brintos

brintos / llvm-project-archived public Read only

0
0
Text · 2.6 KiB · 0dafb44 Raw
46 lines · plain
1// RUN: %clang_cc1 -cl-std=CL2.0 -O0 -triple=amdgcn-amd-amdhsa -target-cpu gfx90a \2// RUN:     -Rpass=si-lower -munsafe-fp-atomics %s -emit-llvm -o - 2>&1 | \3// RUN:     FileCheck %s --check-prefix=GFX90A-HW4 5// RUN: %clang_cc1 -cl-std=CL2.0 -O0 -triple=amdgcn-amd-amdhsa -target-cpu gfx90a \6// RUN:     -Rpass=si-lower -munsafe-fp-atomics %s -S -o - 2>&1 | \7// RUN:     FileCheck %s --check-prefix=GFX90A-HW-REMARK8 9 10// REQUIRES: amdgpu-registered-target11 12typedef enum memory_order {13  memory_order_relaxed = __ATOMIC_RELAXED,14  memory_order_acquire = __ATOMIC_ACQUIRE,15  memory_order_release = __ATOMIC_RELEASE,16  memory_order_acq_rel = __ATOMIC_ACQ_REL,17  memory_order_seq_cst = __ATOMIC_SEQ_CST18} memory_order;19 20typedef enum memory_scope {21  memory_scope_work_item = __OPENCL_MEMORY_SCOPE_WORK_ITEM,22  memory_scope_work_group = __OPENCL_MEMORY_SCOPE_WORK_GROUP,23  memory_scope_device = __OPENCL_MEMORY_SCOPE_DEVICE,24  memory_scope_all_svm_devices = __OPENCL_MEMORY_SCOPE_ALL_SVM_DEVICES,25#if defined(cl_intel_subgroups) || defined(cl_khr_subgroups)26  memory_scope_sub_group = __OPENCL_MEMORY_SCOPE_SUB_GROUP27#endif28} memory_scope;29 30// GFX90A-HW-REMARK: Hardware instruction generated for atomic fadd operation at memory scope wavefront-one-as due to an unsafe request. [-Rpass=si-lower]31// GFX90A-HW-REMARK: Hardware instruction generated for atomic fadd operation at memory scope agent-one-as due to an unsafe request. [-Rpass=si-lower]32// GFX90A-HW-REMARK: Hardware instruction generated for atomic fadd operation at memory scope workgroup-one-as due to an unsafe request. [-Rpass=si-lower]33 34// GFX90A-HW-REMARK: global_atomic_add_f32 v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], v{{[0-9]+}}, off glc35// GFX90A-HW-REMARK: global_atomic_add_f32 v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], v{{[0-9]+}}, off glc36// GFX90A-HW-REMARK: global_atomic_add_f32 v{{[0-9]+}}, v[{{[0-9]+}}:{{[0-9]+}}], v{{[0-9]+}}, off glc37// GFX90A-HW-LABEL: @atomic_unsafe_hw38// GFX90A-HW:   atomicrmw fadd ptr addrspace(1) %{{.*}}, float %{{.*}} syncscope("workgroup-one-as") monotonic, align 439// GFX90A-HW:   atomicrmw fadd ptr addrspace(1) %{{.*}}, float %{{.*}} syncscope("agent-one-as") monotonic, align 440// GFX90A-HW:   atomicrmw fadd ptr addrspace(1) %{{.*}}, float %{{.*}} syncscope("wavefront-one-as") monotonic, align 441void atomic_unsafe_hw(__global atomic_float *d, float a) {42  float ret1 = __opencl_atomic_fetch_add(d, a, memory_order_relaxed, memory_scope_work_group);43  float ret2 = __opencl_atomic_fetch_add(d, a, memory_order_relaxed, memory_scope_device);44  float ret3 = __opencl_atomic_fetch_add(d, a, memory_order_relaxed, memory_scope_sub_group);45}46