clang/test/CodeGenOpenCL/amdgcn-automatic-variable.cl

   1 // RUN: %clang_cc1 -no-opaque-pointers -O0 -cl-std=CL1.2 -triple amdgcn---amdgizcl -emit-llvm %s -o - | FileCheck -check-prefixes=CHECK,CL12 %s
   2 // RUN: %clang_cc1 -no-opaque-pointers -O0 -cl-std=CL2.0 -triple amdgcn---amdgizcl -emit-llvm %s -o - | FileCheck -check-prefixes=CHECK,CL20 %s
   3
   4 // CL12-LABEL: define{{.*}} void @func1(i32 addrspace(5)* noundef %x)
   5 // CL20-LABEL: define{{.*}} void @func1(i32* noundef %x)
   6 void func1(int *x) {
   7   // CL12: %[[x_addr:.*]] = alloca i32 addrspace(5)*{{.*}}addrspace(5)
   8   // CL12: store i32 addrspace(5)* %x, i32 addrspace(5)* addrspace(5)* %[[x_addr]]
   9   // CL12: %[[r0:.*]] = load i32 addrspace(5)*, i32 addrspace(5)* addrspace(5)* %[[x_addr]]
  10   // CL12: store i32 1, i32 addrspace(5)* %[[r0]]
  11   // CL20: %[[x_addr:.*]] = alloca i32*{{.*}}addrspace(5)
  12   // CL20: store i32* %x, i32* addrspace(5)* %[[x_addr]]
  13   // CL20: %[[r0:.*]] = load i32*, i32* addrspace(5)* %[[x_addr]]
  14   // CL20: store i32 1, i32* %[[r0]]
  15   *x = 1;
  16 }
  17
  18 // CHECK-LABEL: define{{.*}} void @func2()
  19 void func2(void) {
  20   // CHECK: %lv1 = alloca i32, align 4, addrspace(5)
  21   // CHECK: %lv2 = alloca i32, align 4, addrspace(5)
  22   // CHECK: %la = alloca [100 x i32], align 4, addrspace(5)
  23   // CL12: %lp1 = alloca i32 addrspace(5)*, align 4, addrspace(5)
  24   // CL12: %lp2 = alloca i32 addrspace(5)*, align 4, addrspace(5)
  25   // CL20: %lp1 = alloca i32*, align 8, addrspace(5)
  26   // CL20: %lp2 = alloca i32*, align 8, addrspace(5)
  27   // CHECK: %lvc = alloca i32, align 4, addrspace(5)
  28
  29   // CHECK: store i32 1, i32 addrspace(5)* %lv1
  30   int lv1;
  31   lv1 = 1;
  32   // CHECK: store i32 2, i32 addrspace(5)* %lv2
  33   int lv2 = 2;
  34
  35   // CHECK: %[[arrayidx:.*]] = getelementptr inbounds [100 x i32], [100 x i32] addrspace(5)* %la, i64 0, i64 0
  36   // CHECK: store i32 3, i32 addrspace(5)* %[[arrayidx]], align 4
  37   int la[100];
  38   la[0] = 3;
  39
  40   // CL12: store i32 addrspace(5)* %lv1, i32 addrspace(5)* addrspace(5)* %lp1, align 4
  41   // CL20: %[[r0:.*]] = addrspacecast i32 addrspace(5)* %lv1 to i32*
  42   // CL20: store i32* %[[r0]], i32* addrspace(5)* %lp1, align 8
  43   int *lp1 = &lv1;
  44
  45   // CHECK: %[[arraydecay:.*]] = getelementptr inbounds [100 x i32], [100 x i32] addrspace(5)* %la, i64 0, i64 0
  46   // CL12: store i32 addrspace(5)* %[[arraydecay]], i32 addrspace(5)* addrspace(5)* %lp2, align 4
  47   // CL20: %[[r1:.*]] = addrspacecast i32 addrspace(5)* %[[arraydecay]] to i32*
  48   // CL20: store i32* %[[r1]], i32* addrspace(5)* %lp2, align 8
  49   int *lp2 = la;
  50
  51   // CL12: call void @func1(i32 addrspace(5)* noundef %lv1)
  52   // CL20: %[[r2:.*]] = addrspacecast i32 addrspace(5)* %lv1 to i32*
  53   // CL20: call void @func1(i32* noundef %[[r2]])
  54   func1(&lv1);
  55
  56   // CHECK: store i32 4, i32 addrspace(5)* %lvc
  57   // CHECK: store i32 4, i32 addrspace(5)* %lv1
  58   const int lvc = 4;
  59   lv1 = lvc;
  60 }
  61
  62 // CHECK-LABEL: define{{.*}} void @func3()
  63 // CHECK: %a = alloca [16 x [1 x float]], align 4, addrspace(5)
  64 // CHECK: %[[CAST:.+]] = bitcast [16 x [1 x float]] addrspace(5)* %a to i8 addrspace(5)*
  65 // CHECK: call void @llvm.memset.p5i8.i64(i8 addrspace(5)* align 4 %[[CAST]], i8 0, i64 64, i1 false)
  66 void func3(void) {
  67   float a[16][1] = {{0.}};
  68 }