================
@@ -106,3 +102,17 @@ float3 test_lerp_uint64_t3(uint64_t3 p0) { return lerp(p0,
p0, p0); }
// CHECK: %hlsl.lerp = call reassoc nnan ninf nsz arcp afn <4 x float>
@llvm.[[TARGET]].lerp.v4f32(<4 x float> %{{.*}}, <4 x float> %{{.*}}, <4 x
float> %{{.*}})
// CHECK: ret <4 x float> %hlsl.lerp
float4 test_lerp_uint64_t4(uint64_t4 p0) { return lerp(p0, p0, p0); }
+
+// CHECK-LABEL: test_lerp_half_scalar
+// NATIVE_HALF: %hlsl.lerp = call reassoc nnan ninf nsz arcp afn <3 x half>
@llvm.[[TARGET]].lerp.v3f16(<3 x half> %{{.*}}, <3 x half> %{{.*}}, <3 x half>
%{{.*}})
+// NATIVE_HALF: ret <3 x half> %hlsl.lerp
+// NO_HALF: %hlsl.lerp = call reassoc nnan ninf nsz arcp afn <3 x float>
@llvm.[[TARGET]].lerp.v3f32(<3 x float> %{{.*}}, <3 x float> %{{.*}}, <3 x
float> %{{.*}})
+// NO_HALF: ret <3 x float> %hlsl.lerp
+half3 test_lerp_half_scalar(half3 x, half3 y, half s) { return lerp(x, y, s); }
----------------
bogner wrote:
Yeah, that's not a bad idea. However, when I started to do that I realized this
whole test file is a bit messed up - it's using `-disable-llvm-passes`, so none
of the logic in the hlsl headers is inlined and we're actually splitting our
checks across multiple functions in most of these cases o_O.
I'll put up an NFC PR to improve these tests and then follow up here shortly.
https://github.com/llvm/llvm-project/pull/137877
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
