[PATCH] D16313: [X86][AVX512] fix dag & add intrinsics for fixupimm

Asaf Badouh via llvm-commits llvm-commits at lists.llvm.org
Tue Jan 19 05:06:25 PST 2016


AsafBadouh marked 3 inline comments as done.

================
Comment at: E:/Work/LLVM/trunk/test/CodeGen/X86/avx512vl-intrinsics.ll:7708
@@ +7707,3 @@
+  %res = call <2 x double> @llvm.x86.avx512.mask.fixupimm.pd.128(<2 x double> %x0, <2 x double> %x1,<2 x i64> %x2, i32 5, i8 %x4)
+  %res1 = call <2 x double> @llvm.x86.avx512.mask.fixupimm.pd.128(<2 x double> zeroinitializer, <2 x double> %x1, <2 x i64> %x2, i32 4, i8 %x4)
+  %res2 = call <2 x double> @llvm.x86.avx512.mask.fixupimm.pd.128(<2 x double> %x0, <2 x double> %x1, <2 x i64> %x2, i32 3, i8 -1)
----------------
igorb wrote:
> could you please remove zeroinitializer from the tests? It only complicate checks.
I think that this zeroinitializer is necessary since you can create fixupimmup {k}{z} with the non-maskz intrinsic version, and that case check it's ok.
unlike the zeroinitializer in line 7726, there, the zeroinitializer is just another argument.

what do you think?


Repository:
  rL LLVM

http://reviews.llvm.org/D16313





More information about the llvm-commits mailing list