aaron/ispc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

More AVX fixes

Browse Source 
Fix RNG state initialization for 16-wide targets
Fix a number of bugs in reduce_add builtin implementations for AVX.
Fix some tests that had incorrect expected results for the 16-wide
  case.
This commit is contained in:
Matt Pharr
2011-09-06 15:53:11 -07:00
parent c76ef7b174
commit 4f451bd041
8 changed files with 24 additions and 22 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
builtins-avx.ll
View File

@@ -294,10 +294,12 @@ define internal double @__reduce_add_double(<8 x double>) nounwind readonly alwa
<4 x i32> <i32 0, i32 1, i32 2, i32 3>
%v1 = shufflevector <8 x double> %0, <8 x double> undef,
<4 x i32> <i32 4, i32 5, i32 6, i32 7>
%sum01 = fadd <4 x double> %v0, %v1
%red0 = call <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double> %sum01, <4 x double> %sum01)
%red1 = call <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double> %red0, <4 x double> %red0)
%sum = extractelement <4 x double> %red1, i32 0
%sum0 = call <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double> %v0, <4 x double> %v1)
%sum1 = call <4 x double> @llvm.x86.avx.hadd.pd.256(<4 x double> %sum0, <4 x double> %sum0)
%final0 = extractelement <4 x double> %sum1, i32 0
%final1 = extractelement <4 x double> %sum1, i32 2
%sum = fadd double %final0, %final1
ret double %sum
}