Merge remote-tracking branch 'public/dev' into dev

8 files changed, 29 insertions, 16 deletions

diff --git a/.gitignore b/.gitignore
index 53fadb5..887b605 100644
--- a/.gitignore
+++ b/.gitignore
@@ -8,6 +8,7 @@ libcuda/cuobjdump_parser.h
 libcuda/cuobjdump_parser.output
 
 lib/*
+doc/doxygen/html
 
 cuobjdump_to_ptxplus/elf_lexer.cc
 cuobjdump_to_ptxplus/elf_parser.cc
@@ -26,4 +27,5 @@ cuobjdump_to_ptxplus/sass_parser.cc
 cuobjdump_to_ptxplus/sass_parser.hh
 cuobjdump_to_ptxplus/sass_parser.output
 
-build/*
\ No newline at end of file
+build/*
+*.swp
diff --git a/Makefile b/Makefile
index e83de9a..2d0466e 100644
--- a/Makefile
+++ b/Makefile
@@ -143,7 +143,7 @@ no_opencl_support:
 	@echo "Warning: gpgpu-sim is building without opencl support. Make sure NVOPENCL_LIBDIR and NVOPENCL_INCDIR are set"
 
 $(SIM_LIB_DIR)/libcudart.so: makedirs $(LIBS) cudalib
-	g++ -shared -Wl,-soname,libcudart_$(GPGPUSIM_BUILD).so \
+	g++ -shared -Wl,-soname,libcudart_$(GPGPUSIM_BUILD).so -Wl,--version-script=linux-so-version.txt\
 			$(SIM_OBJ_FILES_DIR)/libcuda/*.o \
 			$(SIM_OBJ_FILES_DIR)/cuda-sim/*.o \
 			$(SIM_OBJ_FILES_DIR)/cuda-sim/decuda_pred_table/*.o \
@@ -161,6 +161,8 @@ $(SIM_LIB_DIR)/libcudart.so: makedirs $(LIBS) cudalib
 	if [ ! -f $(SIM_LIB_DIR)/libcudart.so.6.5 ]; then ln -s libcudart.so $(SIM_LIB_DIR)/libcudart.so.6.5; fi
 	if [ ! -f $(SIM_LIB_DIR)/libcudart.so.7.5 ]; then ln -s libcudart.so $(SIM_LIB_DIR)/libcudart.so.7.5; fi
 	if [ ! -f $(SIM_LIB_DIR)/libcudart.so.8.0 ]; then ln -s libcudart.so $(SIM_LIB_DIR)/libcudart.so.8.0; fi
+	if [ ! -f $(SIM_LIB_DIR)/libcudart.so.9.0 ]; then ln -s libcudart.so $(SIM_LIB_DIR)/libcudart.so.9.0; fi
+	if [ ! -f $(SIM_LIB_DIR)/libcudart.so.9.1 ]; then ln -s libcudart.so $(SIM_LIB_DIR)/libcudart.so.9.1; fi
 
 $(SIM_LIB_DIR)/libcudart.dylib: makedirs $(LIBS) cudalib
 	g++ -dynamiclib -Wl,-headerpad_max_install_names,-undefined,dynamic_lookup,-compatibility_version,1.1,-current_version,1.1\
diff --git a/libcuda/cuobjdump.l b/libcuda/cuobjdump.l
index f63ee73..0953ea1 100644
--- a/libcuda/cuobjdump.l
+++ b/libcuda/cuobjdump.l
@@ -159,8 +159,6 @@ newlines	{newline}+
 %%
 void cuobjdump_error(const char* message)
 {
-	printf("   "); printf(message); printf(" near \""); printf(yytext); printf("\"");
-	printf(" on line "); 
-	char line[5]; sprintf(line, "%i", yylineno); printf(line);
-	printf("\n");
+	printf("   %s near \"%s\"",message, yytext);
+	printf(" on line %i\n",yylineno); 
 }
diff --git a/linux-so-version.txt b/linux-so-version.txt
new file mode 100644
index 0000000..40f775d
--- /dev/null
+++ b/linux-so-version.txt
@@ -0,0 +1,2 @@
+libcudart.so.9.1{
+};
diff --git a/src/cuda-sim/ptx.y b/src/cuda-sim/ptx.y
index e00aa4b..4edae5d 100644
--- a/src/cuda-sim/ptx.y
+++ b/src/cuda-sim/ptx.y
@@ -47,7 +47,6 @@ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 %token  PTR_DIRECTIVE
 %token  ENTRY_DIRECTIVE
 %token  EXTERN_DIRECTIVE
-%token  WEAK_DIRECTIVE
 %token  FILE_DIRECTIVE
 %token  FUNC_DIRECTIVE
 %token  GLOBAL_DIRECTIVE
@@ -513,7 +512,8 @@ compare_spec:EQ_OPTION { add_option(EQ_OPTION); }
 	| NAN_OPTION { add_option(NAN_OPTION); } 
 	;
 
-operand_list: operand
+operand_list: /* empty*/
+    | operand
 	| operand COMMA operand_list;
 
 operand: IDENTIFIER  { add_scalar_operand( $1 ); }
diff --git a/src/cuda-sim/ptx_parser.cc b/src/cuda-sim/ptx_parser.cc
index baa3bcd..a180da9 100644
--- a/src/cuda-sim/ptx_parser.cc
+++ b/src/cuda-sim/ptx_parser.cc
@@ -433,7 +433,7 @@ void add_identifier( const char *identifier, int array_dim, unsigned array_ident
       assert( (num_bits%8) == 0  );
       addr = g_current_symbol_table->get_shared_next();
       addr_pad = pad_address(addr, num_bits/8, 128);
-      printf("from 0x%x to 0x%lx (shared memory space)\n",
+      printf("from 0x%llx to 0x%llx (shared memory space)\n",
               addr+addr_pad,
               addr+addr_pad + num_bits/8);
          fflush(stdout);
@@ -450,7 +450,7 @@ void add_identifier( const char *identifier, int array_dim, unsigned array_ident
          assert( (num_bits%8) == 0  ); 
          addr = g_current_symbol_table->get_global_next();
          addr_pad = pad_address(addr, num_bits/8, 128);
-         printf("from 0x%x to 0x%lx (global memory space) %u\n",
+         printf("from 0x%llx to 0x%llx (global memory space) %u\n",
               addr+addr_pad,
               addr+addr_pad + num_bits/8,
               g_const_alloc++);
@@ -471,7 +471,7 @@ void add_identifier( const char *identifier, int array_dim, unsigned array_ident
       assert( (num_bits%8) == 0  );
       addr = g_current_symbol_table->get_global_next();
       addr_pad = pad_address(addr, num_bits/8, 128);
-      printf("from 0x%x to 0x%lx (global memory space)\n",
+      printf("from 0x%llx to 0x%llx (global memory space)\n",
               addr+addr_pad,
               addr+addr_pad + num_bits/8);
       fflush(stdout);
@@ -488,7 +488,7 @@ void add_identifier( const char *identifier, int array_dim, unsigned array_ident
          assert( (num_bits%8) == 0  );
          addr = g_current_symbol_table->get_local_next();
          addr_pad = pad_address(addr, num_bits/8, 128);
-         printf("from 0x%x to 0x%lx (local memory space)\n",
+         printf("from 0x%llx to 0x%llx (local memory space)\n",
                  addr+addr_pad,
                  addr+addr_pad + num_bits/8);
          fflush(stdout);
@@ -501,7 +501,7 @@ void add_identifier( const char *identifier, int array_dim, unsigned array_ident
         assert( (num_bits%8) == 0 );
         addr = g_current_symbol_table->get_local_next();
         addr_pad = pad_address(addr, num_bits/8, 128);
-        printf("from 0x%x to 0x%lx\n",
+        printf("from 0x%llx to 0x%llx\n",
                 addr+addr_pad,
                 addr+addr_pad + num_bits/8);
         fflush(stdout);
diff --git a/src/gpgpu-sim/gpu-sim.cc b/src/gpgpu-sim/gpu-sim.cc
index 58a5d16..3829861 100644
--- a/src/gpgpu-sim/gpu-sim.cc
+++ b/src/gpgpu-sim/gpu-sim.cc
@@ -32,6 +32,7 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <math.h>
+#include <signal.h>
 #include "zlib.h"
 
 
@@ -1440,7 +1441,7 @@ void gpgpu_sim::cycle()
 
 
       if( g_single_step && ((gpu_sim_cycle+gpu_tot_sim_cycle) >= g_single_step) ) {
-          asm("int $03");
+          raise(SIGTRAP); // Debug breakpoint
       }
       gpu_sim_cycle++;
       if( g_interactive_debugger_enabled ) 
diff --git a/src/gpgpusim_entrypoint.cc b/src/gpgpusim_entrypoint.cc
index 04845e7..ad4587a 100644
--- a/src/gpgpusim_entrypoint.cc
+++ b/src/gpgpusim_entrypoint.cc
@@ -91,6 +91,7 @@ void *gpgpu_sim_thread_sequential(void*)
 pthread_mutex_t g_sim_lock = PTHREAD_MUTEX_INITIALIZER;
 bool g_sim_active = false;
 bool g_sim_done = true;
+bool break_limit = false;
 
 void *gpgpu_sim_thread_concurrent(void*)
 {
@@ -144,11 +145,13 @@ void *gpgpu_sim_thread_concurrent(void*)
                 if(g_the_gpu->cycle_insn_cta_max_hit()){
                     g_stream_manager->stop_all_running_kernels();
                     g_sim_done = true;
+                    break_limit = true;
                 }
             }
 
             active=g_the_gpu->active() || !g_stream_manager->empty_protected();
-        } while( active );
+
+        } while( active && !g_sim_done);
         if(g_debug_execution >= 3) {
            printf("GPGPU-Sim: ** STOP simulation thread (no work) **\n");
            fflush(stdout);
@@ -166,6 +169,11 @@ void *gpgpu_sim_thread_concurrent(void*)
        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);
+    }
+
     sem_post(&g_sim_signal_exit);
     return NULL;
 }
@@ -179,7 +187,7 @@ void synchronize()
     bool done = false;
     do {
         pthread_mutex_lock(&g_sim_lock);
-        done = g_stream_manager->empty() && !g_sim_active;
+        done = ( g_stream_manager->empty() && !g_sim_active ) || g_sim_done;
         pthread_mutex_unlock(&g_sim_lock);
     } while (!done);
     printf("GPGPU-Sim: detected inactive GPU simulation thread\n");