diff options
| author | Tor Aamodt <[email protected]> | 2010-10-22 14:22:21 -0800 |
|---|---|---|
| committer | Tor Aamodt <[email protected]> | 2010-10-22 14:22:21 -0800 |
| commit | 4da926e61569a069bac229e8ba649e600fc78a04 (patch) | |
| tree | cdb7fb44f15cf844c587bf2c92b37d3d2412c152 /src/abstract_hardware_model.cc | |
| parent | dc93f319051a9a9936a02cd9c1f7843a382a2da0 (diff) | |
enables global loads/stores for ptxplus
passing CUDA 3.1 and ptxplus correlation
correlation back to around 0.89 on ptxplus vs quadro fx5800
[git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 7898]
Diffstat (limited to 'src/abstract_hardware_model.cc')
| -rw-r--r-- | src/abstract_hardware_model.cc | 4 |
1 files changed, 3 insertions, 1 deletions
diff --git a/src/abstract_hardware_model.cc b/src/abstract_hardware_model.cc index ad5c983..dfe4b36 100644 --- a/src/abstract_hardware_model.cc +++ b/src/abstract_hardware_model.cc @@ -40,8 +40,10 @@ void warp_inst_t::generate_mem_accesses() { if( empty() || op == MEMORY_BARRIER_OP || m_mem_accesses_created ) return; - if ( !( (op == LOAD_OP) || (op == STORE_OP) || (op == MEMORY_BARRIER_OP) ) ) + if ( !((op == LOAD_OP) || (op == STORE_OP)) ) return; + if( m_warp_active_mask.count() == 0 ) + return; // predicated off assert( is_load() || is_store() ); assert( m_per_scalar_thread_valid ); // need address information per thread |