Merge branch 'dev-purdue-integration' into dev-purdue-integration

5 files changed, 27 insertions, 19 deletions

diff --git a/Jenkinsfile b/Jenkinsfile
index a8bd7f7..4e4ef8c 100644
--- a/Jenkinsfile
+++ b/Jenkinsfile
@@ -28,7 +28,7 @@ pipeline {
                     cd gpgpu-sim_simulations && \
                     git checkout purdue-cluster && \
                     git pull && \
-                    rm -r ./benchmarks/data_dirs && ln -s /home/tgrogers-raid/a/common/data_dirs benchmarks/'
+                    ln -s /home/tgrogers-raid/a/common/data_dirs benchmarks/'
                 sh 'source /home/tgrogers-raid/a/common/gpgpu-sim-setup/4.2_env_setup.sh &&\
                     source `pwd`/setup_environment &&\
                     cd gpgpu-sim_simulations && \
diff --git a/src/cuda-sim/cuda-math.h b/src/cuda-sim/cuda-math.h
index 4721e8a..f88c526 100644
--- a/src/cuda-sim/cuda-math.h
+++ b/src/cuda-sim/cuda-math.h
@@ -67,6 +67,8 @@
 #ifndef CUDA_MATH
 #define CUDA_MATH
 
+#include <cmath>
+
 // cuda math implementations
 #undef max
 #undef min
@@ -321,7 +323,7 @@ float __internal_accurate_fdividef(float a, float b)
 float __saturatef(float a)
 {
    float b; 
-   if (isnan(a)) b = 0.0f; 
+   if (std::isnan(a)) b = 0.0f; 
    else if (a >= 1.0f) b = 1.0f;
    else if (a <= 0.0f) b = 0.0f; 
    else b = a; 
diff --git a/src/cuda-sim/instructions.cc b/src/cuda-sim/instructions.cc
index 011c285..e3b8970 100644
--- a/src/cuda-sim/instructions.cc
+++ b/src/cuda-sim/instructions.cc
@@ -33,6 +33,7 @@
 #include "ptx.tab.h"
 #include <stdlib.h>
 #include <math.h>
+#include <cmath>
 #include <fenv.h>
 #include "cuda-math.h"
 #include "../abstract_hardware_model.h"
@@ -1961,7 +1962,7 @@ ptx_reg_t d2d( ptx_reg_t x, unsigned from_width, unsigned to_width, int to_sign,
       y.f64 = x.f64;
       break; 
    }
-   if (isnan(y.f64)) {
+   if (std::isnan(y.f64)) {
       y.u64 = 0xfff8000000000000ull;
    } else if (saturation_mode) {
       y.f64 = cuda_math::__saturatef(y.f64); 
@@ -2086,7 +2087,7 @@ void ptx_round(ptx_reg_t& data, int rounding_mode, int type)
       }
    }
    if ((type == F64_TYPE)||(type == FF64_TYPE)) {
-      if (isnan(data.f64)) {
+      if (std::isnan(data.f64)) {
          data.u64 = 0xfff8000000000000ull;
       }
    }
@@ -2648,12 +2649,12 @@ void mad_def( const ptx_instruction *pI, ptx_thread_info *thread, bool use_carry
 
 bool isNaN(float x)
 {
-   return isnan(x);
+   return std::isnan(x);
 }
 
 bool isNaN(double x)
 {
-   return isnan(x);
+   return std::isnan(x);
 }
 
 void max_impl( const ptx_instruction *pI, ptx_thread_info *thread ) 
diff --git a/src/gpgpu-sim/gpu-cache.cc b/src/gpgpu-sim/gpu-cache.cc
index 8191f62..26a1638 100644
--- a/src/gpgpu-sim/gpu-cache.cc
+++ b/src/gpgpu-sim/gpu-cache.cc
@@ -658,15 +658,13 @@ void cache_stats::print_stats(FILE *fout, const char *cache_name) const{
     std::string m_cache_name = cache_name;
     for (unsigned type = 0; type < NUM_MEM_ACCESS_TYPE; ++type) {
         for (unsigned status = 0; status < NUM_CACHE_REQUEST_STATUS; ++status) {
-            if(m_stats[type][status] > 0){
-                fprintf(fout, "\t%s[%s][%s] = %u\n",
-                    m_cache_name.c_str(),
-                    mem_access_type_str((enum mem_access_type)type),
-                    cache_request_status_str((enum cache_request_status)status),
-                    m_stats[type][status]);
-                if(status != RESERVATION_FAIL)
-                	 total_access[type]+= m_stats[type][status];
-            }
+            fprintf(fout, "\t%s[%s][%s] = %u\n",
+                m_cache_name.c_str(),
+                mem_access_type_str((enum mem_access_type)type),
+                cache_request_status_str((enum cache_request_status)status),
+                m_stats[type][status]);
+            if(status != RESERVATION_FAIL)
+            	 total_access[type]+= m_stats[type][status];
         }
     }
     for (unsigned type = 0; type < NUM_MEM_ACCESS_TYPE; ++type) {
diff --git a/src/gpgpusim_entrypoint.cc b/src/gpgpusim_entrypoint.cc
index d94fd8a..a18e956 100644
--- a/src/gpgpusim_entrypoint.cc
+++ b/src/gpgpusim_entrypoint.cc
@@ -93,8 +93,15 @@ bool g_sim_active = false;
 bool g_sim_done = true;
 bool break_limit = false;
 
+static void termination_callback()
+{
+    printf("GPGPU-Sim: *** exit detected ***\n");
+    fflush(stdout);
+}
+
 void *gpgpu_sim_thread_concurrent(void*)
 {
+    atexit(termination_callback);
     // concurrent kernel execution simulation thread
     do {
         if(g_debug_execution >= 3) {
@@ -167,10 +174,10 @@ void *gpgpu_sim_thread_concurrent(void*)
         g_sim_active = false;
         pthread_mutex_unlock(&g_sim_lock);
     } while( !g_sim_done );
-    if(g_debug_execution >= 3) {
-       printf("GPGPU-Sim: *** simulation thread exiting ***\n");
-       fflush(stdout);
-    }
+
+    printf("GPGPU-Sim: *** simulation thread exiting ***\n");
+    fflush(stdout);
+
     if(break_limit) {
     	printf("GPGPU-Sim: ** break due to reaching the maximum cycles (or instructions) **\n");
     	exit(1);