RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-24 03:33:20 +01:00
Code Issues Projects Releases Wiki Activity

[AVX512] Fix load opcode for fast isel.

Browse Source 
Differential Revision: http://reviews.llvm.org/D21067

llvm-svn: 272006
This commit is contained in:
Igor Breger 2016-06-07 13:08:45 +00:00
parent f4980a0fe0
commit 53dd84b6c0
2 changed files with 22 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
lib/Target/X86/X86FastISel.cpp
View File

@ -452,7 +452,7 @@ bool X86FastISel::X86FastEmitLoad(EVT VT, X86AddressMode &AM,
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;
}

21
test/CodeGen/X86/fast-isel-vecload.ll
View File

@ -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
}