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

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

@@ -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;SHELL:-Xclang -mcode-object-version=none" CACHE STRING "")

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 "")

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
+)

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
+)

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
 

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

libcxx/utils/libcxx/test/features.py

@@ -496,6 +496,14 @@ DEFAULT_FEATURES += [
           """,
         ),
     ),
+    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)