================ @@ -0,0 +1,19 @@ +// REQUIRES: nvptx-registered-target + +// Check that the CUDA driver translates the legacy Clang option to the +// shortptr ABI. +// RUN: %clang --target=x86_64-linux-gnu -x cuda --cuda-device-only \ +// RUN: -fcuda-short-ptr -nocudainc -S -o /dev/null %s + +// Check that the NVPTX shortptr ABI can be selected directly. +// RUN: %clang_cc1 -triple nvptx64-nvidia-cuda -target-abi shortptr \ +// RUN: -emit-llvm -o - %s | FileCheck %s --check-prefix=SHORT-DL +// RUN: %clang_cc1 -triple nvptx64-nvidia-cuda -target-abi shortptr \ +// RUN: -S -o - %s | FileCheck %s --check-prefix=PTX + +// SHORT-DL: target datalayout = "e-p3:32:32-p4:32:32-p5:32:32-p6:32:32-p7:32:32-p101:32:32-i64:64-i128:128-i256:256-v16:16-v32:32-n16:32:64" ---------------- justinfargnoli wrote:
I moved this testing to `clang/test/CodeGen/target-data.c`, which has the example without directly above it. https://github.com/llvm/llvm-project/pull/214112 _______________________________________________ cfe-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
