[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 16:34:48 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 1/5] [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 2/5] 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 3/5] 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,
>From 37cd64bc733f06d3caf3c660f57b5d717a59faf7 Mon Sep 17 00:00:00 2001
From: "Joel E. Denny" <jdenny.ornl at gmail.com>
Date: Thu, 8 Aug 2024 18:59:15 -0400
Subject: [PATCH 4/5] Apply reviewer suggestions
---
.../include/clang/Basic/DiagnosticDriverKinds.td | 7 ++++---
clang/include/clang/Driver/ToolChain.h | 2 +-
clang/lib/Driver/ToolChain.cpp | 14 ++++++++------
clang/test/Driver/amdgpu-hip-system-arch.c | 8 ++++++++
clang/test/Driver/nvptx-cuda-system-arch.c | 8 ++++++++
clang/test/Driver/openmp-system-arch.c | 16 ++++++++++++++++
llvm/utils/lit/lit/TestingConfig.py | 2 +-
7 files changed, 46 insertions(+), 11 deletions(-)
diff --git a/clang/include/clang/Basic/DiagnosticDriverKinds.td b/clang/include/clang/Basic/DiagnosticDriverKinds.td
index 05642f803d07de..92a602829933ce 100644
--- a/clang/include/clang/Basic/DiagnosticDriverKinds.td
+++ b/clang/include/clang/Basic/DiagnosticDriverKinds.td
@@ -98,9 +98,10 @@ def warn_drv_amdgpu_cov6: Warning<
"code object v6 is still in development and not ready for production use yet;"
" use at your own risk">;
def err_drv_undetermined_gpu_arch : Error<
- "cannot determine %0 architecture: %1; consider passing it via "
- "'%2' or increasing the tool timeout using the environment variable "
- "'CLANG_TOOL_CHAIN_PROGRAM_WAIT' (in secs, <=0 is inifinite)">;
+ "cannot determine %0 architecture: %1; consider passing it via '%2'; "
+ "environment variable CLANG_TOOLCHAIN_PROGRAM_TIMEOUT specifies the tool "
+ "timeout (integer secs, <=0 is infinite)">;
+
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/include/clang/Driver/ToolChain.h b/clang/include/clang/Driver/ToolChain.h
index ece1384d5d3c02..fbac660548346e 100644
--- a/clang/include/clang/Driver/ToolChain.h
+++ b/clang/include/clang/Driver/ToolChain.h
@@ -206,7 +206,7 @@ class ToolChain {
/// Executes the given \p Executable and returns the stdout.
llvm::Expected<std::unique_ptr<llvm::MemoryBuffer>>
executeToolChainProgram(StringRef Executable,
- unsigned SecondsToWait = 0) const;
+ unsigned DefaultSecondsToWait = 0) const;
void setTripleEnvironment(llvm::Triple::EnvironmentType Env);
diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp
index c6436fb2ee33b4..5fd8af373e1db5 100644
--- a/clang/lib/Driver/ToolChain.cpp
+++ b/clang/lib/Driver/ToolChain.cpp
@@ -119,12 +119,14 @@ 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;
- else
- SecondsToWait = 0; // infinite
+ llvm::sys::Process::GetEnv("CLANG_TOOLCHAIN_PROGRAM_TIMEOUT")) {
+ int SecondsToWait;
+ if (!llvm::to_integer(*Str, SecondsToWait))
+ return llvm::createStringError(std::error_code(),
+ "CLANG_TOOLCHAIN_PROGRAM_TIMEOUT expected "
+ "an integer, got '" +
+ *Str + "'");
+ SecondsToWait = std::min(SecondsToWait, 0);
}
if (llvm::sys::ExecuteAndWait(Executable, {}, {}, Redirects, SecondsToWait,
/*MemoryLimit=*/0, &ErrorMessage))
diff --git a/clang/test/Driver/amdgpu-hip-system-arch.c b/clang/test/Driver/amdgpu-hip-system-arch.c
index a46077b38fac08..f25a4087080f6d 100644
--- a/clang/test/Driver/amdgpu-hip-system-arch.c
+++ b/clang/test/Driver/amdgpu-hip-system-arch.c
@@ -29,3 +29,11 @@
// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib --offload-new-driver --offload-arch=native --amdgpu-arch-tool=%t/amdgpu_arch_gfx906 -x hip %s 2>&1 \
// RUN: | FileCheck %s --check-prefix=ARCH-GFX906
// ARCH-GFX906: "-cc1" "-triple" "amdgcn-amd-amdhsa"{{.*}}"-target-cpu" "gfx906"
+
+// case when CLANG_TOOLCHAIN_PROGRAM_TIMEOUT is malformed.
+// RUN: env CLANG_TOOLCHAIN_PROGRAM_TIMEOUT=foo \
+// RUN: not %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \
+// RUN: --offload-arch=native --amdgpu-arch-tool=%t/amdgpu_arch_gfx906 \
+// RUN: -x hip %s 2>&1 | \
+// RUN: FileCheck %s --check-prefix=BAD-TIMEOUT
+// BAD-TIMEOUT: clang: error: cannot determine amdgcn architecture: CLANG_TOOLCHAIN_PROGRAM_TIMEOUT expected an integer, got 'foo'; consider passing it via '--offload-arch'; environment variable CLANG_TOOLCHAIN_PROGRAM_TIMEOUT specifies the tool timeout (integer secs, <=0 is infinite)
diff --git a/clang/test/Driver/nvptx-cuda-system-arch.c b/clang/test/Driver/nvptx-cuda-system-arch.c
index e6a2fa40f0a038..6a8a218406d139 100644
--- a/clang/test/Driver/nvptx-cuda-system-arch.c
+++ b/clang/test/Driver/nvptx-cuda-system-arch.c
@@ -42,3 +42,11 @@
// RUN: --cuda-path=%S/Inputs/CUDA_102/usr/local/cuda %s 2>&1 | FileCheck %s --check-prefix=MARCH-sm_89
// MARCH-sm_89: warning: multiple nvptx64 architectures are detected: sm_89, sm_80; only the first one is used for '-march' [-Wmulti-gpu]
// MARCH-sm_89: "-cc1" "-triple" "nvptx64-nvidia-cuda"{{.*}}"-target-cpu" "sm_89"
+
+// case when CLANG_TOOLCHAIN_PROGRAM_TIMEOUT is malformed.
+// RUN: env CLANG_TOOLCHAIN_PROGRAM_TIMEOUT=foo \
+// RUN: not %clang -### --target=x86_64-unknown-linux-gnu -nogpulib \
+// RUN: --offload-arch=native --nvptx-arch-tool=%t/nvptx_arch_sm_70 \
+// RUN: --cuda-path=%S/Inputs/CUDA_102/usr/local/cuda -x cuda %s 2>&1 | \
+// RUN: FileCheck %s --check-prefix=BAD-TIMEOUT
+// BAD-TIMEOUT: clang: error: cannot determine nvptx64 architecture: CLANG_TOOLCHAIN_PROGRAM_TIMEOUT expected an integer, got 'foo'; consider passing it via '--offload-arch'; environment variable CLANG_TOOLCHAIN_PROGRAM_TIMEOUT specifies the tool timeout (integer secs, <=0 is infinite)
diff --git a/clang/test/Driver/openmp-system-arch.c b/clang/test/Driver/openmp-system-arch.c
index ea6ec6428592b3..cd49f460099666 100644
--- a/clang/test/Driver/openmp-system-arch.c
+++ b/clang/test/Driver/openmp-system-arch.c
@@ -75,3 +75,19 @@
// RUN: -fopenmp-targets=amdgcn-amd-amdhsa --amdgpu-arch-tool=%t/amdgpu_arch_empty %s 2>&1 \
// RUN: | FileCheck %s --check-prefix=AMDGPU
// AMDGPU: error: cannot determine amdgcn architecture: No AMD GPU detected in the system; consider passing it via '-march'
+
+// case when CLANG_TOOLCHAIN_PROGRAM_TIMEOUT is malformed for nvptx-arch.
+// RUN: env CLANG_TOOLCHAIN_PROGRAM_TIMEOUT=foo \
+// RUN: not %clang -### --target=x86_64-unknown-linux-gnu -fopenmp=libomp \
+// RUN: -fopenmp-targets=nvptx64-nvidia-cuda -nogpulib \
+// RUN: --nvptx-arch-tool=%t/nvptx_arch_sm_70 %s 2>&1 | \
+// RUN: FileCheck %s --check-prefix=BAD-TIMEOUT-NVPTX
+// BAD-TIMEOUT-NVPTX: clang: error: cannot determine nvptx64 architecture: CLANG_TOOLCHAIN_PROGRAM_TIMEOUT expected an integer, got 'foo'; consider passing it via '-march'; environment variable CLANG_TOOLCHAIN_PROGRAM_TIMEOUT specifies the tool timeout (integer secs, <=0 is infinite)
+
+// case when CLANG_TOOLCHAIN_PROGRAM_TIMEOUT is malformed for amdgpu-arch.
+// RUN: env CLANG_TOOLCHAIN_PROGRAM_TIMEOUT= \
+// RUN: not %clang -### --target=x86_64-unknown-linux-gnu -fopenmp=libomp \
+// RUN: -fopenmp-targets=amdgcn-amd-amdhsa -nogpulib \
+// RUN: --amdgpu-arch-tool=%t/amdgpu_arch_gfx906 %s 2>&1 | \
+// RUN: FileCheck %s --check-prefix=BAD-TIMEOUT-AMDGPU
+// BAD-TIMEOUT-AMDGPU: clang: error: cannot determine amdgcn architecture: CLANG_TOOLCHAIN_PROGRAM_TIMEOUT expected an integer, got ''; consider passing it via '-march'; environment variable CLANG_TOOLCHAIN_PROGRAM_TIMEOUT specifies the tool timeout (integer secs, <=0 is infinite)
diff --git a/llvm/utils/lit/lit/TestingConfig.py b/llvm/utils/lit/lit/TestingConfig.py
index 06713429d06b4b..9d8e93460933d8 100644
--- a/llvm/utils/lit/lit/TestingConfig.py
+++ b/llvm/utils/lit/lit/TestingConfig.py
@@ -26,7 +26,7 @@ def fromdefaults(litConfig):
"SYSTEMROOT",
"TERM",
"CLANG",
- "CLANG_TOOL_CHAIN_PROGRAM_WAIT",
+ "CLANG_TOOLCHAIN_PROGRAM_TIMEOUT",
"LLDB",
"LD_PRELOAD",
"LLVM_SYMBOLIZER_PATH",
>From 42ac67eae5c3952083de1fcb68b9344374ba56f6 Mon Sep 17 00:00:00 2001
From: "Joel E. Denny" <jdenny.ornl at gmail.com>
Date: Thu, 8 Aug 2024 19:33:44 -0400
Subject: [PATCH 5/5] I wish I knew how to write timeout tests
---
clang/lib/Driver/ToolChain.cpp | 1 -
1 file changed, 1 deletion(-)
diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp
index 5fd8af373e1db5..dbac6c370e85e6 100644
--- a/clang/lib/Driver/ToolChain.cpp
+++ b/clang/lib/Driver/ToolChain.cpp
@@ -120,7 +120,6 @@ ToolChain::executeToolChainProgram(StringRef Executable,
int SecondsToWait = DefaultSecondsToWait;
if (std::optional<std::string> Str =
llvm::sys::Process::GetEnv("CLANG_TOOLCHAIN_PROGRAM_TIMEOUT")) {
- int SecondsToWait;
if (!llvm::to_integer(*Str, SecondsToWait))
return llvm::createStringError(std::error_code(),
"CLANG_TOOLCHAIN_PROGRAM_TIMEOUT expected "
More information about the cfe-commits
mailing list