1 // Tests the driver when targeting the NVPTX architecture directly without a 2 // host toolchain to perform CUDA mappings. 3 4 // 5 // Test the generated phases when targeting NVPTX. 6 // 7 // RUN: %clang -target nvptx64-nvidia-cuda -ccc-print-phases %s 2>&1 \ 8 // RUN: | FileCheck -check-prefix=PHASES %s 9 10 // PHASES: 0: input, "[[INPUT:.+]]", c 11 // PHASES-NEXT: 1: preprocessor, {0}, cpp-output 12 // PHASES-NEXT: 2: compiler, {1}, ir 13 // PHASES-NEXT: 3: backend, {2}, assembler 14 // PHASES-NEXT: 4: assembler, {3}, object 15 // PHASES-NEXT: 5: linker, {4}, image 16 17 // 18 // Test the generated bindings when targeting NVPTX. 19 // 20 // RUN: %clang -target nvptx64-nvidia-cuda -ccc-print-bindings %s 2>&1 \ 21 // RUN: | FileCheck -check-prefix=BINDINGS %s 22 23 // BINDINGS: "nvptx64-nvidia-cuda" - "clang", inputs: ["[[INPUT:.+]]"], output: "[[PTX:.+]].s" 24 // BINDINGS-NEXT: "nvptx64-nvidia-cuda" - "NVPTX::Assembler", inputs: ["[[PTX]].s"], output: "[[CUBIN:.+]].o" 25 // BINDINGS-NEXT: "nvptx64-nvidia-cuda" - "NVPTX::Linker", inputs: ["[[CUBIN]].o"], output: "a.out" 26 27 // 28 // Test the generated arguments to the CUDA binary utils when targeting NVPTX. 29 // Ensure that the '.o' files are converted to '.cubin' if produced internally. 30 // 31 // RUN: %clang -target nvptx64-nvidia-cuda -march=sm_61 -### %s 2>&1 \ 32 // RUN: | FileCheck -check-prefix=ARGS %s 33 34 // ARGS: -cc1" "-triple" "nvptx64-nvidia-cuda" "-S" {{.*}} "-target-cpu" "sm_61" "-target-feature" "+ptx{{[0-9]+}}" {{.*}} "-o" "[[PTX:.+]].s" 35 // ARGS-NEXT: ptxas{{.*}}"-m64" "-O0" "--gpu-name" "sm_61" "--output-file" "[[CUBIN:.+]].o" "[[PTX]].s" "-c" 36 // ARGS-NEXT: clang-nvlink-wrapper{{.*}}"-o" "a.out" "-arch" "sm_61"{{.*}}"[[CUBIN]].o" 37 38 // 39 // Test the generated arguments to the CUDA binary utils when targeting NVPTX. 40 // Ensure that we emit '.o' files if compiled with '-c' 41 // 42 // RUN: %clang -target nvptx64-nvidia-cuda -march=sm_61 -c -### %s 2>&1 \ 43 // RUN: | FileCheck -check-prefix=OBJECT %s 44 // RUN: %clang -target nvptx64-nvidia-cuda -save-temps -march=sm_61 -c -### %s 2>&1 \ 45 // RUN: | FileCheck -check-prefix=OBJECT %s 46 47 // OBJECT: -cc1" "-triple" "nvptx64-nvidia-cuda" "-S" {{.*}} "-target-cpu" "sm_61" "-target-feature" "+ptx{{[0-9]+}}" {{.*}} "-o" "[[PTX:.+]].s" 48 // OBJECT-NEXT: ptxas{{.*}}"-m64" "-O0" "--gpu-name" "sm_61" "--output-file" "[[OBJ:.+]].o" "[[PTX]].s" "-c" 49 50 // 51 // Test the generated arguments to the CUDA binary utils when targeting NVPTX. 52 // Ensure that we copy input '.o' files to '.cubin' files when linking. 53 // 54 // RUN: touch %t.o 55 // RUN: %clang -target nvptx64-nvidia-cuda -march=sm_61 -### %t.o 2>&1 \ 56 // RUN: | FileCheck -check-prefix=LINK %s 57 58 // LINK: clang-nvlink-wrapper{{.*}}"-o" "a.out" "-arch" "sm_61"{{.*}}[[CUBIN:.+]].o 59 60 // 61 // Test to ensure that we enable handling global constructors in a freestanding 62 // Nvidia compilation. 63 // 64 // RUN: %clang -target nvptx64-nvidia-cuda -march=sm_70 %s -### 2>&1 \ 65 // RUN: | FileCheck -check-prefix=LOWERING %s 66 // RUN: %clang -target nvptx64-nvidia-cuda -march=sm_70 -flto -c %s -### 2>&1 \ 67 // RUN: | FileCheck -check-prefix=LOWERING-LTO %s 68 69 // LOWERING: -cc1" "-triple" "nvptx64-nvidia-cuda" {{.*}} "-mllvm" "--nvptx-lower-global-ctor-dtor" 70 // LOWERING: clang-nvlink-wrapper{{.*}} "-mllvm" "--nvptx-lower-global-ctor-dtor" 71 // LOWERING-LTO-NOT: "--nvptx-lower-global-ctor-dtor" 72 73 // 74 // Test passing arguments directly to nvlink. 75 // 76 // RUN: %clang -target nvptx64-nvidia-cuda -Wl,-v -Wl,a,b -march=sm_52 -### %s 2>&1 \ 77 // RUN: | FileCheck -check-prefix=LINKER-ARGS %s 78 79 // LINKER-ARGS: clang-nvlink-wrapper{{.*}}"-v"{{.*}}"a" "b" 80 81 // Tests for handling a missing architecture. 82 // 83 // RUN: not %clang -target nvptx64-nvidia-cuda %s -### 2>&1 \ 84 // RUN: | FileCheck -check-prefix=MISSING %s 85 // RUN: not %clang -target nvptx64-nvidia-cuda -march=generic %s -### 2>&1 \ 86 // RUN: | FileCheck -check-prefix=MISSING %s 87 88 // MISSING: error: must pass in an explicit nvptx64 gpu architecture to 'ptxas' 89 // MISSING: error: must pass in an explicit nvptx64 gpu architecture to 'nvlink' 90 91 // Do not error when performing LTO. 92 // 93 // RUN: %clang -target nvptx64-nvidia-cuda -flto %s -### 2>&1 \ 94 // RUN: | FileCheck -check-prefix=MISSING-LTO %s 95 96 // MISSING-LTO-NOT: error: must pass in an explicit nvptx64 gpu architecture to 'nvlink' 97 98 // RUN: %clang -target nvptx64-nvidia-cuda -flto -c %s -### 2>&1 \ 99 // RUN: | FileCheck -check-prefix=GENERIC %s 100 // RUN: %clang -target nvptx64-nvidia-cuda -march=sm_52 -march=generic -flto -c %s -### 2>&1 \ 101 // RUN: | FileCheck -check-prefix=GENERIC %s 102 103 // GENERIC-NOT: -cc1" "-triple" "nvptx64-nvidia-cuda" {{.*}} "-target-cpu" 104 105 // 106 // Test forwarding the necessary +ptx feature. 107 // 108 // RUN: %clang -target nvptx64-nvidia-cuda --cuda-feature=+ptx63 -march=sm_52 -### %s 2>&1 \ 109 // RUN: | FileCheck -check-prefix=FEATURE %s 110 111 // FEATURE: clang-nvlink-wrapper{{.*}}"--plugin-opt=-mattr=+ptx63" 112 113 // 114 // Test including the libc startup files and libc 115 // 116 // RUN: %clang -target nvptx64-nvidia-cuda -march=sm_61 -stdlib -startfiles \ 117 // RUN: -nogpulib -nogpuinc -### %s 2>&1 | FileCheck -check-prefix=STARTUP %s 118 119 // STARTUP: clang-nvlink-wrapper{{.*}}"-lc" "-lm" "{{.*}}crt1.o" 120