================
@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -finclude-default-header -x hlsl -triple
dxil-pc-shadermodel6.3-library %s -fnative-half-type -emit-llvm
-disable-llvm-passes -o - | FileCheck %s
+
+
+// CHECK-LABEL: test_asuint4_uint
+// CHECK: ret i32 %0
+export uint test_asuint4_uint(uint p0) {
+ return asuint(p0);
+}
----------------
bogner wrote:
This doesn't really test the function effectively, it just checks that we
return an `i32`. We can use a couple of tools in FileCheck to do better,
notably [capturing
variables](https://llvm.org/docs/CommandGuide/FileCheck.html#filecheck-string-substitution-blocks)
and
[CHECK-NOT](https://llvm.org/docs/CommandGuide/FileCheck.html#the-check-not-directive).
Now, we want to make sure that the input of the function is returned, and that
we don't see a bitcast in between.
Take a look at the output of the test command by running it manually. Assuming
we update the flags as I suggested above, we see something like this:
```llvm
define noundef i32 @"?test_uint@@YAII@Z"(i32 noundef returned %p0)
local_unnamed_addr #0 {
entry:
ret i32 %p0
}
```
Getting the name of the input variable is a little bit awkward because (1) we
mangle the function name, and (2) we get some attributes applied that we don't
really want the test to be sensitive to, so need a somewhat loose regex. Then
we want to check that no bitcast occurs, and that we return the input directly:
```c++
// CHECK: define {{.*}}test_asuint4_uint{{.*}}(i32{{.*}} [[VAL:%.*]])
// CHECK-NOT: bitcast
// CHECK: ret i32 [[VAL]]
```
https://github.com/llvm/llvm-project/pull/107292
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
