clang/test/CodeGenOpenCL/builtins-amdgcn-gfx940.cl

   1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
   2 // RUN: %clang_cc1 -cl-std=CL2.0 -O0 -triple amdgcn-unknown-unknown -target-cpu gfx940 -emit-llvm -o - %s | FileCheck %s
   3 // REQUIRES: amdgpu-registered-target
   4
   5 typedef unsigned int u32;
   6 typedef unsigned short u16;
   7 typedef unsigned char u8;
   8
   9 // CHECK-LABEL: @test_global_load_lds_u32(
  10 // CHECK-NEXT:  entry:
  11 // CHECK-NEXT:    [[SRC_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
  12 // CHECK-NEXT:    [[DST_ADDR:%.*]] = alloca ptr addrspace(3), align 4, addrspace(5)
  13 // CHECK-NEXT:    [[SRC_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SRC_ADDR]] to ptr
  14 // CHECK-NEXT:    [[DST_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DST_ADDR]] to ptr
  15 // CHECK-NEXT:    store ptr addrspace(1) [[SRC:%.*]], ptr [[SRC_ADDR_ASCAST]], align 8
  16 // CHECK-NEXT:    store ptr addrspace(3) [[DST:%.*]], ptr [[DST_ADDR_ASCAST]], align 4
  17 // CHECK-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr [[SRC_ADDR_ASCAST]], align 8
  18 // CHECK-NEXT:    [[TMP1:%.*]] = load ptr addrspace(3), ptr [[DST_ADDR_ASCAST]], align 4
  19 // CHECK-NEXT:    call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) [[TMP0]], ptr addrspace(3) [[TMP1]], i32 4, i32 0, i32 0)
  20 // CHECK-NEXT:    ret void
  21 //
  22 void test_global_load_lds_u32(global u32* src, local u32 *dst) {
  23   __builtin_amdgcn_global_load_lds(src, dst, /*size=*/4, /*offset=*/0, /*aux=*/0);
  24 }
  25
  26 // CHECK-LABEL: @test_global_load_lds_u16(
  27 // CHECK-NEXT:  entry:
  28 // CHECK-NEXT:    [[SRC_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
  29 // CHECK-NEXT:    [[DST_ADDR:%.*]] = alloca ptr addrspace(3), align 4, addrspace(5)
  30 // CHECK-NEXT:    [[SRC_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SRC_ADDR]] to ptr
  31 // CHECK-NEXT:    [[DST_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DST_ADDR]] to ptr
  32 // CHECK-NEXT:    store ptr addrspace(1) [[SRC:%.*]], ptr [[SRC_ADDR_ASCAST]], align 8
  33 // CHECK-NEXT:    store ptr addrspace(3) [[DST:%.*]], ptr [[DST_ADDR_ASCAST]], align 4
  34 // CHECK-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr [[SRC_ADDR_ASCAST]], align 8
  35 // CHECK-NEXT:    [[TMP1:%.*]] = load ptr addrspace(3), ptr [[DST_ADDR_ASCAST]], align 4
  36 // CHECK-NEXT:    call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) [[TMP0]], ptr addrspace(3) [[TMP1]], i32 2, i32 0, i32 0)
  37 // CHECK-NEXT:    ret void
  38 //
  39 void test_global_load_lds_u16(global u16* src, local u16 *dst) {
  40   __builtin_amdgcn_global_load_lds(src, dst, /*size=*/2, /*offset=*/0, /*aux=*/0);
  41 }
  42
  43 // CHECK-LABEL: @test_global_load_lds_u8(
  44 // CHECK-NEXT:  entry:
  45 // CHECK-NEXT:    [[SRC_ADDR:%.*]] = alloca ptr addrspace(1), align 8, addrspace(5)
  46 // CHECK-NEXT:    [[DST_ADDR:%.*]] = alloca ptr addrspace(3), align 4, addrspace(5)
  47 // CHECK-NEXT:    [[SRC_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SRC_ADDR]] to ptr
  48 // CHECK-NEXT:    [[DST_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DST_ADDR]] to ptr
  49 // CHECK-NEXT:    store ptr addrspace(1) [[SRC:%.*]], ptr [[SRC_ADDR_ASCAST]], align 8
  50 // CHECK-NEXT:    store ptr addrspace(3) [[DST:%.*]], ptr [[DST_ADDR_ASCAST]], align 4
  51 // CHECK-NEXT:    [[TMP0:%.*]] = load ptr addrspace(1), ptr [[SRC_ADDR_ASCAST]], align 8
  52 // CHECK-NEXT:    [[TMP1:%.*]] = load ptr addrspace(3), ptr [[DST_ADDR_ASCAST]], align 4
  53 // CHECK-NEXT:    call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) [[TMP0]], ptr addrspace(3) [[TMP1]], i32 1, i32 0, i32 0)
  54 // CHECK-NEXT:    ret void
  55 //
  56 void test_global_load_lds_u8(global u8* src, local u8 *dst) {
  57   __builtin_amdgcn_global_load_lds(src, dst, /*size=*/1, /*offset=*/0, /*aux=*/0);
  58 }