f4381d4644
This patch adds initial support for using aliases when targeting PTX. We perform a pretty strict conversion from the globals referenced to the expected output. as described in the PTX documentation at https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#kernel-and-function-directives-alias These cannot currently be used due to a bug in the `nvlink` implementation that causes aliases to pruned functions to crash the linker. Reviewed By: tra Differential Revision: https://reviews.llvm.org/D155211
10 lines
496 B
LLVM
10 lines
496 B
LLVM
; RUN: not --crash llc < %s -march=nvptx64 -mcpu=sm_30 -mattr=+ptx43 2>&1 | FileCheck %s --check-prefix=ATTR
|
|
; RUN: not --crash llc < %s -march=nvptx64 -mcpu=sm_20 -mattr=+ptx63 2>&1 | FileCheck %s --check-prefix=ATTR
|
|
; RUN: not --crash llc < %s -march=nvptx64 -mcpu=sm_30 -mattr=+ptx63 2>&1 | FileCheck %s --check-prefix=ALIAS
|
|
|
|
; ATTR: .alias requires PTX version >= 6.3 and sm_30
|
|
|
|
; ALIAS: NVPTX aliasee must be a non-kernel function
|
|
@a = global i32 42, align 8
|
|
@b = internal alias i32, ptr @a
|