https://github.com/mmha created https://github.com/llvm/llvm-project/pull/234314
The host RUN line inherits the compiler default triple even though the checks describe an x86 host with an AMDGPU offload target. Clang rejects the incompatible host and device pointer sizes on platforms like ARM32. Assisted-by: Codex >From f0091b01504eaebfaa44becc1a2094de2605153f Mon Sep 17 00:00:00 2001 From: Morris Hafner <[email protected]> Date: Wed, 21 Oct 2026 18:43:12 +0000 Subject: [PATCH] [CIR][OpenMP] Specify x86 host triple in target-parallel test The host RUN line inherits the compiler default triple even though the checks describe an x86 host with an AMDGPU offload target. On the 32-bit ARM two-stage builder, Clang rejects the incompatible host and device pointer sizes before generating CIR. Assisted-by: Codex --- clang/test/CIR/CodeGenOpenMP/target-parallel.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/clang/test/CIR/CodeGenOpenMP/target-parallel.c b/clang/test/CIR/CodeGenOpenMP/target-parallel.c index a1832e7ffb3c2f..c7992665419853 100644 --- a/clang/test/CIR/CodeGenOpenMP/target-parallel.c +++ b/clang/test/CIR/CodeGenOpenMP/target-parallel.c @@ -1,7 +1,8 @@ // REQUIRES: amdgpu-registered-target // Host compilation (x86 host, AMDGPU offload target). -// RUN: %clang_cc1 -fopenmp -fopenmp-targets=amdgcn-amd-amdhsa -emit-cir -fclangir %s -o - \ +// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fopenmp \ +// RUN: -fopenmp-targets=amdgcn-amd-amdhsa -emit-cir -fclangir %s -o - \ // RUN: | FileCheck %s --check-prefix=CIR-HOST // Device compilation (AMDGPU): allocas live in the private address space. _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
