This is an archive of the discontinued LLVM Phabricator instance.

Differential D27788

[CUDA] Add --ptxas-path= flag.
ClosedPublic

Authored by jlebar on Dec 14 2016, 5:25 PM.

Details

Reviewers
tra
Commits
rG3cf25461e05d: [CUDA] Add --ptxas-path= flag.
rC289847: [CUDA] Add --ptxas-path= flag.
rL289847: [CUDA] Add --ptxas-path= flag.
Summary

This lets you build with one CUDA installation but use ptxas from
another install.

This is useful e.g. if you want to avoid bugs in an old ptxas without
actually upgrading wholesale to a newer CUDA version.

Diff Detail

Event Timeline

jlebar updated this revision to Diff 81501.Dec 14 2016, 5:25 PM
jlebar retitled this revision from to [CUDA] Add --ptxas-path= flag..
jlebar updated this object.
jlebar added a reviewer: tra.
jlebar added a subscriber: cfe-commits.
tra accepted this revision.Dec 15 2016, 9:42 AM
tra edited edge metadata.
This revision is now accepted and ready to land.Dec 15 2016, 9:42 AM
Closed by commit rL289847: [CUDA] Add --ptxas-path= flag. (authored by jlebar). · Explain WhyDec 15 2016, 10:55 AM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
clang/
include/
clang/
Driver/
2 lines
lib/
Driver/
6 lines
test/
Driver/
12 lines

Diff 81501

clang/include/clang/Driver/Options.td

Show First 20 Lines • Show All 419 Lines • ▼ Show 20 Lines
def cuda_noopt_device_debug : Flag<["--"], "cuda-noopt-device-debug">, def cuda_noopt_device_debug : Flag<["--"], "cuda-noopt-device-debug">,
HelpText<"Enable device-side debug info generation. Disables ptxas optimizations.">; HelpText<"Enable device-side debug info generation. Disables ptxas optimizations.">;
def no_cuda_version_check : Flag<["--"], "no-cuda-version-check">, def no_cuda_version_check : Flag<["--"], "no-cuda-version-check">,
HelpText<"Don't error out if the detected version of the CUDA install is " HelpText<"Don't error out if the detected version of the CUDA install is "
"too low for the requested CUDA gpu architecture.">; "too low for the requested CUDA gpu architecture.">;
def no_cuda_noopt_device_debug : Flag<["--"], "no-cuda-noopt-device-debug">; def no_cuda_noopt_device_debug : Flag<["--"], "no-cuda-noopt-device-debug">;
def cuda_path_EQ : Joined<["--"], "cuda-path=">, Group<i_Group>, def cuda_path_EQ : Joined<["--"], "cuda-path=">, Group<i_Group>,
HelpText<"CUDA installation path">; HelpText<"CUDA installation path">;
def ptxas_path_EQ : Joined<["--"], "ptxas-path=">, Group<i_Group>,
HelpText<"Path to ptxas (used for compiling CUDA code)">;
def fcuda_flush_denormals_to_zero : Flag<["-"], "fcuda-flush-denormals-to-zero">, def fcuda_flush_denormals_to_zero : Flag<["-"], "fcuda-flush-denormals-to-zero">,
Flags<[CC1Option]>, HelpText<"Flush denormal floating point values to zero in CUDA device mode.">; Flags<[CC1Option]>, HelpText<"Flush denormal floating point values to zero in CUDA device mode.">;
def fno_cuda_flush_denormals_to_zero : Flag<["-"], "fno-cuda-flush-denormals-to-zero">; def fno_cuda_flush_denormals_to_zero : Flag<["-"], "fno-cuda-flush-denormals-to-zero">;
def fcuda_approx_transcendentals : Flag<["-"], "fcuda-approx-transcendentals">, def fcuda_approx_transcendentals : Flag<["-"], "fcuda-approx-transcendentals">,
Flags<[CC1Option]>, HelpText<"Use approximate transcendental functions">; Flags<[CC1Option]>, HelpText<"Use approximate transcendental functions">;
def fno_cuda_approx_transcendentals : Flag<["-"], "fno-cuda-approx-transcendentals">; def fno_cuda_approx_transcendentals : Flag<["-"], "fno-cuda-approx-transcendentals">;
def dA : Flag<["-"], "dA">, Group<d_Group>; def dA : Flag<["-"], "dA">, Group<d_Group>;
def dD : Flag<["-"], "dD">, Group<d_Group>, Flags<[CC1Option]>, def dD : Flag<["-"], "dD">, Group<d_Group>, Flags<[CC1Option]>,
▲ Show 20 Lines • Show All 1,943 Lines • Show Last 20 Lines

clang/lib/Driver/Tools.cpp

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 12,069 Lines • ▼ Show 20 Linesvoid NVPTX::Assembler::ConstructJob(Compilation &C, const JobAction &JA,
CmdArgs.push_back("--output-file"); CmdArgs.push_back("--output-file");
CmdArgs.push_back(Args.MakeArgString(Output.getFilename())); CmdArgs.push_back(Args.MakeArgString(Output.getFilename()));
for (const auto& II : Inputs) for (const auto& II : Inputs)
CmdArgs.push_back(Args.MakeArgString(II.getFilename())); CmdArgs.push_back(Args.MakeArgString(II.getFilename()));
for (const auto& A : Args.getAllArgValues(options::OPT_Xcuda_ptxas)) for (const auto& A : Args.getAllArgValues(options::OPT_Xcuda_ptxas))
CmdArgs.push_back(Args.MakeArgString(A)); CmdArgs.push_back(Args.MakeArgString(A));
const char *Exec = Args.MakeArgString(TC.GetProgramPath("ptxas")); const char *Exec;
if (Arg *A = Args.getLastArg(options::OPT_ptxas_path_EQ))
Exec = A->getValue();
else
Exec = Args.MakeArgString(TC.GetProgramPath("ptxas"));
C.addCommand(llvm::make_unique<Command>(JA, *this, Exec, CmdArgs, Inputs)); C.addCommand(llvm::make_unique<Command>(JA, *this, Exec, CmdArgs, Inputs));
} }
// All inputs to this linker must be from CudaDeviceActions, as we need to look // All inputs to this linker must be from CudaDeviceActions, as we need to look
// at the Inputs' Actions in order to figure out which GPU architecture they // at the Inputs' Actions in order to figure out which GPU architecture they
// correspond to. // correspond to.
void NVPTX::Linker::ConstructJob(Compilation &C, const JobAction &JA, void NVPTX::Linker::ConstructJob(Compilation &C, const JobAction &JA,
const InputInfo &Output, const InputInfo &Output,
Show All 38 Lines

clang/test/Driver/cuda-ptxas-path.cu

  • This file was added.
// REQUIRES: clang-driver
// REQUIRES: x86-registered-target
// REQUIRES: nvptx-registered-target
// RUN: %clang -### --target=i386-unknown-linux \
// RUN: --cuda-path=%S/Inputs/CUDA_80/usr/local/cuda \
// RUN: --ptxas-path=/some/path/to/ptxas %s 2>&1 \
// RUN: | FileCheck %s
// CHECK-NOT: "ptxas"
// CHECK: "/some/path/to/ptxas"
// CHECK-SAME: "--gpu-name" "sm_20"