263 lines · plain
1// RUN: %clang -### -x hip -nogpulib -nogpuinc --gpu-max-threads-per-block=1024 %s 2>&1 | FileCheck %s2 3// Check that there are commands for both host- and device-side compilations.4//5// CHECK: "-cc1" {{.*}} "-fcuda-is-device"6// CHECK-SAME: "--gpu-max-threads-per-block=1024"7 8// RUN: %clang -### -nogpuinc -nogpulib -fgpu-allow-device-init \9// RUN: %s 2>&1 | FileCheck -check-prefix=DEVINIT %s10// DEVINIT: "-cc1" {{.*}}"-fgpu-allow-device-init"11// DEVINIT: "-cc1" {{.*}}"-fgpu-allow-device-init"12 13// Check -fgpu-default-stream=per-thread.14// RUN: %clang -### -nogpuinc -nogpulib -fgpu-default-stream=per-thread \15// RUN: %s -save-temps 2>&1 | FileCheck -check-prefix=PTH %s16// PTH: "-cc1"{{.*}} "-E" {{.*}}"-fgpu-default-stream=per-thread"17// PTH: "-cc1"{{.*}} "-fgpu-default-stream=per-thread" {{.*}}"-x" "hip-cpp-output"18// PTH: "-cc1"{{.*}} "-E" {{.*}}"-fgpu-default-stream=per-thread"19// PTH: "-cc1"{{.*}} "-fgpu-default-stream=per-thread" {{.*}}"-x" "hip-cpp-output"20 21// Check -mprintf-kind=hostcall22// RUN: %clang -### -nogpulib -nogpuinc --target=x86_64-unknown-linux-gnu -mprintf-kind=hostcall %s -save-temps 2>&1 | FileCheck -check-prefix=HOSTC %s23// HOSTC: "-cc1" "-triple" "amdgcn-amd-amdhsa"{{.*}}"-mprintf-kind=hostcall" "-Werror=format-invalid-specifier"{{.*}}"-E" {{.*}}24// HOSTC: "-cc1" "-triple" "amdgcn-amd-amdhsa"{{.*}} "-mprintf-kind=hostcall" "-Werror=format-invalid-specifier" {{.*}}"-x" "hip-cpp-output"25// HOSTC: "-cc1" "-triple" "amdgcn-amd-amdhsa"{{.*}} "-mprintf-kind=hostcall" "-Werror=format-invalid-specifier" {{.*}}"-x" "ir"26// HOSTC: "-cc1" "-triple" "x86_64-unknown-linux-gnu"{{.*}} "-E" {{.*}}27// HOSTC: "-cc1" "-triple" "x86_64-unknown-linux-gnu"{{.*}}"-x" "ir"28 29// Check -mprintf-kind=buffered30// RUN: %clang -### -nogpulib -nogpuinc --target=x86_64-unknown-linux-gnu -mprintf-kind=buffered %s -save-temps 2>&1 | FileCheck -check-prefix=BUFF %s31// BUFF: "-cc1" "-triple" "amdgcn-amd-amdhsa"{{.*}}"-mprintf-kind=buffered" "-Werror=format-invalid-specifier"{{.*}}"-E" {{.*}}32// BUFF: "-cc1" "-triple" "amdgcn-amd-amdhsa"{{.*}} "-mprintf-kind=buffered" "-Werror=format-invalid-specifier" {{.*}}"-x" "hip-cpp-output"33// BUFF: "-cc1" "-triple" "amdgcn-amd-amdhsa"{{.*}} "-mprintf-kind=buffered" "-Werror=format-invalid-specifier" {{.*}}"-x" "ir"34// BUFF: "-cc1" "-triple" "x86_64-unknown-linux-gnu"{{.*}} "-E" {{.*}}35// BUFF: "-cc1" "-triple" "x86_64-unknown-linux-gnu"{{.*}}"-x" "ir"36 37// RUN: %clang -### -x hip -nogpulib -nogpuinc --target=x86_64-pc-windows-msvc -fms-extensions \38// RUN: -mllvm -amdgpu-early-inline-all=true %s 2>&1 | \39// RUN: FileCheck -check-prefix=MLLVM %s40// MLLVM-NOT: "-mllvm"{{.*}}"-amdgpu-early-inline-all=true"{{.*}}"-mllvm"{{.*}}"-amdgpu-early-inline-all=true"41 42// RUN: %clang -### -Xarch_device -g -nogpulib -nogpuinc --cuda-gpu-arch=gfx900 \43// RUN: -Xarch_device -fcf-protection=branch -Xarch_device -mllvm=--inline-threshold=100 \44// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=DEV %s45// DEV: "-cc1"{{.*}} "-fcuda-is-device" {{.*}} "-debug-info-kind={{.*}}" {{.*}} "-fcf-protection=branch" {{.*}}"-mllvm" "--inline-threshold=100"46// DEV: "-cc1"{{.*}} "-fcuda-is-device" {{.*}} "-debug-info-kind={{.*}}" {{.*}} "-fcf-protection=branch" {{.*}}"-mllvm" "--inline-threshold=100"47// DEV-NOT: clang{{.*}} {{.*}} "-debug-info-kind={{.*}}"48 49// RUN: %clang -### -Xarch_host -g -nogpulib -nogpuinc --cuda-gpu-arch=gfx900 \50// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=HOST %s51// HOST-NOT: "-fcuda-is-device" {{.*}} "-debug-info-kind={{.*}}"52// HOST-NOT: "-fcuda-is-device" {{.*}} "-debug-info-kind={{.*}}"53// HOST: clang{{.*}} "-debug-info-kind={{.*}}"54 55// RUN: %clang -### -nogpuinc -nogpulib -munsafe-fp-atomics \56// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=UNSAFE-FP-ATOMICS %s57// UNSAFE-FP-ATOMICS: clang{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-fatomic-ignore-denormal-mode"58 59// RUN: %clang -### -nogpuinc -nogpulib \60// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=DEFAULT-UNSAFE-FP-ATOMICS %s61// DEFAULT-UNSAFE-FP-ATOMICS-NOT: clang{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-fatomic-ignore-denormal-mode"62 63// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib -fgpu-exclude-wrong-side-overloads \64// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=FIX-OVERLOAD %s65// FIX-OVERLOAD: "-cc1"{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-fgpu-exclude-wrong-side-overloads" "-fgpu-defer-diag"66// FIX-OVERLOAD: "-cc1"{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-fgpu-exclude-wrong-side-overloads" "-fgpu-defer-diag"67 68// Check -mconstructor-aliases is not passed to device compilation.69 70// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \71// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=CTA %s72// CTA: "-cc1"{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-mconstructor-aliases"73// CTA-NOT: "-cc1"{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-mconstructor-aliases"74 75// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \76// RUN: --offload-arch=gfx906 -fgpu-inline-threshold=1000 %s 2>&1 | FileCheck -check-prefix=THRESH %s77// THRESH: "-cc1"{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-mllvm" "-inline-threshold=1000"78// THRESH-NOT: "-cc1"{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-inline-threshold=1000"79 80// Check -foffload-lto=thin translated correctly.81 82// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \83// RUN: --cuda-gpu-arch=gfx906 -foffload-lto=thin -fwhole-program-vtables %s 2>&1 \84// RUN: | FileCheck -check-prefix=HIPTHINLTO %s85 86// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \87// RUN: --cuda-gpu-arch=gfx906 -fgpu-rdc -foffload-lto=thin -fwhole-program-vtables %s 2>&1 \88// RUN: | FileCheck -check-prefix=HIPTHINLTO %s89 90// Ensure we don't error about -fwhole-program-vtables for the non-device offload compile.91// HIPTHINLTO-NOT: error: invalid argument '-fwhole-program-vtables' only allowed with '-flto'92// HIPTHINLTO-NOT: "-cc1"{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-flto-unit"93// HIPTHINLTO: "-cc1"{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-flto=thin" "-flto-unit" {{.*}} "-fwhole-program-vtables"94// HIPTHINLTO-NOT: "-cc1"{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-flto-unit"95// HIPTHINLTO: lld{{.*}}"-plugin-opt=mcpu=gfx906" "-plugin-opt=thinlto" "-plugin-opt=-force-import-all"96 97// Check that -flto=thin is handled correctly, particularly with -fwhole-program-vtables.98//99// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \100// RUN: --cuda-gpu-arch=gfx906 -flto=thin -fwhole-program-vtables %s 2>&1 \101// RUN: | FileCheck -check-prefix=THINLTO %s102 103// Ensure we don't error about -fwhole-program-vtables for the device offload compile. We should104// drop -fwhole-program-vtables for the device offload compile and pass it through for the105// non-device offload compile along with -flto=thin.106// THINLTO-NOT: error: invalid argument '-fwhole-program-vtables' only allowed with '-flto'107// THINLTO-NOT: "-cc1"{{.*}}" "-triple" "amdgcn-amd-amdhsa" {{.*}} "-fwhole-program-vtables"108// THINLTO: "-cc1"{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-flto=thin" {{.*}} "-fwhole-program-vtables"109// THINLTO-NOT: "-cc1"{{.*}}" "-triple" "amdgcn-amd-amdhsa" {{.*}} "-fwhole-program-vtables"110 111// Check -fopenmp=libomp is allowed with HIP but -fopenmp-targets= is not allowed.112 113// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \114// RUN: --offload-arch=gfx906 -fopenmp=libomp %s 2>&1 | FileCheck -check-prefix=OMP %s115// OMP-NOT: "-cc1"{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-fopenmp"116// OMP: "-cc1"{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-fopenmp"117 118// Check -Xoffload-linker option is passed to lld.119 120// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \121// RUN: --cuda-gpu-arch=gfx906 -fgpu-rdc -Xoffload-linker --build-id=md5 %s 2>&1 \122// RUN: | FileCheck -check-prefix=OFL-LINK %s123// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \124// RUN: --cuda-gpu-arch=gfx906 -Xoffload-linker --build-id=md5 %s 2>&1 \125// RUN: | FileCheck -check-prefix=OFL-LINK %s126// OFL-LINK: lld{{.*}}"--build-id=md5"127 128// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \129// RUN: --offload-arch=gfx906 -fhip-kernel-arg-name %s 2>&1 \130// RUN: | FileCheck -check-prefix=KAN %s131// KAN: "-cc1"{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-fhip-kernel-arg-name"132 133// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \134// RUN: --offload-arch=gfx906 %s 2>&1 \135// RUN: | FileCheck -check-prefix=KANNEG %s136// KANNEG-NOT: "-fhip-kernel-arg-name"137 138// RUN: %clang -### -nogpuinc -nogpulib -mno-amdgpu-ieee -mamdgpu-ieee \139// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=IEEE-ON %s140// IEEE-ON-NOT: clang{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-m{{(no-)?}}amdgpu-ieee"141 142// RUN: %clang -### -nogpuinc -nogpulib -mamdgpu-ieee -mno-amdgpu-ieee -ffast-math \143// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefixes=IEEE-OFF %s144// IEEE-OFF: clang{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-mno-amdgpu-ieee"145 146// RUN: %clang -### -nogpuinc -nogpulib -mamdgpu-ieee -mno-amdgpu-ieee -ffast-math \147// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefixes=IEEE-OFF-NEG %s148// IEEE-OFF-NEG-NOT: clang{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-mamdgpu-ieee"149 150// Check -fno-hip-fp32-correctly-rounded-divide-sqrt is passed to -cc1 but151// (default) -fhip-fp32-correctly-rounded-divide-sqrt is not.152 153// RUN: %clang -### -nogpuinc -nogpulib -fno-hip-fp32-correctly-rounded-divide-sqrt \154// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefixes=NOCRDS %s155// NOCRDS: clang{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-fno-hip-fp32-correctly-rounded-divide-sqrt"156 157// RUN: %clang -### -nogpuinc -nogpulib \158// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefixes=CRDS %s159// CRDS-NOT: "-f{{(no-)?}}hip-fp32-correctly-rounded-divide-sqrt"160 161// RUN: %clang -### -nogpuinc -nogpulib -fhip-fp32-correctly-rounded-divide-sqrt \162// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefixes=CRDS %s163// CRDS-NOT: "-f{{(no-)?}}hip-fp32-correctly-rounded-divide-sqrt"164 165// Check -fgpu-approx-transcendentals is passed to clang -cc1 but166// (default) -fno-gpu-approx-transcendentals is not.167// -ffast-math implies -fgpu-approx-transcendentals, which can be overridden168// by -fno-gpu-approx-transcendentals.169 170// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib -fgpu-approx-transcendentals \171// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=APPROX %s172 173// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib -ffast-math \174// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=APPROX %s175 176// APPROX: "-cc1"{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-fgpu-approx-transcendentals"177// APPROX: "-cc1"{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-fgpu-approx-transcendentals"178 179// RUN: %clang -### -nogpuinc -nogpulib -fno-gpu-approx-transcendentals \180// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=NOAPPROX %s181 182// RUN: %clang -### -nogpuinc -nogpulib \183// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=NOAPPROX %s184 185// RUN: %clang -### -nogpuinc -nogpulib -ffast-math -fno-fast-math \186// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=NOAPPROX %s187 188// RUN: %clang -### -nogpuinc -nogpulib -ffast-math -fno-gpu-approx-transcendentals \189// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=NOAPPROX %s190 191// NOAPPROX-NOT: "-f{{(no-)?}}gpu-approx-transcendentals"192 193// Check no warnings for -fgpu-approx-transcendentals.194 195// RUN: %clang -fdriver-only -Werror --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib -fgpu-approx-transcendentals \196// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | count 0197 198// RUN: %clang -fdriver-only -Werror --target=x86_64-unknown-linux-gnu -nostdinc -nostdlib -fgpu-approx-transcendentals \199// RUN: -x c++ %s 2>&1 | count 0200/ Check -fno-offload-uniform-block is passed to clang -cc1 but201// (default) -fno-offload-uniform-block is not.202 203// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib -fno-offload-uniform-block \204// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=NOUNIBLK %s205 206// NOUNIBLK: "-cc1"{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-fno-offload-uniform-block"207// NOUNIBLK: "-cc1"{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-fno-offload-uniform-block"208 209// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib -foffload-uniform-block \210// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=UNIBLK %s211 212// UNIBLK: "-cc1"{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-foffload-uniform-block"213// UNIBLK: "-cc1"{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-foffload-uniform-block"214 215// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \216// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=DEFUNIBLK %s217 218// DEFUNIBLK-NOT: "-f{{(no-)?}}offload-uniform-block"219 220// Check no warnings for -f[no-]offload-uniform-block.221 222// RUN: %clang -fdriver-only -c -Werror --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib -fno-offload-uniform-block \223// RUN: -foffload-uniform-block --cuda-gpu-arch=gfx906 %s 2>&1 | count 0224 225// Check -Werror=atomic-alignment is passed for amdpu by default.226 227// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \228// RUN: --cuda-gpu-arch=gfx906 %s 2>&1 | FileCheck -check-prefix=WARN-ATOMIC %s229// WARN-ATOMIC: clang{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-Werror=atomic-alignment"230// WARN-ATOMIC-NOT: clang{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-Werror=atomic-alignment"231 232// Check the default -Werror=atomic-alignment can be overridden.233 234// RUN: %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \235// RUN: --cuda-gpu-arch=gfx906 -Xarch_device -Wno-error=atomic-alignment %s \236// RUN: 2>&1 | FileCheck -check-prefix=NO-WARN-ATOMIC %s237// NO-WARN-ATOMIC: clang{{.*}} "-triple" "amdgcn-amd-amdhsa" {{.*}} "-Werror=atomic-alignment" {{.*}} "-Wno-error=atomic-alignment"238// NO-WARN-ATOMIC-NOT: clang{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-Werror=atomic-alignment"239// NO-WARN-ATOMIC-NOT: clang{{.*}} "-triple" "x86_64-unknown-linux-gnu" {{.*}} "-Wno-error=atomic-alignment"240 241// Check --offload-compress --offload-jobs=N does not cause warning.242// RUN: %clang -### -Werror --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \243// RUN: --offload-arch=gfx1100 --offload-compress --offload-host-only -M %s \244// RUN: --offload-jobs=4 --offload-new-driver245 246// Check --offload-jobs=N option.247 248// RUN: %clang -### -Werror --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \249// RUN: --offload-arch=gfx1100 --offload-new-driver --offload-jobs=4 %s 2>&1 | \250// RUN: FileCheck -check-prefix=JOBS %s251// JOBS: clang-linker-wrapper{{.*}} "--wrapper-jobs=4"252 253// RUN: not %clang -### --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \254// RUN: --offload-arch=gfx1100 --offload-new-driver --offload-jobs=0x4 %s 2>&1 | \255// RUN: FileCheck -check-prefix=INVJOBS %s256// INVJOBS: clang: error: invalid integral value '0x4' in '--offload-jobs=0x4'257 258// RUN: %clang -### -Werror --target=x86_64-unknown-linux-gnu -nogpuinc -nogpulib \259// RUN: --offload-arch=gfx1100 --offload-new-driver --offload-jobs=jobserver %s 2>&1 | \260// RUN: FileCheck -check-prefix=JOBSV %s261// JOBSV: clang-linker-wrapper{{.*}} "--wrapper-jobs=jobserver"262 263