[PATCH] D26725: AMDGPU: Add llvm.amdgcn.interp.mov intrinsic

[Tom Stellard via llvm-commits]

tstellarAMD added inline comments.


================
Comment at: include/llvm/IR/IntrinsicsAMDGPU.td:481
+            [llvm_i32_ty, llvm_i32_ty, llvm_i32_ty, llvm_i32_ty],
+            [IntrNoMem]>;
+
----------------
arsenm wrote:
> Shouldn't this be IntrReadMem? The doc says it reads parameters in LDS
I think technically, this should be inaccessiblememonly,  because but there is no intrinsic attribute for this yet.


================
Comment at: test/CodeGen/AMDGPU/llvm.amdgcn.interp.ll:18
   %p1_1 = call float @llvm.amdgcn.interp.p2(float %p0_1, float %j, i32 1, i32 0, i32 %3)
-  call void @llvm.SI.export(i32 15, i32 1, i32 1, i32 0, i32 1, float %p0_0, float %p0_0, float %p1_1, float %p1_1)
+  %const = call float @llvm.amdgcn.interp.mov(i32 2, i32 0, i32 0, i32 %3)
+  %w = fadd float %p1_1, %const
----------------
arsenm wrote:
> Do the source modifies work? There should be a test using them
There are no source modifiers for the VINTRP instructions.


https://reviews.llvm.org/D26725