clang/test/Driver/linker-wrapper-image.c

   1 // REQUIRES: x86-registered-target
   2 // REQUIRES: nvptx-registered-target
   3 // REQUIRES: amdgpu-registered-target
   4
   5 // RUN: %clang -cc1 %s -triple x86_64-unknown-linux-gnu -emit-obj -o %t.elf.o
   6
   7 // RUN: clang-offload-packager -o %t.out --image=file=%t.elf.o,kind=openmp,triple=nvptx64-nvidia-cuda,arch=sm_70
   8 // RUN: %clang -cc1 %s -triple x86_64-unknown-linux-gnu -emit-obj -o %t.o \
   9 // RUN:   -fembed-offload-object=%t.out
  10 // RUN: clang-linker-wrapper --print-wrapped-module --dry-run --host-triple=x86_64-unknown-linux-gnu \
  11 // RUN:   --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefixes=OPENMP,OPENMP-ELF
  12 // RUN: clang-linker-wrapper --print-wrapped-module --dry-run -r --host-triple=x86_64-unknown-linux-gnu \
  13 // RUN:   --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefixes=OPENMP-ELF,OPENMP-REL
  14 // RUN: clang-linker-wrapper --print-wrapped-module --dry-run --host-triple=x86_64-unknown-windows-gnu \
  15 // RUN:   --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefixes=OPENMP,OPENMP-COFF
  16
  17 //      OPENMP-ELF: @__start_omp_offloading_entries = external hidden constant [0 x %struct.__tgt_offload_entry]
  18 // OPENMP-ELF-NEXT: @__stop_omp_offloading_entries = external hidden constant [0 x %struct.__tgt_offload_entry]
  19 // OPENMP-ELF-NEXT: @__dummy.omp_offloading_entries = internal constant [0 x %struct.__tgt_offload_entry] zeroinitializer, section "omp_offloading_entries"
  20
  21 //      OPENMP-COFF: @__start_omp_offloading_entries = weak_odr hidden constant [0 x %struct.__tgt_offload_entry] zeroinitializer, section "omp_offloading_entries$OA"
  22 // OPENMP-COFF-NEXT: @__stop_omp_offloading_entries = weak_odr hidden constant [0 x %struct.__tgt_offload_entry] zeroinitializer, section "omp_offloading_entries$OZ"
  23
  24 // OPENMP-REL: @.omp_offloading.device_image = internal unnamed_addr constant [[[SIZE:[0-9]+]] x i8] c"\10\FF\10\AD{{.*}}", section ".llvm.offloading.relocatable", align 8
  25
  26 //      OPENMP: @.omp_offloading.device_image = internal unnamed_addr constant [[[SIZE:[0-9]+]] x i8] c"\10\FF\10\AD{{.*}}", section ".llvm.offloading", align 8
  27 // OPENMP-NEXT: @.omp_offloading.device_images = internal unnamed_addr constant [1 x %__tgt_device_image] [%__tgt_device_image { ptr getelementptr ([[[BEGIN:[0-9]+]] x i8], ptr @.omp_offloading.device_image, i64 0, i64 144), ptr getelementptr ([[[END:[0-9]+]] x i8], ptr @.omp_offloading.device_image, i64 0, i64 144), ptr @__start_omp_offloading_entries, ptr @__stop_omp_offloading_entries }]
  28 // OPENMP-NEXT: @.omp_offloading.descriptor = internal constant %__tgt_bin_desc { i32 1, ptr @.omp_offloading.device_images, ptr @__start_omp_offloading_entries, ptr @__stop_omp_offloading_entries }
  29 // OPENMP-NEXT: @llvm.global_ctors = appending global [1 x { i32, ptr, ptr }] [{ i32, ptr, ptr } { i32 101, ptr @.omp_offloading.descriptor_reg, ptr null }]
  30
  31 //      OPENMP: define internal void @.omp_offloading.descriptor_reg() section ".text.startup" {
  32 // OPENMP-NEXT: entry:
  33 // OPENMP-NEXT:   call void @__tgt_register_lib(ptr @.omp_offloading.descriptor)
  34 // OPENMP-NEXT:   %0 = call i32 @atexit(ptr @.omp_offloading.descriptor_unreg)
  35 // OPENMP-NEXT:   ret void
  36 // OPENMP-NEXT: }
  37
  38 //      OPENMP: define internal void @.omp_offloading.descriptor_unreg() section ".text.startup" {
  39 // OPENMP-NEXT: entry:
  40 // OPENMP-NEXT:   call void @__tgt_unregister_lib(ptr @.omp_offloading.descriptor)
  41 // OPENMP-NEXT:   ret void
  42 // OPENMP-NEXT: }
  43
  44 // RUN: clang-offload-packager -o %t.out --image=file=%t.elf.o,kind=cuda,triple=nvptx64-nvidia-cuda,arch=sm_70
  45 // RUN: %clang -cc1 %s -triple x86_64-unknown-linux-gnu -emit-obj -o %t.o \
  46 // RUN:   -fembed-offload-object=%t.out
  47 // RUN: clang-linker-wrapper --print-wrapped-module --dry-run --host-triple=x86_64-unknown-linux-gnu \
  48 // RUN:   --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefixes=CUDA,CUDA-ELF
  49 // RUN: clang-linker-wrapper --print-wrapped-module --dry-run -r --host-triple=x86_64-unknown-linux-gnu \
  50 // RUN:   --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefixes=CUDA,CUDA-ELF
  51 // RUN: clang-linker-wrapper --print-wrapped-module --dry-run --host-triple=x86_64-unknown-windows-gnu \
  52 // RUN:   --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefixes=CUDA,CUDA-COFF
  53
  54 //      CUDA-ELF: @__start_cuda_offloading_entries = external hidden constant [0 x %struct.__tgt_offload_entry]
  55 // CUDA-ELF-NEXT: @__stop_cuda_offloading_entries = external hidden constant [0 x %struct.__tgt_offload_entry]
  56 // CUDA-ELF-NEXT: @__dummy.cuda_offloading_entries = internal constant [0 x %struct.__tgt_offload_entry] zeroinitializer, section "cuda_offloading_entries"
  57
  58 //      CUDA-COFF: @__start_cuda_offloading_entries = weak_odr hidden constant [0 x %struct.__tgt_offload_entry] zeroinitializer, section "cuda_offloading_entries$OA"
  59 // CUDA-COFF-NEXT: @__stop_cuda_offloading_entries = weak_odr hidden constant [0 x %struct.__tgt_offload_entry] zeroinitializer, section "cuda_offloading_entries$OZ"
  60
  61 //      CUDA: @.fatbin_image = internal constant [0 x i8] zeroinitializer, section ".nv_fatbin"
  62 // CUDA-NEXT: @.fatbin_wrapper = internal constant %fatbin_wrapper { i32 1180844977, i32 1, ptr @.fatbin_image, ptr null }, section ".nvFatBinSegment", align 8
  63 // CUDA-NEXT: @.cuda.binary_handle = internal global ptr null
  64
  65 // CUDA: @llvm.global_ctors = appending global [1 x { i32, ptr, ptr }] [{ i32, ptr, ptr } { i32 101, ptr @.cuda.fatbin_reg, ptr null }]
  66
  67 //      CUDA: define internal void @.cuda.fatbin_reg() section ".text.startup" {
  68 // CUDA-NEXT: entry:
  69 // CUDA-NEXT:   %0 = call ptr @__cudaRegisterFatBinary(ptr @.fatbin_wrapper)
  70 // CUDA-NEXT:   store ptr %0, ptr @.cuda.binary_handle, align 8
  71 // CUDA-NEXT:   call void @.cuda.globals_reg(ptr %0)
  72 // CUDA-NEXT:   call void @__cudaRegisterFatBinaryEnd(ptr %0)
  73 // CUDA-NEXT:   %1 = call i32 @atexit(ptr @.cuda.fatbin_unreg)
  74 // CUDA-NEXT:   ret void
  75 // CUDA-NEXT: }
  76
  77 //      CUDA: define internal void @.cuda.fatbin_unreg() section ".text.startup" {
  78 // CUDA-NEXT: entry:
  79 // CUDA-NEXT:   %0 = load ptr, ptr @.cuda.binary_handle, align 8
  80 // CUDA-NEXT:   call void @__cudaUnregisterFatBinary(ptr %0)
  81 // CUDA-NEXT:   ret void
  82 // CUDA-NEXT: }
  83
  84 //      CUDA: define internal void @.cuda.globals_reg(ptr %0) section ".text.startup" {
  85 // CUDA-NEXT: entry:
  86 // CUDA-NEXT:   %1 = icmp ne ptr @__start_cuda_offloading_entries, @__stop_cuda_offloading_entries
  87 // CUDA-NEXT:   br i1 %1, label %while.entry, label %while.end
  88
  89 //      CUDA: while.entry:
  90 // CUDA-NEXT:   %entry1 = phi ptr [ @__start_cuda_offloading_entries, %entry ], [ %12, %if.end ]
  91 // CUDA-NEXT:   %2 = getelementptr inbounds %struct.__tgt_offload_entry, ptr %entry1, i64 0, i32 0
  92 // CUDA-NEXT:   %addr = load ptr, ptr %2, align 8
  93 // CUDA-NEXT:   %3 = getelementptr inbounds %struct.__tgt_offload_entry, ptr %entry1, i64 0, i32 1
  94 // CUDA-NEXT:   %name = load ptr, ptr %3, align 8
  95 // CUDA-NEXT:   %4 = getelementptr inbounds %struct.__tgt_offload_entry, ptr %entry1, i64 0, i32 2
  96 // CUDA-NEXT:   %size = load i64, ptr %4, align 4
  97 // CUDA-NEXT:   %5 = getelementptr inbounds %struct.__tgt_offload_entry, ptr %entry1, i64 0, i32 3
  98 // CUDA-NEXT:   %flags = load i32, ptr %5, align 4
  99 // CUDA-NEXT:   %6 = getelementptr inbounds %struct.__tgt_offload_entry, ptr %entry1, i64 0, i32 4
 100 // CUDA-NEXT:   %textype = load i32, ptr %6, align 4
 101 // CUDA-NEXT:   %type = and i32 %flags, 7
 102 // CUDA-NEXT:   %7 = and i32 %flags, 8
 103 // CUDA-NEXT:   %extern = lshr i32 %7, 3
 104 // CUDA-NEXT:   %8 = and i32 %flags, 16
 105 // CUDA-NEXT:   %constant = lshr i32 %8, 4
 106 // CUDA-NEXT:   %9 = and i32 %flags, 32
 107 // CUDA-NEXT:   %normalized = lshr i32 %9, 5
 108 // CUDA-NEXT:   %10 = icmp eq i64 %size, 0
 109 // CUDA-NEXT:   br i1 %10, label %if.then, label %if.else
 110
 111 //      CUDA: if.then:
 112 // CUDA-NEXT:   %11 = call i32 @__cudaRegisterFunction(ptr %0, ptr %addr, ptr %name, ptr %name, i32 -1, ptr null, ptr null, ptr null, ptr null, ptr null)
 113 // CUDA-NEXT:   br label %if.end
 114
 115 //      CUDA: if.else:
 116 // CUDA-NEXT:   switch i32 %type, label %if.end [
 117 // CUDA-NEXT:     i32 0, label %sw.global
 118 // CUDA-NEXT:     i32 1, label %sw.managed
 119 // CUDA-NEXT:     i32 2, label %sw.surface
 120 // CUDA-NEXT:     i32 3, label %sw.texture
 121 // CUDA-NEXT:   ]
 122
 123 //      CUDA: sw.global:
 124 // CUDA-NEXT:   call void @__cudaRegisterVar(ptr %0, ptr %addr, ptr %name, ptr %name, i32 %extern, i64 %size, i32 %constant, i32 0)
 125 // CUDA-NEXT:   br label %if.end
 126
 127 //      CUDA: sw.managed:
 128 // CUDA-NEXT:   br label %if.end
 129
 130 //      CUDA: sw.surface:
 131 // CUDA-NEXT:   br label %if.end
 132
 133 //      CUDA: sw.texture:
 134 // CUDA-NEXT:   br label %if.end
 135
 136 //      CUDA: if.end:
 137 // CUDA-NEXT:   %12 = getelementptr inbounds %struct.__tgt_offload_entry, ptr %entry1, i64 1
 138 // CUDA-NEXT:   %13 = icmp eq ptr %12, @__stop_cuda_offloading_entries
 139 // CUDA-NEXT:   br i1 %13, label %while.end, label %while.entry
 140
 141 //      CUDA: while.end:
 142 // CUDA-NEXT:   ret void
 143 // CUDA-NEXT: }
 144
 145 // RUN: clang-offload-packager -o %t.out --image=file=%t.elf.o,kind=hip,triple=amdgcn-amd-amdhsa,arch=gfx908
 146 // RUN: %clang -cc1 %s -triple x86_64-unknown-linux-gnu -emit-obj -o %t.o \
 147 // RUN:   -fembed-offload-object=%t.out
 148 // RUN: clang-linker-wrapper --print-wrapped-module --dry-run --host-triple=x86_64-unknown-linux-gnu \
 149 // RUN:   --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefixes=HIP,HIP-ELF
 150 // RUN: clang-linker-wrapper --print-wrapped-module --dry-run --host-triple=x86_64-unknown-linux-gnu -r \
 151 // RUN:   --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefixes=HIP,HIP-ELF
 152 // RUN: clang-linker-wrapper --print-wrapped-module --dry-run --host-triple=x86_64-unknown-windows-gnu \
 153 // RUN:   --linker-path=/usr/bin/ld %t.o -o a.out 2>&1 | FileCheck %s --check-prefixes=HIP,HIP-COFF
 154
 155 //      HIP-ELF: @__start_hip_offloading_entries = external hidden constant [0 x %struct.__tgt_offload_entry]
 156 // HIP-ELF-NEXT: @__stop_hip_offloading_entries = external hidden constant [0 x %struct.__tgt_offload_entry]
 157 // HIP-ELF-NEXT: @__dummy.hip_offloading_entries = internal constant [0 x %struct.__tgt_offload_entry] zeroinitializer, section "hip_offloading_entries"
 158
 159 //      HIP-COFF: @__start_hip_offloading_entries = weak_odr hidden constant [0 x %struct.__tgt_offload_entry] zeroinitializer, section "hip_offloading_entries$OA"
 160 // HIP-COFF-NEXT: @__stop_hip_offloading_entries = weak_odr hidden constant [0 x %struct.__tgt_offload_entry] zeroinitializer, section "hip_offloading_entries$OZ"
 161
 162 //      HIP: @.fatbin_image = internal constant [0 x i8] zeroinitializer, section ".hip_fatbin"
 163 // HIP-NEXT: @.fatbin_wrapper = internal constant %fatbin_wrapper { i32 1212764230, i32 1, ptr @.fatbin_image, ptr null }, section ".hipFatBinSegment", align 8
 164 // HIP-NEXT: @.hip.binary_handle = internal global ptr null
 165
 166 // HIP: @llvm.global_ctors = appending global [1 x { i32, ptr, ptr }] [{ i32, ptr, ptr } { i32 101, ptr @.hip.fatbin_reg, ptr null }]
 167
 168 //      HIP: define internal void @.hip.fatbin_reg() section ".text.startup" {
 169 // HIP-NEXT: entry:
 170 // HIP-NEXT:   %0 = call ptr @__hipRegisterFatBinary(ptr @.fatbin_wrapper)
 171 // HIP-NEXT:   store ptr %0, ptr @.hip.binary_handle, align 8
 172 // HIP-NEXT:   call void @.hip.globals_reg(ptr %0)
 173 // HIP-NEXT:   %1 = call i32 @atexit(ptr @.hip.fatbin_unreg)
 174 // HIP-NEXT:   ret void
 175 // HIP-NEXT: }
 176
 177 //      HIP: define internal void @.hip.fatbin_unreg() section ".text.startup" {
 178 // HIP-NEXT: entry:
 179 // HIP-NEXT:   %0 = load ptr, ptr @.hip.binary_handle, align 8
 180 // HIP-NEXT:   call void @__hipUnregisterFatBinary(ptr %0)
 181 // HIP-NEXT:   ret void
 182 // HIP-NEXT: }
 183
 184 //      HIP: define internal void @.hip.globals_reg(ptr %0) section ".text.startup" {
 185 // HIP-NEXT: entry:
 186 // HIP-NEXT:   %1 = icmp ne ptr @__start_hip_offloading_entries, @__stop_hip_offloading_entries
 187 // HIP-NEXT:   br i1 %1, label %while.entry, label %while.end
 188
 189 //      HIP: while.entry:
 190 // HIP-NEXT:   %entry1 = phi ptr [ @__start_hip_offloading_entries, %entry ], [ %12, %if.end ]
 191 // HIP-NEXT:   %2 = getelementptr inbounds %struct.__tgt_offload_entry, ptr %entry1, i64 0, i32 0
 192 // HIP-NEXT:   %addr = load ptr, ptr %2, align 8
 193 // HIP-NEXT:   %3 = getelementptr inbounds %struct.__tgt_offload_entry, ptr %entry1, i64 0, i32 1
 194 // HIP-NEXT:   %name = load ptr, ptr %3, align 8
 195 // HIP-NEXT:   %4 = getelementptr inbounds %struct.__tgt_offload_entry, ptr %entry1, i64 0, i32 2
 196 // HIP-NEXT:   %size = load i64, ptr %4, align 4
 197 // HIP-NEXT:   %5 = getelementptr inbounds %struct.__tgt_offload_entry, ptr %entry1, i64 0, i32 3
 198 // HIP-NEXT:   %flags = load i32, ptr %5, align 4
 199 // HIP-NEXT:   %6 = getelementptr inbounds %struct.__tgt_offload_entry, ptr %entry1, i64 0, i32 4
 200 // HIP-NEXT:   %textype = load i32, ptr %6, align 4
 201 // HIP-NEXT:   %type = and i32 %flags, 7
 202 // HIP-NEXT:   %7 = and i32 %flags, 8
 203 // HIP-NEXT:   %extern = lshr i32 %7, 3
 204 // HIP-NEXT:   %8 = and i32 %flags, 16
 205 // HIP-NEXT:   %constant = lshr i32 %8, 4
 206 // HIP-NEXT:   %9 = and i32 %flags, 32
 207 // HIP-NEXT:   %normalized = lshr i32 %9, 5
 208 // HIP-NEXT:   %10 = icmp eq i64 %size, 0
 209 // HIP-NEXT:   br i1 %10, label %if.then, label %if.else
 210
 211 //      HIP: if.then:
 212 // HIP-NEXT:   %11 = call i32 @__hipRegisterFunction(ptr %0, ptr %addr, ptr %name, ptr %name, i32 -1, ptr null, ptr null, ptr null, ptr null, ptr null)
 213 // HIP-NEXT:   br label %if.end
 214
 215 //      HIP: if.else:
 216 // HIP-NEXT:   switch i32 %type, label %if.end [
 217 // HIP-NEXT:     i32 0, label %sw.global
 218 // HIP-NEXT:     i32 1, label %sw.managed
 219 // HIP-NEXT:     i32 2, label %sw.surface
 220 // HIP-NEXT:     i32 3, label %sw.texture
 221 // HIP-NEXT:   ]
 222
 223 //      HIP: sw.global:
 224 // HIP-NEXT:   call void @__hipRegisterVar(ptr %0, ptr %addr, ptr %name, ptr %name, i32 %extern, i64 %size, i32 %constant, i32 0)
 225 // HIP-NEXT:   br label %if.end
 226
 227 //      HIP: sw.managed:
 228 // HIP-NEXT:   br label %if.end
 229
 230 //      HIP: sw.surface:
 231 // HIP-NEXT:   call void @__hipRegisterSurface(ptr %0, ptr %addr, ptr %name, ptr %name, i32 %textype, i32 %extern)
 232 // HIP-NEXT:   br label %if.end
 233
 234 //      HIP: sw.texture:
 235 // HIP-NEXT:   call void @__hipRegisterTexture(ptr %0, ptr %addr, ptr %name, ptr %name, i32 %textype, i32 %normalized, i32 %extern)
 236 // HIP-NEXT:   br label %if.end
 237
 238 //      HIP: if.end:
 239 // HIP-NEXT:   %12 = getelementptr inbounds %struct.__tgt_offload_entry, ptr %entry1, i64 1
 240 // HIP-NEXT:   %13 = icmp eq ptr %12, @__stop_hip_offloading_entries
 241 // HIP-NEXT:   br i1 %13, label %while.end, label %while.entry
 242
 243 //      HIP: while.end:
 244 // HIP-NEXT:   ret void
 245 // HIP-NEXT: }