[HIP] Instruct lld to go through all archives

[scchan]

Summary

Add the --whole-archive flag when linking HIP programs to instruct lld
to go through every archive library to link in all the kernel functions
(entry pointers to the GPU program); otherwise, lld may skip some
library files if there are no more symbols that need to be resolved.

Change-Id: Ic7ea61ec3e6925f80a63d04654ac72177ffb27c8

[yaxunl]

clang/lib/Driver/ToolChains/HIPAMD.cpp

168

Better move this before line 155 to allow users to override this option.

Please add a comment that the device linker will not try to resolve a device variable or kernel that is referenced by a host object in the archive which in turn is referenced by a host object not in the archive, therefore --whole-archive is needed.

[scchan]

Updated patch to address review feedback

[scchan]

remove ws

[yaxunl]

LGTM. Thanks.

[scchan]

fix some formatting issue

[MaskRay]

clang/lib/Driver/ToolChains/HIPAMD.cpp

165

Though not strictly required, it's usually better to end --whole-archive with a pairing --no-whole-archive.

[scchan]

added a matching --no-whole-archive

[MaskRay]

clang/lib/Driver/ToolChains/HIPAMD.cpp

175

For complete sentences in comments, capitalize and add a full stop.

Actually, this statement is self-explanatory and I don't think a comment is useful.

[scchan]

Removed redundant comment.

[scchan]

rebased

clang/lib/Driver/ToolChains/HIPAMD.cpp

 //===--- HIPAMD.cpp - HIP Tool and ToolChain Implementations ----*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
for (auto *Arg : Args.filtered(options::OPT_Xoffload_linker)) {
     Arg->claim();
   }
 
   LldArgs.append({"-o", Output.getFilename()});
   for (auto Input : Inputs)
     LldArgs.push_back(Input.getFilename());
 
   // Look for archive of bundled bitcode in arguments, and add temporary files
   // for the extracted archive of bitcode to inputs.
+  // The archive libraries have to be linked with --whole-archive to instruct
+  // the linker to go through every library to look for kernel functions
+  LldArgs.push_back("--whole-archive");
   auto TargetID = Args.getLastArgValue(options::OPT_mcpu_EQ);
   AddStaticDeviceLibsLinking(C, *this, JA, Inputs, Args, LldArgs, "amdgcn",
                              TargetID,
                              /*IsBitCodeSDL=*/true,
                              /*PostClangLink=*/false);
 
   const char *Lld = Args.MakeArgString(getToolChain().GetProgramPath("lld"));
   C.addCommand(std::make_unique<Command>(JA, *this, ResponseFileSupport::None(),
                                          Lld, LldArgs, Inputs, Output));
 }
 
 // For amdgcn the inputs of the linker job are device bitcode and output is
 // either an object file or bitcode (-emit-llvm). It calls llvm-link, opt,
 // llc, then lld steps.
 void AMDGCN::Linker::ConstructJob(Compilation &C, const JobAction &JA,

clang/test/Driver/hip-toolchain-rdc-separate.hip

 // LINK-NOT: "*.llvm-link"
 // LINK-NOT: ".*opt"
 // LINK-NOT: ".*llc"
 // LINK: {{".*lld.*"}} {{.*}} "-plugin-opt=-amdgpu-internalize-symbols"
 // LINK-SAME: "-plugin-opt=mcpu=gfx803"
 // LLD-TMP-SAME: "-o" "[[IMG_DEV1:.*.out]]"
 // LLD-FIN-SAME: "-o" "[[IMG_DEV1:a.out-.*gfx803]]"
 // LINK-SAME "[[A_BC1]]" "[[B_BC1]]"
+// LINK-SAME: "--whole-archive"
 
 // LINK-NOT: "*.llvm-link"
 // LINK-NOT: ".*opt"
 // LINK-NOT: ".*llc"
 // LINK: {{".*lld.*"}} {{.*}} "-plugin-opt=-amdgpu-internalize-symbols"
 // LINK-SAME: "-plugin-opt=mcpu=gfx900"
 // LLD-TMP-SAME: "-o" "[[IMG_DEV2:.*.out]]"
 // LLD-FIN-SAME: "-o" "[[IMG_DEV1:a.out-.*gfx900]]"
 // LINK-SAME "[[A_BC2]]" "[[B_BC2]]"
+// LINK-SAME: "--whole-archive"
 
 // LINK-BUNDLE: [[BUNDLER:".*clang-offload-bundler"]] "-type=o"
 // LINK-BUNDLE-SAME: "-targets={{.*}},hipv4-amdgcn-amd-amdhsa--gfx803,hipv4-amdgcn-amd-amdhsa--gfx900"
 // LINK-BUNDLE-SAME: "-input={{.*}}" "-input=[[IMG_DEV1]]" "-input=[[IMG_DEV2]]" "-output=[[BUNDLE:.*]]"
 // LINK-NOBUNDLE-NOT: {{".*clang-offload-bundler"}} "-type=o"
 
 // LINK-EMBED: {{".*llvm-mc.*"}} "-o" "[[OBJBUNDLE:.*o]]" "{{.*}}.mcin" "--filetype=obj"
 // LINK-NOEMBED-NOT: {{".*llvm-mc.*"}} "-o"
 
 // LINK-EMBED: [[LD:".*ld.*"]] {{.*}} "-o" "a.out" {{.*}} "[[A_OBJ_HOST]]"
 // LINK-EMBED-SAME: "[[B_OBJ_HOST]]" "[[OBJBUNDLE]]"
 // LINK-NOEMBED-NOT: {{".*ld.*"}} {{.*}} "-o" "a.out"

clang/test/Driver/hip-toolchain-rdc-static-lib.hip

 // CHECK-SAME: {{.*}} [[B_SRC]]
 
 // CHECK-NOT: "*.llvm-link"
 // CHECK-NOT: ".*opt"
 // CHECK-NOT: ".*llc"
 // CHECK: [[LLD]] {{.*}} "-plugin-opt=-amdgpu-internalize-symbols"
 // CHECK-SAME: "-plugin-opt=mcpu=gfx900"
 // CHECK-SAME: "-o" "[[IMG_DEV2:.*out]]" [[A_BC2]] [[B_BC2]]
+// CHECK-SAME: "--whole-archive"
 
 // combine images generated into hip fat binary object
 // CHECK: [[BUNDLER:".*clang-offload-bundler"]] "-type=o"
 // CHECK-SAME: "-targets={{.*}},hipv4-amdgcn-amd-amdhsa--gfx803,hipv4-amdgcn-amd-amdhsa--gfx900"
 // CHECK-SAME: "-input=[[IMG_DEV1]]" "-input=[[IMG_DEV2]]" "-output=[[BUNDLE:.*hipfb]]"
 
 // CHECK: [[MC:".*llvm-mc.*"]] "-o" [[OBJBUNDLE:".*o"]] "{{.*}}.mcin" "--filetype=obj"
 
 // CHECK: [[AR:".*llvm-ar.*"]] "rcsD" "{{.*}}.out" [[A_OBJ_HOST]] [[B_OBJ_HOST]] [[OBJBUNDLE]]