https://github.com/ssahasra updated
https://github.com/llvm/llvm-project/pull/73920
>From 8ecb6310a4912de50628cf3db5ff8488fa919bb1 Mon Sep 17 00:00:00 2001
From: Sameer Sahasrabuddhe <sameer.sahasrabud...@amd.com>
Date: Fri, 1 Dec 2023 14:24:30 +0530
Subject: [PATCH] [clang][AMDGPU] precommit test for ballot on Windows
The Clang declaration of the wave-64 builtin uses "UL" as the return type, which
is interpreted as a 32-bit unsigned integer on Windows. This emits an incorrect
LLVM declaration with i32 return type instead of i64. The clang declaration
needs to be fixed to use "WU" instead.
---
clang/test/CodeGenHIP/ballot.cpp | 27 +++++++++++++++++++++++++++
1 file changed, 27 insertions(+)
create mode 100644 clang/test/CodeGenHIP/ballot.cpp
diff --git a/clang/test/CodeGenHIP/ballot.cpp b/clang/test/CodeGenHIP/ballot.cpp
new file mode 100644
index 000000000000000..6e1cbbdfc7af170
--- /dev/null
+++ b/clang/test/CodeGenHIP/ballot.cpp
@@ -0,0 +1,27 @@
+// REQUIRES: amdgpu-registered-target
+// XFAIL: *
+// RUN: %clang_cc1 -triple amdgcn-amd-amdhsa -aux-triple
x86_64-pc-windows-msvc -target-cpu gfx900 -x hip -emit-llvm -fcuda-is-device -o
- %s | FileCheck %s
+// RUN: %clang_cc1 -triple amdgcn-amd-amdhsa -aux-triple
x86_64-pc-windows-msvc -target-cpu gfx900 -x hip -S -fcuda-is-device -o - %s |
FileCheck %s --check-prefix=GFX9
+
+// Unlike OpenCL, HIP depends on the C++ interpration of "unsigned long", which
+// is 64 bits long on Linux and 32 bits long on Windows. The return type of the
+// ballot intrinsic needs to be a 64 bit integer on both platforms. This test
+// cross-compiles to Windows to confirm that the return type is indeed 64 bits
+// on Windows.
+
+// FIXME: The Clang declaration of the wave-64 builtin uses "UL" as the return
+// type, which is interpreted as a 32-bit unsigned integer on Windows. This
+// emits an incorrect LLVM declaration with i32 return type instead of i64. The
+// clang declaration needs to be fixed to use "WU" instead.
+
+// CHECK-LABEL: @_Z3fooi
+// CHECK: call i64 @llvm.amdgcn.ballot.i64
+
+// GFX9-LABEL: _Z3fooi:
+// GFX9: v_cmp_ne_u32_e64
+
+#define __device__ __attribute__((device))
+
+__device__ unsigned long long foo(int p) {
+ return __builtin_amdgcn_ballot_w64(p);
+}
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
