Delete gpu/.depend

Minimap2onGPU · Sep 27, 2023 · 1d5c636 · 1d5c636
2 parents d334c9e + ccddf5c
commit 1d5c636
Show file tree

Hide file tree

Showing 6 changed files with 219 additions and 0 deletions.
diff --git a/.gitignore b/.gitignore
@@ -18,3 +18,4 @@ ncu
 nsys
 profile_output*
 workloads
+.depend
diff --git a/scripts/aac_integrated.sh b/scripts/aac_integrated.sh
@@ -0,0 +1,38 @@
+#!/bin/bash
+WORKSPACE_DIR=/shared/prod/home/liuxs/bioinfo/minimap2
+EXE_PATH=$WORKSPACE_DIR
+CONFIG_PATH=$WORKSPACE_DIR
+DATA_PATH=/shared/prod/home/liuxs/bioinfo/Profile_mm2/data
+# NOTE: currently `-t n_thread`, the n_thread must be equal to the num_streams in gpu_config.json
+N_THREAD=$(sed -n 's/.*"num_streams": \([0-9]*\).*/\1/p' ${CONFIG_PATH}/gpu_config.json)
+# Array of LONG_BLOCK_SIZE values
+MID_BLOCK_SIZES=( 64 )
+MID_CUTS=( 1 )
+LONG_CUTS=( 50 100 )
+GPU_CONFIGS=( gpu_config.json )
+DATA_SETS=( 1kto300k 200kto300k )
+
+# Iterate over LONG_BLOCK_SIZES array
+for DATA_SET in "${DATA_SETS[@]}"
+do
+    QUERY_FILE=$DATA_PATH/random_500MBases_${DATA_SET}.fa
+    for MID_BLOCK_SIZE in "${MID_BLOCK_SIZES[@]}"
+    do
+        for MID_CUT in "${MID_CUTS[@]}"
+        do
+            for LONG_CUT in "${LONG_CUTS[@]}"
+            do
+                for GPU_CONFIG in "${GPU_CONFIGS[@]}"
+                do
+                    echo "Executing with MID_BLOCK_SIZE=${MID_BLOCK_SIZE} MID_CUT=${MID_CUT} LONG_CUT=${LONG_CUT}"
+
+                    make clean 
+
+                    make GPU_CONFIG=${GPU_CONFIG} SHORT_BLOCK_SIZE=64 MID_BLOCK_SIZE=${MID_BLOCK_SIZE} LONG_BLOCK_SIZE=1024 MID_CUT=${MID_CUT} LONG_CUT=${LONG_CUT} 
+
+                    omniperf profile -n integrated_${MID_BLOCK_SIZE}_${LONG_CUT}_${DATA_SET}_report --device 0 -- ${EXE_PATH}/minimap2 ${DATA_PATH}/hg38.mmi -t ${N_THREAD} --max-chain-skip=2147483647 --gpu-chain ${QUERY_FILE} > test_logs.out
+                done
+            done
+        done
+    done
+done
diff --git a/scripts/aac_rocprof.slurm b/scripts/aac_rocprof.slurm
@@ -0,0 +1,36 @@
+#!/bin/bash
+#SBATCH --job-name=minimap2         # Job name
+#SBATCH --partition=1CN128C8G2H_2IB_MI210_RHEL8               # Specify the MI210 GPU partition or queue
+#SBATCH --gres=gpu:1                  # Request 1 GPU
+#SBATCH --nodes=1                     # Number of nodes
+#SBATCH --ntasks-per-node=1           # Number of tasks (processes) per node
+#SBATCH --cpus-per-task=16             # Number of CPU cores per task
+#SBATCH --mem=0                      # Memory per node
+#SBATCH --time=00:40:00               # Maximum execution time (HH:MM:SS)
+#SBATCH --output=slurm_output/sample_sbatch_job.%j.out    # Output file
+#SBATCH --error=slurm_output/sample_sbatch_job.%j.err     # Error file
+
+# Load necessary modules (if required)
+source /etc/profile.d/modules.sh
+scl enable gcc-toolset-11 bash
+module unuse /shared/apps/modules/ubuntu/modulefiles
+module use /shared/apps/modules/rhel8/modulefiles
+module unuse /shared/apps/modules/rhel9/modulefiles
+module unuse /shared/apps/modules/sles15sp4/modulefiles
+module unuse /shared/apps/modules/centos8/modulefiles
+module unuse /shared/apps/modules/rocky9/modulefiles
+
+module load rocm-5.4.3
+# export AMD_LOG_LEVEL=4
+# Replace the following line with the actual command(s) you want to run
+cd /shared/prod/home/liuxs/bioinfo/minimap2/
+make clean
+make GPU_CONFIG=gpu_config.json SHORT_BLOCK_SIZE=64 LONG_BLOCK_SIZE=1024 MID_BLOCK_SIZE=512 MID_CUT=1 LONG_CUT=100
+# ./minimap2 -t 1 --max-chain-skip=2147483647 --gpu-chain /shareddata/umich_folder/data/ONT/hg38.mmi /shareddata/umich_folder/data/ONT/reads_4f452f4a-d82a-4580-981b-32d14b997217.fa
+# ./minimap2 -t 1 --max-chain-skip=2147483647 --gpu-chain /shareddata/umich_folder/data/ONT/hg38.mmi /shareddata/umich_folder/data/ONT/random_500MBases_200kto300k.fa
+rocprof --stats -o rocprof_output/long_seg.${SLURM_JOB_ID}.csv ./minimap2 -t 1 --max-chain-skip=2147483647 --gpu-chain /shareddata/umich_folder/data/ONT/hg38.mmi /shareddata/umich_folder/data/ONT/short_seg_reads_from_1kto10k_distri.fa
+# rocprof --stats -o rocprof_output/long_seg.${SLURM_JOB_ID}.csv ./minimap2 -t 1 --max-chain-skip=2147483647 --gpu-chain /shareddata/umich_folder/data/ONT/hg38.mmi /shareddata/umich_folder/data/ONT/reads_4f452f4a-d82a-4580-981b-32d14b997217.fa 
+
+# Optional: You can add post-processing commands here
+
+# End of the script
diff --git a/scripts/aac_timing_all.sh b/scripts/aac_timing_all.sh
@@ -0,0 +1,42 @@
+#!/bin/bash
+WORKSPACE_DIR=/shared/prod/home/liuxs/bioinfo/minimap2
+EXE_PATH=$WORKSPACE_DIR
+CONFIG_PATH=$WORKSPACE_DIR
+# DATA_PATH=/shared/prod/home/liuxs/bioinfo/Profile_mm2/data
+DATA_PATH=/shareddata/umich_folder/data/ONT
+# NOTE: currently `-t n_thread`, the n_thread must be equal to the num_streams in gpu_config.json
+N_THREAD=$(sed -n 's/.*"num_streams": \([0-9]*\).*/\1/p' ${CONFIG_PATH}/gpu_config.json)
+# Array of LONG_BLOCK_SIZE values
+MID_BLOCK_SIZES=( 512 )
+MID_CUTS=( 1 )
+LONG_CUTS=( 100 )
+GPU_CONFIGS=( gpu_config.json )
+# DATA_SETS=( 50kto100k )
+# DATA_SETS=( 1kto5k 1kto10k 1kto50k 1kto300k 200kto300k 1kto20k 1kto30k 1kto70k 1kto200k 10kto50k 10kto100k 50kto100k)
+# DATA_SETS=( 1kto300k 50kto300k 100kto300k 150kto300k 200kto300k 250kto300k 1kto200k 20kto200k 50kto200k 70kto200k 100kto200k 130kto200k 150kto200k 170kto200k)
+DATA_SETS=( 1kto5k 9kto10k 10kto20k 20kto30k 40kto50k 90kto100k 110kto120k 140kto150k 180kto200k 200kto250k 200kto300k )
+
+# Iterate over LONG_BLOCK_SIZES array
+for MID_BLOCK_SIZE in "${MID_BLOCK_SIZES[@]}"
+do
+    for MID_CUT in "${MID_CUTS[@]}"
+    do
+        for LONG_CUT in "${LONG_CUTS[@]}"
+        do
+            echo "Executing with MID_BLOCK_SIZE=${MID_BLOCK_SIZE} MID_CUT=${MID_CUT} LONG_CUT=${LONG_CUT}"
+            make clean 
+            make GPU_CONFIG=${GPU_CONFIG} SHORT_BLOCK_SIZE=64 MID_BLOCK_SIZE=${MID_BLOCK_SIZE} LONG_BLOCK_SIZE=1024 MID_CUT=${MID_CUT} LONG_CUT=${LONG_CUT} 
+
+            for GPU_CONFIG in "${GPU_CONFIGS[@]}"
+            do
+                for DATA_SET in "${DATA_SETS[@]}"
+                do
+                    QUERY_FILE=$DATA_PATH/random_500MBases_${DATA_SET}.fa
+                    echo "Executing on dataset ${DATA_SET}"    
+                    filename="profile_output/data-${DATA_SET}_profile_${N_THREAD}_midblk-${MID_BLOCK_SIZE}_cut-${LONG_CUT}"                
+                    ${EXE_PATH}/minimap2 ${DATA_PATH}/hg38.mmi -t ${N_THREAD} --max-chain-skip=2147483647 --gpu-chain ${QUERY_FILE} > test.out 2> $filename
+                done
+            done
+        done
+    done
+done
diff --git a/scripts/acc_integrated.slurm b/scripts/acc_integrated.slurm
@@ -0,0 +1,36 @@
+#!/bin/bash
+#SBATCH --job-name=minimap2         # Job name
+#SBATCH --partition=1CN128C8G2H_2IB_MI210_RHEL8               # Specify the MI210 GPU partition or queue
+#SBATCH --gres=gpu:1                  # Request 1 GPU
+#SBATCH --nodes=1                     # Number of nodes
+#SBATCH --ntasks-per-node=1           # Number of tasks (processes) per node
+#SBATCH --cpus-per-task=16             # Number of CPU cores per task
+#SBATCH --mem=0                      # Memory per node
+#SBATCH --time=00:40:00               # Maximum execution time (HH:MM:SS)
+#SBATCH --output=slurm_output/sample_sbatch_job.%j.out    # Output file
+#SBATCH --error=slurm_output/sample_sbatch_job.%j.err     # Error file
+
+# Load necessary modules (if required)
+source /etc/profile.d/modules.sh
+scl enable gcc-toolset-11 bash
+module unuse /shared/apps/modules/ubuntu/modulefiles
+module use /shared/apps/modules/rhel8/modulefiles
+module unuse /shared/apps/modules/rhel9/modulefiles
+module unuse /shared/apps/modules/sles15sp4/modulefiles
+module unuse /shared/apps/modules/centos8/modulefiles
+module unuse /shared/apps/modules/rocky9/modulefiles
+
+module load rocm-5.4.3
+# export AMD_LOG_LEVEL=4
+# Replace the following line with the actual command(s) you want to run
+cd /shared/prod/home/liuxs/bioinfo/minimap2/
+make clean
+make GPU_CONFIG=gpu_config.json SHORT_BLOCK_SIZE=64 LONG_BLOCK_SIZE=1024 MID_BLOCK_SIZE=512 MID_CUT=1 LONG_CUT=100
+# ./minimap2 -t 1 --max-chain-skip=2147483647 --gpu-chain /shareddata/umich_folder/data/ONT/hg38.mmi /shareddata/umich_folder/data/ONT/reads_4f452f4a-d82a-4580-981b-32d14b997217.fa
+./minimap2 -t 1 --max-chain-skip=2147483647 --gpu-chain /shareddata/umich_folder/data/ONT/hg38.mmi /shareddata/umich_folder/data/ONT/random_500MBases_200kto300k.fa
+# ./minimap2 -t 1 --max-chain-skip=2147483647 --gpu-chain /shareddata/umich_folder/data/ONT/hg38.mmi /shareddata/umich_folder/data/ONT/short_seg_reads_from_1kto10k_distri.fa
+
+
+# Optional: You can add post-processing commands here
+
+# End of the script
diff --git a/scripts/parse_seg.py b/scripts/parse_seg.py
@@ -0,0 +1,66 @@
+import re
+import argparse
+import matplotlib.pyplot as plt
+
+# Define a function to extract segments from a line
+def extract_segments(line):
+    match = re.search(r'long segments (\d+)', line)
+    if match:
+        if int(match.group(1)) > 1000000:
+            print(line)
+        return int(match.group(1))
+    return None
+
+# Define a function to extract runtime from a line
+def extract_runtime(line):
+    match = re.search(r'last launch runtime: (\d+\.\d+) ms', line)
+    if match:
+        return float(match.group(1))
+    return None
+
+
+# Initialize variables to store segment counts
+segment_counts = []
+runtimes = []
+
+# Create an argument parser to get the output file name from the command line
+parser = argparse.ArgumentParser(description='Compute and plot a histogram of segments from an output file.')
+parser.add_argument('output_file', help='Path to the output file containing segment data')
+parser.add_argument('runtime_file', help='Path to the file containing runtime data')
+args = parser.parse_args()
+
+# Read the output file specified in the command line argument
+with open(args.output_file, 'r') as file:
+    for line in file:
+        segments = extract_segments(line)
+        if segments is None:
+            continue
+        segment_counts.append(segments)
+
+
+# Read the runtime file specified in the command line argument
+with open(args.runtime_file, 'r') as runtime_file:
+    for line in runtime_file:
+        runtime = extract_runtime(line)
+        if runtime is None:
+            continue
+        runtimes.append(runtime)
+
+# Calculate the total number of segments
+total_segments = sum(segment_counts)
+total_runtime = sum(runtimes)
+throughput = total_segments / total_runtime  # anchors/ms
+
+# Create a histogram
+plt.hist(segment_counts, bins=200, edgecolor='k')
+plt.xlabel('Segments')
+plt.ylabel('Frequency')
+plt.title(f'Total Segments: {total_segments}, throughput: {throughput} anchors/ms')
+plt.grid(True)
+
+# Save the figure with an appropriate name based on the input file name
+output_filename = args.output_file #.split('.')[0]  # Remove the file extension
+plt.savefig(f'{output_filename}_segment_histogram.png')
+
+# Display the histogram
+# plt.show()
-Original file line number
+Diff line change
@@ Expand Up / @@ -18,3 +18,4 @@ ncu @@
     nsys
     profile_output*
     workloads
+    .depend