[llvm] r373706 - [X86] Enable inline memcmp() to use AVX512

[David Zarzycki via llvm-commits]

Author: davezarzycki
Date: Fri Oct  4 00:42:34 2019
New Revision: 373706

URL: http://llvm.org/viewvc/llvm-project?rev=373706&view=rev
Log:
[X86] Enable inline memcmp() to use AVX512

Modified:
    llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp
    llvm/trunk/test/CodeGen/X86/memcmp.ll

Modified: llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp?rev=373706&r1=373705&r2=373706&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86TargetTransformInfo.cpp Fri Oct  4 00:42:34 2019
@@ -3394,9 +3394,8 @@ X86TTIImpl::enableMemCmpExpansion(bool O
   if (IsZeroCmp) {
     // Only enable vector loads for equality comparison. Right now the vector
     // version is not as fast for three way compare (see #33329).
-    // TODO: enable AVX512 when the DAG is ready.
-    // if (ST->hasAVX512()) Options.LoadSizes.push_back(64);
     const unsigned PreferredWidth = ST->getPreferVectorWidth();
+    if (PreferredWidth >= 512 && ST->hasAVX512()) Options.LoadSizes.push_back(64);
     if (PreferredWidth >= 256 && ST->hasAVX2()) Options.LoadSizes.push_back(32);
     if (PreferredWidth >= 128 && ST->hasSSE2()) Options.LoadSizes.push_back(16);
     // All GPR and vector loads can be unaligned. SIMD compare requires integer

Modified: llvm/trunk/test/CodeGen/X86/memcmp.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/memcmp.ll?rev=373706&r1=373705&r2=373706&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/memcmp.ll (original)
+++ llvm/trunk/test/CodeGen/X86/memcmp.ll Fri Oct  4 00:42:34 2019
@@ -5,6 +5,8 @@
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown             | FileCheck %s --check-prefix=X64 --check-prefix=X64-SSE2
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx  | FileCheck %s --check-prefix=X64 --check-prefix=X64-AVX --check-prefix=X64-AVX1
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx2 | FileCheck %s --check-prefix=X64 --check-prefix=X64-AVX --check-prefix=X64-AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx512f | FileCheck %s --check-prefix=X64 --check-prefix=X64-AVX512F
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx512bw | FileCheck %s --check-prefix=X64 --check-prefix=X64-AVX512F --check-prefix=X64-AVX512BW
 
 ; This tests codegen time inlining/optimization of memcmp
 ; rdar://6480398
@@ -1540,6 +1542,15 @@ define i1 @length64_eq(i8* %x, i8* %y) n
 ; X64-AVX2-NEXT:    setne %al
 ; X64-AVX2-NEXT:    vzeroupper
 ; X64-AVX2-NEXT:    retq
+;
+; X64-AVX512F-LABEL: length64_eq:
+; X64-AVX512F:       # %bb.0:
+; X64-AVX512F-NEXT:    vmovdqu64 (%rdi), %zmm0
+; X64-AVX512F-NEXT:    vpcmpeqd (%rsi), %zmm0, %k0
+; X64-AVX512F-NEXT:    kortestw %k0, %k0
+; X64-AVX512F-NEXT:    setae %al
+; X64-AVX512F-NEXT:    vzeroupper
+; X64-AVX512F-NEXT:    retq
   %call = tail call i32 @memcmp(i8* %x, i8* %y, i64 64) nounwind
   %cmp = icmp ne i32 %call, 0
   ret i1 %cmp
@@ -1592,6 +1603,15 @@ define i1 @length64_eq_const(i8* %X) nou
 ; X64-AVX2-NEXT:    sete %al
 ; X64-AVX2-NEXT:    vzeroupper
 ; X64-AVX2-NEXT:    retq
+;
+; X64-AVX512F-LABEL: length64_eq_const:
+; X64-AVX512F:       # %bb.0:
+; X64-AVX512F-NEXT:    vmovdqu64 (%rdi), %zmm0
+; X64-AVX512F-NEXT:    vpcmpeqd {{.*}}(%rip), %zmm0, %k0
+; X64-AVX512F-NEXT:    kortestw %k0, %k0
+; X64-AVX512F-NEXT:    setb %al
+; X64-AVX512F-NEXT:    vzeroupper
+; X64-AVX512F-NEXT:    retq
   %m = tail call i32 @memcmp(i8* %X, i8* getelementptr inbounds ([65 x i8], [65 x i8]* @.str, i32 0, i32 0), i64 64) nounwind
   %c = icmp eq i32 %m, 0
   ret i1 %c