Update yaml (#462)

* upadte yaml with ubench args

* args for atomics

* fix some args

* reduce mem number of blocks

* update workflow

* update workflow

* update hw ssats

* fix path , and dropping Turing

Author: LAhmos

.github/workflows/long-tests.yml

@@ -1,5 +1,7 @@
 # This is a basic workflow to help you get started with Actions
 
+#Turing commands are commented out , we regress QV100,RTX3070,A100 
+
 name: Long Tests
 
 # Controls when the workflow will run
@@ -40,10 +42,17 @@ jobs:
         run: |
           source ./env-setup/12.4_env_setup.sh
           source ./gpu-simulator/setup_environment.sh
-          ./util/job_launching/run_simulations.py -B rodinia_2.0-ft,GPU_Microbenchmark -C QV100-SASS -T ~/../common/accel-sim/traces/volta-tesla-v100/latest/ -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
-          ./util/job_launching/run_simulations.py -B rodinia_2.0-ft,GPU_Microbenchmark -C RTX2060-SASS -T ~/../common/accel-sim/traces/turing-rtx2060/latest/ -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
-          ./util/job_launching/run_simulations.py -B rodinia_2.0-ft,GPU_Microbenchmark -C RTX3070-SASS -T ~/../common/accel-sim/traces/ampere-rtx3070/latest/ -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
-          ./util/job_launching/run_simulations.py -B rodinia_2.0-ft,GPU_Microbenchmark -C A100-SASS -T /scratch/tgrogers-disk01/a/common/for-sharing/accel-sim/A100/traces/device-0/12.8/  -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
+          ./util/job_launching/run_simulations.py -B GPU_Microbenchmark -C QV100-SASS -T /scratch/tgrogers-disk01/a/common/for-sharing/accel-sim/QV100//hw_run/traces/device-0/12.8/ -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
+          ./util/job_launching/run_simulations.py -B rodinia_2.0-ft  -C QV100-SASS -T /scratch/tgrogers-disk01/a/common/for-sharing/accel-sim/QV100/hw_run/traces/device-0/12.8/ -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
+
+          #./util/job_launching/run_simulations.py -B rodinia_2.0-ft -C RTX2060-SASS -T ~/../common/accel-sim/traces/turing-rtx2060/latest/ -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
+          
+          ./util/job_launching/run_simulations.py -B GPU_Microbenchmark -C RTX3070-SASS -T /scratch/tgrogers-disk01/a/common/for-sharing/accel-sim/RTX3070/hw_run/traces/device-0/12.8/  -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
+          ./util/job_launching/run_simulations.py -B rodinia_2.0-ft  -C RTX3070-SASS -T /scratch/tgrogers-disk01/a/common/for-sharing/accel-sim/RTX3070/hw_run/traces/device-0/12.8/  -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
+
+          ./util/job_launching/run_simulations.py -B GPU_Microbenchmark -C A100-SASS -T /scratch/tgrogers-disk01/a/common/for-sharing/accel-sim/A100/hw_run/traces/device-0/12.8/  -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
+          ./util/job_launching/run_simulations.py -B rodinia_2.0-ft -C A100-SASS -T /scratch/tgrogers-disk01/a/common/for-sharing/accel-sim/A100/hw_run/traces/device-0/12.8/  -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
+
           ./util/job_launching/run_simulations.py -B mlperf_inference -C RTX3070-SASS -T /scratch/tgrogers-disk01/a/common/for-sharing/accel-sim/mlperf_rtx3070/traces/device-0/12.8/ -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
 
           ./util/job_launching/monitor_func_test.py -v -s --sleep_time 300 stats-per-app-sass.csv -N sass-short-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
@@ -56,25 +65,25 @@ jobs:
           # either create a new branch or check it out if it already exists
           git -C ./statistics-archive checkout $BRANCH_NAME 2>/dev/null || git -C ./statistics-archive checkout -b $BRANCH_NAME
           ./util/job_launching/get_stats.py -k -K -R -B GPU_Microbenchmark -C QV100-SASS -A | tee v100-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv
-          ./util/job_launching/get_stats.py -k -K -R -B GPU_Microbenchmark -C RTX2060-SASS -A | tee turing-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv
+          #./util/job_launching/get_stats.py -k -K -R -B GPU_Microbenchmark -C RTX2060-SASS -A | tee turing-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv
           ./util/job_launching/get_stats.py -k -K -R -B GPU_Microbenchmark -C RTX3070-SASS -A | tee ampere-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv
           ./util/job_launching/get_stats.py -k -K -R -B GPU_Microbenchmark -C A100-SASS -A | tee ampere-a100-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv
 
           mkdir -p statistics-archive/ubench/
           # First we merge and archive this run to the main csv that contains all previous runs
           ./util/plotting/merge-stats.py -R -c ./statistics-archive/ubench/v100-ubench-sass.csv,v100-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv \
           | tee v100-ubench-sass.csv && mv v100-ubench-sass.csv ./statistics-archive/ubench/
-          ./util/plotting/merge-stats.py -R -c ./statistics-archive/ubench/turing-ubench-sass.csv,turing-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv \
-          | tee turing-ubench-sass.csv && mv turing-ubench-sass.csv ./statistics-archive/ubench/
+          # ./util/plotting/merge-stats.py -R -c ./statistics-archive/ubench/turing-ubench-sass.csv,turing-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv \
+          # | tee turing-ubench-sass.csv && mv turing-ubench-sass.csv ./statistics-archive/ubench/
           ./util/plotting/merge-stats.py -R -c ./statistics-archive/ubench/ampere-ubench-sass.csv,ampere-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv \
           | tee ampere-ubench-sass.csv && mv ampere-ubench-sass.csv ./statistics-archive/ubench/
           ./util/plotting/merge-stats.py -R -c ./statistics-archive/ubench/ampere-a100-ubench-sass.csv,ampere-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv \
           | tee ampere-a100-ubench-sass.csv && mv ampere-a100-ubench-sass.csv ./statistics-archive/ubench/
           # Next we merge the latest run with the current run (used for correlation plots) then archive the current run as the new latest for the next time this action occurs
           ./util/plotting/merge-stats.py -R -c ./statistics-archive/ubench/v100-ubench-sass-latest.csv,v100-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv \
           | tee v100-ubench-sass-latest2.csv && mv v100-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv ./statistics-archive/ubench/v100-ubench-sass-latest.csv
-          ./util/plotting/merge-stats.py -R -c ./statistics-archive/ubench/turing-ubench-sass-latest.csv,turing-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv \
-          | tee turing-ubench-sass-latest2.csv && mv turing-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv ./statistics-archive/ubench/turing-ubench-sass-latest.csv
+          # ./util/plotting/merge-stats.py -R -c ./statistics-archive/ubench/turing-ubench-sass-latest.csv,turing-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv \
+          # | tee turing-ubench-sass-latest2.csv && mv turing-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv ./statistics-archive/ubench/turing-ubench-sass-latest.csv
           ./util/plotting/merge-stats.py -R -c ./statistics-archive/ubench/ampere-ubench-sass-latest.csv,ampere-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv \
           | tee ampere-ubench-sass-latest2.csv && mv ampere-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv ./statistics-archive/ubench/ampere-ubench-sass-latest.csv
           ./util/plotting/merge-stats.py -R -c ./statistics-archive/ubench/ampere-a100-ubench-sass-latest.csv,ampere-a100-ubench-sass-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT.csv \
@@ -90,10 +99,10 @@ jobs:
           source ./env-setup/12.4_env_setup.sh
           ./util/hw_stats/get_hw_data.sh > /dev/null 2>&1
           rm -rf ./util/plotting/correl-html/
-          ./util/plotting/plot-correlation.py -c ./v100-ubench-sass-latest2.csv -H ./hw_run/volta-tesla-v100/11.2/ | tee v100-ubench-correl.txt
-          ./util/plotting/plot-correlation.py -c ./turing-ubench-sass-latest2.csv -H ./hw_run/TURING-RTX2060/10.2/ | tee turing-ubench-correl.txt
-          ./util/plotting/plot-correlation.py -c ./ampere-ubench-sass-latest2.csv -H ./hw_run/AMPERE-RTX3070/11.2/ | tee ampere-ubench-correl.txt
-          ./util/plotting/plot-correlation.py -c ./ampere-a100-ubench-sass-latest2.csv -H /scratch/tgrogers-disk01/a/common/for-sharing/accel-sim/A100/device-0/12.8/ | tee ampere-a100-ubench-correl.txt
+          ./util/plotting/plot-correlation.py -c ./v100-ubench-sass-latest2.csv -H /scratch/tgrogers-disk01/a/common/for-sharing/accel-sim/QV100/hw_run/device-0/12.8/ | tee v100-ubench-correl.txt
+          #./util/plotting/plot-correlation.py -c ./turing-ubench-sass-latest2.csv -H ./hw_run/TURING-RTX2060/10.2/ | tee turing-ubench-correl.txt
+          ./util/plotting/plot-correlation.py -c ./ampere-ubench-sass-latest2.csv -H  /scratch/tgrogers-disk01/a/common/for-sharing/accel-sim/RTX3070/hw_run/device-0/12.8/ | tee ampere-ubench-correl.txt
+          ./util/plotting/plot-correlation.py -c ./ampere-a100-ubench-sass-latest2.csv -H /scratch/tgrogers-disk01/a/common/for-sharing/accel-sim/A100/hw_run/device-0/12.8/ | tee ampere-a100-ubench-correl.txt
           ssh ghci@tgrogers-pc01 mkdir -p /home/ghci/accel-sim/correl/git_${GITHUB_REF}"_"$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT/
           rsync --delete -r ./util/plotting/correl-html/ ghci@tgrogers-pc01:/home/ghci/accel-sim/correl/git_${GITHUB_REF}"_"$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT/
           if [[ $GITHUB_EVENT_NAME == 'push' ]]; then
@@ -142,7 +151,9 @@ jobs:
           srun --time=8:00:00 -c20 make rodinia_2.0-ft GPU_Microbenchmark -j20 -C ./gpu-app-collection/src
           ./gpu-app-collection/get_regression_data.sh
 
-          ./util/job_launching/run_simulations.py -B rodinia_2.0-ft,GPU_Microbenchmark -C QV100-PTX,RTX2060-PTX,RTX3070-PTX,A100-PTX -N short-ptx-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
+          #./util/job_launching/run_simulations.py -B rodinia_2.0-ft,GPU_Microbenchmark -C QV100-PTX,RTX2060-PTX,RTX3070-PTX,A100-PTX -N short-ptx-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
+          ./util/job_launching/run_simulations.py -B rodinia_2.0-ft,GPU_Microbenchmark -C QV100-PTX,RTX3070-PTX,A100-PTX -N short-ptx-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
+
           ./util/job_launching/monitor_func_test.py -v -s stats-per-app-ptx.csv -N short-ptx-$GITHUB_RUN_NUMBER"_"$GITHUB_RUN_ATTEMPT
   Tracer-Tool:
     if: github.repository == 'accel-sim/accel-sim-framework'

util/job_launching/apps/define-all-apps.yml

@@ -54,43 +54,40 @@ GPU_Microbenchmark:
     data_dirs: "$GPUAPPS_ROOT/data_dirs/"
     execs:
         - l1_bw_32f:
-            - args:
+            - args: --tpb 1024 --tpsm 1024 --blocks 1 --ws 32
               accel-sim-mem: 1G
         - l1_bw_64f:
-            - args:
+            - args: --tpb 1024 --tpsm 1024 --blocks 1 --ws 32
               accel-sim-mem: 1G
         - l1_bw_128:
-            - args:
+            - args: --tpb 1024 --tpsm 1024 --blocks 1 --ws 32
               accel-sim-mem: 2G
         - l1_lat:
-            - args:
-              accel-sim-mem: 1G
-        - l1_lat:
-            - args:
+            - args: --blocks 1 --ws 32
               accel-sim-mem: 1G
         - l2_bw_32f:
-            - args:
+            - args: --tpb 1024 --tpsm 1024 --blocks 160 --ws 32
               accel-sim-mem: 6G
         - l2_bw_64f:
-            - args:
+            - args: --tpb 1024 --tpsm 1024 --blocks 160 --l2 786432 --ws 32
               accel-sim-mem: 6G
 #        - l2_bw_128:
 #            - args:
 #              accel-sim-mem: 1G
         - l2_lat:
-            - args:
+            - args: --tpb 1 --tpsm 1 --blocks 1 --l2 786432 --ws 32
               accel-sim-mem: 1G
         - mem_bw:
-            - args:
+            - args: --tpb 1024 --tpsm 1024 --blocks 80 --l2 1572864 --ws 32 --memclk 1132 --membw 64
               accel-sim-mem: 2G
         - mem_lat:
-            - args:
+            - args: --tpb 1024 --tpsm 1024 --blocks 80 --l2 1572864 --ws 32 --memclk 1132 --membw 64
               accel-sim-mem: 1G
         - shared_bw:
-            - args:
+            - args: --tpb 1024 --tpsm 1024 --blocks 1  --ws 32 
               accel-sim-mem: 2G
         - shared_lat:
-            - args:
+            - args:  --blocks 1  --ws 32 
               accel-sim-mem: 1G
         - shared_bank_conflicts:
         ## argument 1 kernel has conflicts
@@ -100,13 +97,13 @@ GPU_Microbenchmark:
             - args: 2
               accel-sim-mem: 1G
         - MaxFlops:
-            - args:
+            - args: --tpb 1024  --blocks 1  --ws 32 
               accel-sim-mem: 1G
         - l1_shared_bw:
-            - args:
+            - args: --tpb 1024  --blocks 1  --ws 32 
               accel-sim-mem: 1G
         - l1_bw_32f_unroll:
-            - args:
+            - args: --tpb 1024 --blocks 1  --ws 32 
               accel-sim-mem: 1G
         - l1_bw_32f_unroll_large:
             - args:
@@ -117,10 +114,10 @@ GPU_Atomic:
     data_dirs: "$GPUAPPS_ROOT/data_dirs/"
     execs:
         - atomic_add_bw:
-            - args: 
+            - args:   --tpb 1 --tpsm 1 --blocks 1  --ws 32 
               accel-sim-mem: 1G
         - atomic_add_bw_conflict:
-            - args:
+            - args: --tpb 1024 --tpsm 2048 --blocks 160  --ws 32 
               accel-sim-mem: 1G
         - atomic_add_bw_profile:
             - args: 16
@@ -1060,4 +1057,4 @@ huggingface:
     execs:
         - helloworld:
             - args:
-              accel-sim-mem: 10G
+              accel-sim-mem: 10G