From 79769a4475b37fa011203c688985e36b972328ff Mon Sep 17 00:00:00 2001
From: Piotr Sobczak <piotr.sobczak@amd.com>
Date: Wed, 16 Oct 2019 11:14:01 +0000
Subject: [InstCombine][AMDGPU] Fix crash with v3i16/v3f16 buffer intrinsics

Summary:
This is something of a workaround to avoid a crash later on in type
legalizer (WidenVectorResult()).
Also added some f16 tests, including a non-working v3f16 case with
a FIXME.

Reviewers: arsenm, tpr, nhaehnle

Reviewed By: arsenm

Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D68865

llvm-svn: 374993
---
 llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp | 7 +++++++
 1 file changed, 7 insertions(+)

(limited to 'llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp')

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
index ee42c730b09..4680cb30060 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp
@@ -971,6 +971,13 @@ InstCombiner::simplifyShrShlDemandedBits(Instruction *Shr, const APInt &ShrOp1,
 Value *InstCombiner::simplifyAMDGCNMemoryIntrinsicDemanded(IntrinsicInst *II,
                                                            APInt DemandedElts,
                                                            int DMaskIdx) {
+
+  // FIXME: Allow v3i16/v3f16 in buffer intrinsics when the types are fully supported.
+  if (DMaskIdx < 0 &&
+      II->getType()->getScalarSizeInBits() != 32 &&
+      DemandedElts.getActiveBits() == 3)
+    return nullptr;
+
   unsigned VWidth = II->getType()->getVectorNumElements();
   if (VWidth == 1)
     return nullptr;
-- 
cgit v1.2.3