clang/test/CodeGenOpenCL/amdgpu-alignment.cl

   1 // REQUIRES: amdgpu-registered-target
   2 // RUN: %clang_cc1 -no-opaque-pointers -triple amdgcn-unknown-unknown -S -disable-llvm-passes -emit-llvm -o - %s | FileCheck %s
   3 // RUN: %clang_cc1 -no-opaque-pointers -triple amdgcn-unknown-unknown-opencl -S -disable-llvm-passes -emit-llvm -o - %s | FileCheck %s
   4
   5 #pragma OPENCL EXTENSION cl_khr_fp64 : enable
   6 #pragma OPENCL EXTENSION cl_khr_fp16 : enable
   7
   8 typedef char __attribute__((ext_vector_type(2))) char2;
   9 typedef char __attribute__((ext_vector_type(3))) char3;
  10 typedef char __attribute__((ext_vector_type(4))) char4;
  11 typedef char __attribute__((ext_vector_type(8))) char8;
  12 typedef char __attribute__((ext_vector_type(16))) char16;
  13
  14 typedef short __attribute__((ext_vector_type(2))) short2;
  15 typedef short __attribute__((ext_vector_type(3))) short3;
  16 typedef short __attribute__((ext_vector_type(4))) short4;
  17 typedef short __attribute__((ext_vector_type(8))) short8;
  18 typedef short __attribute__((ext_vector_type(16))) short16;
  19
  20 typedef int __attribute__((ext_vector_type(2))) int2;
  21 typedef int __attribute__((ext_vector_type(3))) int3;
  22 typedef int __attribute__((ext_vector_type(4))) int4;
  23 typedef int __attribute__((ext_vector_type(8))) int8;
  24 typedef int __attribute__((ext_vector_type(16))) int16;
  25
  26 typedef long __attribute__((ext_vector_type(2))) long2;
  27 typedef long __attribute__((ext_vector_type(3))) long3;
  28 typedef long __attribute__((ext_vector_type(4))) long4;
  29 typedef long __attribute__((ext_vector_type(8))) long8;
  30 typedef long __attribute__((ext_vector_type(16))) long16;
  31
  32 typedef half __attribute__((ext_vector_type(2))) half2;
  33 typedef half __attribute__((ext_vector_type(3))) half3;
  34 typedef half __attribute__((ext_vector_type(4))) half4;
  35 typedef half __attribute__((ext_vector_type(8))) half8;
  36 typedef half __attribute__((ext_vector_type(16))) half16;
  37
  38 typedef float __attribute__((ext_vector_type(2))) float2;
  39 typedef float __attribute__((ext_vector_type(3))) float3;
  40 typedef float __attribute__((ext_vector_type(4))) float4;
  41 typedef float __attribute__((ext_vector_type(8))) float8;
  42 typedef float __attribute__((ext_vector_type(16))) float16;
  43
  44 typedef double __attribute__((ext_vector_type(2))) double2;
  45 typedef double __attribute__((ext_vector_type(3))) double3;
  46 typedef double __attribute__((ext_vector_type(4))) double4;
  47 typedef double __attribute__((ext_vector_type(8))) double8;
  48 typedef double __attribute__((ext_vector_type(16))) double16;
  49
  50 // CHECK: @local_memory_alignment_global.lds_i8 = internal addrspace(3) global [4 x i8] undef, align 1
  51 // CHECK: @local_memory_alignment_global.lds_v2i8 = internal addrspace(3) global [4 x <2 x i8>] undef, align 2
  52 // CHECK: @local_memory_alignment_global.lds_v3i8 = internal addrspace(3) global [4 x <3 x i8>] undef, align 4
  53 // CHECK: @local_memory_alignment_global.lds_v4i8 = internal addrspace(3) global [4 x <4 x i8>] undef, align 4
  54 // CHECK: @local_memory_alignment_global.lds_v8i8 = internal addrspace(3) global [4 x <8 x i8>] undef, align 8
  55 // CHECK: @local_memory_alignment_global.lds_v16i8 = internal addrspace(3) global [4 x <16 x i8>] undef, align 16
  56 // CHECK: @local_memory_alignment_global.lds_i16 = internal addrspace(3) global [4 x i16] undef, align 2
  57 // CHECK: @local_memory_alignment_global.lds_v2i16 = internal addrspace(3) global [4 x <2 x i16>] undef, align 4
  58 // CHECK: @local_memory_alignment_global.lds_v3i16 = internal addrspace(3) global [4 x <3 x i16>] undef, align 8
  59 // CHECK: @local_memory_alignment_global.lds_v4i16 = internal addrspace(3) global [4 x <4 x i16>] undef, align 8
  60 // CHECK: @local_memory_alignment_global.lds_v8i16 = internal addrspace(3) global [4 x <8 x i16>] undef, align 16
  61 // CHECK: @local_memory_alignment_global.lds_v16i16 = internal addrspace(3) global [4 x <16 x i16>] undef, align 32
  62 // CHECK: @local_memory_alignment_global.lds_i32 = internal addrspace(3) global [4 x i32] undef, align 4
  63 // CHECK: @local_memory_alignment_global.lds_v2i32 = internal addrspace(3) global [4 x <2 x i32>] undef, align 8
  64 // CHECK: @local_memory_alignment_global.lds_v3i32 = internal addrspace(3) global [4 x <3 x i32>] undef, align 16
  65 // CHECK: @local_memory_alignment_global.lds_v4i32 = internal addrspace(3) global [4 x <4 x i32>] undef, align 16
  66 // CHECK: @local_memory_alignment_global.lds_v8i32 = internal addrspace(3) global [4 x <8 x i32>] undef, align 32
  67 // CHECK: @local_memory_alignment_global.lds_v16i32 = internal addrspace(3) global [4 x <16 x i32>] undef, align 64
  68 // CHECK: @local_memory_alignment_global.lds_i64 = internal addrspace(3) global [4 x i64] undef, align 8
  69 // CHECK: @local_memory_alignment_global.lds_v2i64 = internal addrspace(3) global [4 x <2 x i64>] undef, align 16
  70 // CHECK: @local_memory_alignment_global.lds_v3i64 = internal addrspace(3) global [4 x <3 x i64>] undef, align 32
  71 // CHECK: @local_memory_alignment_global.lds_v4i64 = internal addrspace(3) global [4 x <4 x i64>] undef, align 32
  72 // CHECK: @local_memory_alignment_global.lds_v8i64 = internal addrspace(3) global [4 x <8 x i64>] undef, align 64
  73 // CHECK: @local_memory_alignment_global.lds_v16i64 = internal addrspace(3) global [4 x <16 x i64>] undef, align 128
  74 // CHECK: @local_memory_alignment_global.lds_f16 = internal addrspace(3) global [4 x half] undef, align 2
  75 // CHECK: @local_memory_alignment_global.lds_v2f16 = internal addrspace(3) global [4 x <2 x half>] undef, align 4
  76 // CHECK: @local_memory_alignment_global.lds_v3f16 = internal addrspace(3) global [4 x <3 x half>] undef, align 8
  77 // CHECK: @local_memory_alignment_global.lds_v4f16 = internal addrspace(3) global [4 x <4 x half>] undef, align 8
  78 // CHECK: @local_memory_alignment_global.lds_v8f16 = internal addrspace(3) global [4 x <8 x half>] undef, align 16
  79 // CHECK: @local_memory_alignment_global.lds_v16f16 = internal addrspace(3) global [4 x <16 x half>] undef, align 32
  80 // CHECK: @local_memory_alignment_global.lds_f32 = internal addrspace(3) global [4 x float] undef, align 4
  81 // CHECK: @local_memory_alignment_global.lds_v2f32 = internal addrspace(3) global [4 x <2 x float>] undef, align 8
  82 // CHECK: @local_memory_alignment_global.lds_v3f32 = internal addrspace(3) global [4 x <3 x float>] undef, align 16
  83 // CHECK: @local_memory_alignment_global.lds_v4f32 = internal addrspace(3) global [4 x <4 x float>] undef, align 16
  84 // CHECK: @local_memory_alignment_global.lds_v8f32 = internal addrspace(3) global [4 x <8 x float>] undef, align 32
  85 // CHECK: @local_memory_alignment_global.lds_v16f32 = internal addrspace(3) global [4 x <16 x float>] undef, align 64
  86 // CHECK: @local_memory_alignment_global.lds_f64 = internal addrspace(3) global [4 x double] undef, align 8
  87 // CHECK: @local_memory_alignment_global.lds_v2f64 = internal addrspace(3) global [4 x <2 x double>] undef, align 16
  88 // CHECK: @local_memory_alignment_global.lds_v3f64 = internal addrspace(3) global [4 x <3 x double>] undef, align 32
  89 // CHECK: @local_memory_alignment_global.lds_v4f64 = internal addrspace(3) global [4 x <4 x double>] undef, align 32
  90 // CHECK: @local_memory_alignment_global.lds_v8f64 = internal addrspace(3) global [4 x <8 x double>] undef, align 64
  91 // CHECK: @local_memory_alignment_global.lds_v16f64 = internal addrspace(3) global [4 x <16 x double>] undef, align 128
  92
  93
  94 // CHECK-LABEL: @local_memory_alignment_global(
  95 // CHECK: store volatile i8 0, i8 addrspace(3)* getelementptr inbounds ([4 x i8], [4 x i8] addrspace(3)* @local_memory_alignment_global.lds_i8, i64 0, i64 0), align 1
  96 // CHECK: store volatile <2 x i8> zeroinitializer, <2 x i8> addrspace(3)* getelementptr inbounds ([4 x <2 x i8>], [4 x <2 x i8>] addrspace(3)* @local_memory_alignment_global.lds_v2i8, i64 0, i64 0), align 2
  97 // CHECK: store volatile <4 x i8> <i8 0, i8 0, i8 0, i8 undef>, <4 x i8> addrspace(3)* bitcast ([4 x <3 x i8>] addrspace(3)* @local_memory_alignment_global.lds_v3i8 to <4 x i8> addrspace(3)*), align 4
  98 // CHECK: store volatile <4 x i8> zeroinitializer, <4 x i8> addrspace(3)* getelementptr inbounds ([4 x <4 x i8>], [4 x <4 x i8>] addrspace(3)* @local_memory_alignment_global.lds_v4i8, i64 0, i64 0), align 4
  99 // CHECK: store volatile <8 x i8> zeroinitializer, <8 x i8> addrspace(3)* getelementptr inbounds ([4 x <8 x i8>], [4 x <8 x i8>] addrspace(3)* @local_memory_alignment_global.lds_v8i8, i64 0, i64 0), align 8
 100 // CHECK: store volatile <16 x i8> zeroinitializer, <16 x i8> addrspace(3)* getelementptr inbounds ([4 x <16 x i8>], [4 x <16 x i8>] addrspace(3)* @local_memory_alignment_global.lds_v16i8, i64 0, i64 0), align 16
 101 // CHECK: store volatile i16 0, i16 addrspace(3)* getelementptr inbounds ([4 x i16], [4 x i16] addrspace(3)* @local_memory_alignment_global.lds_i16, i64 0, i64 0), align 2
 102 // CHECK: store volatile <2 x i16> zeroinitializer, <2 x i16> addrspace(3)* getelementptr inbounds ([4 x <2 x i16>], [4 x <2 x i16>] addrspace(3)* @local_memory_alignment_global.lds_v2i16, i64 0, i64 0), align 4
 103 // CHECK: store volatile <4 x i16> <i16 0, i16 0, i16 0, i16 undef>, <4 x i16> addrspace(3)* bitcast ([4 x <3 x i16>] addrspace(3)* @local_memory_alignment_global.lds_v3i16 to <4 x i16> addrspace(3)*), align 8
 104 // CHECK: store volatile <4 x i16> zeroinitializer, <4 x i16> addrspace(3)* getelementptr inbounds ([4 x <4 x i16>], [4 x <4 x i16>] addrspace(3)* @local_memory_alignment_global.lds_v4i16, i64 0, i64 0), align 8
 105 // CHECK: store volatile <8 x i16> zeroinitializer, <8 x i16> addrspace(3)* getelementptr inbounds ([4 x <8 x i16>], [4 x <8 x i16>] addrspace(3)* @local_memory_alignment_global.lds_v8i16, i64 0, i64 0), align 16
 106 // CHECK: store volatile <16 x i16> zeroinitializer, <16 x i16> addrspace(3)* getelementptr inbounds ([4 x <16 x i16>], [4 x <16 x i16>] addrspace(3)* @local_memory_alignment_global.lds_v16i16, i64 0, i64 0), align 32
 107 // CHECK: store volatile i32 0, i32 addrspace(3)* getelementptr inbounds ([4 x i32], [4 x i32] addrspace(3)* @local_memory_alignment_global.lds_i32, i64 0, i64 0), align 4
 108 // CHECK: store volatile <2 x i32> zeroinitializer, <2 x i32> addrspace(3)* getelementptr inbounds ([4 x <2 x i32>], [4 x <2 x i32>] addrspace(3)* @local_memory_alignment_global.lds_v2i32, i64 0, i64 0), align 8
 109 // CHECK: store volatile <4 x i32> <i32 0, i32 0, i32 0, i32 undef>, <4 x i32> addrspace(3)* bitcast ([4 x <3 x i32>] addrspace(3)* @local_memory_alignment_global.lds_v3i32 to <4 x i32> addrspace(3)*), align 16
 110 // CHECK: store volatile <4 x i32> zeroinitializer, <4 x i32> addrspace(3)* getelementptr inbounds ([4 x <4 x i32>], [4 x <4 x i32>] addrspace(3)* @local_memory_alignment_global.lds_v4i32, i64 0, i64 0), align 16
 111 // CHECK: store volatile <8 x i32> zeroinitializer, <8 x i32> addrspace(3)* getelementptr inbounds ([4 x <8 x i32>], [4 x <8 x i32>] addrspace(3)* @local_memory_alignment_global.lds_v8i32, i64 0, i64 0), align 32
 112 // CHECK: store volatile <16 x i32> zeroinitializer, <16 x i32> addrspace(3)* getelementptr inbounds ([4 x <16 x i32>], [4 x <16 x i32>] addrspace(3)* @local_memory_alignment_global.lds_v16i32, i64 0, i64 0), align 64
 113 // CHECK: store volatile i64 0, i64 addrspace(3)* getelementptr inbounds ([4 x i64], [4 x i64] addrspace(3)* @local_memory_alignment_global.lds_i64, i64 0, i64 0), align 8
 114 // CHECK: store volatile <2 x i64> zeroinitializer, <2 x i64> addrspace(3)* getelementptr inbounds ([4 x <2 x i64>], [4 x <2 x i64>] addrspace(3)* @local_memory_alignment_global.lds_v2i64, i64 0, i64 0), align 16
 115 // CHECK: store volatile <4 x i64> <i64 0, i64 0, i64 0, i64 undef>, <4 x i64> addrspace(3)* bitcast ([4 x <3 x i64>] addrspace(3)* @local_memory_alignment_global.lds_v3i64 to <4 x i64> addrspace(3)*), align 32
 116 // CHECK: store volatile <4 x i64> zeroinitializer, <4 x i64> addrspace(3)* getelementptr inbounds ([4 x <4 x i64>], [4 x <4 x i64>] addrspace(3)* @local_memory_alignment_global.lds_v4i64, i64 0, i64 0), align 32
 117 // CHECK: store volatile <8 x i64> zeroinitializer, <8 x i64> addrspace(3)* getelementptr inbounds ([4 x <8 x i64>], [4 x <8 x i64>] addrspace(3)* @local_memory_alignment_global.lds_v8i64, i64 0, i64 0), align 64
 118 // CHECK: store volatile <16 x i64> zeroinitializer, <16 x i64> addrspace(3)* getelementptr inbounds ([4 x <16 x i64>], [4 x <16 x i64>] addrspace(3)* @local_memory_alignment_global.lds_v16i64, i64 0, i64 0), align 128
 119 // CHECK: store volatile half 0xH0000, half addrspace(3)* getelementptr inbounds ([4 x half], [4 x half] addrspace(3)* @local_memory_alignment_global.lds_f16, i64 0, i64 0), align 2
 120 // CHECK: store volatile <2 x half> zeroinitializer, <2 x half> addrspace(3)* getelementptr inbounds ([4 x <2 x half>], [4 x <2 x half>] addrspace(3)* @local_memory_alignment_global.lds_v2f16, i64 0, i64 0), align 4
 121 // CHECK: store volatile <4 x half> <half 0xH0000, half 0xH0000, half 0xH0000, half undef>, <4 x half> addrspace(3)* bitcast ([4 x <3 x half>] addrspace(3)* @local_memory_alignment_global.lds_v3f16 to <4 x half> addrspace(3)*), align 8
 122 // CHECK: store volatile <4 x half> zeroinitializer, <4 x half> addrspace(3)* getelementptr inbounds ([4 x <4 x half>], [4 x <4 x half>] addrspace(3)* @local_memory_alignment_global.lds_v4f16, i64 0, i64 0), align 8
 123 // CHECK: store volatile <8 x half> zeroinitializer, <8 x half> addrspace(3)* getelementptr inbounds ([4 x <8 x half>], [4 x <8 x half>] addrspace(3)* @local_memory_alignment_global.lds_v8f16, i64 0, i64 0), align 16
 124 // CHECK: store volatile <16 x half> zeroinitializer, <16 x half> addrspace(3)* getelementptr inbounds ([4 x <16 x half>], [4 x <16 x half>] addrspace(3)* @local_memory_alignment_global.lds_v16f16, i64 0, i64 0), align 32
 125 // CHECK: store volatile float 0.000000e+00, float addrspace(3)* getelementptr inbounds ([4 x float], [4 x float] addrspace(3)* @local_memory_alignment_global.lds_f32, i64 0, i64 0), align 4
 126 // CHECK: store volatile <2 x float> zeroinitializer, <2 x float> addrspace(3)* getelementptr inbounds ([4 x <2 x float>], [4 x <2 x float>] addrspace(3)* @local_memory_alignment_global.lds_v2f32, i64 0, i64 0), align 8
 127 // CHECK: store volatile <4 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float undef>, <4 x float> addrspace(3)* bitcast ([4 x <3 x float>] addrspace(3)* @local_memory_alignment_global.lds_v3f32 to <4 x float> addrspace(3)*), align 16
 128 // CHECK: store volatile <4 x float> zeroinitializer, <4 x float> addrspace(3)* getelementptr inbounds ([4 x <4 x float>], [4 x <4 x float>] addrspace(3)* @local_memory_alignment_global.lds_v4f32, i64 0, i64 0), align 16
 129 // CHECK: store volatile <8 x float> zeroinitializer, <8 x float> addrspace(3)* getelementptr inbounds ([4 x <8 x float>], [4 x <8 x float>] addrspace(3)* @local_memory_alignment_global.lds_v8f32, i64 0, i64 0), align 32
 130 // CHECK: store volatile <16 x float> zeroinitializer, <16 x float> addrspace(3)* getelementptr inbounds ([4 x <16 x float>], [4 x <16 x float>] addrspace(3)* @local_memory_alignment_global.lds_v16f32, i64 0, i64 0), align 64
 131 // CHECK: store volatile double 0.000000e+00, double addrspace(3)* getelementptr inbounds ([4 x double], [4 x double] addrspace(3)* @local_memory_alignment_global.lds_f64, i64 0, i64 0), align 8
 132 // CHECK: store volatile <2 x double> zeroinitializer, <2 x double> addrspace(3)* getelementptr inbounds ([4 x <2 x double>], [4 x <2 x double>] addrspace(3)* @local_memory_alignment_global.lds_v2f64, i64 0, i64 0), align 16
 133 // CHECK: store volatile <4 x double> <double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double undef>, <4 x double> addrspace(3)* bitcast ([4 x <3 x double>] addrspace(3)* @local_memory_alignment_global.lds_v3f64 to <4 x double> addrspace(3)*), align 32
 134 // CHECK: store volatile <4 x double> zeroinitializer, <4 x double> addrspace(3)* getelementptr inbounds ([4 x <4 x double>], [4 x <4 x double>] addrspace(3)* @local_memory_alignment_global.lds_v4f64, i64 0, i64 0), align 32
 135 // CHECK: store volatile <8 x double> zeroinitializer, <8 x double> addrspace(3)* getelementptr inbounds ([4 x <8 x double>], [4 x <8 x double>] addrspace(3)* @local_memory_alignment_global.lds_v8f64, i64 0, i64 0), align 64
 136 // CHECK: store volatile <16 x double> zeroinitializer, <16 x double> addrspace(3)* getelementptr inbounds ([4 x <16 x double>], [4 x <16 x double>] addrspace(3)* @local_memory_alignment_global.lds_v16f64, i64 0, i64 0), align 128
 137 kernel void local_memory_alignment_global()
 138 {
 139   volatile local char lds_i8[4];
 140   volatile local char2 lds_v2i8[4];
 141   volatile local char3 lds_v3i8[4];
 142   volatile local char4 lds_v4i8[4];
 143   volatile local char8 lds_v8i8[4];
 144   volatile local char16 lds_v16i8[4];
 145
 146   volatile local short lds_i16[4];
 147   volatile local short2 lds_v2i16[4];
 148   volatile local short3 lds_v3i16[4];
 149   volatile local short4 lds_v4i16[4];
 150   volatile local short8 lds_v8i16[4];
 151   volatile local short16 lds_v16i16[4];
 152
 153   volatile local int lds_i32[4];
 154   volatile local int2 lds_v2i32[4];
 155   volatile local int3 lds_v3i32[4];
 156   volatile local int4 lds_v4i32[4];
 157   volatile local int8 lds_v8i32[4];
 158   volatile local int16 lds_v16i32[4];
 159
 160   volatile local long lds_i64[4];
 161   volatile local long2 lds_v2i64[4];
 162   volatile local long3 lds_v3i64[4];
 163   volatile local long4 lds_v4i64[4];
 164   volatile local long8 lds_v8i64[4];
 165   volatile local long16 lds_v16i64[4];
 166
 167   volatile local half lds_f16[4];
 168   volatile local half2 lds_v2f16[4];
 169   volatile local half3 lds_v3f16[4];
 170   volatile local half4 lds_v4f16[4];
 171   volatile local half8 lds_v8f16[4];
 172   volatile local half16 lds_v16f16[4];
 173
 174   volatile local float lds_f32[4];
 175   volatile local float2 lds_v2f32[4];
 176   volatile local float3 lds_v3f32[4];
 177   volatile local float4 lds_v4f32[4];
 178   volatile local float8 lds_v8f32[4];
 179   volatile local float16 lds_v16f32[4];
 180
 181   volatile local double lds_f64[4];
 182   volatile local double2 lds_v2f64[4];
 183   volatile local double3 lds_v3f64[4];
 184   volatile local double4 lds_v4f64[4];
 185   volatile local double8 lds_v8f64[4];
 186   volatile local double16 lds_v16f64[4];
 187
 188   *lds_i8 = 0;
 189   *lds_v2i8 = 0;
 190   *lds_v3i8 = 0;
 191   *lds_v4i8 = 0;
 192   *lds_v8i8 = 0;
 193   *lds_v16i8 = 0;
 194
 195   *lds_i16 = 0;
 196   *lds_v2i16 = 0;
 197   *lds_v3i16 = 0;
 198   *lds_v4i16 = 0;
 199   *lds_v8i16 = 0;
 200   *lds_v16i16 = 0;
 201
 202   *lds_i32 = 0;
 203   *lds_v2i32 = 0;
 204   *lds_v3i32 = 0;
 205   *lds_v4i32 = 0;
 206   *lds_v8i32 = 0;
 207   *lds_v16i32 = 0;
 208
 209   *lds_i64 = 0;
 210   *lds_v2i64 = 0;
 211   *lds_v3i64 = 0;
 212   *lds_v4i64 = 0;
 213   *lds_v8i64 = 0;
 214   *lds_v16i64 = 0;
 215
 216   *lds_f16 = 0;
 217   *lds_v2f16 = 0;
 218   *lds_v3f16 = 0;
 219   *lds_v4f16 = 0;
 220   *lds_v8f16 = 0;
 221   *lds_v16f16 = 0;
 222
 223   *lds_f32 = 0;
 224   *lds_v2f32 = 0;
 225   *lds_v3f32 = 0;
 226   *lds_v4f32 = 0;
 227   *lds_v8f32 = 0;
 228   *lds_v16f32 = 0;
 229
 230   *lds_f64 = 0;
 231   *lds_v2f64 = 0;
 232   *lds_v3f64 = 0;
 233   *lds_v4f64 = 0;
 234   *lds_v8f64 = 0;
 235   *lds_v16f64 = 0;
 236 }
 237
 238 kernel void local_memory_alignment_arg(
 239   volatile local char* lds_i8,
 240   volatile local char2* lds_v2i8,
 241   volatile local char3* lds_v3i8,
 242   volatile local char4* lds_v4i8,
 243   volatile local char8* lds_v8i8,
 244   volatile local char16* lds_v16i8,
 245
 246   volatile local short* lds_i16,
 247   volatile local short2* lds_v2i16,
 248   volatile local short3* lds_v3i16,
 249   volatile local short4* lds_v4i16,
 250   volatile local short8* lds_v8i16,
 251   volatile local short16* lds_v16i16,
 252
 253   volatile local int* lds_i32,
 254   volatile local int2* lds_v2i32,
 255   volatile local int3* lds_v3i32,
 256   volatile local int4* lds_v4i32,
 257   volatile local int8* lds_v8i32,
 258   volatile local int16* lds_v16i32,
 259
 260   volatile local long* lds_i64,
 261   volatile local long2* lds_v2i64,
 262   volatile local long3* lds_v3i64,
 263   volatile local long4* lds_v4i64,
 264   volatile local long8* lds_v8i64,
 265   volatile local long16* lds_v16i64,
 266
 267   volatile local half* lds_f16,
 268   volatile local half2* lds_v2f16,
 269   volatile local half3* lds_v3f16,
 270   volatile local half4* lds_v4f16,
 271   volatile local half8* lds_v8f16,
 272   volatile local half16* lds_v16f16,
 273
 274   volatile local float* lds_f32,
 275   volatile local float2* lds_v2f32,
 276   volatile local float3* lds_v3f32,
 277   volatile local float4* lds_v4f32,
 278   volatile local float8* lds_v8f32,
 279   volatile local float16* lds_v16f32,
 280
 281   volatile local double* lds_f64,
 282   volatile local double2* lds_v2f64,
 283   volatile local double3* lds_v3f64,
 284   volatile local double4* lds_v4f64,
 285   volatile local double8* lds_v8f64,
 286   volatile local double16* lds_v16f64)
 287 {
 288   *lds_i8 = 0;
 289   *lds_v2i8 = 0;
 290   *lds_v3i8 = 0;
 291   *lds_v4i8 = 0;
 292   *lds_v8i8 = 0;
 293   *lds_v16i8 = 0;
 294
 295   *lds_i16 = 0;
 296   *lds_v2i16 = 0;
 297   *lds_v3i16 = 0;
 298   *lds_v4i16 = 0;
 299   *lds_v8i16 = 0;
 300   *lds_v16i16 = 0;
 301
 302   *lds_i32 = 0;
 303   *lds_v2i32 = 0;
 304   *lds_v3i32 = 0;
 305   *lds_v4i32 = 0;
 306   *lds_v8i32 = 0;
 307   *lds_v16i32 = 0;
 308
 309   *lds_i64 = 0;
 310   *lds_v2i64 = 0;
 311   *lds_v3i64 = 0;
 312   *lds_v4i64 = 0;
 313   *lds_v8i64 = 0;
 314   *lds_v16i64 = 0;
 315
 316   *lds_f16 = 0;
 317   *lds_v2f16 = 0;
 318   *lds_v3f16 = 0;
 319   *lds_v4f16 = 0;
 320   *lds_v8f16 = 0;
 321   *lds_v16f16 = 0;
 322
 323   *lds_f32 = 0;
 324   *lds_v2f32 = 0;
 325   *lds_v3f32 = 0;
 326   *lds_v4f32 = 0;
 327   *lds_v8f32 = 0;
 328   *lds_v16f32 = 0;
 329
 330   *lds_f64 = 0;
 331   *lds_v2f64 = 0;
 332   *lds_v3f64 = 0;
 333   *lds_v4f64 = 0;
 334   *lds_v8f64 = 0;
 335   *lds_v16f64 = 0;
 336 }
 337
 338 // CHECK-LABEL: @private_memory_alignment_alloca(
 339 // CHECK: %private_i8 = alloca [4 x i8], align 1, addrspace(5)
 340 // CHECK: %private_v2i8 = alloca [4 x <2 x i8>], align 2, addrspace(5)
 341 // CHECK: %private_v3i8 = alloca [4 x <3 x i8>], align 4, addrspace(5)
 342 // CHECK: %private_v4i8 = alloca [4 x <4 x i8>], align 4, addrspace(5)
 343 // CHECK: %private_v8i8 = alloca [4 x <8 x i8>], align 8, addrspace(5)
 344 // CHECK: %private_v16i8 = alloca [4 x <16 x i8>], align 16, addrspace(5)
 345 // CHECK: %private_i16 = alloca [4 x i16], align 2, addrspace(5)
 346 // CHECK: %private_v2i16 = alloca [4 x <2 x i16>], align 4, addrspace(5)
 347 // CHECK: %private_v3i16 = alloca [4 x <3 x i16>], align 8, addrspace(5)
 348 // CHECK: %private_v4i16 = alloca [4 x <4 x i16>], align 8, addrspace(5)
 349 // CHECK: %private_v8i16 = alloca [4 x <8 x i16>], align 16, addrspace(5)
 350 // CHECK: %private_v16i16 = alloca [4 x <16 x i16>], align 32, addrspace(5)
 351 // CHECK: %private_i32 = alloca [4 x i32], align 4, addrspace(5)
 352 // CHECK: %private_v2i32 = alloca [4 x <2 x i32>], align 8, addrspace(5)
 353 // CHECK: %private_v3i32 = alloca [4 x <3 x i32>], align 16, addrspace(5)
 354 // CHECK: %private_v4i32 = alloca [4 x <4 x i32>], align 16, addrspace(5)
 355 // CHECK: %private_v8i32 = alloca [4 x <8 x i32>], align 32, addrspace(5)
 356 // CHECK: %private_v16i32 = alloca [4 x <16 x i32>], align 64, addrspace(5)
 357 // CHECK: %private_i64 = alloca [4 x i64], align 8, addrspace(5)
 358 // CHECK: %private_v2i64 = alloca [4 x <2 x i64>], align 16, addrspace(5)
 359 // CHECK: %private_v3i64 = alloca [4 x <3 x i64>], align 32, addrspace(5)
 360 // CHECK: %private_v4i64 = alloca [4 x <4 x i64>], align 32, addrspace(5)
 361 // CHECK: %private_v8i64 = alloca [4 x <8 x i64>], align 64, addrspace(5)
 362 // CHECK: %private_v16i64 = alloca [4 x <16 x i64>], align 128, addrspace(5)
 363 // CHECK: %private_f16 = alloca [4 x half], align 2, addrspace(5)
 364 // CHECK: %private_v2f16 = alloca [4 x <2 x half>], align 4, addrspace(5)
 365 // CHECK: %private_v3f16 = alloca [4 x <3 x half>], align 8, addrspace(5)
 366 // CHECK: %private_v4f16 = alloca [4 x <4 x half>], align 8, addrspace(5)
 367 // CHECK: %private_v8f16 = alloca [4 x <8 x half>], align 16, addrspace(5)
 368 // CHECK: %private_v16f16 = alloca [4 x <16 x half>], align 32, addrspace(5)
 369 // CHECK: %private_f32 = alloca [4 x float], align 4, addrspace(5)
 370 // CHECK: %private_v2f32 = alloca [4 x <2 x float>], align 8, addrspace(5)
 371 // CHECK: %private_v3f32 = alloca [4 x <3 x float>], align 16, addrspace(5)
 372 // CHECK: %private_v4f32 = alloca [4 x <4 x float>], align 16, addrspace(5)
 373 // CHECK: %private_v8f32 = alloca [4 x <8 x float>], align 32, addrspace(5)
 374 // CHECK: %private_v16f32 = alloca [4 x <16 x float>], align 64, addrspace(5)
 375 // CHECK: %private_f64 = alloca [4 x double], align 8, addrspace(5)
 376 // CHECK: %private_v2f64 = alloca [4 x <2 x double>], align 16, addrspace(5)
 377 // CHECK: %private_v3f64 = alloca [4 x <3 x double>], align 32, addrspace(5)
 378 // CHECK: %private_v4f64 = alloca [4 x <4 x double>], align 32, addrspace(5)
 379 // CHECK: %private_v8f64 = alloca [4 x <8 x double>], align 64, addrspace(5)
 380 // CHECK: %private_v16f64 = alloca [4 x <16 x double>], align 128, addrspace(5)
 381
 382 // CHECK: store volatile i8 0, i8 addrspace(5)* %arraydecay, align 1
 383 // CHECK: store volatile <2 x i8> zeroinitializer, <2 x i8> addrspace(5)* %arraydecay{{[0-9]+}}, align 2
 384 // CHECK: store volatile <4 x i8> <i8 0, i8 0, i8 0, i8 undef>, <4 x i8> addrspace(5)* %storetmp, align 4
 385 // CHECK: store volatile <4 x i8> zeroinitializer, <4 x i8> addrspace(5)* %arraydecay{{[0-9]+}}, align 4
 386 // CHECK: store volatile <8 x i8> zeroinitializer, <8 x i8> addrspace(5)* %arraydecay{{[0-9]+}}, align 8
 387 // CHECK: store volatile <16 x i8> zeroinitializer, <16 x i8> addrspace(5)* %arraydecay{{[0-9]+}}, align 16
 388 // CHECK: store volatile i16 0, i16 addrspace(5)* %arraydecay{{[0-9]+}}, align 2
 389 // CHECK: store volatile <2 x i16> zeroinitializer, <2 x i16> addrspace(5)* %arraydecay{{[0-9]+}}, align 4
 390 // CHECK: store volatile <4 x i16> <i16 0, i16 0, i16 0, i16 undef>, <4 x i16> addrspace(5)* %storetmp{{[0-9]+}}, align 8
 391 // CHECK: store volatile <4 x i16> zeroinitializer, <4 x i16> addrspace(5)* %arraydecay{{[0-9]+}}, align 8
 392 // CHECK: store volatile <8 x i16> zeroinitializer, <8 x i16> addrspace(5)* %arraydecay{{[0-9]+}}, align 16
 393 // CHECK: store volatile <16 x i16> zeroinitializer, <16 x i16> addrspace(5)* %arraydecay{{[0-9]+}}, align 32
 394 // CHECK: store volatile i32 0, i32 addrspace(5)* %arraydecay{{[0-9]+}}, align 4
 395 // CHECK: store volatile <2 x i32> zeroinitializer, <2 x i32> addrspace(5)* %arraydecay{{[0-9]+}}, align 8
 396 // CHECK: store volatile <4 x i32> <i32 0, i32 0, i32 0, i32 undef>, <4 x i32> addrspace(5)* %storetmp16, align 16
 397 // CHECK: store volatile <4 x i32> zeroinitializer, <4 x i32> addrspace(5)* %arraydecay{{[0-9]+}}, align 16
 398 // CHECK: store volatile <8 x i32> zeroinitializer, <8 x i32> addrspace(5)* %arraydecay{{[0-9]+}}, align 32
 399 // CHECK: store volatile <16 x i32> zeroinitializer, <16 x i32> addrspace(5)* %arraydecay{{[0-9]+}}, align 64
 400 // CHECK: store volatile i64 0, i64 addrspace(5)* %arraydecay{{[0-9]+}}, align 8
 401 // CHECK: store volatile <2 x i64> zeroinitializer, <2 x i64> addrspace(5)* %arraydecay{{[0-9]+}}, align 16
 402 // CHECK: store volatile <4 x i64> <i64 0, i64 0, i64 0, i64 undef>, <4 x i64> addrspace(5)* %storetmp23, align 32
 403 // CHECK: store volatile <4 x i64> zeroinitializer, <4 x i64> addrspace(5)* %arraydecay{{[0-9]+}}, align 32
 404 // CHECK: store volatile <8 x i64> zeroinitializer, <8 x i64> addrspace(5)* %arraydecay{{[0-9]+}}, align 64
 405 // CHECK: store volatile <16 x i64> zeroinitializer, <16 x i64> addrspace(5)* %arraydecay{{[0-9]+}}, align 128
 406 // CHECK: store volatile half 0xH0000, half addrspace(5)* %arraydecay{{[0-9]+}}, align 2
 407 // CHECK: store volatile <2 x half> zeroinitializer, <2 x half> addrspace(5)* %arraydecay{{[0-9]+}}, align 4
 408 // CHECK: store volatile <4 x half> <half 0xH0000, half 0xH0000, half 0xH0000, half undef>, <4 x half> addrspace(5)* %storetmp{{[0-9]+}}, align 8
 409 // CHECK: store volatile <4 x half> zeroinitializer, <4 x half> addrspace(5)* %arraydecay{{[0-9]+}}, align 8
 410 // CHECK: store volatile <8 x half> zeroinitializer, <8 x half> addrspace(5)* %arraydecay{{[0-9]+}}, align 16
 411 // CHECK: store volatile <16 x half> zeroinitializer, <16 x half> addrspace(5)* %arraydecay{{[0-9]+}}, align 32
 412 // CHECK: store volatile float 0.000000e+00, float addrspace(5)* %arraydecay34, align 4
 413 // CHECK: store volatile <2 x float> zeroinitializer, <2 x float> addrspace(5)* %arraydecay{{[0-9]+}}, align 8
 414 // CHECK: store volatile <4 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float undef>, <4 x float> addrspace(5)* %storetmp{{[0-9]+}}, align 16
 415 // CHECK: store volatile <4 x float> zeroinitializer, <4 x float> addrspace(5)* %arraydecay{{[0-9]+}}, align 16
 416 // CHECK: store volatile <8 x float> zeroinitializer, <8 x float> addrspace(5)* %arraydecay{{[0-9]+}}, align 32
 417 // CHECK: store volatile <16 x float> zeroinitializer, <16 x float> addrspace(5)* %arraydecay{{[0-9]+}}, align 64
 418 // CHECK: store volatile double 0.000000e+00, double addrspace(5)* %arraydecay{{[0-9]+}}, align 8
 419 // CHECK: store volatile <2 x double> zeroinitializer, <2 x double> addrspace(5)* %arraydecay{{[0-9]+}}, align 16
 420 // CHECK: store volatile <4 x double> <double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double undef>, <4 x double> addrspace(5)* %storetmp{{[0-9]+}}, align 32
 421 // CHECK: store volatile <4 x double> zeroinitializer, <4 x double> addrspace(5)* %arraydecay{{[0-9]+}}, align 32
 422 // CHECK: store volatile <8 x double> zeroinitializer, <8 x double> addrspace(5)* %arraydecay{{[0-9]+}}, align 64
 423 // CHECK: store volatile <16 x double> zeroinitializer, <16 x double> addrspace(5)* %arraydecay{{[0-9]+}}, align 128
 424 kernel void private_memory_alignment_alloca()
 425 {
 426   volatile private char private_i8[4];
 427   volatile private char2 private_v2i8[4];
 428   volatile private char3 private_v3i8[4];
 429   volatile private char4 private_v4i8[4];
 430   volatile private char8 private_v8i8[4];
 431   volatile private char16 private_v16i8[4];
 432
 433   volatile private short private_i16[4];
 434   volatile private short2 private_v2i16[4];
 435   volatile private short3 private_v3i16[4];
 436   volatile private short4 private_v4i16[4];
 437   volatile private short8 private_v8i16[4];
 438   volatile private short16 private_v16i16[4];
 439
 440   volatile private int private_i32[4];
 441   volatile private int2 private_v2i32[4];
 442   volatile private int3 private_v3i32[4];
 443   volatile private int4 private_v4i32[4];
 444   volatile private int8 private_v8i32[4];
 445   volatile private int16 private_v16i32[4];
 446
 447   volatile private long private_i64[4];
 448   volatile private long2 private_v2i64[4];
 449   volatile private long3 private_v3i64[4];
 450   volatile private long4 private_v4i64[4];
 451   volatile private long8 private_v8i64[4];
 452   volatile private long16 private_v16i64[4];
 453
 454   volatile private half private_f16[4];
 455   volatile private half2 private_v2f16[4];
 456   volatile private half3 private_v3f16[4];
 457   volatile private half4 private_v4f16[4];
 458   volatile private half8 private_v8f16[4];
 459   volatile private half16 private_v16f16[4];
 460
 461   volatile private float private_f32[4];
 462   volatile private float2 private_v2f32[4];
 463   volatile private float3 private_v3f32[4];
 464   volatile private float4 private_v4f32[4];
 465   volatile private float8 private_v8f32[4];
 466   volatile private float16 private_v16f32[4];
 467
 468   volatile private double private_f64[4];
 469   volatile private double2 private_v2f64[4];
 470   volatile private double3 private_v3f64[4];
 471   volatile private double4 private_v4f64[4];
 472   volatile private double8 private_v8f64[4];
 473   volatile private double16 private_v16f64[4];
 474
 475   *private_i8 = 0;
 476   *private_v2i8 = 0;
 477   *private_v3i8 = 0;
 478   *private_v4i8 = 0;
 479   *private_v8i8 = 0;
 480   *private_v16i8 = 0;
 481
 482   *private_i16 = 0;
 483   *private_v2i16 = 0;
 484   *private_v3i16 = 0;
 485   *private_v4i16 = 0;
 486   *private_v8i16 = 0;
 487   *private_v16i16 = 0;
 488
 489   *private_i32 = 0;
 490   *private_v2i32 = 0;
 491   *private_v3i32 = 0;
 492   *private_v4i32 = 0;
 493   *private_v8i32 = 0;
 494   *private_v16i32 = 0;
 495
 496   *private_i64 = 0;
 497   *private_v2i64 = 0;
 498   *private_v3i64 = 0;
 499   *private_v4i64 = 0;
 500   *private_v8i64 = 0;
 501   *private_v16i64 = 0;
 502
 503   *private_f16 = 0;
 504   *private_v2f16 = 0;
 505   *private_v3f16 = 0;
 506   *private_v4f16 = 0;
 507   *private_v8f16 = 0;
 508   *private_v16f16 = 0;
 509
 510   *private_f32 = 0;
 511   *private_v2f32 = 0;
 512   *private_v3f32 = 0;
 513   *private_v4f32 = 0;
 514   *private_v8f32 = 0;
 515   *private_v16f32 = 0;
 516
 517   *private_f64 = 0;
 518   *private_v2f64 = 0;
 519   *private_v3f64 = 0;
 520   *private_v4f64 = 0;
 521   *private_v8f64 = 0;
 522   *private_v16f64 = 0;
 523 }