Don't prevent a vselect of constants from becoming a single load (PR20648).
[oota-llvm.git] / test / CodeGen / X86 / sse41-blend.ll
2014-08-20 Sanjay PatelDon't prevent a vselect of constants from becoming...
2014-05-31 Filipe CabecinhasMake blend tests more specific
2014-05-29 Filipe CabecinhasAdded tests for shufflevector lowering to blend instrs.
2014-05-27 Filipe CabecinhasConvert some X86 blendv* intrinsics into IR.
2014-05-16 Filipe CabecinhasLower vselects into X86ISD::BLENDI when appropriate.
2014-01-28 Andrea Di Biagio[X86] Add extra rules for combining vselect dag nodes...
2014-01-20 Andrea Di Biagio[X86] Teach how to combine a vselect into a movss/movsd
2013-08-23 Rafael EspindolaReplace more uses of sse41 with sse4.1.
2013-07-18 Stephen LinUpdate to more CodeGen tests to use CHECK-LABEL for...
2012-06-04 Nadav RotemRemove the "-promote-elements" flag. This flag is now...
2012-04-15 Nadav RotemWhen emulating vselect using OR/AND/XOR make sure to...
2012-01-03 Nadav RotemRevert 147426 because it caused pr11696.
2012-01-02 Nadav RotemOptimize the sequence blend(sign_extend(x)) to blend...
2011-09-17 Duncan SandsSynthesize x86 max/min instructions also for vectors...
2011-09-14 Nadav RotemAdd integer promotion support for vselect
2011-09-12 Bruno Cardoso LopesChange testcase commandline to be more strict and silen...
2011-09-12 Bruno Cardoso LopesRevert the wrong part of r139528, and fix testcases.
2011-09-12 Bruno Cardoso LopesNot sure how CMPPS and CMPPD had already ever worked...
2011-09-08 Bruno Cardoso LopesAdd AVX versions of blend vector operations and fix...