Merge branch 'dev' of https://github.com/gpgpu-sim/gpgpu-sim_distribution into dev

24 files changed, 920 insertions, 61 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/.travis.yml b/.travis.yml
index fabfc5b..3f2fb6c 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -19,5 +19,7 @@ matrix:
       env: CONFIG=configs.quadro5800.yml
     - services: docker
       env: CONFIG=configs.teslac2050.yml
+    - services: docker
+      env: CONFIG=configs.gtx1080ti.yml
 
-script: docker run -v `pwd`:/home/runner/gpgpu-sim_distribution:rw tgrogers/gpgpu-sim_regress:latest /bin/bash -c "./start_torque.sh; chown -R runner /home/runner/gpgpu-sim_distribution; su - runner -c 'source /home/runner/gpgpu-sim_distribution/setup_environment && make -j -C /home/runner/gpgpu-sim_distribution && cd /home/runner/gpgpu-sim_simulations/ && git pull && /home/runner/gpgpu-sim_simulations/util/job_launching/run_simulations.py -c /home/runner/gpgpu-sim_simulations/util/job_launching/regression_recipies/rodinia_2.0-ft/$CONFIG -N regress && /home/runner/gpgpu-sim_simulations/util/job_launching/monitor_func_test.py -v -N regress'"
+script: docker run -v `pwd`:/home/runner/gpgpu-sim_distribution:rw tgrogers/gpgpu-sim_regress:latest /bin/bash -c "./start_torque.sh; chown -R runner /home/runner/gpgpu-sim_distribution; su - runner -c 'export CUDA_INSTALL_PATH=/usr/local/cuda-4.2/ && source /home/runner/gpgpu-sim_distribution/setup_environment && make -j -C /home/runner/gpgpu-sim_distribution && cd /home/runner/gpgpu-sim_simulations/ && git pull && /home/runner/gpgpu-sim_simulations/util/job_launching/run_simulations.py -c /home/runner/gpgpu-sim_simulations/util/job_launching/regression_recipies/rodinia_2.0-ft/$CONFIG -N regress && /home/runner/gpgpu-sim_simulations/util/job_launching/monitor_func_test.py -v -N regress'"
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/aerialvision/guiclasses.py b/aerialvision/guiclasses.py
index 0a7013b..45fed26 100644
--- a/aerialvision/guiclasses.py
+++ b/aerialvision/guiclasses.py
@@ -69,7 +69,7 @@ import matplotlib
 matplotlib.use('TkAgg')
 from matplotlib.backends.backend_tkagg import FigureCanvasTkAgg, NavigationToolbar2TkAgg
 from matplotlib.figure import Figure
-from matplotlib import mpl
+import matplotlib as mpl
 from matplotlib.colors import colorConverter
 from matplotlib import pyplot
 import Pmw
diff --git a/configs/GeForceGTX1080Ti/config_fermi_islip.icnt b/configs/GeForceGTX1080Ti/config_fermi_islip.icnt
new file mode 100644
index 0000000..2a69ddd
--- /dev/null
+++ b/configs/GeForceGTX1080Ti/config_fermi_islip.icnt
@@ -0,0 +1,70 @@
+//21*1 fly with 32 flits per packet under gpgpusim injection mode
+use_map = 0;
+flit_size = 32; 
+
+// currently we do not use this, see subnets below
+network_count = 2;
+
+// Topology
+topology = fly;
+k = 50;
+n = 1;
+
+// Routing
+
+routing_function = dest_tag;
+
+// Flow control
+
+num_vcs     = 1;
+vc_buf_size = 8;
+
+wait_for_tail_credit = 0;
+
+// Router architecture
+
+vc_allocator = islip; //separable_input_first;
+sw_allocator = islip; //separable_input_first;
+alloc_iters  = 1;
+
+credit_delay   = 0;
+routing_delay  = 0;
+vc_alloc_delay = 1;
+sw_alloc_delay = 1;
+
+input_speedup     = 2;
+output_speedup    = 1;
+internal_speedup  = 1.0;
+
+// Traffic, GPGPU-Sim does not use this
+
+traffic                = uniform;
+packet_size ={{1,2,3,4},{10,20}};
+packet_size_rate={{1,1,1,1},{2,1}};
+
+// Simulation - Don't change
+
+sim_type       = gpgpusim;
+//sim_type = latency;
+injection_rate = 0.1;
+
+subnets = 2;
+
+// Always use read and write no matter following line
+//use_read_write = 1;
+
+
+read_request_subnet = 0;
+read_reply_subnet = 1;
+write_request_subnet = 0;
+write_reply_subnet = 1;
+
+read_request_begin_vc = 0;
+read_request_end_vc = 0;
+write_request_begin_vc = 0;
+write_request_end_vc = 0;
+read_reply_begin_vc = 0;
+read_reply_end_vc = 0;
+write_reply_begin_vc = 0;
+write_reply_end_vc = 0;
+
diff --git a/configs/GeForceGTX1080Ti/gpgpusim.config b/configs/GeForceGTX1080Ti/gpgpusim.config
new file mode 100644
index 0000000..47c2b6a
--- /dev/null
+++ b/configs/GeForceGTX1080Ti/gpgpusim.config
@@ -0,0 +1,149 @@
+# This config models the Pascal GP102 (GeForceGTX 1080Ti)
+
+# functional simulator specification
+-gpgpu_ptx_instruction_classification 0
+-gpgpu_ptx_sim_mode 0
+-gpgpu_ptx_force_max_capability 20 
+
+# SASS execution (only supported with CUDA >= 4.0)
+-gpgpu_ptx_convert_to_ptxplus 0
+-gpgpu_ptx_save_converted_ptxplus 0
+
+# high level architecture configuration
+-gpgpu_n_clusters 28
+-gpgpu_n_cores_per_cluster 1
+-gpgpu_n_mem 11
+-gpgpu_n_sub_partition_per_mchannel 2 
+
+# Pascal clock domains
+#-gpgpu_clock_domains <Core Clock>:<Interconnect Clock>:<L2 Clock>:<DRAM Clock>
+# Pascal NVIDIA TITAN X clock domains are adopted from 
+# https://en.wikipedia.org/wiki/GeForce_10_series
+-gpgpu_clock_domains 1481.0:2962.0:1481.0:2750.0
+
+# shader core pipeline config
+-gpgpu_shader_registers 65536
+
+# This implies a maximum of 64 warps/SM
+-gpgpu_shader_core_pipeline 2048:32 
+-gpgpu_shader_cta 32
+-gpgpu_simd_model 1 
+
+# Pipeline widths and number of FUs
+# ID_OC_SP,ID_OC_SFU,ID_OC_MEM,OC_EX_SP,OC_EX_SFU,OC_EX_MEM,EX_WB
+## Pascal GP102 has 4 SP SIMD units and 1 SFU unit
+## we need to scale the number of pipeline registers to be equal to the number of SP units
+-gpgpu_pipeline_widths 4,1,1,4,1,1,6
+-gpgpu_num_sp_units 4
+-gpgpu_num_sfu_units 1
+
+# Instruction latencies and initiation intervals
+# "ADD,MAX,MUL,MAD,DIV"
+# SFU is 32-width in pascal, then dp units initiation is 1 cycle
+-ptx_opcode_latency_int 4,13,4,5,145
+-ptx_opcode_initiation_int 1,2,2,2,8
+-ptx_opcode_latency_fp 4,13,4,5,39
+-ptx_opcode_initiation_fp 1,2,1,1,4
+-ptx_opcode_latency_dp 8,19,8,8,330
+-ptx_opcode_initiation_dp 1,2,1,1,130
+
+# <nsets>:<bsize>:<assoc>,<rep>:<wr>:<alloc>:<wr_alloc>:<set_index_fn>,<mshr>:<N>:<merge>,<mq>:**<fifo_entry>
+# ** Optional parameter - Required when mshr_type==Texture Fifo
+# Note: Hashing set index function (H) only applies to a set size of 32 or 64. 
+# Pascal GP102 has 96KB Shared memory
+# Pascal GP102 has 64KB L1 cache
+# The default is to disable the L1 cache, unless cache modifieres is used
+-gpgpu_cache:dl1  64:128:6,L:L:m:N:H,A:128:8,8
+-gpgpu_shmem_size 98304
+-gmem_skip_L1D 1
+
+# 64 sets, each 128 bytes 16-way for each memory sub partition (128 KB per memory sub partition). This gives 3MB L2 cache
+-gpgpu_cache:dl2 64:128:16,L:B:m:W:L,A:1024:1024,4:0,32 # used to be 128:4
+-gpgpu_cache:dl2_texture_only 0 
+
+# 4 KB Inst.
+-gpgpu_cache:il1 8:128:4,L:R:f:N:L,A:2:48,4
+# 48 KB Tex 
+-gpgpu_tex_cache:l1 16:128:24,L:R:m:N:L,F:128:4,128:2
+# 12 KB Const
+-gpgpu_const_cache:l1 128:64:2,L:R:f:N:L,A:2:64,4
+
+# enable operand collector 
+## larger operand collectors and reg_banks are needed for the 4 warp schedulers and 4 SIMD units
+-gpgpu_operand_collector_num_units_sp 20
+-gpgpu_operand_collector_num_units_sfu 4
+-gpgpu_operand_collector_num_units_mem 8
+-gpgpu_operand_collector_num_in_ports_sp 4
+-gpgpu_operand_collector_num_out_ports_sp 4
+-gpgpu_operand_collector_num_in_ports_sfu 1
+-gpgpu_operand_collector_num_out_ports_sfu 1
+-gpgpu_operand_collector_num_in_ports_mem 1
+-gpgpu_operand_collector_num_out_ports_mem 1
+# gpgpu_num_reg_banks should be increased to 32, but it gives an error!
+-gpgpu_num_reg_banks 32
+
+# shared memory bankconflict detection 
+-gpgpu_shmem_num_banks 32
+-gpgpu_shmem_limited_broadcast 0
+-gpgpu_shmem_warp_parts 1
+
+## In Pascal, a warp scheduler can issue 2 insts per cycle
+-gpgpu_max_insn_issue_per_warp 2
+
+# interconnection
+-network_mode 1 
+-inter_config_file config_fermi_islip.icnt
+
+# memory partition latency config 
+-rop_latency 120
+-dram_latency 100
+
+# dram model config
+-gpgpu_dram_scheduler 1
+# The DRAM return queue and the scheduler queue together should provide buffer
+# to sustain the memory level parallelism to tolerate DRAM latency 
+# To allow 100% DRAM utility, there should at least be enough buffer to sustain
+# the minimum DRAM latency (100 core cycles).  I.e. 
+#   Total buffer space required = 100 x 924MHz / 700MHz = 132
+-gpgpu_frfcfs_dram_sched_queue_size 64
+-gpgpu_dram_return_queue_size 116
+
+# for NVIDIA GeForceGTX 1080Ti, bus width is 352bits (11 DRAM chips x 32 bits)
+# 11 memory paritions, 4 bytes (1 DRAM chip) per memory partition
+# the atom size of GDDR5X (the smallest read request) is 32 bytes 
+-gpgpu_n_mem_per_ctrlr 1
+-gpgpu_dram_buswidth 4
+-gpgpu_dram_burst_length 8
+-dram_data_command_freq_ratio 4  # GDDR5X is QDR
+-gpgpu_mem_address_mask 1
+-gpgpu_mem_addr_mapping dramid@8;00000000.00000000.00000000.00000000.0000RRRR.RRRRRRRR.RBBBCCCC.BCCSSSSS
+
+# Use the same GDDR5 timing from hynix H5GQ1H24AFR
+# disable bank groups for now, set nbkgrp to 1 and tCCDL and tRTPL to 0
+-gpgpu_dram_timing_opt "nbk=16:CCD=2:RRD=6:RCD=12:RAS=28:RP=12:RC=40:
+                        CL=12:WL=4:CDLR=5:WR=12:nbkgrp=1:CCDL=0:RTPL=0"
+
+# Pascal has four schedulers per core
+-gpgpu_num_sched_per_core 2
+# Two Level Scheduler with active and pending pools
+#-gpgpu_scheduler two_level_active:6:0:1
+# Loose round robbin scheduler
+#-gpgpu_scheduler lrr
+# Greedy then oldest scheduler
+-gpgpu_scheduler gto
+
+# stat collection
+-gpgpu_memlatency_stat 14 
+-gpgpu_runtime_stat 500
+-enable_ptx_file_line_stats 1
+-visualizer_enabled 0
+
+# power model configs
+-power_simulation_enabled 1
+-gpuwattch_xml_file gpuwattch_gtx1080Ti.xml
+
+# tracing functionality
+#-trace_enabled 1
+#-trace_components WARP_SCHEDULER,SCOREBOARD
+#-trace_sampling_core 0
+
diff --git a/configs/GeForceGTX1080Ti/gpuwattch_gtx1080Ti.xml b/configs/GeForceGTX1080Ti/gpuwattch_gtx1080Ti.xml
new file mode 100755
index 0000000..02619ff
--- /dev/null
+++ b/configs/GeForceGTX1080Ti/gpuwattch_gtx1080Ti.xml
@@ -0,0 +1,538 @@
+<?xml version="1.0" ?>
+<component id="root" name="root">
+	<component id="system" name="system">
+		<!--McPAT will skip the components if number is set to 0 -->
+		<param name="GPU_Architecture" value="1"/><!-- 0-G80; 1-Fermi; others not supported -->
+		<param name="number_of_cores" value="28"/>
+		<param name="architecture" value="1"/> <!-- fermi:1 quadro:2 other: undefined-->
+		<param name="number_of_L1Directories" value="0"/>
+		<param name="number_of_L2Directories" value="0"/>
+		<param name="number_of_L2s" value="1"/> <!-- This number means how many L2 clusters in each cluster there can be multiple banks/ports -->
+		<param name="number_of_L3s" value="0"/> <!-- This number means how many L3 clusters -->
+		<param name="number_of_NoCs" value="1"/> 
+		<param name="homogeneous_cores" value="1"/><!--1 means homo -->
+		<param name="homogeneous_L2s" value="1"/>
+		<param name="homogeneous_L1Directorys" value="1"/>
+		<param name="homogeneous_L2Directorys" value="1"/>
+		<param name="homogeneous_L3s" value="1"/>
+		<param name="homogeneous_ccs" value="1"/><!--cache coherece hardware -->
+		<param name="homogeneous_NoCs" value="1"/>
+		<param name="core_tech_node" value="23"/><!-- nm -->
+		<param name="target_core_clockrate" value="1481"/><!--MHz -->
+		<param name="temperature" value="380"/> <!-- Kelvin -->
+		<param name="number_cache_levels" value="2"/>
+		<param name="interconnect_projection_type" value="0"/><!--0: agressive wire technology; 1: conservative wire technology -->
+		<param name="device_type" value="0"/><!--0: HP(High Performance Type); 1: LSTP(Low standby power) 2: LOP (Low Operating Power)  -->
+		<param name="longer_channel_device" value="1"/><!-- 0 no use; 1 use when possible -->
+		<param name="machine_bits" value="32"/>
+		<param name="virtual_address_width" value="32"/>
+		<param name="physical_address_width" value="32"/>
+		<param name="virtual_memory_page_size" value="4096"/>
+		<param name="idle_core_power" value="1.59"/><!-- idle core power for GTX479 -->
+		<!--param name="scaling_coefficients" value="10,0.0884816,10,10,8,10,4.12782,10,2.48832,10,10,10,4.29982,0.387764,0.0714269,0.14302,0.01,0.546811,0.485351,0.806633,0.818073,1.9207,100,100,100,87.9303,100,10,4.3548,10"/-->
+		<param name="TOT_INST" value="10" />
+		<param name="FP_INT" value="10" />
+		<param name="IC_H" value="0.001" />
+		<param name="IC_M" value="10" />
+		<param name="DC_RH" value="1" />
+		<param name="DC_RM" value="1" />
+		<param name="DC_WH" value="1" />
+		<param name="DC_WM" value="1" />
+		<param name="TC_H" value="0.001" />
+		<param name="TC_M" value="10" />
+		<param name="CC_H" value="4.5071" />
+		<param name="CC_M" value="10" />
+		<param name="SHRD_ACC" value="10" />
+		<param name="REG_RD" value="1.6294" />
+		<param name="REG_WR" value="0.5031" />
+		<param name="NON_REG_OPs" value="0.01" />
+		<param name="SP_ACC" value="10" />
+		<param name="SFU_ACC" value="0.0082" />
+		<param name="FPU_ACC" value="0.4126" />
+		<param name="MEM_RD" value="0.1234" />
+		<param name="MEM_WR" value="0.001" />
+		<param name="MEM_PRE" value="0.001" />
+		<param name="L2_RH" value="100" />
+		<param name="L2_RM" value="100" />
+		<param name="L2_WH" value="100" />
+		<param name="L2_WM" value="42.6966" />
+		<param name="NOC_A" value="100" />
+		<param name="PIPE_A" value="44.8085" />
+		<param name="IDLE_CORE_N" value="2.0382"/>
+		<param name="CONST_DYNAMICN" value="5.0005" />
+		<stat name="num_idle_cores" value="0"/><!-- Average Number of idle cores during this period -->
+		<stat name="total_cycles" value="total_cycles_match_mcpat"/>
+		<stat name="idle_cycles" value="idle_cycles_match_mcpat"/>
+		<stat name="busy_cycles"  value="busy_cycles_match_mcpat"/>
+			<!--This page size(B) is complete different from the page size in Main memo secction. this page size is the size of 
+			virtual memory from OS/Archi perspective; the page size in Main memo secction is the actuall physical line in a DRAM bank  -->
+		<!-- *********************** cores ******************* -->
+		<component id="system.core0" name="core0">
+			<!-- Core property -->
+			<param name="clock_rate" value="1481"/>
+			<param name="instruction_length" value="32"/>
+			<param name="opcode_width" value="9"/>
+			<!-- address width determins the tag_width in Cache, LSQ and buffers in cache controller 
+			default value is machine_bits, if not set --> 
+			<param name="machine_type" value="1"/><!-- 1 inorder; 0 OOO-->
+			<!-- inorder/OoO -->
+			<param name="number_hardware_threads" value="32"/>
+			<!-- number_instruction_fetch_ports(icache ports) is always 1 in single-thread processor,
+			it only may be more than one in SMT processors. BTB ports always equals to fetch ports since 
+			branch information in consective branch instructions in the same fetch group can be read out from BTB once.--> 
+			<param name="fetch_width" value="1"/>
+			<!-- fetch_width determins the size of cachelines of L1 cache block -->
+			<param name="number_instruction_fetch_ports" value="1"/>
+			<param name="decode_width" value="1"/>
+			<!-- decode_width determins the number of ports of the 
+			renaming table (both RAM and CAM) scheme -->
+			<param name="issue_width" value="2"/>
+			<!-- issue_width determins the number of ports of Issue window and other logic 
+			as in the complexity effective proccessors paper; issue_width==dispatch_width -->
+			<param name="commit_width" value="2"/>
+			<!-- commit_width determins the number of ports of register files -->
+			<param name="fp_issue_width" value="1"/>
+			<param name="prediction_width" value="0"/> 
+			<!-- number of branch instructions can be predicted simultannouesl-->
+			<!-- Current version of McPAT does not distinguish int and floating point pipelines 
+			Theses parameters are reserved for future use.--> 
+			<param name="pipelines_per_core" value="1,1"/>
+			<!--integer_pipeline and floating_pipelines, if the floating_pipelines is 0, then the pipeline is shared-->
+			<param name="pipeline_depth" value="8,8"/>
+			<!-- pipeline depth of int and fp, if pipeline is shared, the second number is the average cycles of fp ops -->
+			<!-- issue and exe unit-->
+			<param name="ALU_per_core" value="32"/>
+			<!-- contains an adder, a shifter, and a logical unit -->
+			<param name="MUL_per_core" value="4"/>
+			<!-- For MUL and Div -->
+			<param name="FPU_per_core" value="32"/>		
+			<!-- buffer between IF and ID stage -->
+			<param name="instruction_buffer_size" value="1"/>
+			<!-- buffer between ID and sche/exe stage -->
+			<param name="decoded_stream_buffer_size" value="1"/>
+			<param name="instruction_window_scheme" value="0"/><!-- 0 PHYREG based, 1 RSBASED-->
+			<!-- McPAT support 2 types of OoO cores, RS based and physical reg based-->
+			<param name="instruction_window_size" value="1"/>
+			<param name="fp_instruction_window_size" value="1"/>
+			<!-- the instruction issue Q as in Alpha 21264; The RS as in Intel P6 -->
+			<param name="ROB_size" value="0"/>
+			<!-- each in-flight instruction has an entry in ROB -->
+			<!-- registers -->
+			<!-- SM parameters  Added by Syed Gilani -->
+			<param name="rf_banks" value="32"/>
+			<param name="simd_width" value="32"/>
+			<param name="collector_units" value="32"/>
+			<param name="core_clock_ratio" value="2"/>
+			<param name="warp_size" value="32"/>
+
+			<param name="archi_Regs_IRF_size" value="65536"/>			
+			<param name="archi_Regs_FRF_size" value="32"/>
+			<!--  if OoO processor, phy_reg number is needed for renaming logic, 
+			renaming logic is for both integer and floating point insts.  -->
+			<param name="phy_Regs_IRF_size" value="32"/>
+			<param name="phy_Regs_FRF_size" value="32"/>
+			<!-- rename logic -->
+			<param name="rename_scheme" value="0"/>
+			<!-- can be RAM based(0) or CAM based(1) rename scheme 
+			RAM-based scheme will have free list, status table;
+			CAM-based scheme have the valid bit in the data field of the CAM 
+			both RAM and CAM need RAM-based checkpoint table, checkpoint_depth=# of in_flight instructions;
+			Detailed RAT Implementation see TR -->
+			<param name="register_windows_size" value="0"/>
+			<!-- how many windows in the windowed register file, sun processors;
+			no register windowing is used when this number is 0 -->
+			<!-- In OoO cores, loads and stores can be issued whether inorder(Pentium Pro) or (OoO)out-of-order(Alpha),
+			They will always try to exeute out-of-order though. -->
+			<param name="LSU_order" value="inorder"/>
+			<param name="store_buffer_size" value="32"/>
+			<!-- By default, in-order cores do not have load buffers -->
+			<param name="load_buffer_size" value="32"/>	
+			<!-- number of ports refer to sustainable concurrent memory accesses --> 
+			<param name="memory_ports" value="2"/>	
+			<!-- max_allowed_in_flight_memo_instructions determins the # of ports of load and store buffer
+			as well as the ports of Dcache which is connected to LSU -->	
+			<!-- dual-pumped Dcache can be used to save the extra read/write ports -->
+			<param name="RAS_size" value="1"/>						
+			<!-- general stats, defines simulation periods;require total, idle, and busy cycles for senity check  -->
+			<!-- please note: if target architecture is X86, then all the instrucions refer to (fused) micro-ops -->
+			<stat name="total_instructions" value="total_instructions_match_mcpat"/>
+			<stat name="int_instructions" value="int_instruction_match_mcpat"/>
+			<stat name="fp_instructions" value="flt_instruction_match_mcpat"/>
+			<stat name="branch_instructions" value="branch_instruction_match_mcpat"/>
+			<stat name="branch_mispredictions" value="0"/>
+			<stat name="load_instructions" value="load_instruction_match_mcpat"/>
+			<stat name="store_instructions" value="store_instruction_match_mcpat"/>
+			<stat name="committed_instructions" value="total_instructions_match_mcpat"/>
+			<stat name="committed_int_instructions" value="int_instruction_match_mcpat"/>
+			<stat name="committed_fp_instructions" value="flt_instruction_match_mcpat"/>
+			<stat name="pipeline_duty_cycle" value="0.6"/><!--<=1, runtime_ipc/peak_ipc; averaged for all cores if homogenous -->
+			<!-- the following cycle stats are used for heterogeneouse cores only, 
+				please ignore them if homogeneouse cores -->
+			<stat name="total_cycles" value="total_cycles_match_mcpat"/>
+		    <stat name="idle_cycles" value="idle_cycles_match_mcpat"/>
+		    <stat name="busy_cycles"  value="busy_cycles_match_mcpat"/>
+			<!-- instruction buffer stats -->
+			<!-- ROB stats, both RS and Phy based OoOs have ROB
+			performance simulator should capture the difference on accesses,
+			otherwise, McPAT has to guess based on number of commited instructions. -->
+			<stat name="ROB_reads" value="263886"/>
+			<stat name="ROB_writes" value="263886"/>
+			<!-- RAT accesses -->
+			<stat name="rename_accesses" value="263886"/>
+			<stat name="fp_rename_accesses" value="263886"/>
+			<!-- decode and rename stage use this, should be total ic - nop -->
+			<!-- Inst window stats -->
+			<stat name="inst_window_reads" value="263886"/>
+			<stat name="inst_window_writes" value="263886"/>
+			<stat name="inst_window_wakeup_accesses" value="263886"/>
+			<stat name="fp_inst_window_reads" value="263886"/>
+			<stat name="fp_inst_window_writes" value="263886"/>
+			<stat name="fp_inst_window_wakeup_accesses" value="263886"/>
+			<!--  RF accesses -->
+			<stat name="int_regfile_reads" value="int_register_read_access_match_mcpat"/>
+			<stat name="float_regfile_reads" value="int_register_write_access_match_mcpat"/>
+			<stat name="int_regfile_writes" value="float_register_read_access_match_mcpat"/>
+			<stat name="float_regfile_writes" value="float_register_write_access_match_mcpat"/>
+			
+			<!-- The following stat is for operand collector power - Added by Syed -->
+			<stat name="non_rf_operands" value="0"/>
+
+			<!-- accesses to the working reg -->
+			<stat name="function_calls" value="0"/>
+			<stat name="context_switches" value="0"/> <!--not used in the McPAT -->
+			<!-- Number of Windowes switches (number of function calls and returns)-->
+			<!-- Alu stats by default, the processor has one FPU that includes the divider and 
+			 multiplier. The fpu accesses should include accesses to multiplier and divider  -->
+			<stat name="ialu_accesses" value="ialu_accesses_match_mcpat"/>			
+			<stat name="fpu_accesses" value="fpu_accesses_match_mcpat"/>
+			<stat name="mul_accesses" value="mul_accesses_match_mcpat"/>
+			<stat name="cdb_alu_accesses" value="0"/>
+			<stat name="cdb_mul_accesses" value="0"/>
+			<stat name="cdb_fpu_accesses" value="0"/>
+			<!-- multiple cycle accesses should be counted multiple times, 
+			otherwise, McPAT can use internal counter for different floating point instructions 
+			to get final accesses. But that needs detailed info for floating point inst mix -->
+			<!--  currently the performance simulator should 
+			make sure all the numbers are final numbers, 
+			including the explicit read/write accesses, 
+			and the implicite accesses such as replacements and etc.
+			Future versions of McPAT may be able to reason the implicite access
+			based on param and stats of last level cache
+			The same rule applies to all cache access stats too!  -->
+			<!-- following is AF for max power computation. 
+				Do not change them, unless you understand them-->
+			<stat name="IFU_duty_cycle" value="0.25"/>			
+			<stat name="LSU_duty_cycle" value="0.25"/>
+			<stat name="MemManU_I_duty_cycle" value="1"/>
+			<stat name="MemManU_D_duty_cycle" value="0.25"/>
+			<stat name="ALU_duty_cycle" value="0.9"/>
+			<stat name="MUL_duty_cycle" value="0.5"/>
+			<stat name="FPU_duty_cycle" value="1"/><!-- FPU numbers are already average -->
+			<stat name="ALU_cdb_duty_cycle" value="0.9"/>
+			<stat name="MUL_cdb_duty_cycle" value="0.5"/>
+			<stat name="FPU_cdb_duty_cycle" value="15"/>	
+			<component id="system.core0.predictor" name="PBT">
+				<!-- branch predictor; tournament predictor see Alpha implementation -->
+				<param name="local_predictor_size" value="10,3"/>
+				<param name="local_predictor_entries" value="1024"/>
+				<param name="global_predictor_entries" value="4096"/>
+				<param name="global_predictor_bits" value="2"/>
+				<param name="chooser_predictor_entries" value="4096"/>
+				<param name="chooser_predictor_bits" value="2"/>
+				<!-- These parameters can be combined like below in next version
+				<param name="load_predictor" value="10,3,1024"/>
+				<param name="global_predictor" value="4096,2"/>
+				<param name="predictor_chooser" value="4096,2"/>
+				-->
+			</component>
+			<component id="system.core0.itlb" name="itlb">
+				<param name="number_entries" value="1"/>
+				<stat name="total_accesses" value="0"/>
+				<stat name="total_misses" value="0"/>
+				<stat name="conflicts" value="0"/>	
+				<!-- there is no write requests to itlb although writes happen to itlb after miss, 
+				which is actually a replacement -->
+			</component>
+			<component id="system.core0.icache" name="icache">
+				<!-- there is no write requests to itlb although writes happen to it after miss, 
+				which is actually a replacement -->
+				<param name="icache_config" value="16384,128,4,1,1,3,8,0"/>
+				<!-- the parameters are capacity,block_width, associativity, bank, throughput w.r.t. core clock, latency w.r.t. core clock,output_width, cache policy -->
+				<!-- cache_policy;//0 no write or write-though with non-write allocate;1 write-back with write-allocate -->
+				<param name="buffer_sizes" value="16, 16, 16,0"/>
+				<!-- cache controller buffer sizes: miss_buffer_size(MSHR),fill_buffer_size,prefetch_buffer_size,wb_buffer_size--> 
+				<stat name="read_accesses" value="total_instructions_match_mcpat"/>
+				<stat name="read_misses" value="0"/>
+				<stat name="conflicts" value="0"/>				
+			</component>
+			<component id="system.core0.dtlb" name="dtlb">
+				<param name="number_entries" value="1"/>
+				<stat name="total_accesses" value="0"/>
+				<stat name="total_misses" value="0"/>
+				<stat name="conflicts" value="0"/>	
+			</component>
+			<component id="system.core0.ccache" name="ccache">
+			        <!-- all the buffer related are optional -->
+				<param name="ccache_config" value="16384,64,2,1,1,3,8,0"/>
+				<param name="buffer_sizes" value="16, 16, 16, 0"/>
+				<!-- cache controller buffer sizes: miss_buffer_size(MSHR),fill_buffer_size,prefetch_buffer_size,wb_buffer_size-->	
+				<stat name="read_accesses" value="ccache_read_accesses_match_mcpat"/>
+				<stat name="write_accesses" value="0"/>
+				<stat name="read_misses" value="ccache_read_misses_match_mcpat"/>
+				<stat name="write_misses" value="0"/>
+				<stat name="conflicts" value="0"/>	
+			</component>
+			<component id="system.core0.tcache" name="tcache">
+			        <!-- all the buffer related are optional -->
+				<param name="tcache_config" value="49152,128,8,1,1,3,8,0"/>
+				<param name="buffer_sizes" value="16, 16, 16, 0"/>
+				<!-- cache controller buffer sizes: miss_buffer_size(MSHR),fill_buffer_size,prefetch_buffer_size,wb_buffer_size-->	
+				<stat name="read_accesses" value="tcache_read_accesses_match_mcpat"/>
+				<stat name="write_accesses" value="0"/>
+				<stat name="read_misses" value="tcache_read_misses_match_mcpat"/>
+				<stat name="write_misses" value="0"/>
+				<stat name="conflicts" value="0"/>	
+			</component>
+      <!--model the shared memory by mimicing dcache-->
+			<component id="system.core0.sharedmemory" name="sharedmemory">
+			        <!-- all the buffer related are optional -->
+				<param name="sharedmemory_config" value="98304,16,1,16,1,3,16,0"/>
+				<!-- the parameters are capacity,block_width, associativity, bank, throughput w.r.t. core clock, latency w.r.t. core clock,output_width, cache policy -->
+				<param name="buffer_sizes" value="16, 16, 16, 16"/>
+				<!-- cache controller buffer sizes: miss_buffer_size(MSHR),fill_buffer_size,prefetch_buffer_size,wb_buffer_size-->	
+				<stat name="read_accesses" value="sharedmemory_read_access_match_mcpat"/>
+				<stat name="write_accesses" value="sharedmemory_write_access_match_mcpat"/>
+				<stat name="read_misses" value="0"/>
+				<stat name="write_misses" value="0"/>
+				<stat name="conflicts" value="0"/>	
+			</component>
+			<component id="system.core0.dcache" name="dcache">
+			        <!-- all the buffer related are optional -->
+				<param name="dcache_config" value="16384,32,4,1,1,3,8,0"/>
+				<param name="buffer_sizes" value="16, 16, 16, 0"/>
+				<!-- cache controller buffer sizes: miss_buffer_size(MSHR),fill_buffer_size,prefetch_buffer_size,wb_buffer_size-->	
+				<stat name="read_accesses" value="dcache_read_access_match_mcpat"/>
+				<stat name="write_accesses" value="dcache_write_access_match_mcpat"/>
+				<stat name="read_misses" value="dcache_read_miss_match_mcpat"/>
+				<stat name="write_misses" value="dcache_write_miss_match_mcpat"/>
+				<stat name="conflicts" value="0"/>	
+			</component>
+			<component id="system.core0.BTB" name="BTB">
+			        <!-- all the buffer related are optional -->
+				<param name="BTB_config" value="8192,4,2,1, 1,3"/>
+				<!-- the parameters are capacity,block_width,associativity,bank, throughput w.r.t. core clock, latency w.r.t. core clock,-->
+			</component>
+	</component>
+		<component id="system.L1Directory0" name="L1Directory0">
+				<param name="Directory_type" value="0"/>
+			    <!--0 cam based shadowed tag. 1 directory cache -->	
+				<param name="Dir_config" value="2048,1,0,1, 4, 4,8"/>
+				<!-- the parameters are capacity,block_width, associativity,bank, throughput w.r.t. core clock, latency w.r.t. core clock,-->
+			    <param name="buffer_sizes" value="8, 8, 8, 8"/>	
+				<!-- all the buffer related are optional -->
+			    <param name="clockrate" value="1400"/>
+				<param name="ports" value="1,1,1"/>
+				<!-- number of r, w, and rw search ports -->
+				<param name="device_type" value="0"/>
+				<!-- altough there are multiple access types, 
+				Performance simulator needs to cast them into reads or writes
+				e.g. the invalidates can be considered as writes -->
+				<stat name="read_accesses" value="800000"/>
+				<stat name="write_accesses" value="27276"/>
+				<stat name="read_misses" value="1632"/>
+				<stat name="write_misses" value="183"/>
+				<stat name="conflicts" value="20"/>	
+				<stat name="duty_cycle" value="0.45"/>	
+		</component>
+		<component id="system.L2Directory0" name="L2Directory0">
+				<param name="Directory_type" value="1"/>
+			    <!--0 cam based shadowed tag. 1 directory cache -->	
+				<param name="Dir_config" value="1048576,16,16,1,2, 100"/>
+				<!-- the parameters are capacity,block_width, associativity,bank, throughput w.r.t. core clock, latency w.r.t. core clock,-->
+			    <param name="buffer_sizes" value="8, 8, 8, 8"/>	
+				<!-- all the buffer related are optional -->
+			    <param name="clockrate" value="1400"/>
+				<param name="ports" value="1,1,1"/>
+				<!-- number of r, w, and rw search ports -->
+				<param name="device_type" value="0"/>
+				<!-- altough there are multiple access types, 
+				Performance simulator needs to cast them into reads or writes
+				e.g. the invalidates can be considered as writes -->
+				<stat name="read_accesses" value="0"/>
+				<stat name="write_accesses" value="0"/>
+				<stat name="read_misses" value="0"/>
+				<stat name="write_misses" value="0"/>
+				<stat name="conflicts" value="0"/>
+			    <stat name="duty_cycle" value="0.45"/>		
+		</component>
+		<component id="system.L20" name="L20">
+			<!-- all the buffer related are optional -->
+				<param name="L2_config" value="131072,128,16,1, 4,23, 64, 1"/>
+			    <!-- consider 4-way bank interleaving for Niagara 1 -->
+				<!-- the parameters are capacity,block_width, associativity, bank, throughput w.r.t. core clock, latency w.r.t. core clock,output_width, cache policy -->
+				<param name="buffer_sizes" value="16, 16, 16, 16"/>
+				<!-- cache controller buffer sizes: miss_buffer_size(MSHR),fill_buffer_size,prefetch_buffer_size,wb_buffer_size-->	
+				<param name="clockrate" value="2962"/>
+				<param name="ports" value="1,1,1"/>
+				<!-- number of r, w, and rw ports -->
+				<param name="device_type" value="0"/>
+				<stat name="read_accesses" value="200000"/>
+				<stat name="write_accesses" value="0"/>
+				<stat name="read_misses" value="0"/>
+				<stat name="write_misses" value="0"/>
+				<stat name="conflicts" value="0"/>	
+			    <stat name="duty_cycle" value="0.5"/>	
+		</component>
+		
+<!--**********************************************************************-->
+<component id="system.L30" name="L30">
+				<param name="L3_config" value="1048576,64,16,1, 2,100, 64,1"/>
+				<!-- the parameters are capacity,block_width, associativity, bank, throughput w.r.t. core clock, latency w.r.t. core clock,output_width, cache policy -->
+				<param name="clockrate" value="3500"/>
+				<param name="ports" value="1,1,1"/>
+				<!-- number of r, w, and rw ports -->
+				<param name="device_type" value="0"/>
+				<param name="buffer_sizes" value="16, 16, 16, 16"/>
+				<!-- cache controller buffer sizes: miss_buffer_size(MSHR),fill_buffer_size,prefetch_buffer_size,wb_buffer_size-->	
+				<stat name="read_accesses" value="58824"/>
+				<stat name="write_accesses" value="27276"/>
+				<stat name="read_misses" value="1632"/>
+				<stat name="write_misses" value="183"/>
+				<stat name="conflicts" value="0"/>	
+	            <stat name="duty_cycle" value="0.35"/>	
+		</component>
+
+
+<!--**********************************************************************-->
+		<component id="system.NoC0" name="noc0">
+			<param name="clockrate" value="700"/>
+			<param name="type" value="1"/>
+			<!-- 1 NoC, O bus -->
+			<param name="horizontal_nodes" value="2"/>
+			<param name="vertical_nodes" value="1"/>
+			<param name="has_global_link" value="0"/>
+			<!-- 1 has global link, 0 does not have global link -->
+			<param name="link_throughput" value="1"/><!--w.r.t clock -->
+			<param name="link_latency" value="1"/><!--w.r.t clock -->
+			<!-- througput >= latency -->
+			<!-- Router architecture -->
+			<param name="input_ports" value="6"/>
+			<param name="output_ports" value="6"/>
+			<param name="virtual_channel_per_port" value="1"/>
+			<!-- input buffer; in classic routers only input ports need buffers -->
+			<param name="flit_bits" value="32"/>
+			<param name="input_buffer_entries_per_vc" value="1"/><!--VCs within the same ports share input buffers whose size is propotional to the number of VCs-->
+			<param name="chip_coverage" value="1"/>
+			<!-- When multiple NOC present, one NOC will cover part of the whole chip. chip_coverage <=1 -->
+			<stat name="total_accesses" value="0"/>
+			<!-- This is the number of total accesses within the whole network not for each router -->
+			<stat name="duty_cycle" value="0.6"/>
+		</component>	
+<!--**********************************************************************-->
+<!--**********************************************************************-->
+
+		<component id="system.mem" name="mem">
+			<!-- Main memory property -->
+			<param name="mem_tech_node" value="23"/>
+			<param name="device_clock" value="200"/><!--MHz, this is clock rate of the actual memory device, not the FSB -->
+			<param name="peak_transfer_rate" value="3200"/><!--MB/S-->
+			<param name="internal_prefetch_of_DRAM_chip" value="4"/>
+			<!-- 2 for DDR, 4 for DDR2, 8 for DDR3...-->
+			<!-- the device clock, peak_transfer_rate, and the internal prefetch decide the DIMM property -->
+			<!-- above numbers can be easily found from Wikipedia -->
+			<param name="capacity_per_channel" value="4096"/> <!-- MB -->
+			<!-- capacity_per_Dram_chip=capacity_per_channel/number_of_dimms/number_ranks/Dram_chips_per_rank
+			Current McPAT assumes single DIMMs are used.--> 		
+			<param name="number_ranks" value="2"/>
+			<param name="num_banks_of_DRAM_chip" value="6"/>			
+			<param name="Block_width_of_DRAM_chip" value="64"/> <!-- B -->
+			<param name="output_width_of_DRAM_chip" value="8"/>
+			<!--number of Dram_chips_per_rank=" 72/output_width_of_DRAM_chip-->
+			<!--number of Dram_chips_per_rank=" 72/output_width_of_DRAM_chip-->
+			<param name="page_size_of_DRAM_chip" value="8"/> <!-- 8 or 16 -->
+			<param name="burstlength_of_DRAM_chip" value="8"/>
+			<stat name="memory_accesses" value="1052"/>
+			<stat name="memory_reads" value="1052"/>
+			<stat name="memory_writes" value="1052"/>									
+		</component>
+		<component id="system.mc" name="mc">
+			<!-- Memeory controllers are for DDR(2,3...) DIMMs -->
+			<!-- current version of McPAT uses published values for base parameters of memory controller
+			improvments on MC will be added in later versions. -->
+			<param name="type" value="0"/> <!-- 1: low power; 0 high performance -->
+			<param name="mc_clock" value="1848"/><!--DIMM IO bus clock rate MHz DDR2-400 for Niagara 1--> 
+			<param name="peak_transfer_rate" value="29568"/><!--MB/S  Syed: GTX 470 has 177.4GB/s mem transfer rate with 6 MCs -->
+			<param name="block_size" value="64"/><!--B-->
+			<param name="number_mcs" value="6"/><!-- 6 GDDR5 memory controllers  -->
+			<!-- current McPAT only supports homogeneous memory controllers -->
+			<param name="memory_channels_per_mc" value="2"/>
+			<param name="number_ranks" value="1"/>
+			<param name="withPHY" value="0"/>
+			<!-- # of ranks of each channel-->
+			<param name="req_window_size_per_channel" value="16"/>
+			<param name="IO_buffer_size_per_channel" value="16"/>
+			<param name="databus_width" value="32"/>
+			<param name="addressbus_width" value="32"/>
+			<param name="PRT_entries" value="32"/>
+			<!-- # of empirical DRAM model parameter -->
+			<param name="dram_cmd_coeff" value="0"/>
+			<param name="dram_act_coeff" value="0"/>
+			<param name="dram_nop_coeff" value="0"/>
+			<param name="dram_activity_coeff" value="0"/>
+			<param name="dram_pre_coeff" value="3.8475e-8f"/>
+			<param name="dram_rd_coeff" value="7.74707143e-8f"/>
+			<param name="dram_wr_coeff" value="3.54664286e-8f"/>
+			<param name="dram_req_coeff" value="0"/>
+			<param name="dram_const_coeff" value="0"/>
+
+			<!-- McPAT will add the control bus width to the addressbus width automatically -->
+			<stat name="memory_accesses" value="memory_accesses_match_mcpat"/>
+			<stat name="memory_reads" value="memory_reads_match_mcpat"/>
+			<stat name="memory_writes" value="memory_writes_match_mcpat"/>
+			<!-- McPAT does not track individual mc, instead, it takes the total accesses and calculate 
+			the average power per MC or per channel. This is sufficent for most application. 
+			Further trackdown can be easily added in later versions. -->  			
+		</component>
+<!--**********************************************************************-->
+		<component id="system.niu" name="niu">
+			<!-- On chip 10Gb Ethernet NIC, including XAUI Phy and MAC controller  -->
+			<!-- For a minimum IP packet size of 84B at 10Gb/s, a new packet arrives every 67.2ns. 
+				 the low bound of clock rate of a 10Gb MAC is 150Mhz -->
+			<param name="type" value="0"/> <!-- 1: low power; 0 high performance -->
+			<param name="clockrate" value="350"/>
+			<param name="number_units" value="0"/> <!-- unlike PCIe and memory controllers, each Ethernet controller only have one port -->
+			<stat name="duty_cycle" value="1.0"/> <!-- achievable max load <= 1.0 -->
+			<stat name="total_load_perc" value="0.7"/> <!-- ratio of total achived load to total achivable bandwidth  -->
+			<!-- McPAT does not track individual nic, instead, it takes the total accesses and calculate 
+			the average power per nic or per channel. This is sufficent for most application. -->  			
+		</component>
+<!--**********************************************************************-->
+		<component id="system.pcie" name="pcie">
+			<!-- On chip PCIe controller, including Phy-->
+			<!-- For a minimum PCIe packet size of 84B at 8Gb/s per lane (PCIe 3.0), a new packet arrives every 84ns. 
+				 the low bound of clock rate of a PCIe per lane logic is 120Mhz -->
+			<param name="type" value="0"/> <!-- 1: low power; 0 high performance -->
+			<param name="withPHY" value="1"/>
+			<param name="clockrate" value="350"/>
+			<param name="number_units" value="0"/>
+			<param name="num_channels" value="8"/> <!-- 2 ,4 ,8 ,16 ,32 -->
+			<stat name="duty_cycle" value="1.0"/> <!-- achievable max load <= 1.0 -->
+			<stat name="total_load_perc" value="0.7"/> <!-- Percentage of total achived load to total achivable bandwidth  -->
+			<!-- McPAT does not track individual pcie controllers, instead, it takes the total accesses and calculate 
+			the average power per pcie controller or per channel. This is sufficent for most application. -->  			
+		</component>
+<!--**********************************************************************-->
+		<component id="system.flashc" name="flashc">
+		    <param name="number_flashcs" value="0"/>
+			<param name="type" value="1"/> <!-- 1: low power; 0 high performance -->
+            <param name="withPHY" value="1"/>
+			<param name="peak_transfer_rate" value="200"/><!--Per controller sustainable reak rate MB/S -->
+			<stat name="duty_cycle" value="1.0"/> <!-- achievable max load <= 1.0 -->
+			<stat name="total_load_perc" value="0.7"/> <!-- Percentage of total achived load to total achivable bandwidth  -->
+			<!-- McPAT does not track individual flash controller, instead, it takes the total accesses and calculate 
+			the average power per fc or per channel. This is sufficent for most application -->  			
+		</component>
+<!--**********************************************************************-->
+
+		</component>
+</component>
diff --git a/libcuda/cuda_runtime_api.cc b/libcuda/cuda_runtime_api.cc
index cbe8a11..9bdb993 100644
--- a/libcuda/cuda_runtime_api.cc
+++ b/libcuda/cuda_runtime_api.cc
@@ -1406,7 +1406,7 @@ void extract_code_using_cuobjdump(){
 		cmd << "ldd " << app_binary << " | grep $CUDA_INSTALL_PATH | awk \'{print $3}\' > _tempfile_.txt";
 		int result = system(cmd.str().c_str());
 		if(result){
-			std::cout << "Failed to execute: " << cmd << std::endl;
+			std::cout << "Failed to execute: " << cmd.str() << std::endl;
 			exit(1);
 		}
 		std::ifstream libsf;
@@ -1438,7 +1438,7 @@ void extract_code_using_cuobjdump(){
 			if(result) {printf("ERROR: Failed to execute: %s\n", command); exit(1);}
 			std::cout << "Done" << std::endl;
 
-			std::cout << "Trying to parse " << libcodfn << std::endl;
+			std::cout << "Trying to parse " << libcodfn.str() << std::endl;
 			cuobjdump_in = fopen(libcodfn.str().c_str(), "r");
 			cuobjdump_parse();
 			fclose(cuobjdump_in);
@@ -1540,7 +1540,7 @@ std::list<cuobjdumpSection*> pruneSectionList(std::list<cuobjdumpSection*> cuobj
 
 //! Merge all PTX sections that have a specific identifier into one file
 std::list<cuobjdumpSection*> mergeMatchingSections(std::list<cuobjdumpSection*> cuobjdumpSectionList, std::string identifier){
-	char *ptxcode = "";
+	const char *ptxcode = "";
 	std::list<cuobjdumpSection*>::iterator old_iter;
 	cuobjdumpPTXSection* old_ptxsection = NULL;
 	cuobjdumpPTXSection* ptxsection;
@@ -1689,7 +1689,7 @@ std::map<int, bool>fatbin_registered;
 std::map<std::string, symbol_table*> name_symtab;
 
 //! Keep track of the association between filename and cubin handle
-void cuobjdumpRegisterFatBinary(unsigned int handle, char* filename){
+void cuobjdumpRegisterFatBinary(unsigned int handle, const char* filename){
 	fatbinmap[handle] = filename;
 }
 
@@ -1764,6 +1764,7 @@ void** CUDARTAPI __cudaRegisterFatBinary( void *fatCubin )
 		if (sizeof(void*) == 4) 
 			printf("GPGPU-Sim PTX: FatBin file name extraction has not been tested on 32-bit system.\n"); 
 
+		#if (CUDART_VERSION <= 6000)
 		// FatBin handle from the .fatbin.c file (one of the intermediate files generated by NVCC)
 		typedef struct {int m; int v; const unsigned long long* d; char* f;} __fatDeviceText __attribute__ ((aligned (8))); 
 		__fatDeviceText * fatDeviceText = (__fatDeviceText *) fatCubin;
@@ -1772,12 +1773,11 @@ void** CUDARTAPI __cudaRegisterFatBinary( void *fatCubin )
 		// - Obtains the pointer to the actual fatbin structure from the FatBin handle (fatCubin).
 		// - An integer inside the fatbin structure contains the relative offset to the source code file name.
 		// - This offset differs among different CUDA and GCC versions. 
-		#if (CUDART_VERSION <= 6000)
 		char * pfatbin = (char*) fatDeviceText->d; 
 		int offset = *((int*)(pfatbin+48)); 
 		char * filename = (pfatbin+16+offset); 
 		#else
-		char * filename = "default";
+		const char * filename = "default";
 		#endif
 		// The extracted file name is associated with a fat_cubin_handle passed
 		// into cudaLaunch().  Inside cudaLaunch(), the associated file name is
@@ -1798,7 +1798,7 @@ void** CUDARTAPI __cudaRegisterFatBinary( void *fatCubin )
 
 		return (void**)fat_cubin_handle;
 	} 
-	#if (CUDART_VERSION < 8000)
+#if (CUDART_VERSION < 8000)
 	else {
 		static unsigned source_num=1;
 		unsigned long long fat_cubin_handle = next_fat_bin_handle++;
@@ -1856,7 +1856,12 @@ void** CUDARTAPI __cudaRegisterFatBinary( void *fatCubin )
 		}
 		return (void**)fat_cubin_handle;
 	}
-	#endif
+#else
+        else {
+		printf("ERROR **  __cudaRegisterFatBinary() needs to be updated\n");
+		abort();
+        }
+#endif
 }
 
 void __cudaUnregisterFatBinary(void **fatCubinHandle)
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/setup_environment b/setup_environment
index 854a335..3f2476e 100644
--- a/setup_environment
+++ b/setup_environment
@@ -6,7 +6,12 @@ export GPGPUSIM_SETUP_ENVIRONMENT_WAS_RUN=
 export GPGPUSIM_ROOT="$( cd "$( dirname "$BASH_SOURCE" )" && pwd )"
 
 GPGPUSIM_VERSION_STRING=`cat $GPGPUSIM_ROOT/version | awk '/Version/ {print $8}'`
-GPGPUSIM_BUILD_STRING=`cat $GPGPUSIM_ROOT/version | awk '/Change/ {print $6}'`
+#Detect Git branch and commit #
+GIT_COMMIT=`git log -n 1 | head -1 | sed -re 's/commit (.*)/\1/'`
+GIT_FILES_CHANGED=`git diff --numstat | wc | sed -re 's/^\s+([0-9]+).*/\1./'`
+GIT_FILES_CHANGED+=`git diff --numstat --cached | wc | sed -re 's/^\s+([0-9]+).*/\1/'`
+GPGPUSIM_BUILD_STRING="gpgpu-sim_git-commit-$GIT_COMMIT-modified_$GIT_FILES_CHANGED"
+
 echo -n "GPGPU-Sim version $GPGPUSIM_VERSION_STRING (build $GPGPUSIM_BUILD_STRING) ";
 
 if [ ! -n "$CUDA_INSTALL_PATH" ]; then
@@ -43,11 +48,9 @@ CC_VERSION=`gcc --version | head -1 | awk '{for(i=1;i<=NF;i++){ if(match($i,/^[0
 
 CUDA_VERSION_STRING=`$CUDA_INSTALL_PATH/bin/nvcc --version | awk '/release/ {print $5;}' | sed 's/,//'`;
 CUDA_VERSION_NUMBER=`echo $CUDA_VERSION_STRING | sed 's/\./ /' | awk '{printf("%02u%02u", 10*int($1), 10*$2);}'`
-if [ $CUDA_VERSION_NUMBER -gt 8000 -o $CUDA_VERSION_NUMBER -lt 2030  ]; then
+if [ $CUDA_VERSION_NUMBER -gt 9100 -o $CUDA_VERSION_NUMBER -lt 2030  ]; then
 	echo "ERROR ** GPGPU-Sim version $GPGPUSIM_VERSION_STRING not tested with CUDA version $CUDA_VERSION_STRING (please see README)";
 	return
-elif [ $CUDA_VERSION_NUMBER -gt 4020 ]; then
-	echo "WARNING ** GPGPU-Sim version $GPGPUSIM_VERSION_STRING not fully tested with CUDA version $CUDA_VERSION_STRING (please see README)";
 fi
 
 if [ $# = '1' ] ;
diff --git a/src/abstract_hardware_model.h b/src/abstract_hardware_model.h
index aaa4b00..7125b6b 100644
--- a/src/abstract_hardware_model.h
+++ b/src/abstract_hardware_model.h
@@ -383,7 +383,7 @@ protected:
     std::deque<simt_stack_entry> m_stack;
 };
 
-#define GLOBAL_HEAP_START 0x703E20000
+#define GLOBAL_HEAP_START 0xC0000000
    // start allocating from this address (lower values used for allocating globals in .ptx file)
 #define SHARED_MEM_SIZE_MAX (64*1024)
 #define LOCAL_MEM_SIZE_MAX (8*1024)
diff --git a/src/cuda-sim/cuda-math.h b/src/cuda-sim/cuda-math.h
index 4721e8a..afac330 100644
--- a/src/cuda-sim/cuda-math.h
+++ b/src/cuda-sim/cuda-math.h
@@ -67,6 +67,8 @@
 #ifndef CUDA_MATH
 #define CUDA_MATH
 
+#include <cmath>
+
 // cuda math implementations
 #undef max
 #undef min
@@ -148,6 +150,7 @@ float __ll2float_rd(long long int a) {
 #include <device_types.h>
 #include <fenv.h>
 
+
 // 32-bit integer to float
 float __int2float_rn(int a) {
    int orig_rnd_mode = fegetround();
@@ -321,7 +324,7 @@ float __internal_accurate_fdividef(float a, float b)
 float __saturatef(float a)
 {
    float b; 
-   if (isnan(a)) b = 0.0f; 
+   if (std::isnan(a)) b = 0.0f; 
    else if (a >= 1.0f) b = 1.0f;
    else if (a <= 0.0f) b = 0.0f; 
    else b = a; 
diff --git a/src/cuda-sim/cuda-sim.cc b/src/cuda-sim/cuda-sim.cc
index d4ace76..a34b99b 100644
--- a/src/cuda-sim/cuda-sim.cc
+++ b/src/cuda-sim/cuda-sim.cc
@@ -1408,7 +1408,7 @@ void ptx_thread_info::ptx_exec_inst( warp_inst_t &inst, unsigned lane_id)
    if ( (g_ptx_sim_num_insn % 100000) == 0 ) {
       dim3 ctaid = get_ctaid();
       dim3 tid = get_tid();
-      printf("GPGPU-Sim PTX: %u instructions simulated : ctaid=(%u,%u,%u) tid=(%u,%u,%u)\n",
+      DPRINTF(LIVENESS, "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/instructions.cc b/src/cuda-sim/instructions.cc
index 011c285..71286c9 100644
--- a/src/cuda-sim/instructions.cc
+++ b/src/cuda-sim/instructions.cc
@@ -1961,7 +1961,7 @@ ptx_reg_t d2d( ptx_reg_t x, unsigned from_width, unsigned to_width, int to_sign,
       y.f64 = x.f64;
       break; 
    }
-   if (isnan(y.f64)) {
+   if (std::isnan(y.f64)) {
       y.u64 = 0xfff8000000000000ull;
    } else if (saturation_mode) {
       y.f64 = cuda_math::__saturatef(y.f64); 
@@ -2086,7 +2086,7 @@ void ptx_round(ptx_reg_t& data, int rounding_mode, int type)
       }
    }
    if ((type == F64_TYPE)||(type == FF64_TYPE)) {
-      if (isnan(data.f64)) {
+      if (std::isnan(data.f64)) {
          data.u64 = 0xfff8000000000000ull;
       }
    }
@@ -2648,12 +2648,12 @@ void mad_def( const ptx_instruction *pI, ptx_thread_info *thread, bool use_carry
 
 bool isNaN(float x)
 {
-   return isnan(x);
+   return std::isnan(x);
 }
 
 bool isNaN(double x)
 {
-   return isnan(x);
+   return std::isnan(x);
 }
 
 void max_impl( const ptx_instruction *pI, ptx_thread_info *thread ) 
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-cache.cc b/src/gpgpu-sim/gpu-cache.cc
index 5ea4190..8886398 100644
--- a/src/gpgpu-sim/gpu-cache.cc
+++ b/src/gpgpu-sim/gpu-cache.cc
@@ -517,18 +517,28 @@ void cache_stats::print_stats(FILE *fout, const char *cache_name) const{
     /// the provided name is used.
     /// The printed format is "<cache_name>[<request_type>][<request_status>] = <stat_value>"
     ///
+    std::vector< unsigned > total_access;
+    total_access.resize(NUM_MEM_ACCESS_TYPE, 0);
     std::string m_cache_name = cache_name;
     for (unsigned type = 0; type < NUM_MEM_ACCESS_TYPE; ++type) {
         for (unsigned status = 0; status < NUM_CACHE_REQUEST_STATUS; ++status) {
-            if(m_stats[type][status] > 0){
-                fprintf(fout, "\t%s[%s][%s] = %u\n",
-                    m_cache_name.c_str(),
-                    mem_access_type_str((enum mem_access_type)type),
-                    cache_request_status_str((enum cache_request_status)status),
-                    m_stats[type][status]);
-            }
+            fprintf(fout, "\t%s[%s][%s] = %u\n",
+                m_cache_name.c_str(),
+                mem_access_type_str((enum mem_access_type)type),
+                cache_request_status_str((enum cache_request_status)status),
+                m_stats[type][status]);
+            if(status != RESERVATION_FAIL)
+                 total_access[type]+= m_stats[type][status];
         }
     }
+    for (unsigned type = 0; type < NUM_MEM_ACCESS_TYPE; ++type) {
+         if(total_access[type] > 0)
+          fprintf(fout, "\t%s[%s][%s] = %u\n",
+                m_cache_name.c_str(),
+                mem_access_type_str((enum mem_access_type)type),
+                "TOTAL_ACCESS",
+                total_access[type]);
+    }
 }
 
 void cache_sub_stats::print_port_stats(FILE *fout, const char *cache_name) const
diff --git a/src/gpgpu-sim/gpu-sim.cc b/src/gpgpu-sim/gpu-sim.cc
index 58a5d16..bd4c00c 100644
--- a/src/gpgpu-sim/gpu-sim.cc
+++ b/src/gpgpu-sim/gpu-sim.cc
@@ -32,10 +32,12 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <math.h>
+#include <signal.h>
 #include "zlib.h"
 
 
 #include "shader.h"
+#include "shader_trace.h"
 #include "dram.h"
 #include "mem_fetch.h"
 
@@ -87,6 +89,14 @@ unsigned long long  gpu_tot_sim_cycle = 0;
 // performance counter for stalls due to congestion.
 unsigned int gpu_stall_dramfull = 0; 
 unsigned int gpu_stall_icnt2sh = 0;
+unsigned long long partiton_reqs_in_parallel = 0;
+unsigned long long partiton_reqs_in_parallel_total = 0;
+unsigned long long partiton_reqs_in_parallel_util = 0;
+unsigned long long partiton_reqs_in_parallel_util_total = 0;
+unsigned long long  gpu_sim_cycle_parition_util = 0;
+unsigned long long  gpu_tot_sim_cycle_parition_util = 0;
+unsigned long long partiton_replys_in_parallel = 0;
+unsigned long long partiton_replys_in_parallel_total = 0;
 
 /* Clock Domains */
 
@@ -223,7 +233,7 @@ void shader_core_config::reg_options(class OptionParser * opp)
                    "per-shader L1 data cache config "
                    " {<nsets>:<bsize>:<assoc>,<rep>:<wr>:<alloc>:<wr_alloc>,<mshr>:<N>:<merge>,<mq> | none}",
                    "none" );
-    option_parser_register(opp, "-gpgpu_cache:dl1PreShared", OPT_CSTR, &m_L1D_config.m_config_stringPrefShared,
+    option_parser_register(opp, "-gpgpu_cache:dl1PrefShared", OPT_CSTR, &m_L1D_config.m_config_stringPrefShared,
                    "per-shader L1 data cache config "
                    " {<nsets>:<bsize>:<assoc>,<rep>:<wr>:<alloc>:<wr_alloc>,<mshr>:<N>:<merge>,<mq> | none}",
                    "none" );
@@ -744,6 +754,10 @@ void gpgpu_sim::init()
     gpu_sim_insn = 0;
     last_gpu_sim_insn = 0;
     m_total_cta_launched=0;
+    partiton_reqs_in_parallel = 0;
+    partiton_replys_in_parallel = 0;
+    partiton_reqs_in_parallel_util = 0;
+    gpu_sim_cycle_parition_util = 0;
 
     reinit_clock_domains();
     set_param_gpgpu_num_shaders(m_config.num_shader());
@@ -780,8 +794,16 @@ void gpgpu_sim::update_stats() {
     gpu_tot_sim_cycle += gpu_sim_cycle;
     gpu_tot_sim_insn += gpu_sim_insn;
     gpu_tot_issued_cta += m_total_cta_launched;
+    partiton_reqs_in_parallel_total += partiton_reqs_in_parallel;
+    partiton_replys_in_parallel_total += partiton_replys_in_parallel;
+    partiton_reqs_in_parallel_util_total += partiton_reqs_in_parallel_util;
+    gpu_tot_sim_cycle_parition_util += gpu_sim_cycle_parition_util ;
 
     gpu_sim_cycle = 0;
+    partiton_reqs_in_parallel = 0;
+    partiton_replys_in_parallel = 0;
+    partiton_reqs_in_parallel_util = 0;
+    gpu_sim_cycle_parition_util = 0;
     gpu_sim_insn = 0;
     m_total_cta_launched = 0;
 }
@@ -965,6 +987,21 @@ void gpgpu_sim::gpu_print_stat()
    printf("gpu_stall_dramfull = %d\n", gpu_stall_dramfull);
    printf("gpu_stall_icnt2sh    = %d\n", gpu_stall_icnt2sh );
 
+   printf("partiton_reqs_in_parallel = %lld\n", partiton_reqs_in_parallel);
+   printf("partiton_reqs_in_parallel_total    = %lld\n", partiton_reqs_in_parallel_total );
+   printf("partiton_level_parallism = %12.4f\n", (float)partiton_reqs_in_parallel / gpu_sim_cycle);
+   printf("partiton_level_parallism_total  = %12.4f\n", (float)(partiton_reqs_in_parallel+partiton_reqs_in_parallel_total) / (gpu_tot_sim_cycle+gpu_sim_cycle) );
+   printf("partiton_reqs_in_parallel_util = %lld\n", partiton_reqs_in_parallel_util);
+   printf("partiton_reqs_in_parallel_util_total    = %lld\n", partiton_reqs_in_parallel_util_total );
+   printf("gpu_sim_cycle_parition_util = %lld\n", gpu_sim_cycle_parition_util);
+   printf("gpu_tot_sim_cycle_parition_util    = %lld\n", gpu_tot_sim_cycle_parition_util );
+   printf("partiton_level_parallism_util = %12.4f\n", (float)partiton_reqs_in_parallel_util / gpu_sim_cycle_parition_util);
+   printf("partiton_level_parallism_util_total  = %12.4f\n", (float)(partiton_reqs_in_parallel_util+partiton_reqs_in_parallel_util_total) / (gpu_sim_cycle_parition_util+gpu_tot_sim_cycle_parition_util) );
+   printf("partiton_replys_in_parallel = %lld\n", partiton_replys_in_parallel);
+   printf("partiton_replys_in_parallel_total    = %lld\n", partiton_replys_in_parallel_total );
+   printf("L2_BW  = %12.4f GB/Sec\n", ((float)(partiton_replys_in_parallel * 32) / (gpu_sim_cycle * m_config.icnt_period)) / 1000000000);
+   printf("L2_BW_total  = %12.4f GB/Sec\n", ((float)((partiton_replys_in_parallel+partiton_replys_in_parallel_total) * 32) / ((gpu_tot_sim_cycle+gpu_sim_cycle) * m_config.icnt_period)) / 1000000000 );
+
    time_t curr_time;
    time(&curr_time);
    unsigned long long elapsed_time = MAX( curr_time - g_simulation_starttime, 1 );
@@ -1175,8 +1212,8 @@ bool shader_core_ctx::occupy_shader_resource_1block(kernel_info_t & k, bool occu
        m_occupied_regs += (padded_cta_size * ((kernel_info->regs+3)&~3));
        m_occupied_ctas++;
 
-       printf("GPGPU-Sim uArch: Shader %d occupied %d threads, %d shared mem, %d registers, %d ctas\n",
-            m_sid, m_occupied_n_threads, m_occupied_shmem, m_occupied_regs, m_occupied_ctas);  
+      SHADER_DPRINTF(LIVENESS, "GPGPU-Sim uArch: Occupied %d threads, %d shared mem, %d registers, %d ctas\n",
+            m_occupied_n_threads, m_occupied_shmem, m_occupied_regs, m_occupied_ctas);  
    }
 
    return true;
@@ -1301,8 +1338,8 @@ void shader_core_ctx::issue_block2core( kernel_info_t &kernel )
     m_n_active_cta++;
 
     shader_CTA_count_log(m_sid, 1);
-    printf("GPGPU-Sim uArch: core:%3d, cta:%2u, start_tid:%4u, end_tid:%4u, initialized @(%lld,%lld)\n", 
-        m_sid, free_cta_hw_id, start_thread, end_thread, gpu_sim_cycle, gpu_tot_sim_cycle );
+    SHADER_DPRINTF(LIVENESS, "GPGPU-Sim uArch: cta:%2u, start_tid:%4u, end_tid:%4u, initialized @(%lld,%lld)\n", 
+        free_cta_hw_id, start_thread, end_thread, gpu_sim_cycle, gpu_tot_sim_cycle );
 
 }
 
@@ -1366,6 +1403,7 @@ void gpgpu_sim::cycle()
       for (unsigned i=0;i<m_shader_config->n_simt_clusters;i++) 
          m_cluster[i]->icnt_cycle(); 
    }
+    unsigned partiton_replys_in_parallel_per_cycle = 0;
     if (clock_mask & ICNT) {
         // pop from memory controller to interconnect
         for (unsigned i=0;i<m_memory_config->m_n_mem_sub_partition;i++) {
@@ -1378,6 +1416,7 @@ void gpgpu_sim::cycle()
                     mf->set_status(IN_ICNT_TO_SHADER,gpu_sim_cycle+gpu_tot_sim_cycle);
                     ::icnt_push( m_shader_config->mem2device(i), mf->get_tpc(), mf, response_size );
                     m_memory_sub_partition[i]->pop();
+                    partiton_replys_in_parallel_per_cycle++;
                 } else {
                     gpu_stall_icnt2sh++;
                 }
@@ -1386,6 +1425,7 @@ void gpgpu_sim::cycle()
             }
         }
     }
+    partiton_replys_in_parallel += partiton_replys_in_parallel_per_cycle;
 
    if (clock_mask & DRAM) {
       for (unsigned i=0;i<m_memory_config->m_n_mem;i++){
@@ -1398,6 +1438,7 @@ void gpgpu_sim::cycle()
    }
 
    // L2 operations follow L2 clock domain
+   unsigned partiton_reqs_in_parallel_per_cycle = 0;
    if (clock_mask & L2) {
        m_power_stats->pwr_mem_stat->l2_cache_stats[CURRENT_STAT_IDX].clear();
       for (unsigned i=0;i<m_memory_config->m_n_mem_sub_partition;i++) {
@@ -1408,11 +1449,17 @@ void gpgpu_sim::cycle()
           } else {
               mem_fetch* mf = (mem_fetch*) icnt_pop( m_shader_config->mem2device(i) );
               m_memory_sub_partition[i]->push( mf, gpu_sim_cycle + gpu_tot_sim_cycle );
+              partiton_reqs_in_parallel_per_cycle++;
           }
           m_memory_sub_partition[i]->cache_cycle(gpu_sim_cycle+gpu_tot_sim_cycle);
           m_memory_sub_partition[i]->accumulate_L2cache_stats(m_power_stats->pwr_mem_stat->l2_cache_stats[CURRENT_STAT_IDX]);
        }
    }
+   partiton_reqs_in_parallel += partiton_reqs_in_parallel_per_cycle;
+   if(partiton_reqs_in_parallel_per_cycle > 0){
+	   partiton_reqs_in_parallel_util += partiton_reqs_in_parallel_per_cycle;
+	   gpu_sim_cycle_parition_util++;
+   }
 
    if (clock_mask & ICNT) {
       icnt_transfer();
@@ -1440,7 +1487,7 @@ 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 ) 
@@ -1499,7 +1546,8 @@ void gpgpu_sim::cycle()
             hrs     = elapsed_time/3600 - 24*days;
             minutes = elapsed_time/60 - 60*(hrs + 24*days);
             sec = elapsed_time - 60*(minutes + 60*(hrs + 24*days));
-            printf("GPGPU-Sim uArch: cycles simulated: %lld  inst.: %lld (ipc=%4.1f) sim_rate=%u (inst/sec) elapsed = %u:%u:%02u:%02u / %s", 
+
+            DPRINTF(LIVENESS, "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/gpu-sim.h b/src/gpgpu-sim/gpu-sim.h
index a2d1b9b..7d92c66 100644
--- a/src/gpgpu-sim/gpu-sim.h
+++ b/src/gpgpu-sim/gpu-sim.h
@@ -195,7 +195,7 @@ struct memory_config {
       for (i=0; nbkt>0; i++) {
           nbkt = nbkt>>1;
       }
-      bk_tag_length = i;
+      bk_tag_length = i-1;
       assert(nbkgrp>0 && "Number of bank groups cannot be zero");
       tRCDWR = tRCD-(WL+1);
       tRTW = (CL+(BL/data_command_freq_ratio)+2-WL);
diff --git a/src/gpgpu-sim/shader.cc b/src/gpgpu-sim/shader.cc
index d17e51d..4640d65 100644
--- a/src/gpgpu-sim/shader.cc
+++ b/src/gpgpu-sim/shader.cc
@@ -1969,12 +1969,12 @@ void shader_core_ctx::register_cta_thread_exit( unsigned cta_num, kernel_info_t
       m_barriers.deallocate_barrier(cta_num);
       shader_CTA_count_unlog(m_sid, 1);
 
-      printf("GPGPU-Sim uArch: Shader %d finished CTA #%d (%lld,%lld), %u CTAs running\n", m_sid, cta_num, gpu_sim_cycle, gpu_tot_sim_cycle,
-             m_n_active_cta );
+     SHADER_DPRINTF(LIVENESS, "GPGPU-Sim uArch: Finished CTA #%d (%lld,%lld), %u CTAs running\n",
+        cta_num, gpu_sim_cycle, gpu_tot_sim_cycle, m_n_active_cta);
 
       if( m_n_active_cta == 0 ) {
-          printf("GPGPU-Sim uArch: Shader %u empty (last released kernel %u \'%s\').\n", m_sid, kernel->get_uid(),
-                 kernel->name().c_str() );
+        SHADER_DPRINTF(LIVENESS, "GPGPU-Sim uArch: Empty (last released kernel %u \'%s\').\n",
+            kernel->get_uid(), kernel->name().c_str());
           fflush(stdout);
 
           //Shader can only be empty when no more cta are dispatched
@@ -1989,8 +1989,10 @@ void shader_core_ctx::register_cta_thread_exit( unsigned cta_num, kernel_info_t
       kernel->dec_running();
       if( !m_gpu->kernel_more_cta_left(kernel) ) {
           if( !kernel->running() ) {
-              printf("GPGPU-Sim uArch: GPU detected kernel %u \'%s\' finished on shader %u.\n", kernel->get_uid(),
-                kernel->name().c_str(), m_sid );
+              SHADER_DPRINTF(LIVENESS,
+                "GPGPU-Sim uArch: GPU detected kernel %u \'%s\' finished on shader %u.\n", kernel->get_uid(),
+                kernel->name().c_str(), m_sid);
+
               if(m_kernel == kernel)
                 m_kernel = NULL;
               m_gpu->set_kernel_done( kernel );
diff --git a/src/gpgpusim_entrypoint.cc b/src/gpgpusim_entrypoint.cc
index 04845e7..ede9f20 100644
--- a/src/gpgpusim_entrypoint.cc
+++ b/src/gpgpusim_entrypoint.cc
@@ -91,9 +91,17 @@ 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;
+
+static void termination_callback()
+{
+    printf("GPGPU-Sim: *** exit detected ***\n");
+    fflush(stdout);
+}
 
 void *gpgpu_sim_thread_concurrent(void*)
 {
+    atexit(termination_callback);
     // concurrent kernel execution simulation thread
     do {
         if(g_debug_execution >= 3) {
@@ -144,17 +152,19 @@ 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);
         }
-        g_the_gpu->print_stats();
         if(sim_cycles) {
+            g_the_gpu->print_stats();
             g_the_gpu->update_stats();
             print_simulation_time();
         }
@@ -162,10 +172,14 @@ void *gpgpu_sim_thread_concurrent(void*)
         g_sim_active = false;
         pthread_mutex_unlock(&g_sim_lock);
     } while( !g_sim_done );
-    if(g_debug_execution >= 3) {
-       printf("GPGPU-Sim: *** simulation thread exiting ***\n");
-       fflush(stdout);
+    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 +193,7 @@ 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");
diff --git a/src/intersim2/Makefile b/src/intersim2/Makefile
index ef948d6..bd42000 100644
--- a/src/intersim2/Makefile
+++ b/src/intersim2/Makefile
@@ -125,6 +125,14 @@ endif
 
 # rules to compile simulator
 
+$(OBJDIR)/Makefile.makedepend: depend
+
+ALL_SRCS = $(CPP_SRCS)
+ALL_SRCS += $(shell ls **/*.cpp)
+
+depend:
+	touch $(OBJDIR)/Makefile.makedepend
+	makedepend -f$(OBJDIR)/Makefile.makedepend -I$(INCPATH) -p$(OBJDIR)/ $(ALL_SRCS) 2> /dev/null
 
 ${LEX_OBJS}: $(OBJDIR)/lex.yy.c $(OBJDIR)/y.tab.h
 	$(CC) $(CPPFLAGS) -c $< -o $@
@@ -173,3 +181,5 @@ $(OBJDIR)/y.tab.c $(OBJDIR)/y.tab.h: config.y
 
 $(OBJDIR)/lex.yy.c: config.l
 	$(LEX) -o$@ $<
+
+include $(OBJDIR)/Makefile.makedepend
diff --git a/src/trace_streams.tup b/src/trace_streams.tup
index c41690e..fd68a16 100644
--- a/src/trace_streams.tup
+++ b/src/trace_streams.tup
@@ -29,5 +29,6 @@ TS_TUP_BEGIN( trace_streams_type )
     TS_TUP( WARP_SCHEDULER ),
     TS_TUP( SCOREBOARD ),
     TS_TUP( MEMORY_PARTITION_UNIT ),
+    TS_TUP( LIVENESS ),
     TS_TUP( NUM_TRACE_STREAMS )
 TS_TUP_END( trace_streams_type )