[clang] [llvm] [Clang] Add env var for nvptx-arch/amdgpu-arch timeout (PR #102521)
Joel E. Denny via cfe-commits
cfe-commits at lists.llvm.org
Thu Aug 8 13:26:32 PDT 2024
https://github.com/jdenny-ornl updated https://github.com/llvm/llvm-project/pull/102521
>From 6546428805b52f1b6f350193ab08ff027892710f Mon Sep 17 00:00:00 2001
From: "Joel E. Denny" <jdenny.ornl at gmail.com>
Date: Thu, 8 Aug 2024 15:02:04 -0400
Subject: [PATCH 1/3] [Clang] Add env var for nvptx-arch/amdgpu-arch timeout
When working on very busy systems, check-offload frequently fails many
tests with this diagnostic:
```
clang: error: cannot determine amdgcn architecture: /tmp/llvm/build/bin/amdgpu-arch: Child timed out: ; consider passing it via '-march'
```
The timeout is 10 seconds. This patch accepts the environment
variable `CLANG_TOOL_CHAIN_PROGRAM_WAIT` to increase it.
It should be documented somewhere. Any suggestions on where?
---
clang/lib/Driver/ToolChain.cpp | 10 +++++++++-
clang/lib/Driver/ToolChains/AMDGPU.cpp | 3 ++-
clang/lib/Driver/ToolChains/Cuda.cpp | 3 ++-
llvm/utils/lit/lit/TestingConfig.py | 1 +
4 files changed, 14 insertions(+), 3 deletions(-)
diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp
index 2d50c2cbbc881c..04b281e1bb10cd 100644
--- a/clang/lib/Driver/ToolChain.cpp
+++ b/clang/lib/Driver/ToolChain.cpp
@@ -40,6 +40,7 @@
#include "llvm/Support/FileSystem.h"
#include "llvm/Support/FileUtilities.h"
#include "llvm/Support/Path.h"
+#include "llvm/Support/Process.h"
#include "llvm/Support/VersionTuple.h"
#include "llvm/Support/VirtualFileSystem.h"
#include "llvm/TargetParser/AArch64TargetParser.h"
@@ -105,7 +106,7 @@ ToolChain::ToolChain(const Driver &D, const llvm::Triple &T,
llvm::Expected<std::unique_ptr<llvm::MemoryBuffer>>
ToolChain::executeToolChainProgram(StringRef Executable,
- unsigned SecondsToWait) const {
+ unsigned DefaultSecondsToWait) const {
llvm::SmallString<64> OutputFile;
llvm::sys::fs::createTemporaryFile("toolchain-program", "txt", OutputFile);
llvm::FileRemover OutputRemover(OutputFile.c_str());
@@ -116,6 +117,13 @@ ToolChain::executeToolChainProgram(StringRef Executable,
};
std::string ErrorMessage;
+ int SecondsToWait = DefaultSecondsToWait;
+ if (std::optional<std::string> Str =
+ llvm::sys::Process::GetEnv("CLANG_TOOL_CHAIN_PROGRAM_WAIT")) {
+ int Val = std::atoi(Str->c_str());
+ if (Val > 0)
+ SecondsToWait = Val;
+ }
if (llvm::sys::ExecuteAndWait(Executable, {}, {}, Redirects, SecondsToWait,
/*MemoryLimit=*/0, &ErrorMessage))
return llvm::createStringError(std::error_code(),
diff --git a/clang/lib/Driver/ToolChains/AMDGPU.cpp b/clang/lib/Driver/ToolChains/AMDGPU.cpp
index aa8f9197cfabc3..4ed366d21f5c43 100644
--- a/clang/lib/Driver/ToolChains/AMDGPU.cpp
+++ b/clang/lib/Driver/ToolChains/AMDGPU.cpp
@@ -899,7 +899,8 @@ AMDGPUToolChain::getSystemGPUArchs(const ArgList &Args) const {
else
Program = GetProgramPath("amdgpu-arch");
- auto StdoutOrErr = executeToolChainProgram(Program, /*SecondsToWait=*/10);
+ auto StdoutOrErr = executeToolChainProgram(Program,
+ /*DefaultSecondsToWait=*/10);
if (!StdoutOrErr)
return StdoutOrErr.takeError();
diff --git a/clang/lib/Driver/ToolChains/Cuda.cpp b/clang/lib/Driver/ToolChains/Cuda.cpp
index 17c952c808f725..104217eaf5d849 100644
--- a/clang/lib/Driver/ToolChains/Cuda.cpp
+++ b/clang/lib/Driver/ToolChains/Cuda.cpp
@@ -804,7 +804,8 @@ NVPTXToolChain::getSystemGPUArchs(const ArgList &Args) const {
else
Program = GetProgramPath("nvptx-arch");
- auto StdoutOrErr = executeToolChainProgram(Program, /*SecondsToWait=*/10);
+ auto StdoutOrErr = executeToolChainProgram(Program,
+ /*DefaultSecondsToWait=*/10);
if (!StdoutOrErr)
return StdoutOrErr.takeError();
diff --git a/llvm/utils/lit/lit/TestingConfig.py b/llvm/utils/lit/lit/TestingConfig.py
index eb9f8de2a7f960..06713429d06b4b 100644
--- a/llvm/utils/lit/lit/TestingConfig.py
+++ b/llvm/utils/lit/lit/TestingConfig.py
@@ -26,6 +26,7 @@ def fromdefaults(litConfig):
"SYSTEMROOT",
"TERM",
"CLANG",
+ "CLANG_TOOL_CHAIN_PROGRAM_WAIT",
"LLDB",
"LD_PRELOAD",
"LLVM_SYMBOLIZER_PATH",
>From 711cf93741ba618c7ee8051190b18bba938121fa Mon Sep 17 00:00:00 2001
From: "Joel E. Denny" <jdenny.ornl at gmail.com>
Date: Thu, 8 Aug 2024 16:21:28 -0400
Subject: [PATCH 2/3] Apply reviewer suggestion
---
clang/include/clang/Basic/DiagnosticDriverKinds.td | 3 ++-
clang/lib/Driver/ToolChain.cpp | 2 ++
2 files changed, 4 insertions(+), 1 deletion(-)
diff --git a/clang/include/clang/Basic/DiagnosticDriverKinds.td b/clang/include/clang/Basic/DiagnosticDriverKinds.td
index 3d8240f8357b40..05642f803d07de 100644
--- a/clang/include/clang/Basic/DiagnosticDriverKinds.td
+++ b/clang/include/clang/Basic/DiagnosticDriverKinds.td
@@ -99,7 +99,8 @@ def warn_drv_amdgpu_cov6: Warning<
" use at your own risk">;
def err_drv_undetermined_gpu_arch : Error<
"cannot determine %0 architecture: %1; consider passing it via "
- "'%2'">;
+ "'%2' or increasing the tool timeout using the environment variable "
+ "'CLANG_TOOL_CHAIN_PROGRAM_WAIT' (in secs, <=0 is inifinite)">;
def warn_drv_multi_gpu_arch : Warning<
"multiple %0 architectures are detected: %1; only the first one is used for "
"'%2'">, InGroup<MultiGPU>;
diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp
index 04b281e1bb10cd..22de2bdeaa29f5 100644
--- a/clang/lib/Driver/ToolChain.cpp
+++ b/clang/lib/Driver/ToolChain.cpp
@@ -123,6 +123,8 @@ ToolChain::executeToolChainProgram(StringRef Executable,
int Val = std::atoi(Str->c_str());
if (Val > 0)
SecondsToWait = Val;
+ else if (Val <= 0)
+ SecondsToWait = 0; // infinite
}
if (llvm::sys::ExecuteAndWait(Executable, {}, {}, Redirects, SecondsToWait,
/*MemoryLimit=*/0, &ErrorMessage))
>From 6ec00ff8e8ca2d4a3f7267f3189ee190e211ca05 Mon Sep 17 00:00:00 2001
From: "Joel E. Denny" <jdenny.ornl at gmail.com>
Date: Thu, 8 Aug 2024 16:26:24 -0400
Subject: [PATCH 3/3] Update clang/lib/Driver/ToolChain.cpp
---
clang/lib/Driver/ToolChain.cpp | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp
index 22de2bdeaa29f5..c6436fb2ee33b4 100644
--- a/clang/lib/Driver/ToolChain.cpp
+++ b/clang/lib/Driver/ToolChain.cpp
@@ -123,7 +123,7 @@ ToolChain::executeToolChainProgram(StringRef Executable,
int Val = std::atoi(Str->c_str());
if (Val > 0)
SecondsToWait = Val;
- else if (Val <= 0)
+ else
SecondsToWait = 0; // infinite
}
if (llvm::sys::ExecuteAndWait(Executable, {}, {}, Redirects, SecondsToWait,
More information about the cfe-commits
mailing list