[HIP] Enable -amdgpu-internalize-symbols

Enable -amdgpu-internalize-symbols to eliminate unused functions and global variables
for whole program to speed up compilation and improve performance.

For -fno-gpu-rdc, -amdgpu-internalize-symbols is passed to clang -cc1.

For -fgpu-rdc, -amdgpu-internalize-symbols is passed to lld.

Differential Revision: https://reviews.llvm.org/D81959

clang/lib/Driver/ToolChains/HIP.cpp

@@ -57,8 +57,14 @@ void AMDGCN::Linker::constructLldCommand(Compilation &C, const JobAction &JA,
                                           const llvm::opt::ArgList &Args) const {
   // Construct lld command.
   // The output from ld.lld is an HSA code object file.
-  ArgStringList LldArgs{"-flavor",    "gnu", "--no-undefined",
-                        "-shared",    "-o",  Output.getFilename()};
+  ArgStringList LldArgs{"-flavor",
+                        "gnu",
+                        "--no-undefined",
+                        "-shared",
+                        "-mllvm",
+                        "-amdgpu-internalize-symbols",
+                        "-o",
+                        Output.getFilename()};
   for (auto Input : Inputs)
     LldArgs.push_back(Input.getFilename());
   const char *Lld = Args.MakeArgString(getToolChain().GetProgramPath("lld"));
@@ -143,6 +149,8 @@ void HIPToolChain::addClangTargetOptions(
   if (DriverArgs.hasFlag(options::OPT_fgpu_rdc, options::OPT_fno_gpu_rdc,
                          false))
     CC1Args.push_back("-fgpu-rdc");
+  else
+    CC1Args.append({"-mllvm", "-amdgpu-internalize-symbols"});
 
   StringRef MaxThreadsPerBlock =
       DriverArgs.getLastArgValue(options::OPT_gpu_max_threads_per_block_EQ);

clang/test/Driver/hip-link-save-temps.hip

@@ -23,8 +23,10 @@
 // CHECK-NOT: llvm-link
 // CHECK-NOT: opt
 // CHECK-NOT: llc
-// CHECK: "{{.*lld.*}}" {{.*}} "-o" "a.out-hip-amdgcn-amd-amdhsa-gfx900" "obj1-hip-amdgcn-amd-amdhsa-gfx900.o" "obj2-hip-amdgcn-amd-amdhsa-gfx900.o"
-// CHECK: "{{.*lld.*}}" {{.*}} "-o" "a.out-hip-amdgcn-amd-amdhsa-gfx906" "obj1-hip-amdgcn-amd-amdhsa-gfx906.o" "obj2-hip-amdgcn-amd-amdhsa-gfx906.o"
+// CHECK: "{{.*lld.*}}" {{.*}} "-mllvm" "-amdgpu-internalize-symbols"
+// CHECK-SAME: "-o" "a.out-hip-amdgcn-amd-amdhsa-gfx900" "obj1-hip-amdgcn-amd-amdhsa-gfx900.o" "obj2-hip-amdgcn-amd-amdhsa-gfx900.o"
+// CHECK: "{{.*lld.*}}" {{.*}} "-mllvm" "-amdgpu-internalize-symbols"
+// CHECK-SAME: "-o" "a.out-hip-amdgcn-amd-amdhsa-gfx906" "obj1-hip-amdgcn-amd-amdhsa-gfx906.o" "obj2-hip-amdgcn-amd-amdhsa-gfx906.o"
 // OUT: "{{.*clang-offload-bundler.*}}" {{.*}} "-outputs=executable.hipfb"
 // OUT: "{{.*ld.*}}" {{.*}} "-o" "executable" {{.*}} "-T" "executable.lk"
 // NOUT: "{{.*clang-offload-bundler.*}}" {{.*}} "-outputs=a.out.hipfb"

clang/test/Driver/hip-save-temps.hip

@@ -35,14 +35,22 @@
 // CHECK: {{".*clang.*"}} "-cc1" {{.*}} "-E" {{.*}} [[CPU:"-target-cpu" "gfx900"]] {{.*}}  "-o" "hip-save-temps-hip-amdgcn-amd-amdhsa-gfx900.cui"
 // NORDC: {{".*clang.*"}} "-cc1" {{.*}} "-emit-llvm-bc" {{.*}} [[CPU]] {{.*}} "-disable-llvm-passes" {{.*}} "-o" "hip-save-temps-hip-amdgcn-amd-amdhsa-gfx900.bc"
 // RDC: {{".*clang.*"}} "-cc1" {{.*}} "-emit-llvm-bc" {{.*}} [[CPU]] {{.*}} "-disable-llvm-passes" {{.*}} "-o" "hip-save-temps-hip-amdgcn-amd-amdhsa-gfx900.tmp.bc"
-// NORDC: {{".*clang.*"}} "-cc1" {{.*}} "-S" {{.*}} [[CPU]] {{.*}} "-o" "hip-save-temps-hip-amdgcn-amd-amdhsa-gfx900.s"
+
+// NORDC: {{".*clang.*"}} "-cc1" {{.*}} "-S"
+// NORDC-SAME: "-mllvm" "-amdgpu-internalize-symbols"
+// NORDC-SAME: [[CPU]]
+// NORDC-SAME: "-o" "hip-save-temps-hip-amdgcn-amd-amdhsa-gfx900.s"
+
 // RDC: {{".*clang.*"}} "-cc1" {{.*}} "-emit-llvm-bc" {{.*}} [[CPU]] {{.*}} "-o" "hip-save-temps-hip-amdgcn-amd-amdhsa-gfx900.bc"
 // NORDC: {{".*clang.*"}} "-cc1as" {{.*}} "-filetype" "obj" {{.*}} [[CPU]] {{.*}} "-o" "hip-save-temps-hip-amdgcn-amd-amdhsa-gfx900.o"
 // CHECK-NOT: llvm-link
 // CHECK-NOT: opt
 // CHECK-NOT: llc
 // NORDC: {{.*lld.*}}"-o" "hip-save-temps-hip-amdgcn-amd-amdhsa-gfx900.out"
-// RDCL: "{{.*lld.*}}" {{.*}} "-o" "a.out-hip-amdgcn-amd-amdhsa-gfx900"
+
+// RDCL: "{{.*lld.*}}" {{.*}} "-mllvm" "-amdgpu-internalize-symbols"
+// RDCL-SAME: "-o" "a.out-hip-amdgcn-amd-amdhsa-gfx900"
+
 // NORDC: "{{.*clang-offload-bundler.*}}" {{.*}} "-outputs=hip-save-temps.hip-hip-amdgcn-amd-amdhsa.hipfb"
 // CHECK: "{{.*clang.*}}" "-cc1" {{.*}} "-E" {{.*}} "-o" "hip-save-temps-host-x86_64-unknown-linux-gnu.cui"
 // NORDC: "{{.*clang.*}}" "-cc1" {{.*}} "-emit-llvm-bc" {{.*}}  "-fcuda-include-gpubinary" {{.*}} "-o" "hip-save-temps-host-x86_64-unknown-linux-gnu.bc"

clang/test/Driver/hip-toolchain-no-rdc.hip

@@ -38,7 +38,8 @@
 // CHECK-SAME: "-aux-triple" "x86_64-unknown-linux-gnu"
 // CHECK-SAME: "-emit-obj"
 // CHECK-SAME: {{.*}} "-main-file-name" "a.cu"
-// CHECK-SAME: "-fcuda-is-device" "-fcuda-allow-variadic-functions" "-fvisibility" "hidden"
+// CHECK-SAME: "-fcuda-is-device" "-mllvm" "-amdgpu-internalize-symbols"
+// CHECK-SAME: "-fcuda-allow-variadic-functions" "-fvisibility" "hidden"
 // CHECK-SAME: "-fapply-global-visibility-to-externs"
 // CHECK-SAME: "{{.*}}lib1.bc" "{{.*}}lib2.bc"
 // CHECK-SAME: "-target-cpu" "gfx803"
@@ -60,7 +61,8 @@
 // CHECK-SAME: "-aux-triple" "x86_64-unknown-linux-gnu"
 // CHECK-SAME: "-emit-obj"
 // CHECK-SAME: {{.*}} "-main-file-name" "a.cu"
-// CHECK-SAME: "-fcuda-is-device" "-fcuda-allow-variadic-functions" "-fvisibility" "hidden"
+// CHECK-SAME: "-fcuda-is-device" "-mllvm" "-amdgpu-internalize-symbols"
+// CHECK-SAME: "-fcuda-allow-variadic-functions" "-fvisibility" "hidden"
 // CHECK-SAME: "-fapply-global-visibility-to-externs"
 // CHECK-SAME: "{{.*}}lib1.bc" "{{.*}}lib2.bc"
 // CHECK-SAME: "-target-cpu" "gfx900"
@@ -98,7 +100,8 @@
 // CHECK-SAME: "-aux-triple" "x86_64-unknown-linux-gnu"
 // CHECK-SAME: "-emit-obj"
 // CHECK-SAME: {{.*}} "-main-file-name" "b.hip"
-// CHECK-SAME: "-fcuda-is-device" "-fcuda-allow-variadic-functions" "-fvisibility" "hidden"
+// CHECK-SAME: "-fcuda-is-device" "-mllvm" "-amdgpu-internalize-symbols"
+// CHECK-SAME: "-fcuda-allow-variadic-functions" "-fvisibility" "hidden"
 // CHECK-SAME: "-fapply-global-visibility-to-externs"
 // CHECK-SAME: "{{.*}}lib1.bc" "{{.*}}lib2.bc"
 // CHECK-SAME: "-target-cpu" "gfx803"
@@ -120,7 +123,8 @@
 // CHECK-SAME: "-aux-triple" "x86_64-unknown-linux-gnu"
 // CHECK-SAME: "-emit-obj"
 // CHECK-SAME: {{.*}} "-main-file-name" "b.hip"
-// CHECK-SAME: "-fcuda-is-device" "-fcuda-allow-variadic-functions" "-fvisibility" "hidden"
+// CHECK-SAME: "-fcuda-is-device" "-mllvm" "-amdgpu-internalize-symbols"
+// CHECK-SAME: "-fcuda-allow-variadic-functions" "-fvisibility" "hidden"
 // CHECK-SAME: "-fapply-global-visibility-to-externs"
 // CHECK-SAME: "{{.*}}lib1.bc" "{{.*}}lib2.bc"
 // CHECK-SAME: "-target-cpu" "gfx900"

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

@@ -97,12 +97,14 @@
 // LINK-NOT: "*.llvm-link"
 // LINK-NOT: ".*opt"
 // LINK-NOT: ".*llc"
-// LINK: {{".*lld.*"}} {{.*}} "-o" "[[IMG_DEV1:.*.out]]" "[[A_BC1]]" "[[B_BC1]]"
+// LINK: {{".*lld.*"}} {{.*}} "-mllvm" "-amdgpu-internalize-symbols"
+// LINK-SAME: "-o" "[[IMG_DEV1:.*.out]]" "[[A_BC1]]" "[[B_BC1]]"
 
 // LINK-NOT: "*.llvm-link"
 // LINK-NOT: ".*opt"
 // LINK-NOT: ".*llc"
-// LINK: {{".*lld.*"}} {{.*}} "-o" "[[IMG_DEV2:.*.out]]" "[[A_BC2]]" "[[B_BC2]]"
+// LINK: {{".*lld.*"}} {{.*}} "-mllvm" "-amdgpu-internalize-symbols"
+// LINK-SAME: "-o" "[[IMG_DEV2:.*.out]]" "[[A_BC2]]" "[[B_BC2]]"
 
 // LINK: [[BUNDLER:".*clang-offload-bundler"]] "-type=o"
 // LINK-SAME: "-targets={{.*}},hip-amdgcn-amd-amdhsa-gfx803,hip-amdgcn-amd-amdhsa-gfx900"

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

@@ -37,7 +37,8 @@
 // CHECK-NOT: "*.llvm-link"
 // CHECK-NOT: ".*opt"
 // CHECK-NOT: ".*llc"
-// CHECK: {{".*lld.*"}} {{.*}} "-o" "[[IMG_DEV1:.*.out]]" [[A_BC1]] [[B_BC1]]
+// CHECK: {{".*lld.*"}} {{.*}} "-mllvm" "-amdgpu-internalize-symbols"
+// CHECK-SAME: "-o" "[[IMG_DEV1:.*.out]]" [[A_BC1]] [[B_BC1]]
 
 // CHECK: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa"
 // CHECK-SAME: "-aux-triple" "x86_64-unknown-linux-gnu"
@@ -62,7 +63,8 @@
 // CHECK-NOT: "*.llvm-link"
 // CHECK-NOT: ".*opt"
 // CHECK-NOT: ".*llc"
-// CHECK: {{".*lld.*"}} {{.*}} "-o" "[[IMG_DEV2:.*.out]]" [[A_BC2]] [[B_BC2]]
+// CHECK: {{".*lld.*"}} {{.*}} "-mllvm" "-amdgpu-internalize-symbols"
+// CHECK-SAME: "-o" "[[IMG_DEV2:.*.out]]" [[A_BC2]] [[B_BC2]]
 
 // CHECK: [[CLANG]] "-cc1" "-triple" "x86_64-unknown-linux-gnu"
 // CHECK-SAME: "-aux-triple" "amdgcn-amd-amdhsa"