From 0a2fc781e3f046e334f492581bbcee90703aa0fd Mon Sep 17 00:00:00 2001 From: Craig Topper Date: Sun, 11 Dec 2016 07:42:04 +0000 Subject: [PATCH] [AVX-512][InstCombine] Teach InstCombineCalls how to simplify demanded for scalar cmp intrinsics with masking and rounding. These intrinsics don't read the upper elements of their first and second input. These are slightly different the the SSE version which does use the upper bits of its first element as passthru bits since the result goes to an XMM register. For AVX-512 the result goes to a mask register instead. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@289371 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Transforms/InstCombine/InstCombineCalls.cpp | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/lib/Transforms/InstCombine/InstCombineCalls.cpp b/lib/Transforms/InstCombine/InstCombineCalls.cpp index 3f666bcdc7d..d87ac44b415 100644 --- a/lib/Transforms/InstCombine/InstCombineCalls.cpp +++ b/lib/Transforms/InstCombine/InstCombineCalls.cpp @@ -1738,7 +1738,9 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) { case Intrinsic::x86_sse2_ucomigt_sd: case Intrinsic::x86_sse2_ucomile_sd: case Intrinsic::x86_sse2_ucomilt_sd: - case Intrinsic::x86_sse2_ucomineq_sd: { + case Intrinsic::x86_sse2_ucomineq_sd: + case Intrinsic::x86_avx512_mask_cmp_ss: + case Intrinsic::x86_avx512_mask_cmp_sd: { // These intrinsics only demand the 0th element of their input vectors. If // we can simplify the input based on that, do so now. bool MadeChange = false; -- 2.50.1