Minor changes

Author: lorenzotomada

scripts/profiling_memory_and_time.py

@@ -23,7 +23,7 @@
 
 profile = LineProfiler()
 profile.add_function(parallel_tridiag_eigen)
-
+os.makedirs("Profiling_files", exist_ok=True) # to save the outputs of profiling
 
 seed = 1000
 np.random.seed(seed)
@@ -91,6 +91,7 @@
     mem_after_lanczos = proc.memory_info().rss / 1024 / 1024  # MB
     delta_mem_lanczos = mem_after_lanczos - mem_before_lanczos
     delta_t_lanczos = end_lanczos - begin_lanczos
+    print(f'delta_t_lanczos: {delta_t_lanczos:.4f} s')
 
     print("Done. Now computing eigenvalues...")
 else:
@@ -214,6 +215,15 @@
             label="SciPy",
         )
 
+        lanczos_avg = df.groupby("matrix_size")["mem_lanczos_mb"].mean()
+        plt.plot(
+            lanczos_avg.index,
+            lanczos_avg.values,
+            marker="*",
+            linestyle="-",
+            label="Lanczos",
+        )
+
         for nproc in nproc_values:
             subset = df[df["n_processes"] == nproc].sort_values("matrix_size")
             label = f"Divide et impera ({nproc} proc{'s' if nproc > 1 else ''})"
@@ -265,6 +275,15 @@
             label="SciPy",
         )
 
+        lanczos_avg = df.groupby("matrix_size")["time_lanczos"].mean()
+        plt.plot(
+            lanczos_avg.index,
+            lanczos_avg.values,
+            marker="*",
+            linestyle="-",
+            label="Lanczos",
+        )
+
         for nproc in nproc_values:
             subset = df[df["n_processes"] == nproc].sort_values("matrix_size")
             label = f"Divide et impera ({nproc} proc{'s' if nproc > 1 else ''})"

shell/submit.sbatch

@@ -40,8 +40,10 @@ last_nproc="${n_processes[-1]}"
 
 CONFIG_FILE="experiments/config.yaml"
 
-#rm -rf logs
-echo "IMPORTANT: please remember to delete the logs folder before calling this script."
+mkdir -p logs
+rm logs/*
+mkdir -p Profiling_files
+rm Profiling_files/*
 
 # Backup the original config
 cp $CONFIG_FILE ${CONFIG_FILE}.bak

shell/submit.sh

@@ -9,8 +9,10 @@ last_nproc="${n_processes[-1]}"
 
 CONFIG_FILE="experiments/config.yaml"
 
-#rm -rf logs
-echo "IMPORTANT: please remember to delete the logs folder before calling this script."
+mkdir -p logs
+rm logs/*
+mkdir -p Profiling_files
+rm Profiling_files/*
 
 # Backup the original config
 cp $CONFIG_FILE ${CONFIG_FILE}.bak

shell/time_profile.sh

@@ -3,9 +3,16 @@
 SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )
 python_script="$SCRIPT_DIR/../scripts/profiling_memory_and_time.py"
 profiling_files="$SCRIPT_DIR/../Profiling_files/"*
+CONFIG_FOLDER="$SCRIPT_DIR/../experiments"
+CONFIG_FILE="$CONFIG_FOLDER/config.yaml"
 
+echo "Please pass as input the following arguments: 1) number of processes 2) matrix size"
 echo "Cleaning the Profiling_files folder"
 rm -f $profiling_files
 
-echo "Running with $1 process(es)"
+cp "${CONFIG_FILE}" "${CONFIG_FOLDER}/config.bak"
+sed -i "s/^dim: .*/dim: $2/" $CONFIG_FILE
+mv "${CONFIG_FOLDER}/config.bak" "${CONFIG_FILE}"
+
+echo "Running with $1 process(es) with a matrix of size $2"
 mpirun -np "$1" python "$python_script"

src/pyclassify/parallel_tridiag_eigen.py

@@ -49,7 +49,6 @@ def find_interval_extreme(total_dimension, n_processor):
     return counts, displs
 
 
-# @profile
 def parallel_tridiag_eigen(
     diag,
     off,