1// RUN: %clang_cc1 %s -cl-std=CL2.0 -ffake-address-space-map -O0 -emit-llvm -o - -triple "spir-unknown-unknown" | FileCheck %s --check-prefix=COMMON --check-prefix=B32 2// RUN: %clang_cc1 %s -cl-std=CL2.0 -ffake-address-space-map -O0 -emit-llvm -o - -triple "spir64-unknown-unknown" | FileCheck %s --check-prefix=COMMON --check-prefix=B64 3// RUN: %clang_cc1 %s -cl-std=CL2.0 -ffake-address-space-map -O1 -emit-llvm -o - -triple "spir64-unknown-unknown" | FileCheck %s --check-prefix=CHECK-LIFETIMES 4 5#pragma OPENCL EXTENSION cl_khr_subgroups : enable 6 7typedef void (^bl_t)(local void *); 8typedef struct {int a;} ndrange_t; 9 10// COMMON: %struct.__opencl_block_literal_generic = type { i32, i32, i8 addrspace(4)* } 11 12// For a block global variable, first emit the block literal as a global variable, then emit the block variable itself. 13// COMMON: [[BL_GLOBAL:@__block_literal_global[^ ]*]] = internal addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8 addrspace(3)*)* [[INV_G:@[^ ]+]] to i8*) to i8 addrspace(4)*) } 14// COMMON: @block_G = addrspace(1) constant %struct.__opencl_block_literal_generic addrspace(4)* addrspacecast (%struct.__opencl_block_literal_generic addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BL_GLOBAL]] to %struct.__opencl_block_literal_generic addrspace(1)*) to %struct.__opencl_block_literal_generic addrspace(4)*) 15 16// For anonymous blocks without captures, emit block literals as global variable. 17// COMMON: [[BLG1:@__block_literal_global[^ ]*]] = internal addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8 addrspace(3)*)* {{@[^ ]+}} to i8*) to i8 addrspace(4)*) } 18// COMMON: [[BLG2:@__block_literal_global[^ ]*]] = internal addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8 addrspace(3)*)* {{@[^ ]+}} to i8*) to i8 addrspace(4)*) } 19// COMMON: [[BLG3:@__block_literal_global[^ ]*]] = internal addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8 addrspace(3)*)* {{@[^ ]+}} to i8*) to i8 addrspace(4)*) } 20// COMMON: [[BLG4:@__block_literal_global[^ ]*]] = internal addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8 addrspace(3)*)* {{@[^ ]+}} to i8*) to i8 addrspace(4)*) } 21// COMMON: [[BLG5:@__block_literal_global[^ ]*]] = internal addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8 addrspace(3)*)* {{@[^ ]+}} to i8*) to i8 addrspace(4)*) } 22// COMMON: [[BLG6:@__block_literal_global[^ ]*]] = internal addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8 addrspace(3)*, i8 addrspace(3)*, i8 addrspace(3)*)* {{@[^ ]+}} to i8*) to i8 addrspace(4)*) } 23// COMMON: [[BLG7:@__block_literal_global[^ ]*]] = internal addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8 addrspace(3)*)* {{@[^ ]+}} to i8*) to i8 addrspace(4)*) } 24// COMMON: [[BLG8:@__block_literal_global[^ ]*]] = internal addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*)* [[INVG8:@[^ ]+]] to i8*) to i8 addrspace(4)*) } 25// COMMON: [[BLG9:@__block_literal_global[^ ]*]] = internal addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*, i8 addrspace(3)*)* [[INVG9:@[^ ]+]] to i8*) to i8 addrspace(4)*) } 26// COMMON: [[BLG10:@__block_literal_global[^ ]*]] = internal addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*)* {{@[^ ]+}} to i8*) to i8 addrspace(4)*) } 27// COMMON: [[BLG11:@__block_literal_global[^ ]*]] = internal addrspace(1) constant { i32, i32, i8 addrspace(4)* } { i32 {{[0-9]+}}, i32 {{[0-9]+}}, i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*)* {{@[^ ]+}} to i8*) to i8 addrspace(4)*) } 28 29// Emits block literal [[BL_GLOBAL]], invoke function [[INV_G]] and global block variable @block_G 30// COMMON: define internal spir_func void [[INV_G]](i8 addrspace(4)* %{{.*}}, i8 addrspace(3)* %{{.*}}) 31const bl_t block_G = (bl_t) ^ (local void *a) {}; 32 33void callee(int id, __global int *out) { 34 out[id] = id; 35} 36 37// COMMON-LABEL: define spir_kernel void @device_side_enqueue(i32 addrspace(1)* %{{.*}}, i32 addrspace(1)* %b, i32 %i) 38kernel void device_side_enqueue(global int *a, global int *b, int i) { 39 // COMMON: %default_queue = alloca %opencl.queue_t* 40 queue_t default_queue; 41 // COMMON: %flags = alloca i32 42 unsigned flags = 0; 43 // COMMON: %ndrange = alloca %struct.ndrange_t 44 ndrange_t ndrange; 45 // COMMON: %clk_event = alloca %opencl.clk_event_t* 46 clk_event_t clk_event; 47 // COMMON: %event_wait_list = alloca %opencl.clk_event_t* 48 clk_event_t event_wait_list; 49 // COMMON: %event_wait_list2 = alloca [1 x %opencl.clk_event_t*] 50 clk_event_t event_wait_list2[] = {clk_event}; 51 52 // COMMON: [[NDR:%[a-z0-9]+]] = alloca %struct.ndrange_t, align 4 53 54 // B32: %[[BLOCK_SIZES1:.*]] = alloca [1 x i32] 55 // B64: %[[BLOCK_SIZES1:.*]] = alloca [1 x i64] 56 // CHECK-LIFETIMES: %[[BLOCK_SIZES1:.*]] = alloca [1 x i64] 57 // B32: %[[BLOCK_SIZES2:.*]] = alloca [1 x i32] 58 // B64: %[[BLOCK_SIZES2:.*]] = alloca [1 x i64] 59 // CHECK-LIFETIMES: %[[BLOCK_SIZES2:.*]] = alloca [1 x i64] 60 // B32: %[[BLOCK_SIZES3:.*]] = alloca [1 x i32] 61 // B64: %[[BLOCK_SIZES3:.*]] = alloca [1 x i64] 62 // CHECK-LIFETIMES: %[[BLOCK_SIZES3:.*]] = alloca [1 x i64] 63 // B32: %[[BLOCK_SIZES4:.*]] = alloca [1 x i32] 64 // B64: %[[BLOCK_SIZES4:.*]] = alloca [1 x i64] 65 // CHECK-LIFETIMES: %[[BLOCK_SIZES4:.*]] = alloca [1 x i64] 66 // B32: %[[BLOCK_SIZES5:.*]] = alloca [1 x i32] 67 // B64: %[[BLOCK_SIZES5:.*]] = alloca [1 x i64] 68 // CHECK-LIFETIMES: %[[BLOCK_SIZES5:.*]] = alloca [1 x i64] 69 // B32: %[[BLOCK_SIZES6:.*]] = alloca [3 x i32] 70 // B64: %[[BLOCK_SIZES6:.*]] = alloca [3 x i64] 71 // CHECK-LIFETIMES: %[[BLOCK_SIZES6:.*]] = alloca [3 x i64] 72 // B32: %[[BLOCK_SIZES7:.*]] = alloca [1 x i32] 73 // B64: %[[BLOCK_SIZES7:.*]] = alloca [1 x i64] 74 // CHECK-LIFETIMES: %[[BLOCK_SIZES7:.*]] = alloca [1 x i64] 75 76 // Emits block literal on stack and block kernel [[INVLK1]]. 77 // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*, %opencl.queue_t{{.*}}** %default_queue 78 // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags 79 // COMMON: store i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*)* [[INVL1:@__device_side_enqueue_block_invoke[^ ]*]] to i8*) to i8 addrspace(4)*), i8 addrspace(4)** %block.invoke 80 // B32: [[BL:%[0-9]+]] = bitcast <{ i32, i32, i8 addrspace(4)*, i32 addrspace(1)*, i32, i32 addrspace(1)* }>* %block to %struct.__opencl_block_literal_generic* 81 // B64: [[BL:%[0-9]+]] = bitcast <{ i32, i32, i8 addrspace(4)*, i32 addrspace(1)*, i32 addrspace(1)*, i32 }>* %block to %struct.__opencl_block_literal_generic* 82 // COMMON: [[BL_I8:%[0-9]+]] = addrspacecast %struct.__opencl_block_literal_generic* [[BL]] to i8 addrspace(4)* 83 // COMMON-LABEL: call i32 @__enqueue_kernel_basic( 84 // COMMON-SAME: %opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]], %struct.ndrange_t* byval(%struct.ndrange_t) [[NDR]]{{([0-9]+)?}}, 85 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVLK1:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 86 // COMMON-SAME: i8 addrspace(4)* [[BL_I8]]) 87 enqueue_kernel(default_queue, flags, ndrange, 88 ^(void) { 89 a[i] = b[i]; 90 }); 91 92 // Emits block literal on stack and block kernel [[INVLK2]]. 93 // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*, %opencl.queue_t{{.*}}** %default_queue 94 // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags 95 // COMMON: [[WAIT_EVNT:%[0-9]+]] = addrspacecast %opencl.clk_event_t{{.*}}** %event_wait_list to %opencl.clk_event_t{{.*}}* addrspace(4)* 96 // COMMON: [[EVNT:%[0-9]+]] = addrspacecast %opencl.clk_event_t{{.*}}** %clk_event to %opencl.clk_event_t{{.*}}* addrspace(4)* 97 // COMMON: store i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*)* [[INVL2:@__device_side_enqueue_block_invoke[^ ]*]] to i8*) to i8 addrspace(4)*), i8 addrspace(4)** %block.invoke 98 // COMMON: [[BL:%[0-9]+]] = bitcast <{ i32, i32, i8 addrspace(4)*, i32{{.*}}, i32{{.*}}, i32{{.*}} }>* %block4 to %struct.__opencl_block_literal_generic* 99 // COMMON: [[BL_I8:%[0-9]+]] = addrspacecast %struct.__opencl_block_literal_generic* [[BL]] to i8 addrspace(4)* 100 // COMMON-LABEL: call i32 @__enqueue_kernel_basic_events 101 // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]], %struct.ndrange_t* {{.*}}, i32 2, %opencl.clk_event_t{{.*}}* addrspace(4)* [[WAIT_EVNT]], %opencl.clk_event_t{{.*}}* addrspace(4)* [[EVNT]], 102 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVLK2:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 103 // COMMON-SAME: i8 addrspace(4)* [[BL_I8]]) 104 enqueue_kernel(default_queue, flags, ndrange, 2, &event_wait_list, &clk_event, 105 ^(void) { 106 a[i] = b[i]; 107 }); 108 109 // COMMON-LABEL: call i32 @__enqueue_kernel_basic_events 110 // COMMON-SAME: (%opencl.queue_t{{.*}}* {{%[0-9]+}}, i32 {{%[0-9]+}}, %struct.ndrange_t* {{.*}}, i32 1, %opencl.clk_event_t{{.*}}* addrspace(4)* null, %opencl.clk_event_t{{.*}}* addrspace(4)* null, 111 enqueue_kernel(default_queue, flags, ndrange, 1, 0, 0, 112 ^(void) { 113 return; 114 }); 115 116 // Emits global block literal [[BLG1]] and block kernel [[INVGK1]]. 117 // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*, %opencl.queue_t{{.*}}** %default_queue 118 // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags 119 // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [1 x i64]* %[[BLOCK_SIZES1]] to i8* 120 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* nonnull [[LIFETIME_PTR]]) 121 // CHECK-LIFETIMES-NEXT: getelementptr inbounds [1 x i64], [1 x i64]* %[[BLOCK_SIZES1]], i64 0, i64 0 122 // CHECK-LIFETIMES-LABEL: call i32 @__enqueue_kernel_varargs( 123 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 8, i8* nonnull [[LIFETIME_PTR]]) 124 // B32: %[[TMP:.*]] = getelementptr [1 x i32], [1 x i32]* %[[BLOCK_SIZES1]], i32 0, i32 0 125 // B32: store i32 256, i32* %[[TMP]], align 4 126 // B64: %[[TMP:.*]] = getelementptr [1 x i64], [1 x i64]* %[[BLOCK_SIZES1]], i32 0, i32 0 127 // B64: store i64 256, i64* %[[TMP]], align 8 128 // COMMON-LABEL: call i32 @__enqueue_kernel_varargs( 129 // COMMON-SAME: %opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]], %struct.ndrange_t* [[NDR]]{{([0-9]+)?}}, 130 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK1:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 131 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG1]] to i8 addrspace(1)*) to i8 addrspace(4)*), i32 1, 132 // B32-SAME: i32* %[[TMP]]) 133 // B64-SAME: i64* %[[TMP]]) 134 enqueue_kernel(default_queue, flags, ndrange, 135 ^(local void *p) { 136 return; 137 }, 138 256); 139 140 char c; 141 // Emits global block literal [[BLG2]] and block kernel [[INVGK2]]. 142 // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*, %opencl.queue_t{{.*}}** %default_queue 143 // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags 144 // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [1 x i64]* %[[BLOCK_SIZES2]] to i8* 145 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* nonnull [[LIFETIME_PTR]]) 146 // CHECK-LIFETIMES-NEXT: getelementptr inbounds [1 x i64], [1 x i64]* %[[BLOCK_SIZES2]], i64 0, i64 0 147 // CHECK-LIFETIMES-LABEL: call i32 @__enqueue_kernel_varargs( 148 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 8, i8* nonnull [[LIFETIME_PTR]]) 149 // B32: %[[TMP:.*]] = getelementptr [1 x i32], [1 x i32]* %[[BLOCK_SIZES2]], i32 0, i32 0 150 // B32: store i32 %{{.*}}, i32* %[[TMP]], align 4 151 // B64: %[[TMP:.*]] = getelementptr [1 x i64], [1 x i64]* %[[BLOCK_SIZES2]], i32 0, i32 0 152 // B64: store i64 %{{.*}}, i64* %[[TMP]], align 8 153 // COMMON-LABEL: call i32 @__enqueue_kernel_varargs( 154 // COMMON-SAME: %opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]], %struct.ndrange_t* [[NDR]]{{([0-9]+)?}}, 155 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK2:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 156 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG2]] to i8 addrspace(1)*) to i8 addrspace(4)*), i32 1, 157 // B32-SAME: i32* %[[TMP]]) 158 // B64-SAME: i64* %[[TMP]]) 159 enqueue_kernel(default_queue, flags, ndrange, 160 ^(local void *p) { 161 return; 162 }, 163 c); 164 165 // Emits global block literal [[BLG3]] and block kernel [[INVGK3]]. 166 // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*, %opencl.queue_t{{.*}}** %default_queue 167 // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags 168 // COMMON: [[AD:%arraydecay[0-9]*]] = getelementptr inbounds [1 x %opencl.clk_event_t*], [1 x %opencl.clk_event_t*]* %event_wait_list2, i{{32|64}} 0, i{{32|64}} 0 169 // COMMON: [[WAIT_EVNT:%[0-9]+]] = addrspacecast %opencl.clk_event_t{{.*}}** [[AD]] to %opencl.clk_event_t{{.*}}* addrspace(4)* 170 // COMMON: [[EVNT:%[0-9]+]] = addrspacecast %opencl.clk_event_t{{.*}}** %clk_event to %opencl.clk_event_t{{.*}}* addrspace(4)* 171 // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [1 x i64]* %[[BLOCK_SIZES3]] to i8* 172 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* nonnull [[LIFETIME_PTR]]) 173 // CHECK-LIFETIMES-NEXT: getelementptr inbounds [1 x i64], [1 x i64]* %[[BLOCK_SIZES3]], i64 0, i64 0 174 // CHECK-LIFETIMES-LABEL: call i32 @__enqueue_kernel_events_varargs( 175 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 8, i8* nonnull [[LIFETIME_PTR]]) 176 // B32: %[[TMP:.*]] = getelementptr [1 x i32], [1 x i32]* %[[BLOCK_SIZES3]], i32 0, i32 0 177 // B32: store i32 256, i32* %[[TMP]], align 4 178 // B64: %[[TMP:.*]] = getelementptr [1 x i64], [1 x i64]* %[[BLOCK_SIZES3]], i32 0, i32 0 179 // B64: store i64 256, i64* %[[TMP]], align 8 180 // COMMON-LABEL: call i32 @__enqueue_kernel_events_varargs 181 // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]], %struct.ndrange_t* {{.*}}, i32 2, %opencl.clk_event_t{{.*}} [[WAIT_EVNT]], %opencl.clk_event_t{{.*}} [[EVNT]], 182 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK3:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 183 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG3]] to i8 addrspace(1)*) to i8 addrspace(4)*), i32 1, 184 // B32-SAME: i32* %[[TMP]]) 185 // B64-SAME: i64* %[[TMP]]) 186 enqueue_kernel(default_queue, flags, ndrange, 2, event_wait_list2, &clk_event, 187 ^(local void *p) { 188 return; 189 }, 190 256); 191 192 // Emits global block literal [[BLG4]] and block kernel [[INVGK4]]. 193 // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*, %opencl.queue_t{{.*}}** %default_queue 194 // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags 195 // COMMON: [[AD:%arraydecay[0-9]*]] = getelementptr inbounds [1 x %opencl.clk_event_t*], [1 x %opencl.clk_event_t*]* %event_wait_list2, i{{32|64}} 0, i{{32|64}} 0 196 // COMMON: [[WAIT_EVNT:%[0-9]+]] = addrspacecast %opencl.clk_event_t{{.*}}** [[AD]] to %opencl.clk_event_t{{.*}}* addrspace(4)* 197 // COMMON: [[EVNT:%[0-9]+]] = addrspacecast %opencl.clk_event_t{{.*}}** %clk_event to %opencl.clk_event_t{{.*}}* addrspace(4)* 198 // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [1 x i64]* %[[BLOCK_SIZES4]] to i8* 199 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* nonnull [[LIFETIME_PTR]]) 200 // CHECK-LIFETIMES-NEXT: getelementptr inbounds [1 x i64], [1 x i64]* %[[BLOCK_SIZES4]], i64 0, i64 0 201 // CHECK-LIFETIMES-LABEL: call i32 @__enqueue_kernel_events_varargs( 202 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 8, i8* nonnull [[LIFETIME_PTR]]) 203 // B32: %[[TMP:.*]] = getelementptr [1 x i32], [1 x i32]* %[[BLOCK_SIZES4]], i32 0, i32 0 204 // B32: store i32 %{{.*}}, i32* %[[TMP]], align 4 205 // B64: %[[TMP:.*]] = getelementptr [1 x i64], [1 x i64]* %[[BLOCK_SIZES4]], i32 0, i32 0 206 // B64: store i64 %{{.*}}, i64* %[[TMP]], align 8 207 // COMMON-LABEL: call i32 @__enqueue_kernel_events_varargs 208 // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]], %struct.ndrange_t* {{.*}}, i32 2, %opencl.clk_event_t{{.*}}* addrspace(4)* [[WAIT_EVNT]], %opencl.clk_event_t{{.*}}* addrspace(4)* [[EVNT]], 209 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK4:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 210 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG4]] to i8 addrspace(1)*) to i8 addrspace(4)*), i32 1, 211 // B32-SAME: i32* %[[TMP]]) 212 // B64-SAME: i64* %[[TMP]]) 213 enqueue_kernel(default_queue, flags, ndrange, 2, event_wait_list2, &clk_event, 214 ^(local void *p) { 215 return; 216 }, 217 c); 218 219 long l; 220 // Emits global block literal [[BLG5]] and block kernel [[INVGK5]]. 221 // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*, %opencl.queue_t{{.*}}** %default_queue 222 // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags 223 // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [1 x i64]* %[[BLOCK_SIZES5]] to i8* 224 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* nonnull [[LIFETIME_PTR]]) 225 // CHECK-LIFETIMES-NEXT: getelementptr inbounds [1 x i64], [1 x i64]* %[[BLOCK_SIZES5]], i64 0, i64 0 226 // CHECK-LIFETIMES-LABEL: call i32 @__enqueue_kernel_varargs( 227 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 8, i8* nonnull [[LIFETIME_PTR]]) 228 // B32: %[[TMP:.*]] = getelementptr [1 x i32], [1 x i32]* %[[BLOCK_SIZES5]], i32 0, i32 0 229 // B32: store i32 %{{.*}}, i32* %[[TMP]], align 4 230 // B64: %[[TMP:.*]] = getelementptr [1 x i64], [1 x i64]* %[[BLOCK_SIZES5]], i32 0, i32 0 231 // B64: store i64 %{{.*}}, i64* %[[TMP]], align 8 232 // COMMON-LABEL: call i32 @__enqueue_kernel_varargs 233 // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]], %struct.ndrange_t* [[NDR]]{{([0-9]+)?}}, 234 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK5:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 235 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG5]] to i8 addrspace(1)*) to i8 addrspace(4)*), i32 1, 236 // B32-SAME: i32* %[[TMP]]) 237 // B64-SAME: i64* %[[TMP]]) 238 enqueue_kernel(default_queue, flags, ndrange, 239 ^(local void *p) { 240 return; 241 }, 242 l); 243 244 // Emits global block literal [[BLG6]] and block kernel [[INVGK6]]. 245 // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*, %opencl.queue_t{{.*}}** %default_queue 246 // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags 247 // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [3 x i64]* %[[BLOCK_SIZES6]] to i8* 248 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64 24, i8* nonnull [[LIFETIME_PTR]]) 249 // CHECK-LIFETIMES-NEXT: getelementptr inbounds [3 x i64], [3 x i64]* %[[BLOCK_SIZES6]], i64 0, i64 0 250 // CHECK-LIFETIMES-LABEL: call i32 @__enqueue_kernel_varargs( 251 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 24, i8* nonnull [[LIFETIME_PTR]]) 252 // B32: %[[TMP:.*]] = getelementptr [3 x i32], [3 x i32]* %[[BLOCK_SIZES6]], i32 0, i32 0 253 // B32: store i32 1, i32* %[[TMP]], align 4 254 // B32: %[[BLOCK_SIZES62:.*]] = getelementptr [3 x i32], [3 x i32]* %[[BLOCK_SIZES6]], i32 0, i32 1 255 // B32: store i32 2, i32* %[[BLOCK_SIZES62]], align 4 256 // B32: %[[BLOCK_SIZES63:.*]] = getelementptr [3 x i32], [3 x i32]* %[[BLOCK_SIZES6]], i32 0, i32 2 257 // B32: store i32 4, i32* %[[BLOCK_SIZES63]], align 4 258 // B64: %[[TMP:.*]] = getelementptr [3 x i64], [3 x i64]* %[[BLOCK_SIZES6]], i32 0, i32 0 259 // B64: store i64 1, i64* %[[TMP]], align 8 260 // B64: %[[BLOCK_SIZES62:.*]] = getelementptr [3 x i64], [3 x i64]* %[[BLOCK_SIZES6]], i32 0, i32 1 261 // B64: store i64 2, i64* %[[BLOCK_SIZES62]], align 8 262 // B64: %[[BLOCK_SIZES63:.*]] = getelementptr [3 x i64], [3 x i64]* %[[BLOCK_SIZES6]], i32 0, i32 2 263 // B64: store i64 4, i64* %[[BLOCK_SIZES63]], align 8 264 // COMMON-LABEL: call i32 @__enqueue_kernel_varargs 265 // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]], %struct.ndrange_t* [[NDR]]{{([0-9]+)?}}, 266 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK6:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 267 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG6]] to i8 addrspace(1)*) to i8 addrspace(4)*), i32 3, 268 // B32-SAME: i32* %[[TMP]]) 269 // B64-SAME: i64* %[[TMP]]) 270 enqueue_kernel(default_queue, flags, ndrange, 271 ^(local void *p1, local void *p2, local void *p3) { 272 return; 273 }, 274 1, 2, 4); 275 276 // Emits global block literal [[BLG7]] and block kernel [[INVGK7]]. 277 // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t*, %opencl.queue_t** %default_queue 278 // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags 279 // CHECK-LIFETIMES: [[LIFETIME_PTR:%[0-9]+]] = bitcast [1 x i64]* %[[BLOCK_SIZES7]] to i8* 280 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.start.p0i8(i64 8, i8* nonnull [[LIFETIME_PTR]]) 281 // CHECK-LIFETIMES-NEXT: getelementptr inbounds [1 x i64], [1 x i64]* %[[BLOCK_SIZES7]], i64 0, i64 0 282 // CHECK-LIFETIMES-LABEL: call i32 @__enqueue_kernel_varargs( 283 // CHECK-LIFETIMES-NEXT: call void @llvm.lifetime.end.p0i8(i64 8, i8* nonnull [[LIFETIME_PTR]]) 284 // B32: %[[TMP:.*]] = getelementptr [1 x i32], [1 x i32]* %[[BLOCK_SIZES7]], i32 0, i32 0 285 // B32: store i32 0, i32* %[[TMP]], align 4 286 // B64: %[[TMP:.*]] = getelementptr [1 x i64], [1 x i64]* %[[BLOCK_SIZES7]], i32 0, i32 0 287 // B64: store i64 4294967296, i64* %[[TMP]], align 8 288 // COMMON-LABEL: call i32 @__enqueue_kernel_varargs 289 // COMMON-SAME: (%opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]], %struct.ndrange_t* [[NDR]]{{([0-9]+)?}}, 290 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK7:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 291 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG7]] to i8 addrspace(1)*) to i8 addrspace(4)*), i32 1, 292 // B32-SAME: i32* %[[TMP]]) 293 // B64-SAME: i64* %[[TMP]]) 294 enqueue_kernel(default_queue, flags, ndrange, 295 ^(local void *p) { 296 return; 297 }, 298 4294967296L); 299 300 // Emits global block literal [[BLG8]] and invoke function [[INVG8]]. 301 // The full type of these expressions are long (and repeated elsewhere), so we 302 // capture it as part of the regex for convenience and clarity. 303 // COMMON: store %struct.__opencl_block_literal_generic addrspace(4)* addrspacecast (%struct.__opencl_block_literal_generic addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG8]] to %struct.__opencl_block_literal_generic addrspace(1)*) to %struct.__opencl_block_literal_generic addrspace(4)*), %struct.__opencl_block_literal_generic addrspace(4)** %block_A 304 void (^const block_A)(void) = ^{ 305 return; 306 }; 307 308 // Emits global block literal [[BLG9]] and invoke function [[INVG9]]. 309 // COMMON: store %struct.__opencl_block_literal_generic addrspace(4)* addrspacecast (%struct.__opencl_block_literal_generic addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG9]] to %struct.__opencl_block_literal_generic addrspace(1)*) to %struct.__opencl_block_literal_generic addrspace(4)*), %struct.__opencl_block_literal_generic addrspace(4)** %block_B 310 void (^const block_B)(local void *) = ^(local void *a) { 311 return; 312 }; 313 314 // Uses global block literal [[BLG8]] and invoke function [[INVG8]]. 315 // COMMON: call spir_func void @__device_side_enqueue_block_invoke_11(i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG8]] to i8 addrspace(1)*) to i8 addrspace(4)*)) 316 block_A(); 317 318 // Emits global block literal [[BLG8]] and block kernel [[INVGK8]]. [[INVGK8]] calls [[INVG8]]. 319 // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*, %opencl.queue_t{{.*}}** %default_queue 320 // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags 321 // COMMON-LABEL: call i32 @__enqueue_kernel_basic( 322 // COMMON-SAME: %opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]], %struct.ndrange_t* byval(%struct.ndrange_t) [[NDR]]{{([0-9]+)?}}, 323 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK8:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 324 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG8]] to i8 addrspace(1)*) to i8 addrspace(4)*)) 325 enqueue_kernel(default_queue, flags, ndrange, block_A); 326 327 // Uses block kernel [[INVGK8]] and global block literal [[BLG8]]. 328 // COMMON: call i32 @__get_kernel_work_group_size_impl( 329 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK8]] to i8*) to i8 addrspace(4)*), 330 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG8]] to i8 addrspace(1)*) to i8 addrspace(4)*)) 331 unsigned size = get_kernel_work_group_size(block_A); 332 333 // Uses global block literal [[BLG8]] and invoke function [[INVG8]]. Make sure no redundant block literal and invoke functions are emitted. 334 // COMMON: call spir_func void @__device_side_enqueue_block_invoke_11(i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG8]] to i8 addrspace(1)*) to i8 addrspace(4)*)) 335 block_A(); 336 337 // Make sure that block invoke function is resolved correctly after sequence of assignements. 338 // COMMON: store %struct.__opencl_block_literal_generic addrspace(4)* 339 // COMMON-SAME: addrspacecast (%struct.__opencl_block_literal_generic addrspace(1)* 340 // COMMON-SAME: bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BL_GLOBAL]] to %struct.__opencl_block_literal_generic addrspace(1)*) 341 // COMMON-SAME: to %struct.__opencl_block_literal_generic addrspace(4)*), 342 // COMMON-SAME: %struct.__opencl_block_literal_generic addrspace(4)** %b1, 343 bl_t b1 = block_G; 344 // COMMON: store %struct.__opencl_block_literal_generic addrspace(4)* 345 // COMMON-SAME: addrspacecast (%struct.__opencl_block_literal_generic addrspace(1)* 346 // COMMON-SAME: bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BL_GLOBAL]] to %struct.__opencl_block_literal_generic addrspace(1)*) 347 // COMMON-SAME: to %struct.__opencl_block_literal_generic addrspace(4)*), 348 // COMMON-SAME: %struct.__opencl_block_literal_generic addrspace(4)** %b2, 349 bl_t b2 = b1; 350 // COMMON: call spir_func void @block_G_block_invoke(i8 addrspace(4)* addrspacecast (i8 addrspace(1)* 351 // COMMON-SAME: bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BL_GLOBAL]] to i8 addrspace(1)*) 352 // COOMON-SAME: to i8 addrspace(4)*), i8 addrspace(3)* null) 353 b2(0); 354 // Uses global block literal [[BL_GLOBAL]] and block kernel [[INV_G_K]]. [[INV_G_K]] calls [[INV_G]]. 355 // COMMON: call i32 @__get_kernel_preferred_work_group_size_multiple_impl( 356 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INV_G_K:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 357 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BL_GLOBAL]] to i8 addrspace(1)*) to i8 addrspace(4)*)) 358 size = get_kernel_preferred_work_group_size_multiple(b2); 359 360 void (^block_C)(void) = ^{ 361 callee(i, a); 362 }; 363 // Emits block literal on stack and block kernel [[INVLK3]]. 364 // COMMON: store i8 addrspace(4)* addrspacecast (i8* bitcast (void (i8 addrspace(4)*)* [[INVL3:@__device_side_enqueue_block_invoke[^ ]*]] to i8*) to i8 addrspace(4)*), i8 addrspace(4)** %block.invoke 365 // COMMON: [[DEF_Q:%[0-9]+]] = load %opencl.queue_t{{.*}}*, %opencl.queue_t{{.*}}** %default_queue 366 // COMMON: [[FLAGS:%[0-9]+]] = load i32, i32* %flags 367 // COMMON: [[BL_I8:%[0-9]+]] = addrspacecast %struct.__opencl_block_literal_generic* {{.*}} to i8 addrspace(4)* 368 // COMMON-LABEL: call i32 @__enqueue_kernel_basic( 369 // COMMON-SAME: %opencl.queue_t{{.*}}* [[DEF_Q]], i32 [[FLAGS]], %struct.ndrange_t* byval(%struct.ndrange_t) [[NDR]]{{([0-9]+)?}}, 370 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVLK3:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 371 // COMMON-SAME: i8 addrspace(4)* [[BL_I8]]) 372 enqueue_kernel(default_queue, flags, ndrange, block_C); 373 374 // Emits global block literal [[BLG9]] and block kernel [[INVGK9]]. [[INVGK9]] calls [[INV9]]. 375 // COMMON: call i32 @__get_kernel_work_group_size_impl( 376 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK9:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 377 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG9]] to i8 addrspace(1)*) to i8 addrspace(4)*)) 378 size = get_kernel_work_group_size(block_B); 379 380 // Uses global block literal [[BLG8]] and block kernel [[INVGK8]]. Make sure no redundant block literal ind invoke functions are emitted. 381 // COMMON: call i32 @__get_kernel_preferred_work_group_size_multiple_impl( 382 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK8]] to i8*) to i8 addrspace(4)*), 383 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG8]] to i8 addrspace(1)*) to i8 addrspace(4)*)) 384 size = get_kernel_preferred_work_group_size_multiple(block_A); 385 386 // Uses global block literal [[BL_GLOBAL]] and block kernel [[INV_G_K]]. [[INV_G_K]] calls [[INV_G]]. 387 // COMMON: call i32 @__get_kernel_preferred_work_group_size_multiple_impl( 388 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INV_G_K:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 389 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BL_GLOBAL]] to i8 addrspace(1)*) to i8 addrspace(4)*)) 390 size = get_kernel_preferred_work_group_size_multiple(block_G); 391 392 // Emits global block literal [[BLG10]] and block kernel [[INVGK10]]. 393 // COMMON: call i32 @__get_kernel_max_sub_group_size_for_ndrange_impl(%struct.ndrange_t* {{[^,]+}}, 394 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK10:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 395 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG10]] to i8 addrspace(1)*) to i8 addrspace(4)*)) 396 size = get_kernel_max_sub_group_size_for_ndrange(ndrange, ^(){}); 397 398 // Emits global block literal [[BLG11]] and block kernel [[INVGK11]]. 399 // COMMON: call i32 @__get_kernel_sub_group_count_for_ndrange_impl(%struct.ndrange_t* {{[^,]+}}, 400 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8* bitcast ({{.*}} [[INVGK11:[^ ]+_kernel]] to i8*) to i8 addrspace(4)*), 401 // COMMON-SAME: i8 addrspace(4)* addrspacecast (i8 addrspace(1)* bitcast ({ i32, i32, i8 addrspace(4)* } addrspace(1)* [[BLG11]] to i8 addrspace(1)*) to i8 addrspace(4)*)) 402 size = get_kernel_sub_group_count_for_ndrange(ndrange, ^(){}); 403} 404 405// COMMON: define internal spir_kernel void [[INVLK1]](i8 addrspace(4)* %0) #{{[0-9]+}} { 406// COMMON: entry: 407// COMMON: call void @__device_side_enqueue_block_invoke(i8 addrspace(4)* %0) 408// COMMON: ret void 409// COMMON: } 410// COMMON: define internal spir_kernel void [[INVLK2]](i8 addrspace(4)*{{.*}}) 411// COMMON: define internal spir_kernel void [[INVGK1]](i8 addrspace(4)*{{.*}}, i8 addrspace(3)*{{.*}}) 412// COMMON: define internal spir_kernel void [[INVGK2]](i8 addrspace(4)*{{.*}}, i8 addrspace(3)*{{.*}}) 413// COMMON: define internal spir_kernel void [[INVGK3]](i8 addrspace(4)*{{.*}}, i8 addrspace(3)*{{.*}}) 414// COMMON: define internal spir_kernel void [[INVGK4]](i8 addrspace(4)*{{.*}}, i8 addrspace(3)*{{.*}}) 415// COMMON: define internal spir_kernel void [[INVGK5]](i8 addrspace(4)*{{.*}}, i8 addrspace(3)*{{.*}}) 416// COMMON: define internal spir_kernel void [[INVGK6]](i8 addrspace(4)* %0, i8 addrspace(3)* %1, i8 addrspace(3)* %2, i8 addrspace(3)* %3) #{{[0-9]+}} { 417// COMMON: entry: 418// COMMON: call void @__device_side_enqueue_block_invoke_9(i8 addrspace(4)* %0, i8 addrspace(3)* %1, i8 addrspace(3)* %2, i8 addrspace(3)* %3) 419// COMMON: ret void 420// COMMON: } 421// COMMON: define internal spir_kernel void [[INVGK7]](i8 addrspace(4)*{{.*}}, i8 addrspace(3)*{{.*}}) 422// COMMON: define internal spir_func void [[INVG8]](i8 addrspace(4)*{{.*}}) 423// COMMON: define internal spir_func void [[INVG9]](i8 addrspace(4)*{{.*}}, i8 addrspace(3)* %{{.*}}) 424// COMMON: define internal spir_kernel void [[INVGK8]](i8 addrspace(4)*{{.*}}) 425// COMMON: define internal spir_kernel void [[INV_G_K]](i8 addrspace(4)*{{.*}}, i8 addrspace(3)*{{.*}}) 426// COMMON: define internal spir_kernel void [[INVLK3]](i8 addrspace(4)*{{.*}}) 427// COMMON: define internal spir_kernel void [[INVGK9]](i8 addrspace(4)*{{.*}}, i8 addrspace(3)*{{.*}}) 428// COMMON: define internal spir_kernel void [[INVGK10]](i8 addrspace(4)*{{.*}}) 429// COMMON: define internal spir_kernel void [[INVGK11]](i8 addrspace(4)*{{.*}}) 430