1 files changed, 86 insertions, 92 deletions

diff --git a/src/gpgpusim_entrypoint.cc b/src/gpgpusim_entrypoint.cc
index 9e2bfa2..816159f 100644
--- a/src/gpgpusim_entrypoint.cc
+++ b/src/gpgpusim_entrypoint.cc
@@ -35,32 +35,30 @@
 #include "gpgpu-sim/gpu-sim.h"
 #include "gpgpu-sim/icnt_wrapper.h"
 #include "stream_manager.h"
-
-#include <pthread.h>
-#include <semaphore.h>
+#include "../libcuda/gpgpu_context.h"
 
 #define MAX(a,b) (((a)>(b))?(a):(b))
 
+static int sg_argc = 3;
+static const char *sg_argv[] = {"", "-config","gpgpusim.config"};
 
 
-struct gpgpu_ptx_sim_arg *grid_params;
-
-sem_t g_sim_signal_start;
-sem_t g_sim_signal_finish;
-sem_t g_sim_signal_exit;
-time_t g_simulation_starttime;
-pthread_t g_simulation_thread;
-
-gpgpu_sim_config g_the_gpu_config;
-gpgpu_sim *g_the_gpu;
-stream_manager *g_stream_manager;
-
+GPGPUsim_ctx* the_gpgpusim =  NULL;
 
+GPGPUsim_ctx* GPGPUsim_ctx_ptr(){
+	if(the_gpgpusim == NULL)
+		the_gpgpusim = GPGPU_Context()->the_gpgpusim;
 
-static int sg_argc = 3;
-static const char *sg_argv[] = {"", "-config","gpgpusim.config"};
+	return the_gpgpusim;
+}
 
+class gpgpu_sim* g_the_gpu() {
+	return GPGPUsim_ctx_ptr()->g_the_gpu;
+}
 
+class stream_manager* g_stream_manager()  {
+	return GPGPUsim_ctx_ptr()->g_stream_manager;
+}
 
 static void print_simulation_time();
 
@@ -69,29 +67,26 @@ void *gpgpu_sim_thread_sequential(void*)
    // at most one kernel running at a time
    bool done;
    do {
-      sem_wait(&g_sim_signal_start);
+      sem_wait(&(GPGPUsim_ctx_ptr()->g_sim_signal_start));
       done = true;
-      if( g_the_gpu->get_more_cta_left() ) {
+      if( GPGPUsim_ctx_ptr()->g_the_gpu->get_more_cta_left() ) {
           done = false;
-          g_the_gpu->init();
-          while( g_the_gpu->active() ) {
-              g_the_gpu->cycle();
-              g_the_gpu->deadlock_check();
+          GPGPUsim_ctx_ptr()->g_the_gpu->init();
+          while( GPGPUsim_ctx_ptr()->g_the_gpu->active() ) {
+        	  GPGPUsim_ctx_ptr()->g_the_gpu->cycle();
+        	  GPGPUsim_ctx_ptr()->g_the_gpu->deadlock_check();
           }
-          g_the_gpu->print_stats();
-          g_the_gpu->update_stats();
+          GPGPUsim_ctx_ptr()->g_the_gpu->print_stats();
+          GPGPUsim_ctx_ptr()->g_the_gpu->update_stats();
           print_simulation_time();
       }
-      sem_post(&g_sim_signal_finish);
+      sem_post(&(GPGPUsim_ctx_ptr()->g_sim_signal_finish));
    } while(!done);
-   sem_post(&g_sim_signal_exit);
+   sem_post(&(GPGPUsim_ctx_ptr()->g_sim_signal_exit));
    return NULL;
 }
 
-pthread_mutex_t g_sim_lock = PTHREAD_MUTEX_INITIALIZER;
-bool g_sim_active = false;
-bool g_sim_done = true;
-bool break_limit = false;
+
 
 static void termination_callback()
 {
@@ -108,19 +103,19 @@ void *gpgpu_sim_thread_concurrent(void*)
            printf("GPGPU-Sim: *** simulation thread starting and spinning waiting for work ***\n");
            fflush(stdout);
         }
-        while( g_stream_manager->empty_protected() && !g_sim_done )
+        while( GPGPUsim_ctx_ptr()->g_stream_manager->empty_protected() && !GPGPUsim_ctx_ptr()->g_sim_done )
             ;
         if(g_debug_execution >= 3) {
            printf("GPGPU-Sim: ** START simulation thread (detected work) **\n");
-           g_stream_manager->print(stdout);
+           GPGPUsim_ctx_ptr()->g_stream_manager->print(stdout);
            fflush(stdout);
         }
-        pthread_mutex_lock(&g_sim_lock);
-        g_sim_active = true;
-        pthread_mutex_unlock(&g_sim_lock);
+        pthread_mutex_lock(&(GPGPUsim_ctx_ptr()->g_sim_lock));
+        GPGPUsim_ctx_ptr()->g_sim_active = true;
+        pthread_mutex_unlock(&(GPGPUsim_ctx_ptr()->g_sim_lock));
         bool active = false;
         bool sim_cycles = false;
-        g_the_gpu->init();
+        GPGPUsim_ctx_ptr()->g_the_gpu->init();
         do {
             // check if a kernel has completed
             // launch operation on device if one is pending and can be run
@@ -132,70 +127,70 @@ void *gpgpu_sim_thread_concurrent(void*)
             // another kernel, the gpu is not re-initialized and the inter-kernel
             // behaviour may be incorrect. Check that a kernel has finished and
             // no other kernel is currently running.
-            if(g_stream_manager->operation(&sim_cycles) && !g_the_gpu->active())
+            if(GPGPUsim_ctx_ptr()->g_stream_manager->operation(&sim_cycles) && !GPGPUsim_ctx_ptr()->g_the_gpu->active())
                 break;
 
             //functional simulation
-            if( g_the_gpu->is_functional_sim()) {
-                kernel_info_t * kernel = g_the_gpu->get_functional_kernel();
+            if( GPGPUsim_ctx_ptr()->g_the_gpu->is_functional_sim()) {
+                kernel_info_t * kernel = GPGPUsim_ctx_ptr()->g_the_gpu->get_functional_kernel();
                 assert(kernel);
-                gpgpu_cuda_ptx_sim_main_func(*kernel);
-                g_the_gpu->finish_functional_sim(kernel);
+                GPGPUsim_ctx_ptr()->gpgpu_ctx->func_sim->gpgpu_cuda_ptx_sim_main_func(*kernel);
+                GPGPUsim_ctx_ptr()->g_the_gpu->finish_functional_sim(kernel);
             }
 
             //performance simulation
-            if( g_the_gpu->active() ) {
-                g_the_gpu->cycle();
-                sim_cycles = true;
-                g_the_gpu->deadlock_check();
+            if( GPGPUsim_ctx_ptr()->g_the_gpu->active() ) {
+            	GPGPUsim_ctx_ptr()->g_the_gpu->cycle();
+            	sim_cycles = true;
+            	GPGPUsim_ctx_ptr()->g_the_gpu->deadlock_check();
             }else {
-                if(g_the_gpu->cycle_insn_cta_max_hit()){
-                    g_stream_manager->stop_all_running_kernels();
-                    g_sim_done = true;
-                    break_limit = true;
+                if(GPGPUsim_ctx_ptr()->g_the_gpu->cycle_insn_cta_max_hit()){
+                	GPGPUsim_ctx_ptr()->g_stream_manager->stop_all_running_kernels();
+                	GPGPUsim_ctx_ptr()->g_sim_done = true;
+                	GPGPUsim_ctx_ptr()->break_limit = true;
                 }
             }
 
-            active=g_the_gpu->active() || !g_stream_manager->empty_protected();
+            active=GPGPUsim_ctx_ptr()->g_the_gpu->active() || !(GPGPUsim_ctx_ptr()->g_stream_manager->empty_protected());
 
-        } while( active && !g_sim_done);
+        } while( active && !GPGPUsim_ctx_ptr()->g_sim_done);
         if(g_debug_execution >= 3) {
            printf("GPGPU-Sim: ** STOP simulation thread (no work) **\n");
            fflush(stdout);
         }
         if(sim_cycles) {
-            g_the_gpu->print_stats();
-            g_the_gpu->update_stats();
+        	GPGPUsim_ctx_ptr()->g_the_gpu->print_stats();
+        	GPGPUsim_ctx_ptr()->g_the_gpu->update_stats();
             print_simulation_time();
         }
-        pthread_mutex_lock(&g_sim_lock);
-        g_sim_active = false;
-        pthread_mutex_unlock(&g_sim_lock);
-    } while( !g_sim_done );
+        pthread_mutex_lock(&(GPGPUsim_ctx_ptr()->g_sim_lock));
+        GPGPUsim_ctx_ptr()->g_sim_active = false;
+        pthread_mutex_unlock(&(GPGPUsim_ctx_ptr()->g_sim_lock));
+    } while( !GPGPUsim_ctx_ptr()->g_sim_done );
 
     printf("GPGPU-Sim: *** simulation thread exiting ***\n");
     fflush(stdout);
 
-    if(break_limit) {
+    if(GPGPUsim_ctx_ptr()->break_limit) {
     	printf("GPGPU-Sim: ** break due to reaching the maximum cycles (or instructions) **\n");
     	exit(1);
     }
 
-    sem_post(&g_sim_signal_exit);
+    sem_post(&(GPGPUsim_ctx_ptr()->g_sim_signal_exit));
     return NULL;
 }
 
 void synchronize()
 {
     printf("GPGPU-Sim: synchronize waiting for inactive GPU simulation\n");
-    g_stream_manager->print(stdout);
+    GPGPUsim_ctx_ptr()->g_stream_manager->print(stdout);
     fflush(stdout);
 //    sem_wait(&g_sim_signal_finish);
     bool done = false;
     do {
-        pthread_mutex_lock(&g_sim_lock);
-        done = ( g_stream_manager->empty() && !g_sim_active ) || g_sim_done;
-        pthread_mutex_unlock(&g_sim_lock);
+        pthread_mutex_lock(&(GPGPUsim_ctx_ptr()->g_sim_lock));
+        done = ( GPGPUsim_ctx_ptr()->g_stream_manager->empty() && !GPGPUsim_ctx_ptr()->g_sim_active ) || GPGPUsim_ctx_ptr()->g_sim_done;
+        pthread_mutex_unlock(&(GPGPUsim_ctx_ptr()->g_sim_lock));
     } while (!done);
     printf("GPGPU-Sim: detected inactive GPU simulation thread\n");
     fflush(stdout);
@@ -204,29 +199,28 @@ void synchronize()
 
 void exit_simulation()
 {
-    g_sim_done=true;
+	GPGPUsim_ctx_ptr()->g_sim_done=true;
     printf("GPGPU-Sim: exit_simulation called\n");
     fflush(stdout);
-    sem_wait(&g_sim_signal_exit);
+    sem_wait(&(GPGPUsim_ctx_ptr()->g_sim_signal_exit));
     printf("GPGPU-Sim: simulation thread signaled exit\n");
     fflush(stdout);
 }
 
-extern bool g_cuda_launch_blocking;
-
-gpgpu_sim *gpgpu_ptx_sim_init_perf()
+gpgpu_sim *gpgpu_context::gpgpu_ptx_sim_init_perf()
 {
    srand(1);
    print_splash();
-   read_sim_environment_variables();
-   read_parser_environment_variables();
+   func_sim->read_sim_environment_variables();
+   ptx_parser->read_parser_environment_variables();
    option_parser_t opp = option_parser_create();
 
    ptx_reg_options(opp);
-   ptx_opcocde_latency_options(opp);
+   func_sim->ptx_opcocde_latency_options(opp);
 
    icnt_reg_options(opp);
-   g_the_gpu_config.reg_options(opp); // register GPU microrachitecture options
+   GPGPUsim_ctx_ptr()->g_the_gpu_config = new gpgpu_sim_config(this);
+   GPGPUsim_ctx_ptr()->g_the_gpu_config->reg_options(opp); // register GPU microrachitecture options
 
    option_parser_cmdline(opp, sg_argc, sg_argv); // parse configuration options
    fprintf(stdout, "GPGPU-Sim: Configuration options:\n\n");
@@ -234,28 +228,28 @@ gpgpu_sim *gpgpu_ptx_sim_init_perf()
    // Set the Numeric locale to a standard locale where a decimal point is a "dot" not a "comma"
    // so it does the parsing correctly independent of the system environment variables
    assert(setlocale(LC_NUMERIC,"C"));
-   g_the_gpu_config.init();
+   GPGPUsim_ctx_ptr()->g_the_gpu_config->init();
 
-   g_the_gpu = new gpgpu_sim(g_the_gpu_config);
-   g_stream_manager = new stream_manager(g_the_gpu,g_cuda_launch_blocking);
+   GPGPUsim_ctx_ptr()->g_the_gpu = new gpgpu_sim(*(GPGPUsim_ctx_ptr()->g_the_gpu_config), this);
+   GPGPUsim_ctx_ptr()->g_stream_manager = new stream_manager((GPGPUsim_ctx_ptr()->g_the_gpu), func_sim->g_cuda_launch_blocking);
 
-   g_simulation_starttime = time((time_t *)NULL);
+   GPGPUsim_ctx_ptr()->g_simulation_starttime = time((time_t *)NULL);
 
-   sem_init(&g_sim_signal_start,0,0);
-   sem_init(&g_sim_signal_finish,0,0);
-   sem_init(&g_sim_signal_exit,0,0);
+   sem_init(&(GPGPUsim_ctx_ptr()->g_sim_signal_start),0,0);
+   sem_init(&(GPGPUsim_ctx_ptr()->g_sim_signal_finish),0,0);
+   sem_init(&(GPGPUsim_ctx_ptr()->g_sim_signal_exit),0,0);
 
-   return g_the_gpu;
+   return GPGPUsim_ctx_ptr()->g_the_gpu;
 }
 
 void start_sim_thread(int api)
 {
-    if( g_sim_done ) {
-        g_sim_done = false;
+    if( GPGPUsim_ctx_ptr()->g_sim_done ) {
+    	GPGPUsim_ctx_ptr()->g_sim_done = false;
         if( api == 1 ) {
-           pthread_create(&g_simulation_thread,NULL,gpgpu_sim_thread_concurrent,NULL);
+           pthread_create(&(GPGPUsim_ctx_ptr()->g_simulation_thread),NULL,gpgpu_sim_thread_concurrent,NULL);
         } else {
-           pthread_create(&g_simulation_thread,NULL,gpgpu_sim_thread_sequential,NULL);
+           pthread_create(&(GPGPUsim_ctx_ptr()->g_simulation_thread),NULL,gpgpu_sim_thread_sequential,NULL);
         }
     }
 }
@@ -264,7 +258,7 @@ void print_simulation_time()
 {
    time_t current_time, difference, d, h, m, s;
    current_time = time((time_t *)NULL);
-   difference = MAX(current_time - g_simulation_starttime, 1);
+   difference = MAX(current_time - GPGPUsim_ctx_ptr()->g_simulation_starttime, 1);
 
    d = difference/(3600*24);
    h = difference/3600 - 24*d;
@@ -274,16 +268,16 @@ void print_simulation_time()
    fflush(stderr);
    printf("\n\ngpgpu_simulation_time = %u days, %u hrs, %u min, %u sec (%u sec)\n",
           (unsigned)d, (unsigned)h, (unsigned)m, (unsigned)s, (unsigned)difference );
-   printf("gpgpu_simulation_rate = %u (inst/sec)\n", (unsigned)(g_the_gpu->gpu_tot_sim_insn / difference) );
-   printf("gpgpu_simulation_rate = %u (cycle/sec)\n", (unsigned)(gpu_tot_sim_cycle / difference) );
+   printf("gpgpu_simulation_rate = %u (inst/sec)\n", (unsigned)(GPGPUsim_ctx_ptr()->g_the_gpu->gpu_tot_sim_insn / difference) );
+   printf("gpgpu_simulation_rate = %u (cycle/sec)\n", (unsigned)(GPGPUsim_ctx_ptr()->g_the_gpu->gpu_tot_sim_cycle / difference) );
    fflush(stdout);
 }
 
 int gpgpu_opencl_ptx_sim_main_perf( kernel_info_t *grid )
 {
-   g_the_gpu->launch(grid);
-   sem_post(&g_sim_signal_start);
-   sem_wait(&g_sim_signal_finish);
+   GPGPUsim_ctx_ptr()->g_the_gpu->launch(grid);
+   sem_post(&(GPGPUsim_ctx_ptr()->g_sim_signal_start));
+   sem_wait(&(GPGPUsim_ctx_ptr()->g_sim_signal_finish));
    return 0;
 }
 
@@ -291,7 +285,7 @@ int gpgpu_opencl_ptx_sim_main_perf( kernel_info_t *grid )
 /*!
  * This function call the CUDA PTX functional simulator
  */
-int gpgpu_opencl_ptx_sim_main_func( kernel_info_t *grid )
+int cuda_sim::gpgpu_opencl_ptx_sim_main_func( kernel_info_t *grid )
 {
     //calling the CUDA PTX simulator, sending the kernel by reference and a flag set to true,
     //the flag used by the function to distinguish OpenCL calls from the CUDA simulation calls which