Subzero: fix assert for sqrt of vector float types

Subzero's sqrt intrinsic implementation was assuming only scalar float args when moving the result to the destination, despite its sqrt implementation supporting vector float. Bug: b/130459196 Change-Id: I1c5baf2dc72448785add421c87900964f877ce6f Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/38872 Kokoro-Presubmit: kokoro <noreply+kokoro@google.com> Reviewed-by: Alexis Hétu <sugoi@google.com> Reviewed-by: Ben Clayton <bclayton@google.com> Tested-by: Antonio Maiorano <amaiorano@google.com>

Subzero: fix assert for sqrt of vector float types
b34b558a · Antonio Maiorano · bb222a10 · b34b558a
Commit b34b558a authored Nov 28, 2019 by Antonio Maiorano
Hide whitespace changes
Inline Side-by-side

Showing with 5 additions and 1 deletion

IceTargetLoweringX86BaseImpl.h third_party/subzero/src/IceTargetLoweringX86BaseImpl.h +5 -1

No files found.
--- a/third_party/subzero/src/IceTargetLoweringX86BaseImpl.h
+++ b/third_party/subzero/src/IceTargetLoweringX86BaseImpl.h
@@ -4439,7 +4439,11 @@ void TargetX86Base<TraitsType>::lowerIntrinsicCall(
    Variable *Dest = Instr->getDest();
    Variable *T = makeReg(Dest->getType());
    _sqrt(T, Src);
-    _mov(Dest, T);
+    if (isVectorType(Dest->getType())) {
+      _movp(Dest, T);
+    } else {
+      _mov(Dest, T);
+    }
    return;
  }
  case Intrinsics::Stacksave: {