[llvm] r272006 - [AVX512] Fix load opcode for fast isel.
Igor Breger via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 7 06:08:45 PDT 2016
Author: ibreger
Date: Tue Jun 7 08:08:45 2016
New Revision: 272006
URL: http://llvm.org/viewvc/llvm-project?rev=272006&view=rev
Log:
[AVX512] Fix load opcode for fast isel.
Differential Revision: http://reviews.llvm.org/D21067
Modified:
llvm/trunk/lib/Target/X86/X86FastISel.cpp
llvm/trunk/test/CodeGen/X86/fast-isel-vecload.ll
Modified: llvm/trunk/lib/Target/X86/X86FastISel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86FastISel.cpp?rev=272006&r1=272005&r2=272006&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86FastISel.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86FastISel.cpp Tue Jun 7 08:08:45 2016
@@ -452,7 +452,7 @@ bool X86FastISel::X86FastEmitLoad(EVT VT
assert(Subtarget->hasAVX512());
// Note: There are a lot more choices based on type with AVX-512, but
// there's really no advantage when the load isn't masked.
- Opc = (Alignment >= 64) ? X86::VMOVDQA64Zmr : X86::VMOVDQU64Zmr;
+ Opc = (Alignment >= 64) ? X86::VMOVDQA64Zrm : X86::VMOVDQU64Zrm;
RC = &X86::VR512RegClass;
break;
}
Modified: llvm/trunk/test/CodeGen/X86/fast-isel-vecload.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/fast-isel-vecload.ll?rev=272006&r1=272005&r2=272006&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/fast-isel-vecload.ll (original)
+++ llvm/trunk/test/CodeGen/X86/fast-isel-vecload.ll Tue Jun 7 08:08:45 2016
@@ -1,5 +1,6 @@
; RUN: llc -O0 -fast-isel -fast-isel-abort=1 -mtriple=x86_64-unknown-unknown -mattr=+sse2 < %s | FileCheck %s --check-prefix=SSE --check-prefix=ALL
; RUN: llc -O0 -fast-isel -fast-isel-abort=1 -mtriple=x86_64-unknown-unknown -mattr=+avx < %s | FileCheck %s --check-prefix=AVX --check-prefix=ALL
+; RUN: llc -O0 -fast-isel -fast-isel-abort=1 -mtriple=x86_64-unknown-unknown -mattr=+avx512f < %s | FileCheck %s --check-prefix=KNL
; Verify that fast-isel knows how to select aligned/unaligned vector loads.
; Also verify that the selected load instruction is in the correct domain.
@@ -183,3 +184,23 @@ entry:
%0 = load <2 x double>, <2 x double>* %V
ret <2 x double> %0
}
+
+define <8 x i64> @test_v8i64_alignment(<8 x i64>* %V) {
+; KNL-LABEL: test_v8i64_alignment:
+; KNL: # BB#0: # %entry
+; KNL-NEXT: vmovdqa64 (%rdi), %zmm0
+; KNL-NEXT: retq
+entry:
+ %0 = load <8 x i64>, <8 x i64>* %V, align 64
+ ret <8 x i64> %0
+}
+
+define <8 x i64> @test_v8i64(<8 x i64>* %V) {
+; KNL-LABEL: test_v8i64:
+; KNL: # BB#0: # %entry
+; KNL-NEXT: vmovdqu64 (%rdi), %zmm0
+; KNL-NEXT: retq
+entry:
+ %0 = load <8 x i64>, <8 x i64>* %V, align 4
+ ret <8 x i64> %0
+}
More information about the llvm-commits
mailing list