[flang-commits] [flang] [Flang] Stop building multiple verisons of the OpenMP library (PR #126657)

[Joseph Huber via flang-commits]

https://github.com/jhuber6 updated https://github.com/llvm/llvm-project/pull/126657

>From 6a05bc07d665a512fbf1b7d75e342d5ab7b16cbf Mon Sep 17 00:00:00 2001
From: Joseph Huber <huberjn at outlook.com>
Date: Mon, 10 Feb 2025 20:44:13 -0600
Subject: [PATCH] [Flang] Stop building multiple verisons of the OpenMP library

Summary:
We do not need to build for all of these architectures. Since the
infrastructure uses the natively created packager binary we can rely on
the support for generic targets. This is the same thing the OpenMP /
libc projects use now.

Depends on https://github.com/llvm/llvm-project/pull/126655
---
 .../modules/AddFlangOffloadRuntime.cmake      | 30 +++----------------
 1 file changed, 4 insertions(+), 26 deletions(-)

diff --git a/flang/cmake/modules/AddFlangOffloadRuntime.cmake b/flang/cmake/modules/AddFlangOffloadRuntime.cmake
index 8e4f47d18535d..7b598241e1681 100644
--- a/flang/cmake/modules/AddFlangOffloadRuntime.cmake
+++ b/flang/cmake/modules/AddFlangOffloadRuntime.cmake
@@ -96,36 +96,14 @@ macro(enable_omp_offload_compilation files)
     if ("${CMAKE_CXX_COMPILER_ID}" MATCHES "Clang" AND
         "${CMAKE_C_COMPILER_ID}" MATCHES "Clang")
 
-      set(all_amdgpu_architectures
-        "gfx700;gfx701;gfx801;gfx803;gfx900;gfx902;gfx906"
-        "gfx908;gfx90a;gfx90c;gfx940;gfx1010;gfx1030"
-        "gfx1031;gfx1032;gfx1033;gfx1034;gfx1035;gfx1036"
-        "gfx1100;gfx1101;gfx1102;gfx1103;gfx1150;gfx1151"
-        "gfx1152;gfx1153"
-        )
-      set(all_nvptx_architectures
-        "sm_35;sm_37;sm_50;sm_52;sm_53;sm_60;sm_61;sm_62"
-        "sm_70;sm_72;sm_75;sm_80;sm_86;sm_89;sm_90"
-        )
-      set(all_gpu_architectures
-        "${all_amdgpu_architectures};${all_nvptx_architectures}"
-        )
-      # TODO: support auto detection on the build system.
-      if (FLANG_OMP_DEVICE_ARCHITECTURES STREQUAL "all")
-        set(FLANG_OMP_DEVICE_ARCHITECTURES ${all_gpu_architectures})
-      endif()
-      list(REMOVE_DUPLICATES FLANG_OMP_DEVICE_ARCHITECTURES)
-
-      string(REPLACE ";" "," compile_for_architectures
-        "${FLANG_OMP_DEVICE_ARCHITECTURES}"
-        )
-
       set(OMP_COMPILE_OPTIONS
         -fopenmp
         -fvisibility=hidden
         -fopenmp-cuda-mode
-        --offload-arch=${compile_for_architectures}
-        # Force LTO for the device part.
+        -fopenmp-targets=amdgcn-amd-amdhsa,nvptx64-nvidia-cuda
+        -Xopenmp-target=amdgcn-amd-amdhsa -march=''
+        -Xopenmp-target=nvptx64-nvidia-cuda -march=''
+        -nogpulib
         -foffload-lto
         )
       set_source_files_properties(${files} PROPERTIES COMPILE_OPTIONS