RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-12-12 22:17:47 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[X86][AVX512] Add patterns that match the AVX512 floating point register vbroadcast intrinsics.

Browse Source 
Patch by Cameron McInally <cameron.mcinally@nyu.edu>

llvm-svn: 193422
This commit is contained in:
Quentin Colombet 2013-10-25 18:04:12 +00:00
parent 5650890143
commit 5d88e45af6
2 changed files with 19 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
lib/Target/X86/X86InstrAVX512.td
View File

@ -465,6 +465,11 @@ def : Pat<(v16f32 (X86VBroadcast (v4f32 VR128X:$src))),
(VBROADCASTSSZrr VR128X:$src)>;
def : Pat<(v8f64 (X86VBroadcast (v2f64 VR128X:$src))),
(VBROADCASTSDZrr VR128X:$src)>;
def : Pat<(v16f32 (int_x86_avx512_vbroadcast_ss_ps_512 (v4f32 VR128X:$src))),
(VBROADCASTSSZrr VR128X:$src)>;
def : Pat<(v8f64 (int_x86_avx512_vbroadcast_sd_pd_512 (v2f64 VR128X:$src))),
(VBROADCASTSDZrr VR128X:$src)>;
// Provide fallback in case the load node that is used in the patterns above
// is used by additional users, which prevents the pattern selection.

14
test/CodeGen/X86/avx512-intrinsics.ll
View File

@ -219,3 +219,17 @@ define <8 x double> @test_x86_vbroadcast_sd_512(i8* %a0) {
ret <8 x double> %res
}
declare <8 x double> @llvm.x86.avx512.vbroadcast.sd.512(i8*) nounwind readonly
define <16 x float> @test_x86_vbroadcast_ss_ps_512(<4 x float> %a0) {
; CHECK: vbroadcastss
%res = call <16 x float> @llvm.x86.avx512.vbroadcast.ss.ps.512(<4 x float> %a0) ; <<16 x float>> [#uses=1]
ret <16 x float> %res
}
declare <16 x float> @llvm.x86.avx512.vbroadcast.ss.ps.512(<4 x float>) nounwind readonly
define <8 x double> @test_x86_vbroadcast_sd_pd_512(<2 x double> %a0) {
; CHECK: vbroadcastsd
%res = call <8 x double> @llvm.x86.avx512.vbroadcast.sd.pd.512(<2 x double> %a0) ; <<8 x double>> [#uses=1]
ret <8 x double> %res
}
declare <8 x double> @llvm.x86.avx512.vbroadcast.sd.pd.512(<2 x double>) nounwind readonly