0.9756 correlation. Set L1T line size to 128 bytes... problem was

stalling to send four requests per warp into L1T tag lookup.

If L1T is really 32B blocks (as per Henry's paper), this suggests 
banking of L1T needs to be modeled. 

Other changes:
1. bug fix in memory access generation for texture/const cache access
2. adding back memory latency measurement for visualizer

[git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 7913]

1 files changed, 30 insertions, 11 deletions

diff --git a/src/gpgpu-sim/visualizer.cc b/src/gpgpu-sim/visualizer.cc
index a6f836c..c2b1d16 100644
--- a/src/gpgpu-sim/visualizer.cc
+++ b/src/gpgpu-sim/visualizer.cc
@@ -95,20 +95,10 @@ void gpgpu_sim::visualizer_printstat()
    cflog_visualizer_gzprint(visualizer_file);
    shader_CTA_count_visualizer_gzprint(visualizer_file);
 
-   // per shader core cache miss rate 
-
    for (unsigned i=0;i<m_memory_config->m_n_mem;i++) 
       m_memory_partition_unit[i]->visualizer_print(visualizer_file);
    m_shader_stats->visualizer_print(visualizer_file);
-
-   gzprintf(visualizer_file, "Ltwowritemiss: %d\n", m_memory_stats->L2_write_miss);
-   gzprintf(visualizer_file, "Ltwowritehit: %d\n",  m_memory_stats->L2_write_hit);
-   gzprintf(visualizer_file, "Ltworeadmiss: %d\n", m_memory_stats->L2_read_miss);
-   gzprintf(visualizer_file, "Ltworeadhit: %d\n", m_memory_stats->L2_read_hit);
-
-   // latency stats
-   if (m_memory_stats->num_mfs) 
-      gzprintf(visualizer_file, "averagemflatency: %lld\n", m_memory_stats->mf_total_lat/m_memory_stats->num_mfs);
+   m_memory_stats->visualizer_print(visualizer_file);
 
    // other parameters for graphing
    gzprintf(visualizer_file, "globalcyclecount: %lld\n", gpu_sim_cycle);
@@ -118,6 +108,35 @@ void gpgpu_sim::visualizer_printstat()
    time_vector_print_interval2gzfile(visualizer_file);
 
    gzclose(visualizer_file);
+/*
+   gzprintf(visualizer_file, "CacheMissRate_GlobalLocalL1_All: ");
+   for (unsigned i=0;i<m_n_shader;i++) 
+      gzprintf(visualizer_file, "%0.4f ", m_sc[i]->L1_windowed_cache_miss_rate(0));
+   gzprintf(visualizer_file, "\n");
+   gzprintf(visualizer_file, "CacheMissRate_TextureL1_All: ");
+   for (unsigned i=0;i<m_n_shader;i++) 
+      gzprintf(visualizer_file, "%0.4f ", m_sc[i]->L1tex_windowed_cache_miss_rate(0));
+   gzprintf(visualizer_file, "\n");
+   gzprintf(visualizer_file, "CacheMissRate_ConstL1_All: ");
+   for (unsigned i=0;i<m_n_shader;i++) 
+      gzprintf(visualizer_file, "%0.4f ", m_sc[i]->L1const_windowed_cache_miss_rate(0));
+   gzprintf(visualizer_file, "\n");
+   gzprintf(visualizer_file, "CacheMissRate_GlobalLocalL1_noMgHt: ");
+   for (unsigned i=0;i<m_n_shader;i++) 
+      gzprintf(visualizer_file, "%0.4f ", m_sc[i]->L1_windowed_cache_miss_rate(1));
+   gzprintf(visualizer_file, "\n");
+   gzprintf(visualizer_file, "CacheMissRate_TextureL1_noMgHt: ");
+   for (unsigned i=0;i<m_n_shader;i++) 
+      gzprintf(visualizer_file, "%0.4f ", m_sc[i]->L1tex_windowed_cache_miss_rate(1));
+   gzprintf(visualizer_file, "\n");
+   gzprintf(visualizer_file, "CacheMissRate_ConstL1_noMgHt: ");
+   for (unsigned i=0;i<m_n_shader;i++) 
+      gzprintf(visualizer_file, "%0.4f ", m_sc[i]->L1const_windowed_cache_miss_rate(1));
+   gzprintf(visualizer_file, "\n");
+   // reset for next interval
+   for (unsigned i=0;i<m_n_shader;i++) 
+      m_sc[i]->new_cache_window();
+*/
 }
 
 #include <list>