[clang] clang/AMDGPU: Add missing field to implicit kernarg struct definition (PR #185209)
via cfe-commits
cfe-commits at lists.llvm.org
Sat Mar 7 09:35:07 PST 2026
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-x86
@llvm/pr-subscribers-clang
Author: Matt Arsenault (arsenm)
<details>
<summary>Changes</summary>
I missed this one when figuring out all the field offsets.
---
Patch is 24.10 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/185209.diff
2 Files Affected:
- (modified) clang/lib/Headers/amdhsa_abi.h (+4-1)
- (modified) clang/test/Headers/amdhsa_abi.cl (+76-70)
``````````diff
diff --git a/clang/lib/Headers/amdhsa_abi.h b/clang/lib/Headers/amdhsa_abi.h
index 7dcf5b7f8624a..f76fe55a6cc21 100644
--- a/clang/lib/Headers/amdhsa_abi.h
+++ b/clang/lib/Headers/amdhsa_abi.h
@@ -19,7 +19,8 @@ typedef struct __attribute__((aligned(8))) amdhsa_implicit_kernarg_v5 {
char reserved0[16];
uint64_t global_offset[3];
uint16_t grid_dims;
- char reserved1[14];
+ char reserved1[6];
+ __attribute__((opencl_global)) void *printf_buffer;
__attribute__((opencl_global)) void *hostcall_buffer;
__attribute__((opencl_global)) void *multigrid_sync_arg;
__attribute__((opencl_global)) void *heap_v1;
@@ -60,6 +61,8 @@ _Static_assert(offsetof(amdhsa_implicit_kernarg_v5, global_offset[2]) == 56,
"wrong offset");
_Static_assert(offsetof(amdhsa_implicit_kernarg_v5, grid_dims) == 64,
"wrong offset");
+_Static_assert(offsetof(amdhsa_implicit_kernarg_v5, printf_buffer) == 72,
+ "wrong offset");
_Static_assert(offsetof(amdhsa_implicit_kernarg_v5, hostcall_buffer) == 80,
"wrong offset");
_Static_assert(offsetof(amdhsa_implicit_kernarg_v5, multigrid_sync_arg) == 88,
diff --git a/clang/test/Headers/amdhsa_abi.cl b/clang/test/Headers/amdhsa_abi.cl
index 170ae925c8c41..258a08d2f1843 100644
--- a/clang/test/Headers/amdhsa_abi.cl
+++ b/clang/test/Headers/amdhsa_abi.cl
@@ -17,99 +17,103 @@ void use_ptr(__global void* p);
// CHECK-NEXT: [[IMPLICIT_ARGS:%.*]] = alloca ptr addrspace(4), align 8, addrspace(5)
// CHECK-NEXT: call void @llvm.lifetime.start.p5(ptr addrspace(5) [[IMPLICIT_ARGS]]) #[[ATTR4:[0-9]+]]
// CHECK-NEXT: [[TMP0:%.*]] = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr()
-// CHECK-NEXT: store ptr addrspace(4) [[TMP0]], ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8:![0-9]+]]
-// CHECK-NEXT: [[TMP1:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
+// CHECK-NEXT: store ptr addrspace(4) [[TMP0]], ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7:![0-9]+]]
+// CHECK-NEXT: [[TMP1:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
// CHECK-NEXT: [[BLOCK_COUNT:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5:%.*]], ptr addrspace(4) [[TMP1]], i32 0, i32 0
// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x i32], ptr addrspace(4) [[BLOCK_COUNT]], i64 0, i64 0
-// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(4) [[ARRAYIDX]], align 8, !tbaa [[INT_TBAA4:![0-9]+]]
+// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(4) [[ARRAYIDX]], align 8, !tbaa [[INT_TBAA3:![0-9]+]]
// CHECK-NEXT: call void @use_u32(i32 noundef [[TMP2]]) #[[ATTR5:[0-9]+]]
-// CHECK-NEXT: [[TMP3:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
+// CHECK-NEXT: [[TMP3:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
// CHECK-NEXT: [[BLOCK_COUNT1:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP3]], i32 0, i32 0
// CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds [3 x i32], ptr addrspace(4) [[BLOCK_COUNT1]], i64 0, i64 1
-// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(4) [[ARRAYIDX2]], align 4, !tbaa [[INT_TBAA4]]
+// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(4) [[ARRAYIDX2]], align 4, !tbaa [[INT_TBAA3]]
// CHECK-NEXT: call void @use_u32(i32 noundef [[TMP4]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP5:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
+// CHECK-NEXT: [[TMP5:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
// CHECK-NEXT: [[BLOCK_COUNT3:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP5]], i32 0, i32 0
// CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds [3 x i32], ptr addrspace(4) [[BLOCK_COUNT3]], i64 0, i64 2
-// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr addrspace(4) [[ARRAYIDX4]], align 8, !tbaa [[INT_TBAA4]]
+// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr addrspace(4) [[ARRAYIDX4]], align 8, !tbaa [[INT_TBAA3]]
// CHECK-NEXT: call void @use_u32(i32 noundef [[TMP6]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP7:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
+// CHECK-NEXT: [[TMP7:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
// CHECK-NEXT: [[GROUP_SIZE:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP7]], i32 0, i32 1
// CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds [3 x i16], ptr addrspace(4) [[GROUP_SIZE]], i64 0, i64 0
-// CHECK-NEXT: [[TMP8:%.*]] = load i16, ptr addrspace(4) [[ARRAYIDX5]], align 4, !tbaa [[SHORT_TBAA11:![0-9]+]]
+// CHECK-NEXT: [[TMP8:%.*]] = load i16, ptr addrspace(4) [[ARRAYIDX5]], align 4, !tbaa [[SHORT_TBAA10:![0-9]+]]
// CHECK-NEXT: call void @use_u16(i16 noundef zeroext [[TMP8]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP9:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
+// CHECK-NEXT: [[TMP9:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
// CHECK-NEXT: [[GROUP_SIZE6:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP9]], i32 0, i32 1
// CHECK-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds [3 x i16], ptr addrspace(4) [[GROUP_SIZE6]], i64 0, i64 1
-// CHECK-NEXT: [[TMP10:%.*]] = load i16, ptr addrspace(4) [[ARRAYIDX7]], align 2, !tbaa [[SHORT_TBAA11]]
+// CHECK-NEXT: [[TMP10:%.*]] = load i16, ptr addrspace(4) [[ARRAYIDX7]], align 2, !tbaa [[SHORT_TBAA10]]
// CHECK-NEXT: call void @use_u16(i16 noundef zeroext [[TMP10]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP11:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
+// CHECK-NEXT: [[TMP11:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
// CHECK-NEXT: [[GROUP_SIZE8:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP11]], i32 0, i32 1
// CHECK-NEXT: [[ARRAYIDX9:%.*]] = getelementptr inbounds [3 x i16], ptr addrspace(4) [[GROUP_SIZE8]], i64 0, i64 2
-// CHECK-NEXT: [[TMP12:%.*]] = load i16, ptr addrspace(4) [[ARRAYIDX9]], align 4, !tbaa [[SHORT_TBAA11]]
+// CHECK-NEXT: [[TMP12:%.*]] = load i16, ptr addrspace(4) [[ARRAYIDX9]], align 4, !tbaa [[SHORT_TBAA10]]
// CHECK-NEXT: call void @use_u16(i16 noundef zeroext [[TMP12]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP13:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
+// CHECK-NEXT: [[TMP13:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
// CHECK-NEXT: [[REMAINDER:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP13]], i32 0, i32 2
// CHECK-NEXT: [[ARRAYIDX10:%.*]] = getelementptr inbounds [3 x i16], ptr addrspace(4) [[REMAINDER]], i64 0, i64 0
-// CHECK-NEXT: [[TMP14:%.*]] = load i16, ptr addrspace(4) [[ARRAYIDX10]], align 2, !tbaa [[SHORT_TBAA11]]
+// CHECK-NEXT: [[TMP14:%.*]] = load i16, ptr addrspace(4) [[ARRAYIDX10]], align 2, !tbaa [[SHORT_TBAA10]]
// CHECK-NEXT: call void @use_u16(i16 noundef zeroext [[TMP14]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP15:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
+// CHECK-NEXT: [[TMP15:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
// CHECK-NEXT: [[REMAINDER11:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP15]], i32 0, i32 2
// CHECK-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds [3 x i16], ptr addrspace(4) [[REMAINDER11]], i64 0, i64 1
-// CHECK-NEXT: [[TMP16:%.*]] = load i16, ptr addrspace(4) [[ARRAYIDX12]], align 2, !tbaa [[SHORT_TBAA11]]
+// CHECK-NEXT: [[TMP16:%.*]] = load i16, ptr addrspace(4) [[ARRAYIDX12]], align 2, !tbaa [[SHORT_TBAA10]]
// CHECK-NEXT: call void @use_u16(i16 noundef zeroext [[TMP16]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP17:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
+// CHECK-NEXT: [[TMP17:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
// CHECK-NEXT: [[REMAINDER13:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP17]], i32 0, i32 2
// CHECK-NEXT: [[ARRAYIDX14:%.*]] = getelementptr inbounds [3 x i16], ptr addrspace(4) [[REMAINDER13]], i64 0, i64 2
-// CHECK-NEXT: [[TMP18:%.*]] = load i16, ptr addrspace(4) [[ARRAYIDX14]], align 2, !tbaa [[SHORT_TBAA11]]
+// CHECK-NEXT: [[TMP18:%.*]] = load i16, ptr addrspace(4) [[ARRAYIDX14]], align 2, !tbaa [[SHORT_TBAA10]]
// CHECK-NEXT: call void @use_u16(i16 noundef zeroext [[TMP18]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP19:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
+// CHECK-NEXT: [[TMP19:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
// CHECK-NEXT: [[GLOBAL_OFFSET:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP19]], i32 0, i32 4
// CHECK-NEXT: [[ARRAYIDX15:%.*]] = getelementptr inbounds [3 x i64], ptr addrspace(4) [[GLOBAL_OFFSET]], i64 0, i64 0
-// CHECK-NEXT: [[TMP20:%.*]] = load i64, ptr addrspace(4) [[ARRAYIDX15]], align 8, !tbaa [[LONG_TBAA13:![0-9]+]]
+// CHECK-NEXT: [[TMP20:%.*]] = load i64, ptr addrspace(4) [[ARRAYIDX15]], align 8, !tbaa [[LONG_TBAA12:![0-9]+]]
// CHECK-NEXT: call void @use_u64(i64 noundef [[TMP20]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP21:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
+// CHECK-NEXT: [[TMP21:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
// CHECK-NEXT: [[GLOBAL_OFFSET16:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP21]], i32 0, i32 4
// CHECK-NEXT: [[ARRAYIDX17:%.*]] = getelementptr inbounds [3 x i64], ptr addrspace(4) [[GLOBAL_OFFSET16]], i64 0, i64 1
-// CHECK-NEXT: [[TMP22:%.*]] = load i64, ptr addrspace(4) [[ARRAYIDX17]], align 8, !tbaa [[LONG_TBAA13]]
+// CHECK-NEXT: [[TMP22:%.*]] = load i64, ptr addrspace(4) [[ARRAYIDX17]], align 8, !tbaa [[LONG_TBAA12]]
// CHECK-NEXT: call void @use_u64(i64 noundef [[TMP22]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP23:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
+// CHECK-NEXT: [[TMP23:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
// CHECK-NEXT: [[GLOBAL_OFFSET18:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP23]], i32 0, i32 4
// CHECK-NEXT: [[ARRAYIDX19:%.*]] = getelementptr inbounds [3 x i64], ptr addrspace(4) [[GLOBAL_OFFSET18]], i64 0, i64 2
-// CHECK-NEXT: [[TMP24:%.*]] = load i64, ptr addrspace(4) [[ARRAYIDX19]], align 8, !tbaa [[LONG_TBAA13]]
+// CHECK-NEXT: [[TMP24:%.*]] = load i64, ptr addrspace(4) [[ARRAYIDX19]], align 8, !tbaa [[LONG_TBAA12]]
// CHECK-NEXT: call void @use_u64(i64 noundef [[TMP24]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP25:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
-// CHECK-NEXT: [[HOSTCALL_BUFFER:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP25]], i32 0, i32 7
-// CHECK-NEXT: [[TMP26:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[HOSTCALL_BUFFER]], align 8, !tbaa [[ANYPTR_TBAA15:![0-9]+]]
+// CHECK-NEXT: [[TMP25:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
+// CHECK-NEXT: [[PRINTF_BUFFER:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP25]], i32 0, i32 7
+// CHECK-NEXT: [[TMP26:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[PRINTF_BUFFER]], align 8, !tbaa [[ANYPTR_TBAA14:![0-9]+]]
// CHECK-NEXT: call void @use_ptr(ptr addrspace(1) noundef [[TMP26]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP27:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
-// CHECK-NEXT: [[MULTIGRID_SYNC_ARG:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP27]], i32 0, i32 8
-// CHECK-NEXT: [[TMP28:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[MULTIGRID_SYNC_ARG]], align 8, !tbaa [[ANYPTR_TBAA17:![0-9]+]]
+// CHECK-NEXT: [[TMP27:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
+// CHECK-NEXT: [[HOSTCALL_BUFFER:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP27]], i32 0, i32 8
+// CHECK-NEXT: [[TMP28:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[HOSTCALL_BUFFER]], align 8, !tbaa [[ANYPTR_TBAA16:![0-9]+]]
// CHECK-NEXT: call void @use_ptr(ptr addrspace(1) noundef [[TMP28]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP29:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
-// CHECK-NEXT: [[HEAP_V1:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP29]], i32 0, i32 9
-// CHECK-NEXT: [[TMP30:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[HEAP_V1]], align 8, !tbaa [[ANYPTR_TBAA18:![0-9]+]]
+// CHECK-NEXT: [[TMP29:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
+// CHECK-NEXT: [[MULTIGRID_SYNC_ARG:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP29]], i32 0, i32 9
+// CHECK-NEXT: [[TMP30:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[MULTIGRID_SYNC_ARG]], align 8, !tbaa [[ANYPTR_TBAA17:![0-9]+]]
// CHECK-NEXT: call void @use_ptr(ptr addrspace(1) noundef [[TMP30]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP31:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
-// CHECK-NEXT: [[DEFAULT_QUEUE:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP31]], i32 0, i32 10
-// CHECK-NEXT: [[TMP32:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[DEFAULT_QUEUE]], align 8, !tbaa [[ANYPTR_TBAA19:![0-9]+]]
+// CHECK-NEXT: [[TMP31:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
+// CHECK-NEXT: [[HEAP_V1:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP31]], i32 0, i32 10
+// CHECK-NEXT: [[TMP32:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[HEAP_V1]], align 8, !tbaa [[ANYPTR_TBAA18:![0-9]+]]
// CHECK-NEXT: call void @use_ptr(ptr addrspace(1) noundef [[TMP32]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP33:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
-// CHECK-NEXT: [[COMPLETION_ACTION:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP33]], i32 0, i32 11
-// CHECK-NEXT: [[TMP34:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[COMPLETION_ACTION]], align 8, !tbaa [[ANYPTR_TBAA20:![0-9]+]]
+// CHECK-NEXT: [[TMP33:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
+// CHECK-NEXT: [[DEFAULT_QUEUE:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP33]], i32 0, i32 11
+// CHECK-NEXT: [[TMP34:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[DEFAULT_QUEUE]], align 8, !tbaa [[ANYPTR_TBAA19:![0-9]+]]
// CHECK-NEXT: call void @use_ptr(ptr addrspace(1) noundef [[TMP34]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP35:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
-// CHECK-NEXT: [[QUEUE_PTR:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP35]], i32 0, i32 15
-// CHECK-NEXT: [[TMP36:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[QUEUE_PTR]], align 8, !tbaa [[ANYPTR_TBAA21:![0-9]+]]
+// CHECK-NEXT: [[TMP35:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
+// CHECK-NEXT: [[COMPLETION_ACTION:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP35]], i32 0, i32 12
+// CHECK-NEXT: [[TMP36:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[COMPLETION_ACTION]], align 8, !tbaa [[ANYPTR_TBAA20:![0-9]+]]
// CHECK-NEXT: call void @use_ptr(ptr addrspace(1) noundef [[TMP36]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP37:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
-// CHECK-NEXT: [[PRIVATE_BASE:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP37]], i32 0, i32 13
-// CHECK-NEXT: [[TMP38:%.*]] = load i32, ptr addrspace(4) [[PRIVATE_BASE]], align 8, !tbaa [[INT_TBAA22:![0-9]+]]
-// CHECK-NEXT: call void @use_u32(i32 noundef [[TMP38]]) #[[ATTR5]]
-// CHECK-NEXT: [[TMP39:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA8]]
-// CHECK-NEXT: [[SHARED_BASE:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP39]], i32 0, i32 14
-// CHECK-NEXT: [[TMP40:%.*]] = load i32, ptr addrspace(4) [[SHARED_BASE]], align 4, !tbaa [[INT_TBAA23:![0-9]+]]
+// CHECK-NEXT: [[TMP37:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
+// CHECK-NEXT: [[QUEUE_PTR:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP37]], i32 0, i32 16
+// CHECK-NEXT: [[TMP38:%.*]] = load ptr addrspace(1), ptr addrspace(4) [[QUEUE_PTR]], align 8, !tbaa [[ANYPTR_TBAA21:![0-9]+]]
+// CHECK-NEXT: call void @use_ptr(ptr addrspace(1) noundef [[TMP38]]) #[[ATTR5]]
+// CHECK-NEXT: [[TMP39:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
+// CHECK-NEXT: [[PRIVATE_BASE:%.*]] = getelementptr inbounds nuw [[STRUCT_AMDHSA_IMPLICIT_KERNARG_V5]], ptr addrspace(4) [[TMP39]], i32 0, i32 14
+// CHECK-NEXT: [[TMP40:%.*]] = load i32, ptr addrspace(4) [[PRIVATE_BASE]], align 8, !tbaa [[INT_TBAA22:![0-9]+]]
// CHECK-NEXT: call void @use_u32(i32 noundef [[TMP40]]) #[[ATTR5]]
+// CHECK-NEXT: [[TMP41:%.*]] = load ptr addrspace(4), ptr addrspace(5) [[IMPLICIT_ARGS]], align 8, !tbaa [[_ZTS26AMDHSA_IMPLICIT_KERNARG_V5PTR_TBAA7]]
+// CHECK-NEXT: [[SHARED_BASE:%.*]] = getelementptr inbounds nuw [[STRUCT...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/185209
More information about the cfe-commits
mailing list