Fixed the stat collection for gpgpu_n_shmem_insn. See Bug 128 for more detail. For verification, I added a directed test with a pre-calculated number of shared memory instructions.

[git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 11454]

3 files changed, 15 insertions, 8 deletions

diff --git a/src/gpgpu-sim/gpu-sim.cc b/src/gpgpu-sim/gpu-sim.cc
index 6b448ab..3200c9c 100644
--- a/src/gpgpu-sim/gpu-sim.cc
+++ b/src/gpgpu-sim/gpu-sim.cc
@@ -639,30 +639,31 @@ unsigned gpgpu_sim::threads_per_core() const
 
 void shader_core_ctx::mem_instruction_stats(const warp_inst_t &inst)
 {
+    unsigned active_count = inst.active_count(); 
     //this breaks some encapsulation: the is_[space] functions, if you change those, change this.
     switch (inst.space.get_type()) {
     case undefined_space:
     case reg_space:
         break;
     case shared_space:
-        m_stats->gpgpu_n_shmem_insn++;
+        m_stats->gpgpu_n_shmem_insn += active_count; 
         break;
     case const_space:
-        m_stats->gpgpu_n_const_insn++;
+        m_stats->gpgpu_n_const_insn += active_count;
         break;
     case param_space_kernel:
     case param_space_local:
-        m_stats->gpgpu_n_param_insn++;
+        m_stats->gpgpu_n_param_insn += active_count;
         break;
     case tex_space:
-        m_stats->gpgpu_n_tex_insn++;
+        m_stats->gpgpu_n_tex_insn += active_count;
         break;
     case global_space:
     case local_space:
         if( inst.is_store() )
-            m_stats->gpgpu_n_store_insn++;
+            m_stats->gpgpu_n_store_insn += active_count;
         else 
-            m_stats->gpgpu_n_load_insn++;
+            m_stats->gpgpu_n_load_insn += active_count;
         break;
     default:
         abort();
diff --git a/src/gpgpu-sim/shader.cc b/src/gpgpu-sim/shader.cc
index 42f43fe..82e4540 100644
--- a/src/gpgpu-sim/shader.cc
+++ b/src/gpgpu-sim/shader.cc
@@ -1026,6 +1026,13 @@ unsigned ldst_unit::clock_multiplier() const
     return m_config->mem_warp_parts; 
 }
 
+void ldst_unit::issue( warp_inst_t *&inst ) 
+{ 
+   // stat collection
+   m_core->mem_instruction_stats(*inst); 
+   pipelined_simd_unit::issue(inst); 
+}
+
 void ldst_unit::cycle()
 {
    writeback();
diff --git a/src/gpgpu-sim/shader.h b/src/gpgpu-sim/shader.h
index 43121a1..d367b7f 100644
--- a/src/gpgpu-sim/shader.h
+++ b/src/gpgpu-sim/shader.h
@@ -834,6 +834,7 @@ public:
                unsigned sid, unsigned tpc );
 
     // modifiers
+    virtual void issue( warp_inst_t *&inst ); 
     virtual void cycle();
      
     void fill( mem_fetch *mf );
@@ -1283,8 +1284,6 @@ public:
     }
     virtual void push(mem_fetch *mf) 
     {
-    	if( !mf->get_inst().empty() ) 
-    	    m_core->mem_instruction_stats(mf->get_inst()); // not I$-fetch
         m_cluster->icnt_inject_request_packet(mf);
     }
 private: