maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[HIP] Add option -fgpu-inline-threshold 

Add option -fgpu-inline-threshold for inline threshold for device compilation only.

Reviewed by: Artem Belevich

Differential Revision: https://reviews.llvm.org/D99233
This commit is contained in:
Yaxun (Sam) Liu 2021-04-21 14:34:37 -04:00
parent ac8b71227c
commit 5a2d78b163
3 changed files with 18 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
clang/include/clang/Driver/Options.td
View File

@ -956,6 +956,9 @@ def gpu_max_threads_per_block_EQ : Joined<["--"], "gpu-max-threads-per-block=">,
HelpText<"Default max threads per block for kernel launch bounds for HIP">, HelpText<"Default max threads per block for kernel launch bounds for HIP">,
MarshallingInfoInt<LangOpts<"GPUMaxThreadsPerBlock">, "1024">, MarshallingInfoInt<LangOpts<"GPUMaxThreadsPerBlock">, "1024">,
ShouldParseIf<hip.KeyPath>; ShouldParseIf<hip.KeyPath>;
def fgpu_inline_threshold_EQ : Joined<["-"], "fgpu-inline-threshold=">,
Flags<[HelpHidden]>,
HelpText<"Inline threshold for device compilation for CUDA/HIP">;
def gpu_instrument_lib_EQ : Joined<["--"], "gpu-instrument-lib=">, def gpu_instrument_lib_EQ : Joined<["--"], "gpu-instrument-lib=">,
HelpText<"Instrument device library for HIP, which is a LLVM bitcode containing " HelpText<"Instrument device library for HIP, which is a LLVM bitcode containing "
"__cyg_profile_func_enter and __cyg_profile_func_exit">; "__cyg_profile_func_enter and __cyg_profile_func_exit">;

10
clang/lib/Driver/ToolChains/Clang.cpp
View File

@ -6473,6 +6473,16 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
if (IsHIP) if (IsHIP)
CmdArgs.push_back("-fcuda-allow-variadic-functions"); CmdArgs.push_back("-fcuda-allow-variadic-functions");
if (IsCudaDevice || IsHIPDevice) {
StringRef InlineThresh =
Args.getLastArgValue(options::OPT_fgpu_inline_threshold_EQ);
if (!InlineThresh.empty()) {
std::string ArgStr =
std::string("-inline-threshold=") + InlineThresh.str();
CmdArgs.append({"-mllvm", Args.MakeArgStringRef(ArgStr)});
}
}
// OpenMP offloading device jobs take the argument -fopenmp-host-ir-file-path // OpenMP offloading device jobs take the argument -fopenmp-host-ir-file-path
// to specify the result of the compile phase on the host, so the meaningful // to specify the result of the compile phase on the host, so the meaningful
// device declarations can be identified. Also, -fopenmp-is-device is passed // device declarations can be identified. Also, -fopenmp-is-device is passed

5
clang/test/Driver/hip-options.hip
View File

@ -51,3 +51,8 @@
// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=CTA %s // RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=CTA %s
// CTA: clang{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-mconstructor-aliases" // CTA: clang{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-mconstructor-aliases"
// CTA-NOT: clang{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-mconstructor-aliases" // CTA-NOT: clang{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-mconstructor-aliases"
// RUN: %clang -### -target x86_64-unknown-linux-gnu -nogpuinc -nogpulib \
// RUN: --offload-arch=gfx906 -fgpu-inline-threshold=1000 %s 2>&1 | FileCheck -check-prefix=THRESH %s
// THRESH: clang{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-mllvm" "-inline-threshold=1000"
// THRESH-NOT: clang{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-inline-threshold=1000"