, %ymm1, %ymm1
+; CHECK-SKX-NOVBMI-NEXT:    vpsllw $8, %ymm1, %ymm1
+; CHECK-SKX-NOVBMI-NEXT:    vpternlogq {{.*#+}} ymm1 = ymm1 | (ymm5 & ymm4)
+; CHECK-SKX-NOVBMI-NEXT:    vpand %ymm4, %ymm2, %ymm3
+; CHECK-SKX-NOVBMI-NEXT:    vpmaddubsw %ymm3, %ymm0, %ymm3
+; CHECK-SKX-NOVBMI-NEXT:    vpandn %ymm2, %ymm4, %ymm2
+; CHECK-SKX-NOVBMI-NEXT:    vpmaddubsw %ymm2, %ymm0, %ymm0
+; CHECK-SKX-NOVBMI-NEXT:    vpsllw $8, %ymm0, %ymm0
+; CHECK-SKX-NOVBMI-NEXT:    vpternlogq {{.*#+}} ymm0 = ymm0 | (ymm3 & ymm4)
+; CHECK-SKX-NOVBMI-NEXT:    vmovdqa %ymm0, (%rdx)
+; CHECK-SKX-NOVBMI-NEXT:    vmovdqa %ymm1, 32(%rdx)
+; CHECK-SKX-NOVBMI-NEXT:    vzeroupper
+; CHECK-SKX-NOVBMI-NEXT:    retq
+;
 ; CHECK-SKX-VBMI-LABEL: mul256:
 ; CHECK-SKX-VBMI:       # %bb.0:
 ; CHECK-SKX-VBMI-NEXT:    vmovdqa (%rdi), %ymm0
@@ -960,6 +984,21 @@ define dso_local void @mul256(ptr %a, ptr %b, ptr %c) "min-legal-vector-width"="
 }
 
 define dso_local void @mul512(ptr %a, ptr %b, ptr %c) "min-legal-vector-width"="512" {
+; CHECK-SKX-NOVBMI-LABEL: mul512:
+; CHECK-SKX-NOVBMI:       # %bb.0:
+; CHECK-SKX-NOVBMI-NEXT:    vmovdqa64 (%rdi), %zmm0
+; CHECK-SKX-NOVBMI-NEXT:    vmovdqa64 (%rsi), %zmm1
+; CHECK-SKX-NOVBMI-NEXT:    vpbroadcastd {{.*#+}} zmm2 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; CHECK-SKX-NOVBMI-NEXT:    vpandq %zmm2, %zmm1, %zmm3
+; CHECK-SKX-NOVBMI-NEXT:    vpmaddubsw %zmm3, %zmm0, %zmm3
+; CHECK-SKX-NOVBMI-NEXT:    vpandnq %zmm1, %zmm2, %zmm1
+; CHECK-SKX-NOVBMI-NEXT:    vpmaddubsw %zmm1, %zmm0, %zmm0
+; CHECK-SKX-NOVBMI-NEXT:    vpsllw $8, %zmm0, %zmm0
+; CHECK-SKX-NOVBMI-NEXT:    vpternlogq {{.*#+}} zmm0 = zmm0 | (zmm3 & zmm2)
+; CHECK-SKX-NOVBMI-NEXT:    vmovdqa64 %zmm0, (%rdx)
+; CHECK-SKX-NOVBMI-NEXT:    vzeroupper
+; CHECK-SKX-NOVBMI-NEXT:    retq
+;
 ; CHECK-SKX-VBMI-LABEL: mul512:
 ; CHECK-SKX-VBMI:       # %bb.0:
 ; CHECK-SKX-VBMI-NEXT:    vmovdqa64 (%rdi), %zmm0
@@ -1137,6 +1176,14 @@ define <16 x i16> @trunc_v16i32_v16i16_zeroes(ptr %x) nounwind "min-legal-vector
 }
 
 define <32 x i8> @trunc_v32i16_v32i8_zeroes(ptr %x) nounwind "min-legal-vector-width"="256" {
+; CHECK-SKX-NOVBMI-LABEL: trunc_v32i16_v32i8_zeroes:
+; CHECK-SKX-NOVBMI:       # %bb.0:
+; CHECK-SKX-NOVBMI-NEXT:    vpsrlw $8, 32(%rdi), %ymm0
+; CHECK-SKX-NOVBMI-NEXT:    vpsrlw $8, (%rdi), %ymm1
+; CHECK-SKX-NOVBMI-NEXT:    vpackuswb %ymm0, %ymm1, %ymm0
+; CHECK-SKX-NOVBMI-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
+; CHECK-SKX-NOVBMI-NEXT:    retq
+;
 ; CHECK-SKX-VBMI-LABEL: trunc_v32i16_v32i8_zeroes:
 ; CHECK-SKX-VBMI:       # %bb.0:
 ; CHECK-SKX-VBMI-NEXT:    vmovdqa (%rdi), %ymm1
@@ -1192,6 +1239,14 @@ define <16 x i16> @trunc_v16i32_v16i16_sign(ptr %x) nounwind "min-legal-vector-w
 }
 
 define <32 x i8> @trunc_v32i16_v32i8_sign(ptr %x) nounwind "min-legal-vector-width"="256" {
+; CHECK-SKX-NOVBMI-LABEL: trunc_v32i16_v32i8_sign:
+; CHECK-SKX-NOVBMI:       # %bb.0:
+; CHECK-SKX-NOVBMI-NEXT:    vpsrlw $8, 32(%rdi), %ymm0
+; CHECK-SKX-NOVBMI-NEXT:    vpsrlw $8, (%rdi), %ymm1
+; CHECK-SKX-NOVBMI-NEXT:    vpackuswb %ymm0, %ymm1, %ymm0
+; CHECK-SKX-NOVBMI-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
+; CHECK-SKX-NOVBMI-NEXT:    retq
+;
 ; CHECK-SKX-VBMI-LABEL: trunc_v32i16_v32i8_sign:
 ; CHECK-SKX-VBMI:       # %bb.0:
 ; CHECK-SKX-VBMI-NEXT:    vmovdqa (%rdi), %ymm1