[libcxx-commits] [libcxx] [libcxx] Add testing configuration for GPU targets (PR #104515)

Joseph Huber via libcxx-commits libcxx-commits at lists.llvm.org
Mon Oct 14 14:57:22 PDT 2024


https://github.com/jhuber6 updated https://github.com/llvm/llvm-project/pull/104515

>From 83c207e6ab99c689354b5ee37298d985be1d25c0 Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn at outlook.com>
Date: Fri, 2 Aug 2024 10:43:06 -0500
Subject: [PATCH] [libcxx] Add testing configuration for GPU targets

Summary:
The GPU runs these tests using the files built from the `libc` project.
These will be placed in `include/<triple>` and `lib/<triple>`. We use
the `amdhsa-loader` and `nvptx-loader` tools, which are also provided by
`libc`. These launch a kernel called `_start` which calls `main` so we
can pretend like GPU programs are normal terminal applications.

We force serial exeuction here, because `llvm-lit` runs way too many
processes in parallel, which has a bad habit of making the GPU drivers
hang or run out of resources. This allows the compilation to be run in
parallel while the jobs themselves are serialized via a file lock.

In the future this can likely be refined to accept user specified
architectures, or better handle including the root directory by exposing
that instead of just `include/<triple>/c++/v1/`.

This currently fails ~1% of the tests on AMDGPU and ~3% of the tests on
NVPTX. This will hopefully be reduced further, and later patches can
XFAIL a lot of them once it's down to a reasonable number.

Future support will likely want to allow passing in a custom
architecture instead of simply relying on `-mcpu=native`.
---
 libcxx/cmake/caches/AMDGPU.cmake              |  4 +++
 libcxx/cmake/caches/NVPTX.cmake               |  4 +++
 .../test/configs/amdgpu-libc++-shared.cfg.in  | 29 ++++++++++++++++++
 .../test/configs/nvptx-libc++-shared.cfg.in   | 30 +++++++++++++++++++
 .../deque.modifiers/insert_range.pass.cpp     |  4 +++
 .../replace_with_range.pass.cpp               |  4 +++
 libcxx/utils/libcxx/test/features.py          |  8 +++++
 7 files changed, 83 insertions(+)
 create mode 100644 libcxx/test/configs/amdgpu-libc++-shared.cfg.in
 create mode 100644 libcxx/test/configs/nvptx-libc++-shared.cfg.in

diff --git a/libcxx/cmake/caches/AMDGPU.cmake b/libcxx/cmake/caches/AMDGPU.cmake
index c7d6afc854a54c..f563e1a12c4c19 100644
--- a/libcxx/cmake/caches/AMDGPU.cmake
+++ b/libcxx/cmake/caches/AMDGPU.cmake
@@ -27,6 +27,10 @@ set(LIBCXXABI_ENABLE_SHARED OFF CACHE BOOL "")
 set(LIBCXXABI_ENABLE_THREADS OFF CACHE BOOL "")
 set(LIBCXXABI_USE_LLVM_UNWINDER OFF CACHE BOOL "")
 
+# Test configuration.
+set(LIBCXX_TEST_CONFIG "amdgpu-libc++-shared.cfg.in" CACHE STRING "")
+set(LIBCXX_TEST_PARAMS "optimization=none;long_tests=False;executor=amdhsa-loader" CACHE STRING "")
+
 # Necessary compile flags for AMDGPU.
 set(LIBCXX_ADDITIONAL_COMPILE_FLAGS
     "-nogpulib;-flto;-fconvergent-functions;-Xclang;-mcode-object-version=none" CACHE STRING "")
diff --git a/libcxx/cmake/caches/NVPTX.cmake b/libcxx/cmake/caches/NVPTX.cmake
index 231b457d3017e8..b6ec6c04ea700c 100644
--- a/libcxx/cmake/caches/NVPTX.cmake
+++ b/libcxx/cmake/caches/NVPTX.cmake
@@ -27,6 +27,10 @@ set(LIBCXXABI_ENABLE_SHARED OFF CACHE BOOL "")
 set(LIBCXXABI_ENABLE_THREADS OFF CACHE BOOL "")
 set(LIBCXXABI_USE_LLVM_UNWINDER OFF CACHE BOOL "")
 
+# Test configuration.
+set(LIBCXX_TEST_CONFIG "nvptx-libc++-shared.cfg.in" CACHE STRING "")
+set(LIBCXX_TEST_PARAMS "optimization=none;long_tests=False;executor=nvptx-loader" CACHE STRING "")
+
 # Necessary compile flags for NVPTX.
 set(LIBCXX_ADDITIONAL_COMPILE_FLAGS
     "-nogpulib;-flto;-fconvergent-functions;--cuda-feature=+ptx63" CACHE STRING "")
diff --git a/libcxx/test/configs/amdgpu-libc++-shared.cfg.in b/libcxx/test/configs/amdgpu-libc++-shared.cfg.in
new file mode 100644
index 00000000000000..80c80ebaa8bd0d
--- /dev/null
+++ b/libcxx/test/configs/amdgpu-libc++-shared.cfg.in
@@ -0,0 +1,29 @@
+lit_config.load_config(config, '@CMAKE_CURRENT_BINARY_DIR@/cmake-bridge.cfg')
+
+config.substitutions.append(('%{flags}',
+  f'--target={config.target_triple} -Wno-multi-gpu -flto -mcpu=native'))
+config.substitutions.append(('%{compile_flags}',
+    '-nogpulib -fno-builtin-printf -I %{include-dir} '
+    '-I %{target-include-dir} -I %{libcxx-dir}/test/support'
+))
+config.substitutions.append(('%{link_flags}',
+  '-nostdlib++ -startfiles -stdlib '
+  '-Wl,-mllvm,-amdgpu-lower-global-ctor-dtor=0 '
+  '-L %{lib-dir} -lc++ -lc++abi -lclang_rt.builtins'
+))
+
+config.substitutions.append(('%{exec}',
+    '%{executor}'
+))
+
+config.stdlib = 'llvm-libc++'
+
+import os, site
+site.addsitedir(os.path.join('@LIBCXX_SOURCE_DIR@', 'utils'))
+import libcxx.test.params, libcxx.test.config
+libcxx.test.config.configure(
+    libcxx.test.params.DEFAULT_PARAMETERS,
+    libcxx.test.features.DEFAULT_FEATURES,
+    config,
+    lit_config
+)
diff --git a/libcxx/test/configs/nvptx-libc++-shared.cfg.in b/libcxx/test/configs/nvptx-libc++-shared.cfg.in
new file mode 100644
index 00000000000000..9a3ca9c8da9509
--- /dev/null
+++ b/libcxx/test/configs/nvptx-libc++-shared.cfg.in
@@ -0,0 +1,30 @@
+lit_config.load_config(config, '@CMAKE_CURRENT_BINARY_DIR@/cmake-bridge.cfg')
+
+config.substitutions.append(('%{flags}',
+  f'--target={config.target_triple} -Wno-multi-gpu -flto -march=native'))
+config.substitutions.append(('%{compile_flags}',
+    '-nogpulib -fno-builtin-printf -I %{include-dir} '
+    '-I %{target-include-dir} -I %{libcxx-dir}/test/support'
+))
+config.substitutions.append(('%{link_flags}',
+   '-nostdlib++ -startfiles -stdlib '
+   '-L %{lib-dir} -lc++ -lc++abi '
+   '-Wl,--suppress-stack-size-warning '
+   '-Wl,-mllvm,-nvptx-lower-global-ctor-dtor=1 '
+   '-Wl,-mllvm,-nvptx-emit-init-fini-kernel'
+))
+config.substitutions.append(('%{exec}',
+    '%{executor} --no-parallelism'
+))
+
+config.stdlib = 'llvm-libc++'
+
+import os, site
+site.addsitedir(os.path.join('@LIBCXX_SOURCE_DIR@', 'utils'))
+import libcxx.test.params, libcxx.test.config
+libcxx.test.config.configure(
+    libcxx.test.params.DEFAULT_PARAMETERS,
+    libcxx.test.features.DEFAULT_FEATURES,
+    config,
+    lit_config
+)
diff --git a/libcxx/test/std/containers/sequences/deque/deque.modifiers/insert_range.pass.cpp b/libcxx/test/std/containers/sequences/deque/deque.modifiers/insert_range.pass.cpp
index a5f5455297ad44..7681eb63b90761 100644
--- a/libcxx/test/std/containers/sequences/deque/deque.modifiers/insert_range.pass.cpp
+++ b/libcxx/test/std/containers/sequences/deque/deque.modifiers/insert_range.pass.cpp
@@ -6,6 +6,10 @@
 //
 //===----------------------------------------------------------------------===//
 
+// FIXME: This takes over an hour to compile, disable for now.
+// UNSUPPORTED: LIBCXX-AMDGPU-FIXME
+// UNSUPPORTED: LIBCXX-NVPTX-FIXME
+
 // UNSUPPORTED: c++03, c++11, c++14, c++17, c++20
 // UNSUPPORTED: GCC-ALWAYS_INLINE-FIXME
 
diff --git a/libcxx/test/std/strings/basic.string/string.modifiers/string_replace/replace_with_range.pass.cpp b/libcxx/test/std/strings/basic.string/string.modifiers/string_replace/replace_with_range.pass.cpp
index 03e82590ed4ef6..2bd27f63e751a6 100644
--- a/libcxx/test/std/strings/basic.string/string.modifiers/string_replace/replace_with_range.pass.cpp
+++ b/libcxx/test/std/strings/basic.string/string.modifiers/string_replace/replace_with_range.pass.cpp
@@ -6,6 +6,10 @@
 //
 //===----------------------------------------------------------------------===//
 
+// FIXME: This takes over an hour to compile, disable for now.
+// UNSUPPORTED: LIBCXX-AMDGPU-FIXME
+// UNSUPPORTED: LIBCXX-NVPTX-FIXME
+
 // UNSUPPORTED: c++03, c++11, c++14, c++17, c++20
 // ADDITIONAL_COMPILE_FLAGS(has-fconstexpr-steps): -fconstexpr-steps=10000000
 // ADDITIONAL_COMPILE_FLAGS(has-fconstexpr-ops-limit): -fconstexpr-ops-limit=70000000
diff --git a/libcxx/utils/libcxx/test/features.py b/libcxx/utils/libcxx/test/features.py
index 735eb5ac949dc0..2381296009f3a7 100644
--- a/libcxx/utils/libcxx/test/features.py
+++ b/libcxx/utils/libcxx/test/features.py
@@ -496,6 +496,14 @@ def _mingwSupportsModules(cfg):
           """,
         ),
     ),
+    Feature(
+        name="LIBCXX-AMDGPU-FIXME",
+        when=lambda cfg: "__AMDGPU__" in compilerMacros(cfg),
+    ),
+    Feature(
+        name="LIBCXX-NVPTX-FIXME",
+        when=lambda cfg: "__NVPTX__" in compilerMacros(cfg),
+    ),
     Feature(
         name="can-create-symlinks",
         when=lambda cfg: "_WIN32" not in compilerMacros(cfg)



More information about the libcxx-commits mailing list