caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[libomptarget] Build plugins-nextgen for SystemZ (#83978)

Browse Source 
The plugin was not getting built as the build_generic_elf64 macro
assumes the LLVM triple processor name matches the CMake processor name,
which is unfortunately not the case for SystemZ.

Fix this by providing two separate arguments instead.

Actually building the plugin exposed a number of other issues causing
various test failures. Specifically, I've had to add the SystemZ target
to
- CompilerInvocation::ParseLangArgs
- linkDevice in ClangLinuxWrapper.cpp
- OMPContext::OMPContext (to set the device_kind_cpu trait)
- LIBOMPTARGET_ALL_TARGETS in libomptarget/CMakeLists.txt
- a check_plugin_target call in libomptarget/src/CMakeLists.txt

Finally, I've had to set a number of test cases to UNSUPPORTED on
s390x-ibm-linux-gnu; all these tests were already marked as UNSUPPORTED
for x86_64-pc-linux-gnu and aarch64-unknown-linux-gnu and are failing on
s390x for what seem to be the same reason.

In addition, this also requires support for BE ELF files in
plugins-nextgen: https://github.com/llvm/llvm-project/pull/83976
This commit is contained in:
Ulrich Weigand
2024-03-06 20:50:01 +01:00
committed by GitHub
parent 15b7b3182c
commit 3ecd38c8e1
37 changed files with 71 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
clang/lib/Frontend/CompilerInvocation.cpp
View File

@@ -4012,6 +4012,7 @@ bool CompilerInvocation::ParseLangArgs(LangOptions &Opts, ArgList &Args,
if (TT.getArch() == llvm::Triple::UnknownArch ||
!(TT.getArch() == llvm::Triple::aarch64 || TT.isPPC() ||
TT.getArch() == llvm::Triple::systemz ||
TT.getArch() == llvm::Triple::nvptx ||
TT.getArch() == llvm::Triple::nvptx64 ||
TT.getArch() == llvm::Triple::amdgcn ||

1
clang/tools/clang-linker-wrapper/ClangLinkerWrapper.cpp
View File

@@ -552,6 +552,7 @@ Expected<StringRef> linkDevice(ArrayRef<StringRef> InputFiles,
case Triple::aarch64_be:
case Triple::ppc64:
case Triple::ppc64le:
case Triple::systemz:
return generic::clang(InputFiles, Args);
default:
return createStringError(inconvertibleErrorCode(),

1
llvm/lib/Frontend/OpenMP/OMPContext.cpp
View File

@@ -44,6 +44,7 @@ OMPContext::OMPContext(bool IsDeviceCompilation, Triple TargetTriple) {
case Triple::ppcle:
case Triple::ppc64:
case Triple::ppc64le:
case Triple::systemz:
case Triple::x86:
case Triple::x86_64:
ActiveTraits.set(unsigned(TraitProperty::device_kind_cpu));

2
openmp/libomptarget/CMakeLists.txt
View File

@@ -56,6 +56,8 @@ set (LIBOMPTARGET_ALL_TARGETS "${LIBOMPTARGET_ALL_TARGETS} x86_64-pc-linux-gnu-L
set (LIBOMPTARGET_ALL_TARGETS "${LIBOMPTARGET_ALL_TARGETS} nvptx64-nvidia-cuda")
set (LIBOMPTARGET_ALL_TARGETS "${LIBOMPTARGET_ALL_TARGETS} nvptx64-nvidia-cuda-LTO")
set (LIBOMPTARGET_ALL_TARGETS "${LIBOMPTARGET_ALL_TARGETS} nvptx64-nvidia-cuda-JIT-LTO")
set (LIBOMPTARGET_ALL_TARGETS "${LIBOMPTARGET_ALL_TARGETS} s390x-ibm-linux-gnu")
set (LIBOMPTARGET_ALL_TARGETS "${LIBOMPTARGET_ALL_TARGETS} s390x-ibm-linux-gnu-LTO")
# Once the plugins for the different targets are validated, they will be added to
# the list of supported targets in the current system.

9
openmp/libomptarget/plugins-nextgen/CMakeLists.txt
View File

@@ -12,12 +12,15 @@
add_subdirectory(common)
# void build_generic_elf64(string tmachine, string tmachine_name, string tmachine_libname, string elf_machine_id);
# void build_generic_elf64(string tmachine, string tmachine_name, string tmachine_libname,
# string tmachine_llvm, string tmachine_triple, string elf_machine_id);
# - build a plugin for an ELF based generic 64-bit target based on libffi.
# - tmachine: name of the machine processor as used in the cmake build system.
# - tmachine_name: name of the machine to be printed with the debug messages.
# - tmachine_libname: machine name to be appended to the plugin library name.
macro(build_generic_elf64 tmachine tmachine_name tmachine_libname tmachine_triple elf_machine_id)
# - tmachine_llvm: LLVM triple for the processor
# - tmachine_triple: GNU target triple
macro(build_generic_elf64 tmachine tmachine_name tmachine_libname tmachine_llvm tmachine_triple elf_machine_id)
if(CMAKE_SYSTEM_PROCESSOR MATCHES "${tmachine}$")
# Define macro to be used as prefix of the runtime messages for this target.
add_definitions("-DTARGET_NAME=${tmachine_name}")
@@ -30,7 +33,7 @@ if(CMAKE_SYSTEM_PROCESSOR MATCHES "${tmachine}$")
add_definitions("-DTARGET_ELF_ID=${elf_machine_id}")
# Define target triple
add_definitions("-DLIBOMPTARGET_NEXTGEN_GENERIC_PLUGIN_TRIPLE=${tmachine}")
add_definitions("-DLIBOMPTARGET_NEXTGEN_GENERIC_PLUGIN_TRIPLE=${tmachine_llvm}")
add_llvm_library("omptarget.rtl.${tmachine_libname}"
SHARED

2
openmp/libomptarget/plugins-nextgen/aarch64/CMakeLists.txt
View File

@@ -11,7 +11,7 @@
##===----------------------------------------------------------------------===##
if(CMAKE_SYSTEM_NAME MATCHES "Linux")
build_generic_elf64("aarch64" "aarch64" "aarch64" "aarch64-unknown-linux-gnu" "183")
build_generic_elf64("aarch64" "aarch64" "aarch64" "aarch64" "aarch64-unknown-linux-gnu" "183")
else()
libomptarget_say("Not building aarch64 NextGen offloading plugin: machine not found in the system.")
endif()

2
openmp/libomptarget/plugins-nextgen/ppc64/CMakeLists.txt
View File

@@ -11,7 +11,7 @@
##===----------------------------------------------------------------------===##
if(CMAKE_SYSTEM_NAME MATCHES "Linux")
build_generic_elf64("ppc64" "PPC64" "ppc64" "powerpc64-ibm-linux-gnu" "21")
build_generic_elf64("ppc64" "PPC64" "ppc64" "ppc64" "powerpc64-ibm-linux-gnu" "21")
else()
libomptarget_say("Not building ppc64 NextGen offloading plugin: machine not found in the system.")
endif()

2
openmp/libomptarget/plugins-nextgen/ppc64le/CMakeLists.txt
View File

@@ -11,7 +11,7 @@
##===----------------------------------------------------------------------===##
if(CMAKE_SYSTEM_NAME MATCHES "Linux")
build_generic_elf64("ppc64le" "PPC64le" "ppc64" "powerpc64le-ibm-linux-gnu" "21")
build_generic_elf64("ppc64le" "PPC64le" "ppc64" "ppc64le" "powerpc64le-ibm-linux-gnu" "21")
else()
libomptarget_say("Not building ppc64le NextGen offloading plugin: machine not found in the system.")
endif()

2
openmp/libomptarget/plugins-nextgen/s390x/CMakeLists.txt
View File

@@ -11,7 +11,7 @@
##===----------------------------------------------------------------------===##
if(CMAKE_SYSTEM_NAME MATCHES "Linux")
build_generic_elf64("SystemZ" "S390X" "s390x" "s390x-ibm-linux-gnu" "22")
build_generic_elf64("s390x" "S390X" "s390x" "systemz" "s390x-ibm-linux-gnu" "22")
else()
libomptarget_say("Not building s390x NextGen offloading plugin: machine not found in the system.")
endif()

2
openmp/libomptarget/plugins-nextgen/x86_64/CMakeLists.txt
View File

@@ -11,7 +11,7 @@
##===----------------------------------------------------------------------===##
if(CMAKE_SYSTEM_NAME MATCHES "Linux")
build_generic_elf64("x86_64" "x86_64" "x86_64" "x86_64-pc-linux-gnu" "62")
build_generic_elf64("x86_64" "x86_64" "x86_64" "x86_64" "x86_64-pc-linux-gnu" "62")
else()
libomptarget_say("Not building x86_64 NextGen offloading plugin: machine not found in the system.")
endif()

1
openmp/libomptarget/src/CMakeLists.txt
View File

@@ -73,6 +73,7 @@ if (NOT LIBOMPTARGET_PLUGINS_TO_LOAD)
check_plugin_target(cuda)
check_plugin_target(aarch64)
check_plugin_target(amdgpu)
check_plugin_target(s390x)
endif()
list(TRANSFORM LIBOMPTARGET_PLUGINS_TO_LOAD PREPEND "\"libomptarget.rtl.")

2
openmp/libomptarget/test/api/omp_dynamic_shared_memory.c
View File

@@ -10,6 +10,8 @@
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: aarch64-unknown-linux-gnu
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
#include <omp.h>
#include <stdio.h>

2
openmp/libomptarget/test/jit/empty_kernel_lvl1.c
View File

@@ -32,5 +32,7 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
#include "empty_kernel.inc"

2
openmp/libomptarget/test/jit/empty_kernel_lvl2.c
View File

@@ -92,5 +92,7 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
#include "empty_kernel.inc"

2
openmp/libomptarget/test/jit/type_punning.c
View File

@@ -12,6 +12,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
// Ensure that there is only the kernel function left, not any outlined
// parallel regions.

2
openmp/libomptarget/test/mapping/auto_zero_copy.cpp
View File

@@ -13,6 +13,8 @@
// UNSUPPORTED: nvptx64-nvidia-cuda-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
// REQUIRES: unified_shared_memory

2
openmp/libomptarget/test/mapping/auto_zero_copy_globals.cpp
View File

@@ -9,6 +9,8 @@
// UNSUPPORTED: nvptx64-nvidia-cuda-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
// REQUIRES: unified_shared_memory

2
openmp/libomptarget/test/offloading/barrier_fence.c
View File

@@ -7,6 +7,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
#include <omp.h>
#include <stdio.h>

2
openmp/libomptarget/test/offloading/bug49334.cpp
View File

@@ -9,6 +9,8 @@
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: aarch64-unknown-linux-gnu
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
// UNSUPPORTED: amdgcn-amd-amdhsa
// UNSUPPORTED: nvptx64-nvidia-cuda
// UNSUPPORTED: nvptx64-nvidia-cuda-LTO

2
openmp/libomptarget/test/offloading/default_thread_limit.c
View File

@@ -9,6 +9,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
__attribute__((optnone)) int optnone() { return 1; }

5
openmp/libomptarget/test/offloading/ompx_bare.c
View File

@@ -1,10 +1,13 @@
// RUN: %libomptarget-compile-generic
// RUN: env LIBOMPTARGET_INFO=63 %libomptarget-run-generic 2>&1 | %fcheck-generic
// RUN: env LIBOMPTARGET_INFO=63 %libomptarget-run-generic 2>&1 | \
// RUN: %fcheck-generic
//
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: aarch64-unknown-linux-gnu
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
#include <assert.h>
#include <ompx.h>

2
openmp/libomptarget/test/offloading/ompx_coords.c
View File

@@ -4,6 +4,8 @@
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: aarch64-unknown-linux-gnu
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
#include <omp.h>
#include <ompx.h>

2
openmp/libomptarget/test/offloading/ompx_saxpy_mixed.c
View File

@@ -4,6 +4,8 @@
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: aarch64-unknown-linux-gnu
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
#include <math.h>
#include <omp.h>

2
openmp/libomptarget/test/offloading/parallel_target_teams_reduction.cpp
View File

@@ -6,6 +6,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
#include <iostream>
#include <vector>

2
openmp/libomptarget/test/offloading/small_trip_count.c
View File

@@ -9,6 +9,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
#define N 128

2
openmp/libomptarget/test/offloading/small_trip_count_thread_limit.cpp
View File

@@ -7,6 +7,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
int main(int argc, char *argv[]) {
constexpr const int block_size = 256;

2
openmp/libomptarget/test/offloading/spmdization.c
View File

@@ -11,6 +11,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
#include <omp.h>
#include <stdio.h>

2
openmp/libomptarget/test/offloading/target_critical_region.cpp
View File

@@ -6,6 +6,8 @@
// UNSUPPORTED: nvptx64-nvidia-cuda-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
// UNSUPPORTED: amdgcn-amd-amdhsa
#include <omp.h>

2
openmp/libomptarget/test/offloading/thread_limit.c
View File

@@ -9,6 +9,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
int main() {
int n = 1 << 20;

2
openmp/libomptarget/test/ompt/target_memcpy.c
View File

@@ -4,6 +4,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
/*
* Verify that for the target OpenMP APIs, the return address is non-null and

2
openmp/libomptarget/test/ompt/target_memcpy_emi.c
View File

@@ -4,6 +4,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
/*
* Verify all three data transfer directions: H2D, D2D and D2H

2
openmp/libomptarget/test/ompt/veccopy.c
View File

@@ -4,6 +4,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
/*
* Example OpenMP program that registers non-EMI callbacks

2
openmp/libomptarget/test/ompt/veccopy_data.c
View File

@@ -4,6 +4,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
/*
* Example OpenMP program that registers EMI callbacks.

2
openmp/libomptarget/test/ompt/veccopy_disallow_both.c
View File

@@ -4,6 +4,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
/*
* Example OpenMP program that shows that both EMI and non-EMI

2
openmp/libomptarget/test/ompt/veccopy_emi.c
View File

@@ -4,6 +4,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
/*
* Example OpenMP program that registers EMI callbacks

2
openmp/libomptarget/test/ompt/veccopy_emi_map.c
View File

@@ -4,6 +4,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
/*
* Example OpenMP program that shows that map-EMI callbacks are not supported.

2
openmp/libomptarget/test/ompt/veccopy_map.c
View File

@@ -4,6 +4,8 @@
// UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
// UNSUPPORTED: x86_64-pc-linux-gnu
// UNSUPPORTED: x86_64-pc-linux-gnu-LTO
// UNSUPPORTED: s390x-ibm-linux-gnu
// UNSUPPORTED: s390x-ibm-linux-gnu-LTO
/*
* Example OpenMP program that shows that map callbacks are not supported.