- parameter memory and active threads now part of kernel_info_t:

  Parameters are finalized at kernel launch, which means the contents
  of parameter memory are initialized. Kernel arguement names have a
  fixed order, hence same address should be assigned on subsequent
  kernel launches of same kernel in other streams provided the data size
  param_t::size of arguments for each kernel launch is identical (an
  assertion has been added to check this is true).

- passing regression

[git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 8303]

4 files changed, 17 insertions, 13 deletions

diff --git a/src/cuda-sim/cuda-sim.cc b/src/cuda-sim/cuda-sim.cc
index b6a15e8..e33181f 100644
--- a/src/cuda-sim/cuda-sim.cc
+++ b/src/cuda-sim/cuda-sim.cc
@@ -1075,7 +1075,8 @@ unsigned ptx_sim_init_thread( kernel_info_t &kernel,
                               unsigned hw_warp_id,
                               gpgpu_t *gpu )
 {
-   static std::list<ptx_thread_info *> active_threads;
+   std::list<ptx_thread_info *> &active_threads = kernel.active_threads();
+
    static std::map<unsigned,memory_space*> shared_memory_lookup;
    static std::map<unsigned,ptx_cta_info*> ptx_cta_lookup;
    static std::map<unsigned,std::map<unsigned,memory_space*> > local_memory_lookup;
@@ -1096,7 +1097,7 @@ unsigned ptx_sim_init_thread( kernel_info_t &kernel,
       *thread_info = NULL;
    }
 
-   if ( !active_threads.empty() ) { //if g_active_threads not empty...
+   if ( !active_threads.empty() ) {
       assert( active_threads.size() <= threads_left );
       ptx_thread_info *thd = active_threads.front(); 
       active_threads.pop_front();
@@ -1156,7 +1157,7 @@ unsigned ptx_sim_init_thread( kernel_info_t &kernel,
       dim3 tid3d = kernel.get_next_thread_id_3d();
       kernel.increment_thread_id();
       new_tid += tid;
-      ptx_thread_info *thd = new ptx_thread_info();
+      ptx_thread_info *thd = new ptx_thread_info(kernel);
    
       memory_space *local_mem = NULL;
       std::map<unsigned,memory_space*>::iterator l = local_mem_lookup.find(new_tid);
@@ -1210,23 +1211,24 @@ size_t get_kernel_code_size( class function_info *entry )
 }
 
 
-kernel_info_t gpgpu_opencl_ptx_sim_init_grid(class function_info *entry,
+kernel_info_t *gpgpu_opencl_ptx_sim_init_grid(class function_info *entry,
                                              gpgpu_ptx_sim_arg_list_t args, 
                                              struct dim3 gridDim,
                                              struct dim3 blockDim,
                                              gpgpu_t *gpu )
 {
+   kernel_info_t *result = new kernel_info_t(gridDim,blockDim,entry);
    unsigned argcount=args.size();
    unsigned argn=1;
    for( gpgpu_ptx_sim_arg_list_t::iterator a = args.begin(); a != args.end(); a++ ) {
       entry->add_param_data(argcount-argn,&(*a));
       argn++;
    }
-   entry->finalize(gpu->get_param_memory());
+   entry->finalize(result->get_param_memory());
    g_ptx_kernel_count++; 
    fflush(stdout);
 
-   return kernel_info_t(gridDim,blockDim,entry);
+   return result;
 }
 
 const char *g_gpgpusim_version_string = "2.1.1b (beta)";
@@ -1395,7 +1397,7 @@ ptx_cta_info *g_func_cta_info = NULL;
 
 #define MAX(a,b) (((a)>(b))?(a):(b))
 
-void gpgpu_cuda_ptx_sim_main_func( kernel_info_t kernel )
+void gpgpu_cuda_ptx_sim_main_func( kernel_info_t &kernel )
 {
    printf("GPGPU-Sim: Performing Functional Simulation...\n");
 
@@ -1416,7 +1418,7 @@ void gpgpu_cuda_ptx_sim_main_func( kernel_info_t kernel )
             g_func_cta_info->check_cta_thread_status_and_reset();
             while( kernel.more_threads_in_cta() ) {
                      memory_space *local_mem = NULL;
-                     ptx_thread_info *thd = new ptx_thread_info();
+                     ptx_thread_info *thd = new ptx_thread_info(kernel);
                      dim3 tid3d = kernel.get_next_thread_id_3d();
                      unsigned lm_idx = kernel.get_next_thread_id();
                      kernel.increment_thread_id();
diff --git a/src/cuda-sim/cuda-sim.h b/src/cuda-sim/cuda-sim.h
index ef045fb..3bc915d 100644
--- a/src/cuda-sim/cuda-sim.h
+++ b/src/cuda-sim/cuda-sim.h
@@ -20,12 +20,12 @@ extern void ** g_inst_op_classification_stat;
 extern int g_ptx_kernel_count; // used for classification stat collection purposes 
 
 
-extern class kernel_info_t gpgpu_opencl_ptx_sim_init_grid(class function_info *entry,
+extern class kernel_info_t *gpgpu_opencl_ptx_sim_init_grid(class function_info *entry,
                                             gpgpu_ptx_sim_arg_list_t args, 
                                             struct dim3 gridDim, 
                                             struct dim3 blockDim, 
                                                           class gpgpu_t *gpu );
-extern void gpgpu_cuda_ptx_sim_main_func( kernel_info_t kernel );
+extern void gpgpu_cuda_ptx_sim_main_func( kernel_info_t &kernel );
 extern void   print_splash();
 extern void   gpgpu_ptx_sim_register_const_variable(void*, const char *deviceName, size_t size );
 extern void   gpgpu_ptx_sim_register_global_variable(void *hostVar, const char *deviceName, size_t size );
diff --git a/src/cuda-sim/ptx_sim.cc b/src/cuda-sim/ptx_sim.cc
index 59f4522..3ab0594 100644
--- a/src/cuda-sim/ptx_sim.cc
+++ b/src/cuda-sim/ptx_sim.cc
@@ -211,7 +211,8 @@ ptx_thread_info::~ptx_thread_info()
    g_ptx_thread_info_delete_count++;
 }
 
-ptx_thread_info::ptx_thread_info()
+ptx_thread_info::ptx_thread_info( kernel_info_t &kernel )
+    : m_kernel(kernel)
 {
    m_uid = g_ptx_thread_info_uid_next++;
    m_core = NULL;
diff --git a/src/cuda-sim/ptx_sim.h b/src/cuda-sim/ptx_sim.h
index d60955b..cc555cf 100644
--- a/src/cuda-sim/ptx_sim.h
+++ b/src/cuda-sim/ptx_sim.h
@@ -274,7 +274,7 @@ private:
 class ptx_thread_info {
 public:
    ~ptx_thread_info();
-   ptx_thread_info();
+   ptx_thread_info( kernel_info_t &kernel );
 
    void init(gpgpu_t *gpu, core_t *core, unsigned sid, unsigned cta_id, unsigned wid, unsigned tid ) 
    { 
@@ -429,7 +429,7 @@ public:
    memory_space *get_global_memory() { return m_gpu->get_global_memory(); }
    memory_space *get_tex_memory() { return m_gpu->get_tex_memory(); }
    memory_space *get_surf_memory() { return m_gpu->get_surf_memory(); }
-   memory_space *get_param_memory() { return m_gpu->get_param_memory(); }
+   memory_space *get_param_memory() { return m_kernel.get_param_memory(); }
    const gpgpu_functional_sim_config &get_config() const { return m_gpu->get_config(); }
 
 public:
@@ -446,6 +446,7 @@ private:
 
 
    unsigned m_uid;
+   kernel_info_t &m_kernel;
    core_t *m_core;
    gpgpu_t *m_gpu;
    bool   m_valid;