watchpoint facility for ptx debugging

[git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 7160]

13 files changed, 553 insertions, 205 deletions

diff --git a/src/abstract_hardware_model.h b/src/abstract_hardware_model.h
index 98d934c..3b3d07e 100644
--- a/src/abstract_hardware_model.h
+++ b/src/abstract_hardware_model.h
@@ -67,62 +67,6 @@ private:
    unsigned m_bank; // n in ".const[n]"; note .const == .const[0] (see PTX 2.1 manual, sec. 5.1.3)
 };
 
-#include <string>
-
-class brk_pt {
-public:
-   brk_pt() { m_valid=false; }
-   brk_pt( const char *fileline, unsigned uid )
-   {
-      m_valid = true;
-      m_watch = false;
-      m_fileline = std::string(fileline);
-      m_thread_uid=uid;
-   }
-   brk_pt( unsigned addr, unsigned value )
-   {
-      m_valid = true;
-      m_watch = true;
-      m_addr = addr;
-      m_value = value;
-   }
-
-   unsigned get_value() const { return m_value; }
-   addr_t get_addr() const { return m_addr; }
-   bool is_valid() const { return m_valid; }
-   bool is_watchpoint() const { return m_watch; }
-   bool is_equal( const std::string &fileline, unsigned uid ) const
-   {
-      if( m_watch ) 
-         return false; 
-      if( (m_thread_uid != (unsigned)-1) && (uid != m_thread_uid) ) 
-         return false;
-      return m_fileline == fileline;
-   }
-   std::string location() const
-   {
-      char buffer[1024];
-      sprintf(buffer,"%s thread uid = %u", m_fileline.c_str(), m_thread_uid);
-      return buffer;
-   }
-
-   unsigned set_value( unsigned val ) { return m_value=val; }
-private:
-   bool         m_valid;
-   bool         m_watch;
-
-   // break point
-   std::string  m_fileline;
-   unsigned     m_thread_uid;
-
-   // watch point
-   unsigned     m_addr;
-   unsigned     m_value;
-};
-
-bool thread_at_brkpt( void *ptx_thd_info, const struct brk_pt &b );
-unsigned read_location( addr_t addr );
- 
 #endif
 
 #endif
diff --git a/src/cuda-sim/cuda-sim.cc b/src/cuda-sim/cuda-sim.cc
index adc9e98..5f568b4 100644
--- a/src/cuda-sim/cuda-sim.cc
+++ b/src/cuda-sim/cuda-sim.cc
@@ -587,7 +587,7 @@ int load_static_globals( symbol_table *symtab, unsigned min_gaddr, unsigned max_
             operand_info op = *i;
             ptx_reg_t value = op.get_literal_value();
             assert( (addr+offset+nbytes) < min_gaddr ); // min_gaddr is start of "heap" for cudaMalloc
-            g_global_mem->write(addr+offset,nbytes,&value); // assuming little endian here
+            g_global_mem->write(addr+offset,nbytes,&value,NULL,NULL); // assuming little endian here
             offset+=nbytes;
             ng_bytes+=nbytes;
          }
@@ -631,7 +631,7 @@ int load_constants( symbol_table *symtab, addr_t min_gaddr )
             unsigned addr=constant->get_address() + nbytes_written;
             assert( addr+nbytes < min_gaddr );
 
-            g_global_mem->write(addr,nbytes,&value); // assume little endian (so u8 is the first byte in u32)
+            g_global_mem->write(addr,nbytes,&value,NULL,NULL); // assume little endian (so u8 is the first byte in u32)
             nc_bytes+=nbytes;
             nbytes_written += nbytes;
          }
@@ -755,7 +755,7 @@ void gpgpu_ptx_sim_memcpy_to_gpu( size_t dst_start_addr, const void *src, size_t
    fflush(stdout);
    char *src_data = (char*)src;
    for (unsigned n=0; n < count; n ++ ) 
-      g_global_mem->write(dst_start_addr+n,1, src_data+n);
+      g_global_mem->write(dst_start_addr+n,1, src_data+n,NULL,NULL);
    printf( " done.\n");
    fflush(stdout);
 }
@@ -779,7 +779,7 @@ void gpgpu_ptx_sim_memcpy_gpu_to_gpu( size_t dst, size_t src, size_t count )
    for (unsigned n=0; n < count; n ++ ) {
       unsigned char tmp;
       g_global_mem->read(src+n,1,&tmp); 
-      g_global_mem->write(dst+n,1, &tmp);
+      g_global_mem->write(dst+n,1, &tmp,NULL,NULL);
    }
    printf( " done.\n");
    fflush(stdout);
@@ -792,7 +792,7 @@ void gpgpu_ptx_sim_memset( size_t dst_start_addr, int c, size_t count )
    fflush(stdout);
    unsigned char c_value = (unsigned char)c;
    for (unsigned n=0; n < count; n ++ ) 
-      g_global_mem->write(dst_start_addr+n,1,&c_value);
+      g_global_mem->write(dst_start_addr+n,1,&c_value,NULL,NULL);
    printf( " done.\n");
    fflush(stdout);
 }
@@ -1058,7 +1058,7 @@ void function_info::finalize( memory_space *param_mem, symbol_table *symtab  )
       const size_t word_size = 4; 
       for (size_t idx = 0; idx < size; idx += word_size) {
          const char *pdata = reinterpret_cast<const char*>(param_value.pdata) + idx; // cast to char * for ptr arithmetic
-         param_mem->write(param_address + idx, word_size, pdata); 
+         param_mem->write(param_address + idx, word_size, pdata,NULL,NULL); 
       }
       param->set_address(param_address);
       param_address += size; 
@@ -1681,7 +1681,7 @@ void gpgpu_ptx_sim_memcpy_symbol(const char *hostVar, const void *src, size_t co
    printf("GPGPU-Sim PTX: gpgpu_ptx_sim_memcpy_symbol: copying %zu bytes %s symbol %s+%zu @0x%x ...\n", 
           count, (to?" to ":"from"), sym_name.c_str(), offset, dst );
    for ( unsigned n=0; n < count; n++ ) {
-      if( to ) mem->write(dst+n,1,((char*)src)+n); 
+      if( to ) mem->write(dst+n,1,((char*)src)+n,NULL,NULL); 
       else mem->read(dst+n,1,((char*)src)+n); 
    }
    fflush(stdout);
@@ -2357,15 +2357,3 @@ void *gpgpusim_opencl_getkernel_Object( const char *kernel_name )
    return i->second;
 }
 
-bool thread_at_brkpt( void *thd, const struct brk_pt &b )
-{
-   ptx_thread_info *thread = (ptx_thread_info *) thd;
-   return b.is_equal(thread->get_location(),thread->get_uid());
-}
-
-unsigned read_location( addr_t addr )
-{
-   unsigned result=0;
-   g_global_mem->read(addr,4,&result);
-   return result;
-}
diff --git a/src/cuda-sim/cuda-sim.h b/src/cuda-sim/cuda-sim.h
new file mode 100644
index 0000000..83e274a
--- /dev/null
+++ b/src/cuda-sim/cuda-sim.h
@@ -0,0 +1,8 @@
+#ifndef CUDASIM_H_INCLUDED
+#define CUDASIM_H_INCLUDED
+
+class memory_space;
+
+extern memory_space *g_global_mem;
+
+#endif
diff --git a/src/cuda-sim/instructions.cc b/src/cuda-sim/instructions.cc
index 3ce0a2f..d663d0a 100644
--- a/src/cuda-sim/instructions.cc
+++ b/src/cuda-sim/instructions.cc
@@ -635,7 +635,7 @@ void atom_callback( void* ptx_inst, void* thd )
 
    // Write operation result into global memory
    // (i.e. copy src1_data to dst)
-   g_global_mem->write(src1_data.u32,size/8,&op_result.s64);
+   g_global_mem->write(src1_data.u32,size/8,&op_result.s64,thread,pI);
 }
 
 // atom_impl will now result in a callback being called in mem_ctrl_pop (gpu-sim.c)
@@ -2638,30 +2638,30 @@ void st_impl( const ptx_instruction *pI, ptx_thread_info *thread )
 
    if (!vector_spec) {
       data = thread->get_operand_value(src1);
-      mem->write(addr,size/8,&data.s64);
+      mem->write(addr,size/8,&data.s64,thread,pI);
    } else {
       if (vector_spec == V2_TYPE) {
          ptx_reg_t* ptx_regs = new ptx_reg_t[2]; 
          thread->get_vector_operand_values(src1, ptx_regs, 2); 
-         mem->write(addr,size/8,&ptx_regs[0].s64);
-         mem->write(addr+size/8,size/8,&ptx_regs[1].s64);
+         mem->write(addr,size/8,&ptx_regs[0].s64,thread,pI);
+         mem->write(addr+size/8,size/8,&ptx_regs[1].s64,thread,pI);
          free(ptx_regs);
       }
       if (vector_spec == V3_TYPE) {
          ptx_reg_t* ptx_regs = new ptx_reg_t[3]; 
          thread->get_vector_operand_values(src1, ptx_regs, 3); 
-         mem->write(addr,size/8,&ptx_regs[0].s64);
-         mem->write(addr+size/8,size/8,&ptx_regs[1].s64);
-         mem->write(addr+2*size/8,size/8,&ptx_regs[2].s64);
+         mem->write(addr,size/8,&ptx_regs[0].s64,thread,pI);
+         mem->write(addr+size/8,size/8,&ptx_regs[1].s64,thread,pI);
+         mem->write(addr+2*size/8,size/8,&ptx_regs[2].s64,thread,pI);
          free(ptx_regs);
       }
       if (vector_spec == V4_TYPE) {
          ptx_reg_t* ptx_regs = new ptx_reg_t[4]; 
          thread->get_vector_operand_values(src1, ptx_regs, 4); 
-         mem->write(addr,size/8,&ptx_regs[0].s64);
-         mem->write(addr+size/8,size/8,&ptx_regs[1].s64);
-         mem->write(addr+2*size/8,size/8,&ptx_regs[2].s64);
-         mem->write(addr+3*size/8,size/8,&ptx_regs[3].s64);
+         mem->write(addr,size/8,&ptx_regs[0].s64,thread,pI);
+         mem->write(addr+size/8,size/8,&ptx_regs[1].s64,thread,pI);
+         mem->write(addr+2*size/8,size/8,&ptx_regs[2].s64,thread,pI);
+         mem->write(addr+3*size/8,size/8,&ptx_regs[3].s64,thread,pI);
          free(ptx_regs);
       }
    }
diff --git a/src/cuda-sim/memory.cc b/src/cuda-sim/memory.cc
index 09c9e81..b825f12 100644
--- a/src/cuda-sim/memory.cc
+++ b/src/cuda-sim/memory.cc
@@ -63,6 +63,7 @@
 
 #include "memory.h"
 #include <stdlib.h>
+#include "../debug.h"
 
 template<unsigned BSIZE> memory_space_impl<BSIZE>::memory_space_impl( std::string name, unsigned hash_size )
 {
@@ -79,13 +80,21 @@ template<unsigned BSIZE> memory_space_impl<BSIZE>::memory_space_impl( std::strin
    assert( m_log2_block_size != (unsigned)-1 );
 }
 
-template<unsigned BSIZE> void memory_space_impl<BSIZE>::write( mem_addr_t addr, size_t length, const void *data )
+template<unsigned BSIZE> void memory_space_impl<BSIZE>::write( mem_addr_t addr, size_t length, const void *data, class ptx_thread_info *thd, const ptx_instruction *pI)
 {
    mem_addr_t index = addr >> m_log2_block_size;
    unsigned offset = addr & (BSIZE-1);
    unsigned nbytes = length;
    assert( (addr+length) <= (index+1)*BSIZE );
    m_data[index].write(offset,nbytes,(const unsigned char*)data);
+   if( !m_watchpoints.empty() ) {
+      std::map<unsigned,mem_addr_t>::iterator i;
+      for( i=m_watchpoints.begin(); i!=m_watchpoints.end(); i++ ) {
+         mem_addr_t wa = i->second;
+         if( ((addr<=wa) && ((addr+length)>wa)) || ((addr>wa) && (addr < (wa+4))) ) 
+            hit_watchpoint(i->first,thd,pI);
+      }
+   }
 }
 
 template<unsigned BSIZE> void memory_space_impl<BSIZE>::read( mem_addr_t addr, size_t length, void *data ) const
@@ -119,6 +128,11 @@ template<unsigned BSIZE> void memory_space_impl<BSIZE>::print( const char *forma
    }
 }
 
+template<unsigned BSIZE> void memory_space_impl<BSIZE>::set_watch( addr_t addr, unsigned watchpoint ) 
+{
+   m_watchpoints[watchpoint]=addr;
+}
+
 template class memory_space_impl<32>;
 template class memory_space_impl<64>;
 template class memory_space_impl<8192>;
@@ -137,7 +151,7 @@ int main(int argc, char *argv[] )
    memory_space *mem = new memory_space_impl<32>("test",4);
    // write address to [address]
    for( mem_addr_t addr=0; addr < 16*1024; addr+=4) 
-      mem->write(addr,4,&addr);
+      mem->write(addr,4,&addr,NULL,NULL);
 
    for( mem_addr_t addr=0; addr < 16*1024; addr+=4) {
       unsigned tmp=0;
@@ -150,7 +164,7 @@ int main(int argc, char *argv[] )
 
    for( mem_addr_t addr=0; addr < 16*1024; addr+=1) {
       unsigned char val = (addr + 128) % 256;
-      mem->write(addr,1,&val);
+      mem->write(addr,1,&val,NULL,NULL);
    }
 
    for( mem_addr_t addr=0; addr < 16*1024; addr+=1) {
diff --git a/src/cuda-sim/memory.h b/src/cuda-sim/memory.h
index 78e7db7..bac4d24 100644
--- a/src/cuda-sim/memory.h
+++ b/src/cuda-sim/memory.h
@@ -62,6 +62,8 @@
 #ifndef memory_h_INCLUDED
 #define memory_h_INCLUDED
 
+#include "../abstract_hardware_model.h"
+
 #ifdef __GNUC__
 #if __GNUC__ >= 4 && __GNUC_MINOR__ >= 3
    #include <unordered_map>
@@ -85,7 +87,8 @@
 #include <string.h>
 #include <stdio.h>
 #include <string>
-#include "../abstract_hardware_model.h"
+#include <map>
+#include <stdlib.h>
 
 typedef address_type mem_addr_t;
 
@@ -95,16 +98,16 @@ template<unsigned BSIZE> class mem_storage {
 public:
    mem_storage( const mem_storage &another )
    {
-      m_data = new unsigned char[ BSIZE ];
+      m_data = (unsigned char*)calloc(1,BSIZE);
       memcpy(m_data,another.m_data,BSIZE);
    }
    mem_storage()
    {
-      m_data = new unsigned char[ BSIZE ];
+      m_data = (unsigned char*)calloc(1,BSIZE);
    }
    ~mem_storage()
    {
-      delete[] m_data;
+      free(m_data);
    }
 
    void write( unsigned offset, size_t length, const unsigned char *data )
@@ -138,28 +141,34 @@ private:
    unsigned char *m_data;
 };
 
+class ptx_thread_info;
+class ptx_instruction;
+
 class memory_space
 {
 public:
    virtual ~memory_space() {}
-   virtual void write( mem_addr_t addr, size_t length, const void *data ) = 0;
+   virtual void write( mem_addr_t addr, size_t length, const void *data, ptx_thread_info *thd, const ptx_instruction *pI ) = 0;
    virtual void read( mem_addr_t addr, size_t length, void *data ) const = 0;
    virtual void print( const char *format, FILE *fout ) const = 0;
+   virtual void set_watch( addr_t addr, unsigned watchpoint ) = 0;
 };
 
 template<unsigned BSIZE> class memory_space_impl : public memory_space {
 public:
    memory_space_impl( std::string name, unsigned hash_size );
 
-   virtual void write( mem_addr_t addr, size_t length, const void *data );
+   virtual void write( mem_addr_t addr, size_t length, const void *data, ptx_thread_info *thd, const ptx_instruction *pI );
    virtual void read( mem_addr_t addr, size_t length, void *data ) const;
    virtual void print( const char *format, FILE *fout ) const;
+   virtual void set_watch( addr_t addr, unsigned watchpoint ); 
 
 private:
    std::string m_name;
    unsigned m_log2_block_size;
    typedef mem_map<mem_addr_t,mem_storage<BSIZE> > map_t;
    map_t m_data;
+   std::map<unsigned,mem_addr_t> m_watchpoints;
 };
 
 #endif
diff --git a/src/cuda-sim/ptx_ir.cc b/src/cuda-sim/ptx_ir.cc
index cfa6650..e0f2316 100644
--- a/src/cuda-sim/ptx_ir.cc
+++ b/src/cuda-sim/ptx_ir.cc
@@ -1416,7 +1416,7 @@ void copy_buffer_to_frame(ptx_thread_info * thread, const arg_buffer_t &a)
       const symbol *dst = a.get_dst();
       addr_t frame_offset = dst->get_address();
       addr_t to_addr = thread->get_local_mem_stack_pointer() + frame_offset;
-      thread->m_local_mem->write(to_addr,size,buffer);
+      thread->m_local_mem->write(to_addr,size,buffer,NULL,NULL);
    }
 }
 
diff --git a/src/debug.cc b/src/debug.cc
new file mode 100644
index 0000000..3e71209
--- /dev/null
+++ b/src/debug.cc
@@ -0,0 +1,176 @@
+#include "debug.h"
+#include "gpgpu-sim/shader.h"
+#include "gpgpu-sim/gpu-sim.h"
+#include "cuda-sim/ptx_sim.h"
+#include "cuda-sim/cuda-sim.h"
+#include "cuda-sim/ptx_ir.h"
+
+#include <map>
+#include <stdio.h>
+#include <string.h>
+
+class watchpoint_event {
+public:
+   watchpoint_event()
+   {
+      m_thread=NULL;
+      m_inst=NULL;
+   }
+   watchpoint_event(const ptx_thread_info *thd, const ptx_instruction *pI) 
+   {
+      m_thread=thd;
+      m_inst = pI;
+   }
+   const ptx_thread_info *thread() const { return m_thread; }
+   const ptx_instruction *inst() const { return m_inst; }
+private:
+   const ptx_thread_info *m_thread;
+   const ptx_instruction *m_inst;
+};
+
+std::map<unsigned,watchpoint_event> g_watchpoint_hits;
+
+void hit_watchpoint( unsigned watchpoint_num, ptx_thread_info *thd, const ptx_instruction *pI )
+{
+   g_watchpoint_hits[watchpoint_num]=watchpoint_event(thd,pI);
+}
+
+/// interactive debugger 
+
+void gpgpu_debug()
+{
+   bool done=true;
+
+   static bool single_step=true;
+   static unsigned next_brkpt=1;
+   static std::map<unsigned,brk_pt> breakpoints;
+
+   /// if single stepping, go to interactive debugger
+
+   if( single_step ) 
+      done=false;
+
+   /// check if we've reached a breakpoint
+
+   for( std::map<unsigned,brk_pt>::iterator i=breakpoints.begin(); i!=breakpoints.end(); i++) {
+      unsigned num=i->first;
+      brk_pt &b=i->second;
+      if( b.is_watchpoint() ) {
+         unsigned addr = b.get_addr();
+         unsigned new_value = read_location(addr);
+         if( new_value != b.get_value() ) {
+            printf( "GPGPU-Sim DBG: watch point %u triggered (old value=%x, new value=%x)\n",
+                     num,b.get_value(),new_value );
+            std::map<unsigned,watchpoint_event>::iterator w=g_watchpoint_hits.find(num);
+            if( w==g_watchpoint_hits.end() ) 
+               printf( "GPGPU-Sim DBG: memory transfer modified value\n");
+            else {
+               watchpoint_event wa = w->second;
+               const ptx_thread_info *thd = wa.thread();
+               const ptx_instruction *pI = wa.inst();
+               printf( "GPGPU-Sim DBG: modified by thread uid=%u, sid=%u, hwtid=%u\n",
+                       thd->get_uid(),thd->get_hw_sid(), thd->get_hw_tid() );
+               printf( "GPGPU-Sim DBG: ");
+               pI->print_insn(stdout);
+               printf( "\n" );
+               g_watchpoint_hits.erase(w);
+            }
+            b.set_value(new_value);
+            done = false; 
+         }
+      } else {
+         for( unsigned sid=0; sid < gpu_n_shader; sid++ ) { 
+            inst_t *fvi = first_valid_thread(sc[sid]->pipeline_reg[IF_ID]);
+            if( !fvi ) continue;
+            if( thread_at_brkpt(fvi->ptx_thd_info, b) ) {
+               done = false;
+               printf("GPGPU-Sim DBG: reached breakpoint %u at %s (sm=%u, hwtid=%u)\n", 
+                      num, b.location().c_str(), sid, fvi->hw_thread_id );
+            }
+         }
+      }
+   }
+
+   if( done ) 
+      assert( g_watchpoint_hits.empty() );
+
+   /// enter interactive debugger loop
+
+   while (!done) {
+      printf("(gpgpu-sim dbg) ");
+      fflush(stdout);
+      
+      char line[1024];
+      fgets(line,1024,stdin);
+
+      char *tok = strtok(line," \t\n");
+      if( !strcmp(tok,"dp") ) {
+         int shader_num = 0;
+         tok = strtok(NULL," \t\n");
+         sscanf(tok,"%d",&shader_num);
+         dump_pipeline_impl((0x40|0x4|0x1),shader_num,0);
+         printf("\n");
+         fflush(stdout);
+      } else if( !strcmp(tok,"q") || !strcmp(tok,"quit") ) {
+         printf("\nreally quit GPGPU-Sim (y/n)?\n");
+         fgets(line,1024,stdin);
+         tok = strtok(NULL," \t\n");
+         if( !strcmp(tok,"y") ) {
+            exit(0);
+         } else {
+            printf("not quiting.\n");
+         }
+      } else if( !strcmp(tok,"b") ) {
+         tok = strtok(NULL," \t\n");
+         char brkpt[1024];
+         sscanf(tok,"%s",brkpt);
+         tok = strtok(NULL," \t\n");
+         unsigned uid;
+         sscanf(tok,"%u",&uid);
+         breakpoints[next_brkpt++] = brk_pt(brkpt,uid);
+      } else if( !strcmp(tok,"d") ) {
+         tok = strtok(NULL," \t\n");
+         unsigned uid;
+         sscanf(tok,"%u",&uid);
+         breakpoints.erase(uid);
+      } else if( !strcmp(tok,"s") ) {
+         done = true;
+      } else if( !strcmp(tok,"c") ) {
+         single_step=false;
+         done = true;
+      } else if( !strcmp(tok,"w") ) {
+         tok = strtok(NULL," \t\n");
+         unsigned addr;
+         sscanf(tok,"%x",&addr);
+         unsigned value = read_location(addr);
+         g_global_mem->set_watch(addr,next_brkpt); 
+         breakpoints[next_brkpt++] = brk_pt(addr,value);
+      } else if( !strcmp(tok,"h") ) {
+         printf("commands:\n");
+         printf("  q                           - quit GPGPU-Sim\n");
+         printf("  b <file>:<line> <thead uid> - set breakpoint\n");
+         printf("  w <global address>          - set watchpoint\n");
+         printf("  del <n>                     - delete breakpoint\n");
+         printf("  s                           - single step one shader cycle (all cores)\n");
+         printf("  c                           - continue simulation without single stepping\n");
+         printf("  dp <n>                      - display pipeline contents on SM <n>\n");
+         printf("  h                           - print this message\n");
+      } else {
+         printf("\ncommand not understood.\n");
+      }
+      fflush(stdout);
+   }
+}
+
+bool thread_at_brkpt( void *thd, const struct brk_pt &b )
+{
+   ptx_thread_info *thread = (ptx_thread_info *) thd;
+   return b.is_equal(thread->get_location(),thread->get_uid());
+}
+
+unsigned read_location( addr_t addr )
+{
+   unsigned result=0;
+   g_global_mem->read(addr,4,&result);
+   return result;
+}
diff --git a/src/debug.h b/src/debug.h
new file mode 100644
index 0000000..463780a
--- /dev/null
+++ b/src/debug.h
@@ -0,0 +1,66 @@
+#ifndef PTX_DEBUG_INCLUDED
+#define PTX_DEBUG_INCLUDED
+
+#include "abstract_hardware_model.h"
+
+#include <string>
+
+class brk_pt {
+public:
+   brk_pt() { m_valid=false; }
+   brk_pt( const char *fileline, unsigned uid )
+   {
+      m_valid = true;
+      m_watch = false;
+      m_fileline = std::string(fileline);
+      m_thread_uid=uid;
+   }
+   brk_pt( unsigned addr, unsigned value )
+   {
+      m_valid = true;
+      m_watch = true;
+      m_addr = addr;
+      m_value = value;
+   }
+
+   unsigned get_value() const { return m_value; }
+   addr_t get_addr() const { return m_addr; }
+   bool is_valid() const { return m_valid; }
+   bool is_watchpoint() const { return m_watch; }
+   bool is_equal( const std::string &fileline, unsigned uid ) const
+   {
+      if( m_watch ) 
+         return false; 
+      if( (m_thread_uid != (unsigned)-1) && (uid != m_thread_uid) ) 
+         return false;
+      return m_fileline == fileline;
+   }
+   std::string location() const
+   {
+      char buffer[1024];
+      sprintf(buffer,"%s thread uid = %u", m_fileline.c_str(), m_thread_uid);
+      return buffer;
+   }
+
+   unsigned set_value( unsigned val ) { return m_value=val; }
+private:
+   bool         m_valid;
+   bool         m_watch;
+
+   // break point
+   std::string  m_fileline;
+   unsigned     m_thread_uid;
+
+   // watch point
+   unsigned     m_addr;
+   unsigned     m_value;
+};
+
+class ptx_thread_info;
+class ptx_instruction;
+bool thread_at_brkpt( void *ptx_thd_info, const struct brk_pt &b );
+unsigned read_location( addr_t addr );
+void hit_watchpoint( unsigned watchpoint_num, ptx_thread_info *thd, const ptx_instruction *pI );
+void gpgpu_debug();
+
+#endif
diff --git a/src/gpgpu-sim/Makefile b/src/gpgpu-sim/Makefile
index 2b1412a..96b1a5c 100644
--- a/src/gpgpu-sim/Makefile
+++ b/src/gpgpu-sim/Makefile
@@ -114,3 +114,251 @@ dram_sched.$(OEXT): ../cuda-sim/ptx.tab.h
 
 # DO NOT DELETE
 
+addrdec.o: /usr/include/string.h /usr/include/features.h
+addrdec.o: /usr/include/sys/cdefs.h /usr/include/bits/wordsize.h
+addrdec.o: /usr/include/gnu/stubs.h /usr/include/gnu/stubs-64.h addrdec.h
+addrdec.o: /usr/include/stdio.h /usr/include/bits/types.h
+addrdec.o: /usr/include/bits/typesizes.h /usr/include/libio.h
+addrdec.o: /usr/include/_G_config.h /usr/include/wchar.h
+addrdec.o: /usr/include/bits/stdio_lim.h /usr/include/bits/sys_errlist.h
+addrdec.o: /usr/include/stdlib.h /usr/include/sys/types.h /usr/include/time.h
+addrdec.o: /usr/include/endian.h /usr/include/bits/endian.h
+addrdec.o: /usr/include/bits/byteswap.h /usr/include/sys/select.h
+addrdec.o: /usr/include/bits/select.h /usr/include/bits/sigset.h
+addrdec.o: /usr/include/bits/time.h /usr/include/sys/sysmacros.h
+addrdec.o: /usr/include/bits/pthreadtypes.h /usr/include/alloca.h
+addrdec.o: /usr/include/assert.h ../option_parser.h
+delayqueue.o: delayqueue.h /usr/include/stdio.h /usr/include/features.h
+delayqueue.o: /usr/include/sys/cdefs.h /usr/include/bits/wordsize.h
+delayqueue.o: /usr/include/gnu/stubs.h /usr/include/gnu/stubs-64.h
+delayqueue.o: /usr/include/bits/types.h /usr/include/bits/typesizes.h
+delayqueue.o: /usr/include/libio.h /usr/include/_G_config.h
+delayqueue.o: /usr/include/wchar.h /usr/include/bits/stdio_lim.h
+delayqueue.o: /usr/include/bits/sys_errlist.h /usr/include/assert.h
+delayqueue.o: /usr/include/stdlib.h /usr/include/sys/types.h
+delayqueue.o: /usr/include/time.h /usr/include/endian.h
+delayqueue.o: /usr/include/bits/endian.h /usr/include/bits/byteswap.h
+delayqueue.o: /usr/include/sys/select.h /usr/include/bits/select.h
+delayqueue.o: /usr/include/bits/sigset.h /usr/include/bits/time.h
+delayqueue.o: /usr/include/sys/sysmacros.h /usr/include/bits/pthreadtypes.h
+delayqueue.o: /usr/include/alloca.h gpu-misc.h ../intersim/statwraper.h
+dram.o: gpu-misc.h dram.h /usr/include/stdio.h /usr/include/features.h
+dram.o: /usr/include/sys/cdefs.h /usr/include/bits/wordsize.h
+dram.o: /usr/include/gnu/stubs.h /usr/include/gnu/stubs-64.h
+dram.o: /usr/include/bits/types.h /usr/include/bits/typesizes.h
+dram.o: /usr/include/libio.h /usr/include/_G_config.h /usr/include/wchar.h
+dram.o: /usr/include/bits/stdio_lim.h /usr/include/bits/sys_errlist.h
+dram.o: /usr/include/stdlib.h /usr/include/sys/types.h /usr/include/time.h
+dram.o: /usr/include/endian.h /usr/include/bits/endian.h
+dram.o: /usr/include/bits/byteswap.h /usr/include/sys/select.h
+dram.o: /usr/include/bits/select.h /usr/include/bits/sigset.h
+dram.o: /usr/include/bits/time.h /usr/include/sys/sysmacros.h
+dram.o: /usr/include/bits/pthreadtypes.h /usr/include/alloca.h delayqueue.h
+dram.o: /usr/include/assert.h ../cuda-sim/dram_callback.h
+dram_sched.o: dram_sched.h dram.h /usr/include/stdio.h
+dram_sched.o: /usr/include/features.h /usr/include/sys/cdefs.h
+dram_sched.o: /usr/include/bits/wordsize.h /usr/include/gnu/stubs.h
+dram_sched.o: /usr/include/gnu/stubs-64.h /usr/include/bits/types.h
+dram_sched.o: /usr/include/bits/typesizes.h /usr/include/libio.h
+dram_sched.o: /usr/include/_G_config.h /usr/include/wchar.h
+dram_sched.o: /usr/include/bits/stdio_lim.h /usr/include/bits/sys_errlist.h
+dram_sched.o: /usr/include/stdlib.h /usr/include/sys/types.h
+dram_sched.o: /usr/include/time.h /usr/include/endian.h
+dram_sched.o: /usr/include/bits/endian.h /usr/include/bits/byteswap.h
+dram_sched.o: /usr/include/sys/select.h /usr/include/bits/select.h
+dram_sched.o: /usr/include/bits/sigset.h /usr/include/bits/time.h
+dram_sched.o: /usr/include/sys/sysmacros.h /usr/include/bits/pthreadtypes.h
+dram_sched.o: /usr/include/alloca.h delayqueue.h /usr/include/assert.h
+dram_sched.o: ../cuda-sim/dram_callback.h shader.h /usr/include/math.h
+dram_sched.o: /usr/include/bits/huge_val.h /usr/include/bits/mathdef.h
+dram_sched.o: /usr/include/bits/mathcalls.h /usr/include/limits.h
+dram_sched.o: /usr/include/bits/posix1_lim.h /usr/include/bits/local_lim.h
+dram_sched.o: /usr/include/linux/limits.h /usr/include/bits/posix2_lim.h
+dram_sched.o: ../cuda-sim/ptx.tab.h gpu-cache.h ../abstract_hardware_model.h
+dram_sched.o: stack.h gpu-sim.h /usr/include/zlib.h /usr/include/zconf.h
+dram_sched.o: /usr/include/unistd.h /usr/include/bits/posix_opt.h
+dram_sched.o: /usr/include/bits/confname.h /usr/include/getopt.h
+dram_sched.o: ../option_parser.h mem_fetch.h addrdec.h gpu-misc.h
+dwf.o: dwf.h ../abstract_hardware_model.h histogram.h /usr/include/string.h
+dwf.o: /usr/include/features.h /usr/include/sys/cdefs.h
+dwf.o: /usr/include/bits/wordsize.h /usr/include/gnu/stubs.h
+dwf.o: /usr/include/gnu/stubs-64.h
+gpu-cache.o: gpu-cache.h /usr/include/stdio.h /usr/include/features.h
+gpu-cache.o: /usr/include/sys/cdefs.h /usr/include/bits/wordsize.h
+gpu-cache.o: /usr/include/gnu/stubs.h /usr/include/gnu/stubs-64.h
+gpu-cache.o: /usr/include/bits/types.h /usr/include/bits/typesizes.h
+gpu-cache.o: /usr/include/libio.h /usr/include/_G_config.h
+gpu-cache.o: /usr/include/wchar.h /usr/include/bits/stdio_lim.h
+gpu-cache.o: /usr/include/bits/sys_errlist.h /usr/include/stdlib.h
+gpu-cache.o: /usr/include/sys/types.h /usr/include/time.h
+gpu-cache.o: /usr/include/endian.h /usr/include/bits/endian.h
+gpu-cache.o: /usr/include/bits/byteswap.h /usr/include/sys/select.h
+gpu-cache.o: /usr/include/bits/select.h /usr/include/bits/sigset.h
+gpu-cache.o: /usr/include/bits/time.h /usr/include/sys/sysmacros.h
+gpu-cache.o: /usr/include/bits/pthreadtypes.h /usr/include/alloca.h
+gpu-cache.o: ../abstract_hardware_model.h gpu-misc.h addrdec.h
+gpu-cache.o: /usr/include/assert.h ../option_parser.h /usr/include/string.h
+gpu-misc.o: gpu-misc.h
+gpu-sim.o: gpu-sim.h /usr/include/stdio.h /usr/include/features.h
+gpu-sim.o: /usr/include/sys/cdefs.h /usr/include/bits/wordsize.h
+gpu-sim.o: /usr/include/gnu/stubs.h /usr/include/gnu/stubs-64.h
+gpu-sim.o: /usr/include/bits/types.h /usr/include/bits/typesizes.h
+gpu-sim.o: /usr/include/libio.h /usr/include/_G_config.h /usr/include/wchar.h
+gpu-sim.o: /usr/include/bits/stdio_lim.h /usr/include/bits/sys_errlist.h
+gpu-sim.o: /usr/include/stdlib.h /usr/include/sys/types.h /usr/include/time.h
+gpu-sim.o: /usr/include/endian.h /usr/include/bits/endian.h
+gpu-sim.o: /usr/include/bits/byteswap.h /usr/include/sys/select.h
+gpu-sim.o: /usr/include/bits/select.h /usr/include/bits/sigset.h
+gpu-sim.o: /usr/include/bits/time.h /usr/include/sys/sysmacros.h
+gpu-sim.o: /usr/include/bits/pthreadtypes.h /usr/include/alloca.h
+gpu-sim.o: /usr/include/math.h /usr/include/bits/huge_val.h
+gpu-sim.o: /usr/include/bits/mathdef.h /usr/include/bits/mathcalls.h
+gpu-sim.o: /usr/include/zlib.h /usr/include/zconf.h /usr/include/unistd.h
+gpu-sim.o: /usr/include/bits/posix_opt.h /usr/include/bits/confname.h
+gpu-sim.o: /usr/include/getopt.h ../option_parser.h shader.h
+gpu-sim.o: /usr/include/limits.h /usr/include/bits/posix1_lim.h
+gpu-sim.o: /usr/include/bits/local_lim.h /usr/include/linux/limits.h
+gpu-sim.o: /usr/include/bits/posix2_lim.h /usr/include/assert.h
+gpu-sim.o: ../cuda-sim/ptx.tab.h ../cuda-sim/dram_callback.h gpu-cache.h
+gpu-sim.o: ../abstract_hardware_model.h delayqueue.h stack.h dram.h
+gpu-sim.o: mem_fetch.h addrdec.h gpu-misc.h icnt_wrapper.h dwf.h
+gpu-sim.o: warp_tracker.h cflogger.h l2cache.h ../cuda-sim/ptx-stats.h
+gpu-sim.o: ../intersim/statwraper.h ../debug.h ../abstract_hardware_model.h
+gpu-sim.o: /usr/include/string.h mem_latency_stat.h
+icnt_wrapper.o: icnt_wrapper.h /usr/include/assert.h /usr/include/features.h
+icnt_wrapper.o: /usr/include/sys/cdefs.h /usr/include/bits/wordsize.h
+icnt_wrapper.o: /usr/include/gnu/stubs.h /usr/include/gnu/stubs-64.h
+icnt_wrapper.o: ../intersim/interconnect_interface.h ../intersim/flit.hpp
+icnt_wrapper.o: ../intersim/booksim.hpp ../intersim/trafficmanager.hpp
+icnt_wrapper.o: ../intersim/module.hpp ../intersim/config_utils.hpp
+icnt_wrapper.o: /usr/include/stdio.h /usr/include/bits/types.h
+icnt_wrapper.o: /usr/include/bits/typesizes.h /usr/include/libio.h
+icnt_wrapper.o: /usr/include/_G_config.h /usr/include/wchar.h
+icnt_wrapper.o: /usr/include/bits/stdio_lim.h /usr/include/bits/sys_errlist.h
+icnt_wrapper.o: ../intersim/network.hpp ../intersim/credit.hpp
+icnt_wrapper.o: ../intersim/router.hpp ../intersim/buffer_state.hpp
+icnt_wrapper.o: ../intersim/stats.hpp ../intersim/traffic.hpp
+icnt_wrapper.o: ../intersim/routefunc.hpp ../intersim/outputset.hpp
+icnt_wrapper.o: ../intersim/injection.hpp
+l2cache.o: /usr/include/stdlib.h /usr/include/features.h
+l2cache.o: /usr/include/sys/cdefs.h /usr/include/bits/wordsize.h
+l2cache.o: /usr/include/gnu/stubs.h /usr/include/gnu/stubs-64.h
+l2cache.o: /usr/include/sys/types.h /usr/include/bits/types.h
+l2cache.o: /usr/include/bits/typesizes.h /usr/include/time.h
+l2cache.o: /usr/include/endian.h /usr/include/bits/endian.h
+l2cache.o: /usr/include/bits/byteswap.h /usr/include/sys/select.h
+l2cache.o: /usr/include/bits/select.h /usr/include/bits/sigset.h
+l2cache.o: /usr/include/bits/time.h /usr/include/sys/sysmacros.h
+l2cache.o: /usr/include/bits/pthreadtypes.h /usr/include/alloca.h
+l2cache.o: /usr/include/stdio.h /usr/include/libio.h /usr/include/_G_config.h
+l2cache.o: /usr/include/wchar.h /usr/include/bits/stdio_lim.h
+l2cache.o: /usr/include/bits/sys_errlist.h /usr/include/string.h
+l2cache.o: ../tr1_hash_map.h ../option_parser.h mem_fetch.h shader.h
+l2cache.o: /usr/include/math.h /usr/include/bits/huge_val.h
+l2cache.o: /usr/include/bits/mathdef.h /usr/include/bits/mathcalls.h
+l2cache.o: /usr/include/limits.h /usr/include/bits/posix1_lim.h
+l2cache.o: /usr/include/bits/local_lim.h /usr/include/linux/limits.h
+l2cache.o: /usr/include/bits/posix2_lim.h /usr/include/assert.h
+l2cache.o: ../cuda-sim/ptx.tab.h ../cuda-sim/dram_callback.h gpu-cache.h
+l2cache.o: ../abstract_hardware_model.h delayqueue.h stack.h dram.h addrdec.h
+l2cache.o: histogram.h l2cache.h ../intersim/statwraper.h
+shader.o: shader.h /usr/include/stdio.h /usr/include/features.h
+shader.o: /usr/include/sys/cdefs.h /usr/include/bits/wordsize.h
+shader.o: /usr/include/gnu/stubs.h /usr/include/gnu/stubs-64.h
+shader.o: /usr/include/bits/types.h /usr/include/bits/typesizes.h
+shader.o: /usr/include/libio.h /usr/include/_G_config.h /usr/include/wchar.h
+shader.o: /usr/include/bits/stdio_lim.h /usr/include/bits/sys_errlist.h
+shader.o: /usr/include/stdlib.h /usr/include/sys/types.h /usr/include/time.h
+shader.o: /usr/include/endian.h /usr/include/bits/endian.h
+shader.o: /usr/include/bits/byteswap.h /usr/include/sys/select.h
+shader.o: /usr/include/bits/select.h /usr/include/bits/sigset.h
+shader.o: /usr/include/bits/time.h /usr/include/sys/sysmacros.h
+shader.o: /usr/include/bits/pthreadtypes.h /usr/include/alloca.h
+shader.o: /usr/include/math.h /usr/include/bits/huge_val.h
+shader.o: /usr/include/bits/mathdef.h /usr/include/bits/mathcalls.h
+shader.o: /usr/include/limits.h /usr/include/bits/posix1_lim.h
+shader.o: /usr/include/bits/local_lim.h /usr/include/linux/limits.h
+shader.o: /usr/include/bits/posix2_lim.h /usr/include/assert.h
+shader.o: ../cuda-sim/ptx.tab.h ../cuda-sim/dram_callback.h gpu-cache.h
+shader.o: ../abstract_hardware_model.h delayqueue.h stack.h dram.h gpu-sim.h
+shader.o: /usr/include/zlib.h /usr/include/zconf.h /usr/include/unistd.h
+shader.o: /usr/include/bits/posix_opt.h /usr/include/bits/confname.h
+shader.o: /usr/include/getopt.h ../option_parser.h mem_fetch.h addrdec.h
+shader.o: dwf.h warp_tracker.h cflogger.h gpu-misc.h ../cuda-sim/ptx_sim.h
+shader.o: ../cuda-sim/dram_callback.h ../cuda-sim/opcodes.h
+shader.o: ../cuda-sim/opcodes.def ../cuda-sim/ptx-stats.h
+shader.o: /usr/include/string.h
+stack.o: stack.h ../abstract_hardware_model.h /usr/include/stdlib.h
+stack.o: /usr/include/features.h /usr/include/sys/cdefs.h
+stack.o: /usr/include/bits/wordsize.h /usr/include/gnu/stubs.h
+stack.o: /usr/include/gnu/stubs-64.h /usr/include/sys/types.h
+stack.o: /usr/include/bits/types.h /usr/include/bits/typesizes.h
+stack.o: /usr/include/time.h /usr/include/endian.h /usr/include/bits/endian.h
+stack.o: /usr/include/bits/byteswap.h /usr/include/sys/select.h
+stack.o: /usr/include/bits/select.h /usr/include/bits/sigset.h
+stack.o: /usr/include/bits/time.h /usr/include/sys/sysmacros.h
+stack.o: /usr/include/bits/pthreadtypes.h /usr/include/alloca.h
+stack.o: /usr/include/assert.h
+stat-tool.o: /usr/include/stdio.h /usr/include/features.h
+stat-tool.o: /usr/include/sys/cdefs.h /usr/include/bits/wordsize.h
+stat-tool.o: /usr/include/gnu/stubs.h /usr/include/gnu/stubs-64.h
+stat-tool.o: /usr/include/bits/types.h /usr/include/bits/typesizes.h
+stat-tool.o: /usr/include/libio.h /usr/include/_G_config.h
+stat-tool.o: /usr/include/wchar.h /usr/include/bits/stdio_lim.h
+stat-tool.o: /usr/include/bits/sys_errlist.h /usr/include/stdlib.h
+stat-tool.o: /usr/include/sys/types.h /usr/include/time.h
+stat-tool.o: /usr/include/endian.h /usr/include/bits/endian.h
+stat-tool.o: /usr/include/bits/byteswap.h /usr/include/sys/select.h
+stat-tool.o: /usr/include/bits/select.h /usr/include/bits/sigset.h
+stat-tool.o: /usr/include/bits/time.h /usr/include/sys/sysmacros.h
+stat-tool.o: /usr/include/bits/pthreadtypes.h /usr/include/alloca.h
+stat-tool.o: /usr/include/assert.h /usr/include/zlib.h /usr/include/zconf.h
+stat-tool.o: /usr/include/unistd.h /usr/include/bits/posix_opt.h
+stat-tool.o: /usr/include/bits/confname.h /usr/include/getopt.h histogram.h
+stat-tool.o: ../abstract_hardware_model.h cflogger.h
+visualizer.o: gpu-sim.h /usr/include/stdio.h /usr/include/features.h
+visualizer.o: /usr/include/sys/cdefs.h /usr/include/bits/wordsize.h
+visualizer.o: /usr/include/gnu/stubs.h /usr/include/gnu/stubs-64.h
+visualizer.o: /usr/include/bits/types.h /usr/include/bits/typesizes.h
+visualizer.o: /usr/include/libio.h /usr/include/_G_config.h
+visualizer.o: /usr/include/wchar.h /usr/include/bits/stdio_lim.h
+visualizer.o: /usr/include/bits/sys_errlist.h /usr/include/stdlib.h
+visualizer.o: /usr/include/sys/types.h /usr/include/time.h
+visualizer.o: /usr/include/endian.h /usr/include/bits/endian.h
+visualizer.o: /usr/include/bits/byteswap.h /usr/include/sys/select.h
+visualizer.o: /usr/include/bits/select.h /usr/include/bits/sigset.h
+visualizer.o: /usr/include/bits/time.h /usr/include/sys/sysmacros.h
+visualizer.o: /usr/include/bits/pthreadtypes.h /usr/include/alloca.h
+visualizer.o: /usr/include/math.h /usr/include/bits/huge_val.h
+visualizer.o: /usr/include/bits/mathdef.h /usr/include/bits/mathcalls.h
+visualizer.o: /usr/include/zlib.h /usr/include/zconf.h /usr/include/unistd.h
+visualizer.o: /usr/include/bits/posix_opt.h /usr/include/bits/confname.h
+visualizer.o: /usr/include/getopt.h ../option_parser.h shader.h
+visualizer.o: /usr/include/limits.h /usr/include/bits/posix1_lim.h
+visualizer.o: /usr/include/bits/local_lim.h /usr/include/linux/limits.h
+visualizer.o: /usr/include/bits/posix2_lim.h /usr/include/assert.h
+visualizer.o: ../cuda-sim/ptx.tab.h ../cuda-sim/dram_callback.h gpu-cache.h
+visualizer.o: ../abstract_hardware_model.h delayqueue.h stack.h dram.h
+visualizer.o: mem_fetch.h addrdec.h /usr/include/string.h
+visualizer.o: ../gpgpu-sim/shader.h ../gpgpu-sim/mem_fetch.h
+warp_tracker.o: warp_tracker.h ../abstract_hardware_model.h shader.h
+warp_tracker.o: /usr/include/stdio.h /usr/include/features.h
+warp_tracker.o: /usr/include/sys/cdefs.h /usr/include/bits/wordsize.h
+warp_tracker.o: /usr/include/gnu/stubs.h /usr/include/gnu/stubs-64.h
+warp_tracker.o: /usr/include/bits/types.h /usr/include/bits/typesizes.h
+warp_tracker.o: /usr/include/libio.h /usr/include/_G_config.h
+warp_tracker.o: /usr/include/wchar.h /usr/include/bits/stdio_lim.h
+warp_tracker.o: /usr/include/bits/sys_errlist.h /usr/include/stdlib.h
+warp_tracker.o: /usr/include/sys/types.h /usr/include/time.h
+warp_tracker.o: /usr/include/endian.h /usr/include/bits/endian.h
+warp_tracker.o: /usr/include/bits/byteswap.h /usr/include/sys/select.h
+warp_tracker.o: /usr/include/bits/select.h /usr/include/bits/sigset.h
+warp_tracker.o: /usr/include/bits/time.h /usr/include/sys/sysmacros.h
+warp_tracker.o: /usr/include/bits/pthreadtypes.h /usr/include/alloca.h
+warp_tracker.o: /usr/include/math.h /usr/include/bits/huge_val.h
+warp_tracker.o: /usr/include/bits/mathdef.h /usr/include/bits/mathcalls.h
+warp_tracker.o: /usr/include/limits.h /usr/include/bits/posix1_lim.h
+warp_tracker.o: /usr/include/bits/local_lim.h /usr/include/linux/limits.h
+warp_tracker.o: /usr/include/bits/posix2_lim.h /usr/include/assert.h
+warp_tracker.o: ../cuda-sim/ptx.tab.h ../cuda-sim/dram_callback.h gpu-cache.h
+warp_tracker.o: delayqueue.h stack.h dram.h
diff --git a/src/gpgpu-sim/gpu-sim.cc b/src/gpgpu-sim/gpu-sim.cc
index 2b4ebe3..1bec72f 100644
--- a/src/gpgpu-sim/gpu-sim.cc
+++ b/src/gpgpu-sim/gpu-sim.cc
@@ -82,6 +82,7 @@
 #include "../cuda-sim/ptx-stats.h"
 #include "../intersim/statwraper.h"
 #include "../abstract_hardware_model.h"
+#include "../debug.h"
 
 #include <stdio.h>
 #include <string.h>
@@ -206,7 +207,6 @@ void check_time_vector_update(unsigned int uid,int slot ,long int latency,int ty
 void node_req_hist_clear(void *p);
 void node_req_hist_dump(void *p);
 void node_req_hist_update(void * p,int node, long long cycle);
-static void gpgpu_debug();
 
 /* functionally simulated memory */
 extern struct mem_t *mem;
@@ -1765,111 +1765,3 @@ void dump_pipeline()
    dump_pipeline_impl(0,-1,-1);
 }
 
-/// interactive debugger 
-
-static void gpgpu_debug()
-{
-   bool done=true;
-
-   static bool single_step=true;
-   static unsigned next_brkpt=1;
-   static std::map<unsigned,brk_pt> breakpoints;
-
-   /// if single stepping, go to interactive debugger
-
-   if( single_step ) 
-      done=false;
-
-   /// check if we've reached a breakpoint
-
-   for( std::map<unsigned,brk_pt>::iterator i=breakpoints.begin(); i!=breakpoints.end(); i++) {
-      unsigned num=i->first;
-      brk_pt &b=i->second;
-      if( b.is_watchpoint() ) {
-         unsigned addr = b.get_addr();
-         unsigned new_value = read_location(addr);
-         if( new_value != b.get_value() ) {
-            printf( "GPGPU-Sim DBG: watch point %u triggered (old value=%x, new value=%x)\n",
-                     num,b.get_value(),new_value );
-            b.set_value(new_value);
-            done = false; 
-         }
-      } else {
-         for( unsigned sid=0; sid < gpu_n_shader; sid++ ) { 
-            inst_t *fvi = first_valid_thread(sc[sid]->pipeline_reg[IF_ID]);
-            if( !fvi ) continue;
-            if( thread_at_brkpt(fvi->ptx_thd_info, b) ) {
-               done = false;
-               printf("GPGPU-Sim DBG: reached breakpoint %u at %s (sm=%u, hwtid=%u)\n", 
-                      num, b.location().c_str(), sid, fvi->hw_thread_id );
-            }
-         }
-      }
-   }
-
-   /// enter interactive debugger loop
-
-   while (!done) {
-      printf("(gpgpu-sim dbg) ");
-      fflush(stdout);
-      
-      char line[1024];
-      fgets(line,1024,stdin);
-
-      char *tok = strtok(line," \t\n");
-      if( !strcmp(tok,"dp") ) {
-         int shader_num = 0;
-         tok = strtok(NULL," \t\n");
-         sscanf(tok,"%d",&shader_num);
-         dump_pipeline_impl((0x40|0x4|0x1),shader_num,0);
-         printf("\n");
-         fflush(stdout);
-      } else if( !strcmp(tok,"q") || !strcmp(tok,"quit") ) {
-         printf("\nreally quit GPGPU-Sim (y/n)?\n");
-         fgets(line,1024,stdin);
-         tok = strtok(NULL," \t\n");
-         if( !strcmp(tok,"y") ) {
-            exit(0);
-         } else {
-            printf("not quiting.\n");
-         }
-      } else if( !strcmp(tok,"b") ) {
-         tok = strtok(NULL," \t\n");
-         char brkpt[1024];
-         sscanf(tok,"%s",brkpt);
-         tok = strtok(NULL," \t\n");
-         unsigned uid;
-         sscanf(tok,"%u",&uid);
-         breakpoints[next_brkpt++] = brk_pt(brkpt,uid);
-      } else if( !strcmp(tok,"d") ) {
-         tok = strtok(NULL," \t\n");
-         unsigned uid;
-         sscanf(tok,"%u",&uid);
-         breakpoints.erase(uid);
-      } else if( !strcmp(tok,"s") ) {
-         done = true;
-      } else if( !strcmp(tok,"c") ) {
-         single_step=false;
-         done = true;
-      } else if( !strcmp(tok,"w") ) {
-         tok = strtok(NULL," \t\n");
-         unsigned addr;
-         sscanf(tok,"%x",&addr);
-         unsigned value = read_location(addr);
-         breakpoints[next_brkpt++] = brk_pt(addr,value);
-      } else if( !strcmp(tok,"h") ) {
-         printf("commands:\n");
-         printf("  q                           - quit GPGPU-Sim\n");
-         printf("  b <file>:<line> <thead uid> - set breakpoint\n");
-         printf("  w <global address>          - set watchpoint\n");
-         printf("  del <n>                     - delete breakpoint\n");
-         printf("  s                           - single step one shader cycle (all cores)\n");
-         printf("  c                           - continue simulation without single stepping\n");
-         printf("  dp <n>                      - display pipeline contents on SM <n>\n");
-         printf("  h                           - print this message\n");
-      } else {
-         printf("\ncommand not understood.\n");
-      }
-      fflush(stdout);
-   }
-}
diff --git a/src/gpgpu-sim/gpu-sim.h b/src/gpgpu-sim/gpu-sim.h
index e30797e..27d83ad 100644
--- a/src/gpgpu-sim/gpu-sim.h
+++ b/src/gpgpu-sim/gpu-sim.h
@@ -112,5 +112,6 @@ void gpu_print_stat();
 int mem_ctrl_full( int mc_id );
 
 void dramqueue_latency_log_dump();
+void dump_pipeline_impl( int mask, int s, int m );
 
 #endif
diff --git a/src/gpgpu-sim/shader.h b/src/gpgpu-sim/shader.h
index 9f64da5..9dd7dc8 100644
--- a/src/gpgpu-sim/shader.h
+++ b/src/gpgpu-sim/shader.h
@@ -896,5 +896,7 @@ unsigned int max_cta_per_shader( shader_core_ctx_t *shader);
 #define WB_RT 6
 #define ID_OC 7
 
+extern unsigned int gpu_n_shader;
+extern shader_core_ctx_t **sc;
 
 #endif /* SHADER_H */