- adding CL_DEVICE_SINGLE_FP_CONFIG (need to get info on what it should be set to)

- bug fix for shader_core object allocation/initialization
- bug fix for ret instruction w/ OpenCL + new PTX ABI
- some formatting of output 

[git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 6910]

6 files changed, 12 insertions, 10 deletions

diff --git a/libopencl/opencl_runtime_api.cc b/libopencl/opencl_runtime_api.cc
index d49cc6e..c5297a3 100644
--- a/libopencl/opencl_runtime_api.cc
+++ b/libopencl/opencl_runtime_api.cc
@@ -1064,6 +1064,7 @@ clGetDeviceInfo(cl_device_id    device,
    case CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG: CL_INT_CASE(1); break;
    case CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT: CL_INT_CASE(1); break;
    case CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE: CL_INT_CASE(0); break;
+   case CL_DEVICE_SINGLE_FP_CONFIG: CL_INT_CASE(0); break;
    default:
       opencl_not_implemented(__my_func__,__LINE__);
    }
diff --git a/src/cuda-sim/cuda-sim.cc b/src/cuda-sim/cuda-sim.cc
index 8703922..2a41e4b 100644
--- a/src/cuda-sim/cuda-sim.cc
+++ b/src/cuda-sim/cuda-sim.cc
@@ -1239,7 +1239,7 @@ void function_info::ptx_exec_inst( ptx_thread_info *thread,
    if ( (g_ptx_sim_num_insn % 100000) == 0 ) {
       dim3 ctaid = thread->get_ctaid();
       dim3 tid = thread->get_tid();
-      printf("%u instructions simulated : ctaid=(%u,%u,%u) tid=(%u,%u,%u)\n",
+      printf("GPGPU-Sim PTX: %u instructions simulated : ctaid=(%u,%u,%u) tid=(%u,%u,%u)\n",
              g_ptx_sim_num_insn, ctaid.x,ctaid.y,ctaid.z,tid.x,tid.y,tid.z );
       fflush(stdout);
    }
diff --git a/src/cuda-sim/ptx_sim.cc b/src/cuda-sim/ptx_sim.cc
index 3974408..bf7b7b6 100644
--- a/src/cuda-sim/ptx_sim.cc
+++ b/src/cuda-sim/ptx_sim.cc
@@ -356,6 +356,7 @@ void ptx_thread_info::callstack_push( unsigned pc, unsigned rpc, const symbol *r
    m_RPC = -1;
    m_RPC_updated = true;
    m_last_was_call = true;
+   assert( m_func_info != NULL );
    m_callstack.push_back( stack_entry(m_symbol_table,m_func_info,pc,rpc,return_var_src,return_var_dst,call_uid) );
    m_regs.push_back( reg_map_t() );
    m_local_mem_stack_pointer += m_func_info->local_mem_framesize(); 
@@ -368,8 +369,6 @@ extern void set_operand_value( const symbol *dst, const ptx_reg_t &data );
 
 bool ptx_thread_info::callstack_pop()
 {
-   assert( !m_callstack.empty() );
-   assert( m_local_mem_stack_pointer >= m_callstack.back().m_func_info->local_mem_framesize() );
    const symbol *rv_src = m_callstack.back().m_return_var_src;
    const symbol *rv_dst = m_callstack.back().m_return_var_dst;
    assert( !((rv_src != NULL) ^ (rv_dst != NULL)) ); // ensure caller and callee agree on whether there is a return value
@@ -384,8 +383,11 @@ bool ptx_thread_info::callstack_pop()
    m_RPC_updated = true;
    m_last_was_call = false;
    m_RPC = m_callstack.back().m_RPC;
+   if( m_callstack.back().m_func_info ) {
+      assert( m_local_mem_stack_pointer >= m_callstack.back().m_func_info->local_mem_framesize() );
+      m_local_mem_stack_pointer -= m_func_info->local_mem_framesize(); 
+   }
    m_func_info = m_callstack.back().m_func_info;
-   m_local_mem_stack_pointer -= m_func_info->local_mem_framesize(); 
    m_callstack.pop_back();
    m_regs.pop_back();
 
diff --git a/src/gpgpu-sim/gpu-sim.cc b/src/gpgpu-sim/gpu-sim.cc
index e7e5067..d07cac6 100644
--- a/src/gpgpu-sim/gpu-sim.cc
+++ b/src/gpgpu-sim/gpu-sim.cc
@@ -855,7 +855,7 @@ unsigned int run_gpu_sim(int grid_num)
 #define DEADLOCK 0
    if (gpu_deadlock_detect && gpu_deadlock) {
       fflush(stdout);
-      printf("ERROR ** deadlock detected: last writeback @ gpu_sim_cycle %u (+ gpu_tot_sim_cycle %u) (%u cycles ago)\n", 
+      printf("GPGPU-Sim uArch: ERROR ** deadlock detected: last writeback @ gpu_sim_cycle %u (+ gpu_tot_sim_cycle %u) (%u cycles ago)\n", 
              (unsigned) gpu_sim_insn_last_update, (unsigned) (gpu_tot_sim_cycle-gpu_sim_cycle),
              (unsigned) (gpu_sim_cycle - gpu_sim_insn_last_update )); 
       fflush(stdout);
@@ -1252,7 +1252,7 @@ int issue_block2core( shader_core_ctx_t *shdr, int grid_num  )
    if (gpgpu_spread_blocks_across_cores) {
       nthreads_2beissued += start_thread;
    }
-   printf("Shader %d initializing CTA #%d with hw tids from %d to %d @(%lld,%lld)", 
+   printf("GPGPU-Sim uArch: Shader %d initializing CTA #%d with hw tids from %d to %d @(%lld,%lld)", 
           shdr->sid, cta_id, start_thread, nthreads_2beissued, gpu_sim_cycle, gpu_tot_sim_cycle );
    printf(" shdr->not_completed = %d\n", shdr->not_completed);
 
@@ -1593,7 +1593,7 @@ void gpu_sim_loop( int grid_num )
          hrs     = elapsed_time/3600 - 24*days;
          minutes = elapsed_time/60 - 60*(hrs + 24*days);
          sec = elapsed_time - 60*(minutes + 60*(hrs + 24*days));
-         printf("cycles: %lld  inst.: %lld (ipc=%4.1f) sim_rate=%u (inst/sec) elapsed = %u:%u:%02u:%02u / %s", 
+         printf("GPGPU-Sim uArch: cycles simulated: %lld  inst.: %lld (ipc=%4.1f) sim_rate=%u (inst/sec) elapsed = %u:%u:%02u:%02u / %s", 
                 gpu_tot_sim_cycle + gpu_sim_cycle, gpu_tot_sim_insn + gpu_sim_insn, 
                 (double)gpu_sim_insn/(double)gpu_sim_cycle,
                 (unsigned)((gpu_tot_sim_insn+gpu_sim_insn) / elapsed_time),
diff --git a/src/gpgpu-sim/shader.cc b/src/gpgpu-sim/shader.cc
index 2271d79..4901d51 100644
--- a/src/gpgpu-sim/shader.cc
+++ b/src/gpgpu-sim/shader.cc
@@ -570,7 +570,8 @@ shader_core_ctx_t* shader_create( const char *name, int sid,
       }
    }
 
-   sc = new shader_core_ctx_t( gpu_n_warp_per_shader, gpgpu_shader_cta );
+   sc = (shader_core_ctx_t*)calloc(sizeof(shader_core_ctx_t),1);
+   sc = new (sc) shader_core_ctx_t( gpu_n_warp_per_shader, gpgpu_shader_cta );
 
    sc->name = name;
    sc->sid = sid;
diff --git a/src/gpgpu-sim/shader.h b/src/gpgpu-sim/shader.h
index 1394481..f67e8fd 100644
--- a/src/gpgpu-sim/shader.h
+++ b/src/gpgpu-sim/shader.h
@@ -446,8 +446,6 @@ typedef struct shader_core_ctx : public core_t
    unsigned int n_registers;   //registers available in the shader core 
    unsigned int n_cta;      //Limit on number of concurrent CTAs in shader core
 
-   //void *req_hist; //not used anywhere
-
    mshr_shader_unit *mshr_unit;
 } shader_core_ctx_t;