From 4da926e61569a069bac229e8ba649e600fc78a04 Mon Sep 17 00:00:00 2001 From: Tor Aamodt Date: Fri, 22 Oct 2010 14:22:21 -0800 Subject: 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] --- src/abstract_hardware_model.cc | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src/abstract_hardware_model.cc') 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 -- cgit v1.3