186 lines · plain
1// RUN: %clang -### --target=x86_64-linux-gnu -v \2// RUN: -x hip --cuda-gpu-arch=gfx803 --cuda-gpu-arch=gfx900 \3// RUN: --hip-device-lib=lib1.bc --hip-device-lib=lib2.bc \4// RUN: --hip-device-lib-path=%S/Inputs/hip_multiple_inputs/lib1 \5// RUN: --hip-device-lib-path=%S/Inputs/hip_multiple_inputs/lib2 \6// RUN: -fuse-ld=lld -B%S/Inputs/lld -fgpu-rdc -nogpuinc \7// RUN: --no-offload-new-driver -fhip-dump-offload-linker-script \8// RUN: %S/Inputs/hip_multiple_inputs/a.cu \9// RUN: %S/Inputs/hip_multiple_inputs/b.hip \10// RUN: 2>&1 | FileCheck -check-prefixes=CHECK,LNX %s11 12// RUN: %clang -### --target=x86_64-pc-windows-msvc -v \13// RUN: -x hip --cuda-gpu-arch=gfx803 --cuda-gpu-arch=gfx900 \14// RUN: --hip-device-lib=lib1.bc --hip-device-lib=lib2.bc \15// RUN: --hip-device-lib-path=%S/Inputs/hip_multiple_inputs/lib1 \16// RUN: --hip-device-lib-path=%S/Inputs/hip_multiple_inputs/lib2 \17// RUN: -fuse-ld=lld -B%S/Inputs/lld -fgpu-rdc -nogpuinc \18// RUN: --no-offload-new-driver -fhip-dump-offload-linker-script \19// RUN: %S/Inputs/hip_multiple_inputs/a.cu \20// RUN: %S/Inputs/hip_multiple_inputs/b.hip \21// RUN: 2>&1 | FileCheck -check-prefixes=CHECK,MSVC %s22 23// Test fatbin symbol search for -l libraries.24 25// RUN: touch librdctest.a rdctest2.bin26// RUN: %clang -### --target=x86_64-linux-gnu -v \27// RUN: -fgpu-rdc -nogpuinc -nogpulib \28// RUN: --no-offload-new-driver -L. -lrdctest -l:rdctest2.bin \29// RUN: %s \30// RUN: 2>&1 | FileCheck -check-prefixes=LIB,LNX-LIB %s31// RUN: rm librdctest.a rdctest2.bin32 33// RUN: touch rdctest.lib rdctest2.bin34// RUN: %clang -### --target=x86_64-pc-windows-msvc -v \35// RUN: -fgpu-rdc -nogpuinc -nogpulib \36// RUN: --no-offload-new-driver -L. -lrdctest -l:rdctest2.bin \37// RUN: %s \38// RUN: 2>&1 | FileCheck -check-prefixes=LIB,MSVC-LIB %s39// RUN: rm rdctest.lib rdctest2.bin40 41// LIB: HIP fatbin symbol search uses library path: .42// LIB: HIP fatbin symbol search found library: .{{/|\\}}rdctest2.bin43// LNX-LIB: HIP fatbin symbol search found library: .{{/|\\}}librdctest.a44// MSVC-LIB: HIP fatbin symbol search found library: .{{/|\\}}rdctest.lib45 46// check HIP fatbin and gpubin handle symbols and code object alignment in dumped llvm-mc input47// CHECK: Found undefined HIP fatbin symbol: __hip_fatbin_[[ID1:[0-9a-f]+]]48// CHECK: Found undefined HIP fatbin symbol: __hip_fatbin_[[ID2:[0-9a-f]+]]49// CHECK: Found undefined HIP gpubin handle symbol: __hip_gpubin_handle_[[ID1]]50// CHECK: Found undefined HIP gpubin handle symbol: __hip_gpubin_handle_[[ID2]]51// LNX: .protected __hip_gpubin_handle_[[ID1]]52// LNX: .type __hip_gpubin_handle_[[ID1]]53// LNX-LABEL: .section .hip_gpubin_handle,"aw"54// MSVC-LABEL: .section .hip_gpubin_handle,"dw"55// CHECK: .globl __hip_gpubin_handle_[[ID1]]56// CHECK-NEXT: .p2align 357// CHECK-NEXT:__hip_gpubin_handle_[[ID1]]:58// CHECK-NEXT: .zero 859// CHECK-NEXT: .globl __hip_gpubin_handle_[[ID2]]60// CHECK-NEXT: .set __hip_gpubin_handle_[[ID2]],__hip_gpubin_handle_[[ID1]]61// LNX: .protected __hip_fatbin_[[ID1]]62// LNX: .type __hip_fatbin_[[ID1]],@object63// LNX-LABEL: .section .hip_fatbin,"a",@progbits64// MSVC-LABEL: .section .hip_fatbin,"dw"65// CHECK: .globl __hip_fatbin_[[ID1]]66// CHECK-NEXT: .p2align 1267// CHECK-NEXT: .globl __hip_fatbin_[[ID2]]68// CHECK-NEXT: .set __hip_fatbin_[[ID2]],__hip_fatbin_[[ID1]]69// CHECK-NEXT: __hip_fatbin_[[ID1]]:70// CHECK-NEXT: .incbin "[[BUNDLE:.*hipfb]]"71// LNX: .section .note.GNU-stack, "", @progbits72// MSVC-NOT: .note.GNU-stack73 74// emit objects for host side path75// CHECK: [[CLANG:".*clang.*"]] "-cc1" "-triple" [[HOST:"x86_64-[^"]+"]]76// CHECK-SAME: "-aux-triple" "amdgcn-amd-amdhsa"77// CHECK-SAME: "-emit-obj"78// CHECK-SAME: {{.*}} "-main-file-name" "a.cu"79// CHECK-SAME: "-fgpu-rdc"80// CHECK-SAME: {{.*}} "-o" [[A_OBJ_HOST:".*o"]] "-x" "hip"81// CHECK-SAME: {{.*}} [[A_SRC:".*a.cu"]]82 83// CHECK: [[CLANG]] "-cc1" "-triple" [[HOST]]84// CHECK-SAME: "-aux-triple" "amdgcn-amd-amdhsa"85// CHECK-SAME: "-emit-obj"86// CHECK-SAME: {{.*}} "-main-file-name" "b.hip"87// CHECK-SAME: "-fgpu-rdc"88// CHECK-SAME: {{.*}} "-o" [[B_OBJ_HOST:".*o"]] "-x" "hip"89// CHECK-SAME: {{.*}} [[B_SRC:".*b.hip"]]90 91// generate image for device side path on gfx80392// CHECK: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa"93// CHECK-SAME: "-aux-triple" [[HOST:"x86_64-[^"]+"]]94// CHECK-SAME: "-emit-llvm-bc"95// CHECK-SAME: {{.*}} "-main-file-name" "a.cu"96// CHECK-SAME: "-fcuda-is-device" "-fno-threadsafe-statics" "-fcuda-allow-variadic-functions" "-fvisibility=hidden"97// CHECK-SAME: "-fapply-global-visibility-to-externs"98// CHECK-SAME: "{{.*}}lib1.bc" "{{.*}}lib2.bc"99// CHECK-SAME: "-target-cpu" "gfx803"100// CHECK-SAME: "-fgpu-rdc"101// CHECK-SAME: {{.*}} "-o" [[A_BC1:".*bc"]] "-x" "hip"102// CHECK-SAME: {{.*}} [[A_SRC]]103 104// CHECK: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa"105// CHECK-SAME: "-aux-triple" [[HOST]]106// CHECK-SAME: "-emit-llvm-bc"107// CHECK-SAME: {{.*}} "-main-file-name" "b.hip"108// CHECK-SAME: "-fcuda-is-device" "-fno-threadsafe-statics" "-fcuda-allow-variadic-functions" "-fvisibility=hidden"109// CHECK-SAME: "-fapply-global-visibility-to-externs"110// CHECK-SAME: "{{.*}}lib1.bc" "{{.*}}lib2.bc"111// CHECK-SAME: "-target-cpu" "gfx803"112// CHECK-SAME: "-fgpu-rdc"113// CHECK-SAME: {{.*}} "-o" [[B_BC1:".*bc"]] "-x" "hip"114// CHECK-SAME: {{.*}} [[B_SRC]]115 116// CHECK-NOT: "*.llvm-link"117// CHECK-NOT: ".*opt"118// CHECK-NOT: ".*llc"119// CHECK: {{".*lld.*"}} {{.*}} "-plugin-opt=-amdgpu-internalize-symbols"120// CHECK-SAME: "-plugin-opt=mcpu=gfx803"121// CHECK-SAME: "-o" "[[IMG_DEV1:.*.out]]" [[A_BC1]] [[B_BC1]]122 123// generate image for device side path on gfx900124// CHECK: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa"125// CHECK-SAME: "-aux-triple" [[HOST]]126// CHECK-SAME: "-emit-llvm-bc"127// CHECK-SAME: {{.*}} "-main-file-name" "a.cu"128// CHECK-SAME: "-fcuda-is-device"129// CHECK-SAME: "{{.*}}lib1.bc" "{{.*}}lib2.bc"130// CHECK-SAME: "-target-cpu" "gfx900"131// CHECK-SAME: {{.*}} "-o" [[A_BC2:".*bc"]] "-x" "hip"132// CHECK-SAME: {{.*}} [[A_SRC]]133 134// CHECK: [[CLANG]] "-cc1" "-triple" "amdgcn-amd-amdhsa"135// CHECK-SAME: "-aux-triple" [[HOST]]136// CHECK-SAME: "-emit-llvm-bc"137// CHECK-SAME: {{.*}} "-main-file-name" "b.hip"138// CHECK-SAME: "-fcuda-is-device"139// CHECK-SAME: "{{.*}}lib1.bc" "{{.*}}lib2.bc"140// CHECK-SAME: "-target-cpu" "gfx900"141// CHECK-SAME: "-fgpu-rdc"142// CHECK-SAME: {{.*}} "-o" [[B_BC2:".*bc"]] "-x" "hip"143// CHECK-SAME: {{.*}} [[B_SRC]]144 145// CHECK-NOT: "*.llvm-link"146// CHECK-NOT: ".*opt"147// CHECK-NOT: ".*llc"148// CHECK: {{".*lld.*"}} {{.*}} "-plugin-opt=-amdgpu-internalize-symbols"149// CHECK-SAME: "--lto-partitions={{[0-9]+}}"150// CHECK-SAME: "-plugin-opt=mcpu=gfx900"151// CHECK-SAME: "-o" "[[IMG_DEV2:.*.out]]" [[A_BC2]] [[B_BC2]]152 153// combine images generated into hip fat binary object154// CHECK: [[BUNDLER:".*clang-offload-bundler"]] "-type=o"155// CHECK-SAME: "-bundle-align=4096"156// CHECK-SAME: "-targets={{.*}},hipv4-amdgcn-amd-amdhsa--gfx803,hipv4-amdgcn-amd-amdhsa--gfx900"157// CHECK-SAME: "-input={{.*}}" "-input=[[IMG_DEV1]]" "-input=[[IMG_DEV2]]" "-output=[[BUNDLE]]"158 159// CHECK: [[MC:".*clang.*"]] "-target" [[HOST]] "-o" [[OBJBUNDLE:".*o"]] "{{.*}}.mcin"160 161// output the executable162// LNX: [[LD:".*ld.*"]] {{.*}}"-o" "a.out" {{.*}} [[A_OBJ_HOST]] [[B_OBJ_HOST]] [[OBJBUNDLE]]163// MSVC: [[LD:".*lld-link.*"]] {{.*}}"-out:a.exe" {{.*}} [[A_OBJ_HOST]] [[B_OBJ_HOST]] [[OBJBUNDLE]]164 165// Check -flto-partitions166 167// RUN: %clang -### -fgpu-rdc --offload-arch=gfx90a -nogpulib -nogpuinc --no-offload-new-driver \168// RUN: -L. -foffload-lto %s 2>&1 | FileCheck -check-prefix=LTO_DEFAULT %s169// LTO_DEFAULT: lld{{.*}}"--lto-partitions=8"170 171// RUN: %clang -### -fgpu-rdc --offload-arch=gfx90a -nogpulib -nogpuinc --offload-new-driver \172// RUN: -L. -foffload-lto %s 2>&1 | FileCheck -check-prefix=LTO_DEFAULT_NEW %s173// LTO_DEFAULT_NEW: clang-linker-wrapper{{.*}}"--device-compiler=amdgcn-amd-amdhsa=-flto-partitions=8"174 175// RUN: %clang -### -fgpu-rdc --offload-arch=gfx90a -nogpulib -nogpuinc --no-offload-new-driver \176// RUN: -L. -foffload-lto -flto-partitions=42 %s 2>&1 | FileCheck -check-prefix=LTO_PARTS %s177// LTO_PARTS: lld{{.*}}"--lto-partitions=42"178 179// RUN: not %clang -### -fgpu-rdc --offload-arch=gfx90a -nogpulib -nogpuinc --no-offload-new-driver \180// RUN: -L. -foffload-lto -flto-partitions=a %s 2>&1 | FileCheck -check-prefix=LTO_PARTS_INV0 %s181// LTO_PARTS_INV0: clang: error: invalid integral value 'a' in '-flto-partitions=a'182 183// RUN: not %clang -### -fgpu-rdc --offload-arch=gfx90a -nogpulib -nogpuinc --no-offload-new-driver \184// RUN: -L. -foffload-lto -flto-partitions=0 %s 2>&1 | FileCheck -check-prefix=LTO_PARTS_INV1 %s185// LTO_PARTS_INV1: clang: error: invalid integral value '0' in '-flto-partitions=0'186