From 19fbc3bf016d656e7c4da5f247f87c05bbf02daf Mon Sep 17 00:00:00 2001
From: Wilson Fung <wlfung@gmail.com>
Date: Tue, 19 Jun 2012 03:31:27 -0800
Subject: Fixed DRAM performance statistics to display information for more
 than 4 banks.

[git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 13139]
---
 CHANGES                           |  2 ++
 src/gpgpu-sim/mem_latency_stat.cc | 16 ++++++++--------
 version                           |  2 +-
 3 files changed, 11 insertions(+), 9 deletions(-)

diff --git a/CHANGES b/CHANGES
index 5cc73a2..fdcc42a 100644
--- a/CHANGES
+++ b/CHANGES
@@ -25,6 +25,8 @@ Version 3.1.0+edits (development branch) versus 3.1.0
         frequency ratio between the DRAM data bus and command bus.  This allows
         GPGPU-Sim to support both GDDR3 (data rate = 2X command rate) and GDDR5
         (data rate = 4X command rate). 
+    - Fixed the DRAM performance statistics log to display information properly
+      when there are more than 4 banks in each partition (reported by Jungrae). 
 
 Version 3.1.0 versus 3.0.2
 - Support for CUDA 4.0 for both PTX and PTXPlus.
diff --git a/src/gpgpu-sim/mem_latency_stat.cc b/src/gpgpu-sim/mem_latency_stat.cc
index 284ad74..8f04dee 100644
--- a/src/gpgpu-sim/mem_latency_stat.cc
+++ b/src/gpgpu-sim/mem_latency_stat.cc
@@ -260,7 +260,7 @@ void memory_stats_t::memlatstat_print( unsigned n_mem, unsigned gpu_mem_n_bk )
       printf("maximum concurrent accesses to same row:\n");
       for (i=0;i<n_mem ;i++ ) {
          printf("dram[%d]: ", i);
-         for (j=0;j<4 ;j++ ) {
+         for (j=0;j<gpu_mem_n_bk;j++ ) {
             printf("%9d ",max_conc_access2samerow[i][j]);
          }
          printf("\n");
@@ -270,7 +270,7 @@ void memory_stats_t::memlatstat_print( unsigned n_mem, unsigned gpu_mem_n_bk )
       printf("maximum service time to same row:\n");
       for (i=0;i<n_mem ;i++ ) {
          printf("dram[%d]: ", i);
-         for (j=0;j<4 ;j++ ) {
+         for (j=0;j<gpu_mem_n_bk;j++ ) {
             printf("%9d ",max_servicetime2samerow[i][j]);
          }
          printf("\n");
@@ -282,7 +282,7 @@ void memory_stats_t::memlatstat_print( unsigned n_mem, unsigned gpu_mem_n_bk )
       printf("average row accesses per activate:\n");
       for (i=0;i<n_mem ;i++ ) {
          printf("dram[%d]: ", i);
-         for (j=0;j<4 ;j++ ) {
+         for (j=0;j<gpu_mem_n_bk;j++ ) {
             total_row_accesses += row_access[i][j];
             total_num_activates += num_activates[i][j];
             printf("%9f ",(float) row_access[i][j]/num_activates[i][j]);
@@ -301,7 +301,7 @@ void memory_stats_t::memlatstat_print( unsigned n_mem, unsigned gpu_mem_n_bk )
       printf("number of total memory accesses made:\n");
       for (i=0;i<n_mem ;i++ ) {
          printf("dram[%d]: ", i);
-         for (j=0;j<4 ;j++ ) {
+         for (j=0;j<gpu_mem_n_bk;j++ ) {
             l = totalbankaccesses[i][j];
             if (l < min_bank_accesses)
                min_bank_accesses = l;
@@ -339,7 +339,7 @@ void memory_stats_t::memlatstat_print( unsigned n_mem, unsigned gpu_mem_n_bk )
       printf("number of total read accesses:\n");
       for (i=0;i<n_mem ;i++ ) {
          printf("dram[%d]: ", i);
-         for (j=0;j<4 ;j++ ) {
+         for (j=0;j<gpu_mem_n_bk;j++ ) {
             l = totalbankreads[i][j];
             if (l < min_bank_accesses)
                min_bank_accesses = l;
@@ -377,7 +377,7 @@ void memory_stats_t::memlatstat_print( unsigned n_mem, unsigned gpu_mem_n_bk )
       printf("number of total write accesses:\n");
       for (i=0;i<n_mem ;i++ ) {
          printf("dram[%d]: ", i);
-         for (j=0;j<4 ;j++ ) {
+         for (j=0;j<gpu_mem_n_bk;j++ ) {
             l = totalbankwrites[i][j];
             if (l < min_bank_accesses)
                min_bank_accesses = l;
@@ -409,7 +409,7 @@ void memory_stats_t::memlatstat_print( unsigned n_mem, unsigned gpu_mem_n_bk )
       printf("average mf latency per bank:\n");
       for (i=0;i<n_mem ;i++ ) {
          printf("dram[%d]: ", i);
-         for (j=0;j<4 ;j++ ) {
+         for (j=0;j<gpu_mem_n_bk;j++ ) {
             k = totalbankwrites[i][j] + totalbankreads[i][j];
             if (k)
                printf("%10lld", mf_total_lat_table[i][j] / k);
@@ -423,7 +423,7 @@ void memory_stats_t::memlatstat_print( unsigned n_mem, unsigned gpu_mem_n_bk )
       printf("maximum mf latency per bank:\n");
       for (i=0;i<n_mem ;i++ ) {
          printf("dram[%d]: ", i);
-         for (j=0;j<4 ;j++ ) {
+         for (j=0;j<gpu_mem_n_bk;j++ ) {
             printf("%10d", mf_max_lat_table[i][j]);
          }
          printf("\n");
diff --git a/version b/version
index da8015d..05ae64e 100644
--- a/version
+++ b/version
@@ -1,2 +1,2 @@
 const char *g_gpgpusim_version_string = "GPGPU-Sim Simulator Version 3.1.0+edits (development branch)";
-const char *g_gpgpusim_build_string   = "$Change: 13138 $";
+const char *g_gpgpusim_build_string   = "$Change: 13139 $";
-- 
cgit v1.3