[clang] [clang][AMDGPU] fix the return type for ballot (PR #73906)

[Sameer Sahasrabuddhe via cfe-commits]

================
@@ -0,0 +1,15 @@
+// REQUIRES: amdgpu-registered-target
+// RUN: %clang_cc1 -triple amdgcn-amd-amdhsa -target-cpu gfx900 -x hip -emit-llvm -fcuda-is-device -o - %s | FileCheck %s
+// RUN: %clang_cc1 -triple amdgcn-amd-amdhsa -target-cpu gfx900 -x hip -S -fcuda-is-device -o - %s | FileCheck %s --check-prefix=GFX9
----------------
ssahasra wrote:

You're right. The precommit test needed fixing. The updated version uses an aux-triple to specify a Windows host. Now the test exposes the problem even on a Linux build when trying cross-compile to Windows.

Added another note in the test: The same problem does not arise with existing OpenCL tests because the language fully specifies that "long" is 64 bits on all platforms.

https://github.com/llvm/llvm-project/pull/73906