================
@@ -0,0 +1,76 @@
+; RUN: llc -O0 -mtriple=spirv-vulkan-unknown %s -o - | FileCheck %s
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv-vulkan-unknown %s -o -
-filetype=obj | spirv-val %}
+
+; This file generated from the following HLSL:
+; clang -cc1 -triple spirv-vulkan-library -x hlsl -emit-llvm
-disable-llvm-passes -finclude-default-header -o - DispatchThreadID.hlsl
+;
+; [shader("compute")]
----------------
llvm-beanz wrote:
The LLVM Offload project has been using library calls instead of intrinsics
where it makes sense to reduce the need to duplicate intrinsics. We can do that
as well, but in cases like this specific change I think having target specific
intrinsics has minimal cost and simplifies the lowering path.
There is no reasonable way to have DXIL and SPIR-V share a target. They are
wildly different and I suspect any abstraction layers we would need to add to
make that reasonable would be more complex than they are worth.
https://github.com/llvm/llvm-project/pull/82536
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
