diff options
| -rw-r--r-- | src/abstract_hardware_model.h | 1 | ||||
| -rw-r--r-- | src/cuda-sim/cuda-sim.cc | 1 | ||||
| -rw-r--r-- | src/cuda-sim/ptx_ir.cc | 1 | ||||
| -rw-r--r-- | src/gpgpu-sim/shader.cc | 2 |
4 files changed, 4 insertions, 1 deletions
diff --git a/src/abstract_hardware_model.h b/src/abstract_hardware_model.h index 1b764e2..6df7b89 100644 --- a/src/abstract_hardware_model.h +++ b/src/abstract_hardware_model.h @@ -664,6 +664,7 @@ enum cache_operator_type { CACHE_ALL, // .ca CACHE_LAST_USE, // .lu CACHE_VOLATILE, // .cv + CACHE_L1, // .nc // loads and stores CACHE_STREAMING, // .cs diff --git a/src/cuda-sim/cuda-sim.cc b/src/cuda-sim/cuda-sim.cc index 9246613..93bbc1d 100644 --- a/src/cuda-sim/cuda-sim.cc +++ b/src/cuda-sim/cuda-sim.cc @@ -898,6 +898,7 @@ void ptx_instruction::pre_decode() switch( m_cache_option ) { case CA_OPTION: cache_op = CACHE_ALL; break; + case NC_OPTION: cache_op = CACHE_L1; break; case CG_OPTION: cache_op = CACHE_GLOBAL; break; case CS_OPTION: cache_op = CACHE_STREAMING; break; case LU_OPTION: cache_op = CACHE_LAST_USE; break; diff --git a/src/cuda-sim/ptx_ir.cc b/src/cuda-sim/ptx_ir.cc index ee36957..ba6d7ed 100644 --- a/src/cuda-sim/ptx_ir.cc +++ b/src/cuda-sim/ptx_ir.cc @@ -1211,6 +1211,7 @@ ptx_instruction::ptx_instruction( int opcode, case EXTP_OPTION: break; case NC_OPTION: + m_cache_option = last_ptx_inst_option; break; case UP_OPTION: case DOWN_OPTION: diff --git a/src/gpgpu-sim/shader.cc b/src/gpgpu-sim/shader.cc index d2f40a1..533c6f9 100644 --- a/src/gpgpu-sim/shader.cc +++ b/src/gpgpu-sim/shader.cc @@ -1537,7 +1537,7 @@ bool ldst_unit::memory_cycle( warp_inst_t &inst, mem_stage_stall_type &stall_rea bypassL1D = true; } else if (inst.space.is_global()) { // global memory access // skip L1 cache if the option is enabled - if (m_core->get_config()->gmem_skip_L1D) + if (m_core->get_config()->gmem_skip_L1D && (CACHE_L1 != inst.cache_op)) bypassL1D = true; } |