tra created this revision.
tra added a reviewer: echristo.
tra added a subscriber: cfe-commits.
Propagates AuxTriple throughout job pipeline construction in driver and on to
Tool::ConstructJob().
It in turn passes target triple of opposite side of CUDA compilation as
-aux-triple option to sub-compilations.
http://reviews.llvm.org/D13144
Files:
include/clang/Driver/Driver.h
include/clang/Driver/Tool.h
lib/Driver/Driver.cpp
lib/Driver/Tools.cpp
lib/Driver/Tools.h
test/Driver/cuda-options.cu
test/SemaCUDA/function-target-hd.cu
Index: test/SemaCUDA/function-target-hd.cu
===================================================================
--- test/SemaCUDA/function-target-hd.cu
+++ test/SemaCUDA/function-target-hd.cu
@@ -8,9 +8,9 @@
// host device functions are not allowed to call device functions.
// RUN: %clang_cc1 -fsyntax-only -verify %s
-// RUN: %clang_cc1 -fsyntax-only -fcuda-is-device -verify %s
+// RUN: %clang_cc1 -fsyntax-only -fcuda-is-device -triple nvptx-unknown-cuda -verify %s
// RUN: %clang_cc1 -fsyntax-only -fcuda-allow-host-calls-from-host-device -verify %s -DTEST_WARN_HD
-// RUN: %clang_cc1 -fsyntax-only -fcuda-is-device -fcuda-allow-host-calls-from-host-device -verify %s -DTEST_WARN_HD
+// RUN: %clang_cc1 -fsyntax-only -fcuda-is-device -triple nvptx-unknown-cuda -fcuda-allow-host-calls-from-host-device -verify %s -DTEST_WARN_HD
#include "Inputs/cuda.h"
Index: test/Driver/cuda-options.cu
===================================================================
--- test/Driver/cuda-options.cu
+++ test/Driver/cuda-options.cu
@@ -111,14 +111,6 @@
// Make sure we don't link anything.
// RUN: -check-prefix CUDA-NL %s
-// Match device-side preprocessor, and compiler phases with -save-temps
-// CUDA-D1S: "-cc1" "-triple" "nvptx{{(64)?}}-nvidia-cuda"
-// CUDA-D1S-SAME: "-fcuda-is-device"
-// CUDA-D1S-SAME: "-x" "cuda"
-// CUDA-D1S: "-cc1" "-triple" "nvptx{{(64)?}}-nvidia-cuda"
-// CUDA-D1S-SAME: "-fcuda-is-device"
-// CUDA-D1S-SAME: "-x" "cuda-cpp-output"
-
// --cuda-host-only should never trigger unused arg warning.
// RUN: %clang -### -target x86_64-linux-gnu --cuda-host-only -c %s 2>&1 | \
// RUN: FileCheck -check-prefix CUDA-NO-UNUSED-CHO %s
@@ -133,34 +125,47 @@
// RUN: %clang -### -target x86_64-linux-gnu --cuda-device-only -x c -c %s 2>&1 | \
// RUN: FileCheck -check-prefix CUDA-UNUSED-CDO %s
+// Match device-side preprocessor, and compiler phases with -save-temps
+// CUDA-D1S: "-cc1" "-triple" "nvptx64-nvidia-cuda"
+// CUDA-D1S-SAME: "-aux-triple" "x86_64--linux-gnu"
+// CUDA-D1S-SAME: "-fcuda-is-device"
+// CUDA-D1S-SAME: "-x" "cuda"
+
+// CUDA-D1S: "-cc1" "-triple" "nvptx64-nvidia-cuda"
+// CUDA-D1S-SAME: "-aux-triple" "x86_64--linux-gnu"
+// CUDA-D1S-SAME: "-fcuda-is-device"
+// CUDA-D1S-SAME: "-x" "cuda-cpp-output"
+
// Match the job that produces PTX assembly
-// CUDA-D1: "-cc1" "-triple" "nvptx{{(64)?}}-nvidia-cuda"
+// CUDA-D1: "-cc1" "-triple" "nvptx64-nvidia-cuda"
+// CUDA-D1NS-SAME: "-aux-triple" "x86_64--linux-gnu"
// CUDA-D1-SAME: "-fcuda-is-device"
// CUDA-D1-SM35-SAME: "-target-cpu" "sm_35"
// CUDA-D1-SAME: "-o" "[[GPUBINARY1:[^"]*]]"
// CUDA-D1NS-SAME: "-x" "cuda"
// CUDA-D1S-SAME: "-x" "ir"
-// Match anothe device-side compilation
-// CUDA-D2: "-cc1" "-triple" "nvptx{{(64)?}}-nvidia-cuda"
+// Match another device-side compilation
+// CUDA-D2: "-cc1" "-triple" "nvptx64-nvidia-cuda"
+// CUDA-D2-SAME: "-aux-triple" "x86_64--linux-gnu"
// CUDA-D2-SAME: "-fcuda-is-device"
// CUDA-D2-SM30-SAME: "-target-cpu" "sm_30"
// CUDA-D2-SAME: "-o" "[[GPUBINARY2:[^"]*]]"
// CUDA-D2-SAME: "-x" "cuda"
// Match no device-side compilation
-// CUDA-ND-NOT: "-cc1" "-triple" "nvptx{{(64)?}}-nvidia-cuda"
+// CUDA-ND-NOT: "-cc1" "-triple" "nvptx64-nvidia-cuda"
// CUDA-ND-SAME-NOT: "-fcuda-is-device"
// Match host-side preprocessor job with -save-temps
-// CUDA-HS: "-cc1" "-triple"
-// CUDA-HS-SAME-NOT: "nvptx{{(64)?}}-nvidia-cuda"
+// CUDA-HS: "-cc1" "-triple" "x86_64--linux-gnu"
+// CUDA-HS-SAME: "-aux-triple" "nvptx64-nvidia-cuda"
// CUDA-HS-SAME-NOT: "-fcuda-is-device"
// CUDA-HS-SAME: "-x" "cuda"
// Match host-side compilation
-// CUDA-H: "-cc1" "-triple"
-// CUDA-H-SAME-NOT: "nvptx{{(64)?}}-nvidia-cuda"
+// CUDA-H: "-cc1" "-triple" "x86_64--linux-gnu"
+// CUDA-H-SAME: "-aux-triple" "nvptx64-nvidia-cuda"
// CUDA-H-SAME-NOT: "-fcuda-is-device"
// CUDA-H-SAME: "-o" "[[HOSTOUTPUT:[^"]*]]"
// CUDA-HNS-SAME: "-x" "cuda"
Index: lib/Driver/Tools.h
===================================================================
--- lib/Driver/Tools.h
+++ lib/Driver/Tools.h
@@ -106,8 +106,15 @@
void ConstructJob(Compilation &C, const JobAction &JA,
const InputInfo &Output, const InputInfoList &Inputs,
+ const llvm::opt::ArgList &TCArgs, const char *LinkingOutput,
+ const char *AuxTriple) const override;
+
+ void ConstructJob(Compilation &C, const JobAction &JA,
+ const InputInfo &Output, const InputInfoList &Inputs,
const llvm::opt::ArgList &TCArgs,
- const char *LinkingOutput) const override;
+ const char *LinkingOutput) const override {
+ llvm_unreachable("Should never be called");
+ };
};
/// \brief Clang integrated assembler tool.
Index: lib/Driver/Tools.cpp
===================================================================
--- lib/Driver/Tools.cpp
+++ lib/Driver/Tools.cpp
@@ -3169,7 +3169,8 @@
void Clang::ConstructJob(Compilation &C, const JobAction &JA,
const InputInfo &Output, const InputInfoList &Inputs,
- const ArgList &Args, const char *LinkingOutput) const {
+ const ArgList &Args, const char *LinkingOutput,
+ const char *AuxTriple) const {
std::string TripleStr = getToolChain().ComputeEffectiveClangTriple(Args);
const llvm::Triple Triple(TripleStr);
@@ -3201,6 +3202,11 @@
CmdArgs.push_back("-triple");
CmdArgs.push_back(Args.MakeArgString(TripleStr));
+ if (IsCuda && AuxTriple != nullptr) {
+ CmdArgs.push_back("-aux-triple");
+ CmdArgs.push_back(AuxTriple);
+ }
+
if (Triple.isOSWindows() && (Triple.getArch() == llvm::Triple::arm ||
Triple.getArch() == llvm::Triple::thumb)) {
unsigned Offset = Triple.getArch() == llvm::Triple::arm ? 4 : 6;
Index: lib/Driver/Driver.cpp
===================================================================
--- lib/Driver/Driver.cpp
+++ lib/Driver/Driver.cpp
@@ -1762,21 +1762,22 @@
void Driver::BuildJobsForAction(Compilation &C, const Action *A,
const ToolChain *TC, const char *BoundArch,
bool AtTopLevel, bool MultipleArchs,
- const char *LinkingOutput,
- InputInfo &Result) const {
+ const char *LinkingOutput, InputInfo &Result,
+ const char *AuxTriple) const {
llvm::PrettyStackTraceString CrashInfo("Building compilation jobs");
InputInfoList CudaDeviceInputInfos;
if (const CudaHostAction *CHA = dyn_cast<CudaHostAction>(A)) {
InputInfo II;
// Append outputs of device jobs to the input list.
for (const Action *DA : CHA->getDeviceActions()) {
- BuildJobsForAction(C, DA, TC, "", AtTopLevel,
+ BuildJobsForAction(C, DA, TC, nullptr, AtTopLevel,
/*MultipleArchs*/ false, LinkingOutput, II);
CudaDeviceInputInfos.push_back(II);
}
// Override current action with a real host compile action and continue
// processing it.
+ AuxTriple = CHA->getDeviceTriple();
A = *CHA->begin();
}
@@ -1806,16 +1807,20 @@
TC = &C.getDefaultToolChain();
BuildJobsForAction(C, *BAA->begin(), TC, ArchName, AtTopLevel,
- MultipleArchs, LinkingOutput, Result);
+ MultipleArchs, LinkingOutput, Result, AuxTriple);
return;
}
if (const CudaDeviceAction *CDA = dyn_cast<CudaDeviceAction>(A)) {
+ // Initial processing of CudaDeviceAction carries host params.
+ // Call BuildJobsForAction() again, now with correct device parameters.
+ assert(CDA->getGpuArchName() && "No GPU name in device action.");
BuildJobsForAction(
C, *CDA->begin(),
&getToolChain(C.getArgs(), llvm::Triple(CDA->getDeviceTriple())),
CDA->getGpuArchName(), CDA->isAtTopLevel(),
- /*MultipleArchs*/ true, LinkingOutput, Result);
+ /*MultipleArchs*/ true, LinkingOutput, Result,
+ TC->getTriple().str().c_str());
return;
}
@@ -1834,9 +1839,10 @@
InputInfo II;
for (const Action *DA : CollapsedCHA->getDeviceActions()) {
BuildJobsForAction(C, DA, TC, "", AtTopLevel,
- /*MultipleArchs*/ false, LinkingOutput, II);
+ /*MultipleArchs*/ false, LinkingOutput, II, AuxTriple);
CudaDeviceInputInfos.push_back(II);
}
+ AuxTriple = CollapsedCHA->getDeviceTriple();
}
// Only use pipes when there is exactly one input.
@@ -1851,7 +1857,7 @@
InputInfo II;
BuildJobsForAction(C, Input, TC, BoundArch, SubJobAtTopLevel, MultipleArchs,
- LinkingOutput, II);
+ LinkingOutput, II, AuxTriple);
InputInfos.push_back(II);
}
@@ -1886,7 +1892,8 @@
llvm::errs() << "], output: " << Result.getAsString() << "\n";
} else {
T->ConstructJob(C, *JA, Result, InputInfos,
- C.getArgsForToolChain(TC, BoundArch), LinkingOutput);
+ C.getArgsForToolChain(TC, BoundArch), LinkingOutput,
+ AuxTriple);
}
}
Index: include/clang/Driver/Tool.h
===================================================================
--- include/clang/Driver/Tool.h
+++ include/clang/Driver/Tool.h
@@ -124,11 +124,20 @@
/// tool chain specific translations applied.
/// \param LinkingOutput - If this output will eventually feed the
/// linker, then this is the final output name of the linked image.
+ /// \param AuxTriple - Additional triple to pass during CUDA compilation.
virtual void ConstructJob(Compilation &C, const JobAction &JA,
const InputInfo &Output,
const InputInfoList &Inputs,
const llvm::opt::ArgList &TCArgs,
const char *LinkingOutput) const = 0;
+
+ /// \brief default implementation for classes that don't care about AuxTriple.
+ virtual void
+ ConstructJob(Compilation &C, const JobAction &JA, const InputInfo &Output,
+ const InputInfoList &Inputs, const llvm::opt::ArgList &TCArgs,
+ const char *LinkingOutput, const char *AuxTriple) const {
+ ConstructJob(C, JA, Output, Inputs, TCArgs, LinkingOutput);
+ }
};
} // end namespace driver
Index: include/clang/Driver/Driver.h
===================================================================
--- include/clang/Driver/Driver.h
+++ include/clang/Driver/Driver.h
@@ -369,7 +369,8 @@
bool AtTopLevel,
bool MultipleArchs,
const char *LinkingOutput,
- InputInfo &Result) const;
+ InputInfo &Result,
+ const char *AuxTriple = nullptr) const;
/// Returns the default name for linked images (e.g., "a.out").
const char *getDefaultImageName() const;
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
