| ; RUN: opt -mtriple=amdgpu7.00-amd-amdhsa -passes=amdgpu-attributor -o %t.gfx7.bc %s | |
| ; RUN: opt -mtriple=amdgpu8.03-amd-amdhsa -passes=amdgpu-attributor -o %t.gfx8.bc %s | |
| ; RUN: opt -mtriple=amdgpu9.00-amd-amdhsa -passes=amdgpu-attributor -o %t.gfx9.bc %s | |
| ; RUN: llc -mtriple=amdgpu7.00-amd-amdhsa < %t.gfx7.bc | FileCheck --check-prefixes=CHECK,PRE-GFX9 %s | |
| ; RUN: llc -mtriple=amdgpu8.03-amd-amdhsa < %t.gfx8.bc | FileCheck --check-prefixes=CHECK,PRE-GFX9 %s | |
| ; RUN: llc -mtriple=amdgpu9.00-amd-amdhsa < %t.gfx9.bc | FileCheck --check-prefixes=CHECK,GFX9 %s | |
| ; CHECK: addrspacecast_requires_queue_ptr | |
| ; PRE-GFX9: .amdhsa_user_sgpr_queue_ptr 1 | |
| ; GFX9: .amdhsa_user_sgpr_queue_ptr 0 | |
| ; CHECK: is_shared_requires_queue_ptr | |
| ; PRE-GFX9: .amdhsa_user_sgpr_queue_ptr 1 | |
| ; GFX9: .amdhsa_user_sgpr_queue_ptr 0 | |
| ; CHECK: is_private_requires_queue_ptr | |
| ; PRE-GFX9: .amdhsa_user_sgpr_queue_ptr 1 | |
| ; GFX9: .amdhsa_user_sgpr_queue_ptr 0 | |
| ; CHECK: trap_requires_queue_ptr | |
| ; PRE-GFX9: .amdhsa_user_sgpr_queue_ptr 1 | |
| ; GFX9: .amdhsa_user_sgpr_queue_ptr 0 | |
| ; CHECK: debugtrap_requires_queue_ptr | |
| ; PRE-GFX9: .amdhsa_user_sgpr_queue_ptr 1 | |
| ; GFX9: .amdhsa_user_sgpr_queue_ptr 0 | |
| ; CHECK: ubsantrap_requires_queue_ptr | |
| ; PRE-GFX9: .amdhsa_user_sgpr_queue_ptr 1 | |
| ; GFX9: .amdhsa_user_sgpr_queue_ptr 0 | |
| ; CHECK: amdgcn_queue_ptr_requires_queue_ptr | |
| ; CHECK: .amdhsa_user_sgpr_queue_ptr 1 | |
| ; On gfx8, the queue ptr is required for this addrspacecast. | |
| ; CHECK: - .args: | |
| ; CHECK-NOT: hidden_queue_ptr | |
| ; CHECK-LABEL: .name: addrspacecast_requires_queue_ptr | |
| define amdgpu_kernel void @addrspacecast_requires_queue_ptr(ptr addrspace(5) %ptr.private, ptr addrspace(3) %ptr.local) { | |
| %flat.private = addrspacecast ptr addrspace(5) %ptr.private to ptr | |
| %flat.local = addrspacecast ptr addrspace(3) %ptr.local to ptr | |
| store volatile i32 1, ptr %flat.private | |
| store volatile i32 2, ptr %flat.local | |
| ret void | |
| } | |
| ; CHECK: - .args: | |
| ; CHECK-NOT: hidden_shared_base | |
| ; CHECK-LABEL: .name: is_shared_requires_queue_ptr | |
| define amdgpu_kernel void @is_shared_requires_queue_ptr(ptr %ptr) { | |
| %is.shared = call i1 @llvm.amdgcn.is.shared(ptr %ptr) | |
| %zext = zext i1 %is.shared to i32 | |
| store volatile i32 %zext, ptr addrspace(1) poison | |
| ret void | |
| } | |
| ; CHECK: - .args: | |
| ; CHECK-NOT: hidden_shared_base | |
| ; CHECK-NOT: hidden_private_base | |
| ; CHECK-NOT: hidden_queue_ptr | |
| ; CHECK-LABEL: .name: is_private_requires_queue_ptr | |
| define amdgpu_kernel void @is_private_requires_queue_ptr(ptr %ptr) { | |
| %is.private = call i1 @llvm.amdgcn.is.private(ptr %ptr) | |
| %zext = zext i1 %is.private to i32 | |
| store volatile i32 %zext, ptr addrspace(1) poison | |
| ret void | |
| } | |
| ; CHECK: - .args: | |
| ; CHECK-NOT: hidden_shared_base | |
| ; CHECK-NOT: hidden_private_base | |
| ; CHECK-NOT: hidden_queue_ptr | |
| ; CHECK-LABEL: .name: trap_requires_queue_ptr | |
| define amdgpu_kernel void @trap_requires_queue_ptr() { | |
| call void @llvm.trap() | |
| unreachable | |
| } | |
| ; CHECK: - .args: | |
| ; CHECK-NOT: hidden_shared_base | |
| ; CHECK-NOT: hidden_private_base | |
| ; CHECK-NOT: hidden_queue_ptr | |
| ; CHECK-LABEL: .name: debugtrap_requires_queue_ptr | |
| define amdgpu_kernel void @debugtrap_requires_queue_ptr() { | |
| call void @llvm.debugtrap() | |
| unreachable | |
| } | |
| ; CHECK: - .args: | |
| ; CHECK-NOT: hidden_shared_base | |
| ; CHECK-NOT: hidden_private_base | |
| ; CHECK-NOT: hidden_queue_ptr | |
| ; CHECK-LABEL: .name: ubsantrap_requires_queue_ptr | |
| define amdgpu_kernel void @ubsantrap_requires_queue_ptr() { | |
| call void @llvm.ubsantrap(i8 0) | |
| unreachable | |
| } | |
| ; CHECK: - .args: | |
| ; CHECK-NOT: hidden_queue_ptr | |
| ; CHECK-NOT: hidden_shared_base | |
| ; CHECK-NOT: hidden_private_base | |
| ; CHECK-LABEL: .name: amdgcn_queue_ptr_requires_queue_ptr | |
| define amdgpu_kernel void @amdgcn_queue_ptr_requires_queue_ptr(ptr addrspace(1) %ptr) { | |
| %queue.ptr = call ptr addrspace(4) @llvm.amdgcn.queue.ptr() | |
| %implicitarg.ptr = call ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr() | |
| %dispatch.ptr = call ptr addrspace(4) @llvm.amdgcn.dispatch.ptr() | |
| %dispatch.id = call i64 @llvm.amdgcn.dispatch.id() | |
| %queue.load = load volatile i8, ptr addrspace(4) %queue.ptr | |
| %implicitarg.load = load volatile i8, ptr addrspace(4) %implicitarg.ptr | |
| %dispatch.load = load volatile i8, ptr addrspace(4) %dispatch.ptr | |
| store volatile i64 %dispatch.id, ptr addrspace(1) %ptr | |
| ret void | |
| } | |
| declare noalias ptr addrspace(4) @llvm.amdgcn.queue.ptr() | |
| declare noalias ptr addrspace(4) @llvm.amdgcn.implicitarg.ptr() | |
| declare i64 @llvm.amdgcn.dispatch.id() | |
| declare noalias ptr addrspace(4) @llvm.amdgcn.dispatch.ptr() | |
| declare i1 @llvm.amdgcn.is.shared(ptr) | |
| declare i1 @llvm.amdgcn.is.private(ptr) | |
| declare void @llvm.trap() | |
| declare void @llvm.debugtrap() | |
| declare void @llvm.ubsantrap(i8 immarg) | |
| !llvm.module.flags = !{!0} | |
| !0 = !{i32 1, !"amdhsa_code_object_version", i32 400} |