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