Files
clang-p2996/clang/test/CodeGenCUDA/kernel-args-alignment.cu
Tim Northover c46827c7ed LLVM IR: Generate new-style byval-with-Type from Clang
LLVM IR recently added a Type parameter to the byval Attribute, so that
when pointers become opaque and no longer have an element type the
information will still be present in IR.

For now the Type parameter is optional (which is why Clang didn't need
this change at the time), but it will become mandatory soon.

llvm-svn: 362652
2019-06-05 21:12:14 +00:00

41 lines
1.3 KiB
Plaintext

// New CUDA kernel launch sequence does not require explicit specification of
// size/offset for each argument, so only the old way is tested.
//
// RUN: %clang_cc1 --std=c++11 -triple x86_64-unknown-linux-gnu -emit-llvm \
// RUN: -target-sdk-version=8.0 -o - %s \
// RUN: | FileCheck -check-prefixes=HOST-OLD,CHECK %s
// RUN: %clang_cc1 --std=c++11 -fcuda-is-device -triple nvptx64-nvidia-cuda \
// RUN: -emit-llvm -o - %s | FileCheck -check-prefixes=DEVICE,CHECK %s
#include "Inputs/cuda.h"
struct U {
short x;
} __attribute__((packed));
struct S {
int *ptr;
char a;
U u;
};
// Clang should generate a packed LLVM struct for S (denoted by the <>s),
// otherwise this test isn't interesting.
// CHECK: %struct.S = type <{ i32*, i8, %struct.U, [5 x i8] }>
static_assert(alignof(S) == 8, "Unexpected alignment.");
// HOST-LABEL: @_Z6kernelc1SPi
// Marshalled kernel args should be:
// 1. offset 0, width 1
// 2. offset 8 (because alignof(S) == 8), width 16
// 3. offset 24, width 8
// HOST-OLD: call i32 @cudaSetupArgument({{[^,]*}}, i64 1, i64 0)
// HOST-OLD: call i32 @cudaSetupArgument({{[^,]*}}, i64 16, i64 8)
// HOST-OLD: call i32 @cudaSetupArgument({{[^,]*}}, i64 8, i64 24)
// DEVICE-LABEL: @_Z6kernelc1SPi
// DEVICE-SAME: i8{{[^,]*}}, %struct.S* byval(%struct.S) align 8{{[^,]*}}, i32*
__global__ void kernel(char a, S s, int *b) {}