summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--Jenkinsfile4
-rw-r--r--configs/3.x-cfgs/SM6_TITANX/gpgpusim.config1
-rw-r--r--configs/4.x-cfgs/SM6_TITANX/gpgpusim.config6
-rw-r--r--src/abstract_hardware_model.h1
-rw-r--r--src/cuda-sim/cuda-sim.cc1
-rw-r--r--src/cuda-sim/ptx_ir.cc1
-rw-r--r--src/gpgpu-sim/shader.cc2
7 files changed, 10 insertions, 6 deletions
diff --git a/Jenkinsfile b/Jenkinsfile
index 16e932e..11a9fa6 100644
--- a/Jenkinsfile
+++ b/Jenkinsfile
@@ -55,7 +55,7 @@ pipeline {
}, "9.1-functest": {
sh 'source /home/tgrogers-raid/a/common/gpgpu-sim-setup/9.1_env_setup.sh &&\
source `pwd`/setup_environment &&\
- ./gpgpu-sim_simulations/util/job_launching/run_simulations.py -B rodinia_2.0-ft,sdk-4.2 -C TITANX,TITANX-L1ON,P100,TITANV -N regress-$$ && \
+ ./gpgpu-sim_simulations/util/job_launching/run_simulations.py -B rodinia_2.0-ft,sdk-4.2 -C TITANX,TITANX-L1ON -N regress-$$ && \
PLOTDIR="jenkins/${JOB_NAME}/${BUILD_NUMBER}/9.1-rodinia" && ssh [email protected] mkdir -p /home/dynamo/a/tgrogers/website/gpgpu-sim-plots/$PLOTDIR && \
./gpgpu-sim_simulations/util/job_launching/monitor_func_test.py -v -s stats-$$.csv -N regress-$$ && \
./gpgpu-sim_simulations/util/plotting/plot-get-stats.py -c stats-$$.csv -p [email protected]:~/website/gpgpu-sim-plots/$PLOTDIR -w https://engineering.purdue.edu/tgrogers/gpgpu-sim-plots/$PLOTDIR -n $PLOTDIR'
@@ -75,7 +75,7 @@ pipeline {
sh 'source /home/tgrogers-raid/a/common/gpgpu-sim-setup/9.1_env_setup.sh &&\
source `pwd`/setup_environment &&\
PLOTDIR="jenkins/${JOB_NAME}" &&\
- ./gpgpu-sim_simulations/util/plotting/correlate_and_publish.sh TITANX,TITANX-L1ON,P100,TITANV $PLOTDIR ${BUILD_NUMBER}'
+ ./gpgpu-sim_simulations/util/plotting/correlate_and_publish.sh TITANX,TITANX-L1ON $PLOTDIR ${BUILD_NUMBER}'
}
}
}
diff --git a/configs/3.x-cfgs/SM6_TITANX/gpgpusim.config b/configs/3.x-cfgs/SM6_TITANX/gpgpusim.config
index f78bd02..28912a3 100644
--- a/configs/3.x-cfgs/SM6_TITANX/gpgpusim.config
+++ b/configs/3.x-cfgs/SM6_TITANX/gpgpusim.config
@@ -25,6 +25,7 @@
# shader core pipeline config
-gpgpu_shader_registers 65536
+-gpgpu_occupancy_sm_number 61
# This implies a maximum of 64 warps/SM
-gpgpu_shader_core_pipeline 2048:32
diff --git a/configs/4.x-cfgs/SM6_TITANX/gpgpusim.config b/configs/4.x-cfgs/SM6_TITANX/gpgpusim.config
index 4c0586d..8e93723 100644
--- a/configs/4.x-cfgs/SM6_TITANX/gpgpusim.config
+++ b/configs/4.x-cfgs/SM6_TITANX/gpgpusim.config
@@ -65,9 +65,9 @@
# Pascal GP102 has 96KB Shared memory divided over 2 cores, each has 48KB
# Pascal GP102 has 2 banks L1 cache, where each is 24KB L1 cache
# The defulat is to disable the L1 cache, unless cache modifieres are used
--gpgpu_cache:dl1 S:48:128:4,L:L:s:N:H,A:256:8,16:0,32
--gpgpu_cache:dl1PrefL1 S:48:128:4,L:L:s:N:H,A:256:8,16:0,32
--gpgpu_cache:dl1PrefShared S:48:128:4,L:L:s:N:H,A:256:8,16:0,32
+-gpgpu_cache:dl1 S:48:128:4,L:L:s:N:L,A:256:8,16:0,32
+-gpgpu_cache:dl1PrefL1 S:48:128:4,L:L:s:N:L,A:256:8,16:0,32
+-gpgpu_cache:dl1PrefShared S:48:128:4,L:L:s:N:L,A:256:8,16:0,32
-gpgpu_shmem_size 49152
-gpgpu_shmem_size_PrefL1 49152
-gpgpu_shmem_size_PrefShared 49152
diff --git a/src/abstract_hardware_model.h b/src/abstract_hardware_model.h
index a70b077..e708fa7 100644
--- a/src/abstract_hardware_model.h
+++ b/src/abstract_hardware_model.h
@@ -665,6 +665,7 @@ enum cache_operator_type {
CACHE_ALL, // .ca
CACHE_LAST_USE, // .lu
CACHE_VOLATILE, // .cv
+ CACHE_L1, // .nc
// loads and stores
CACHE_STREAMING, // .cs
diff --git a/src/cuda-sim/cuda-sim.cc b/src/cuda-sim/cuda-sim.cc
index 9246613..93bbc1d 100644
--- a/src/cuda-sim/cuda-sim.cc
+++ b/src/cuda-sim/cuda-sim.cc
@@ -898,6 +898,7 @@ void ptx_instruction::pre_decode()
switch( m_cache_option ) {
case CA_OPTION: cache_op = CACHE_ALL; break;
+ case NC_OPTION: cache_op = CACHE_L1; break;
case CG_OPTION: cache_op = CACHE_GLOBAL; break;
case CS_OPTION: cache_op = CACHE_STREAMING; break;
case LU_OPTION: cache_op = CACHE_LAST_USE; break;
diff --git a/src/cuda-sim/ptx_ir.cc b/src/cuda-sim/ptx_ir.cc
index ee36957..ba6d7ed 100644
--- a/src/cuda-sim/ptx_ir.cc
+++ b/src/cuda-sim/ptx_ir.cc
@@ -1211,6 +1211,7 @@ ptx_instruction::ptx_instruction( int opcode,
case EXTP_OPTION:
break;
case NC_OPTION:
+ m_cache_option = last_ptx_inst_option;
break;
case UP_OPTION:
case DOWN_OPTION:
diff --git a/src/gpgpu-sim/shader.cc b/src/gpgpu-sim/shader.cc
index 51689e3..0e2e1c2 100644
--- a/src/gpgpu-sim/shader.cc
+++ b/src/gpgpu-sim/shader.cc
@@ -1642,7 +1642,7 @@ bool ldst_unit::memory_cycle( warp_inst_t &inst, mem_stage_stall_type &stall_rea
bypassL1D = true;
} else if (inst.space.is_global()) { // global memory access
// skip L1 cache if the option is enabled
- if (m_core->get_config()->gmem_skip_L1D)
+ if (m_core->get_config()->gmem_skip_L1D && (CACHE_L1 != inst.cache_op))
bypassL1D = true;
}