2025-12-05 14:29:58 :: Benchmarking

1. Description

This report analyzes the performance of matrix-vector multiplication for different matrix sizes (N) and levels of task parallelism. We compare fill and compute phases and evaluate how execution time and speedup evolve with increasing workloads.

1.1. Session

Hostname: feelpp1
User: feelpp1
Time Start: 20251205T143002+0100
Time End: 20251205T143015+0100

1.2. Cases

Total: 12
Failures: 0
Runs: 1

2. Parametrization

	Hash	tasks	elements	Total Time (s)
pass	73dec7f2	1.0	1000.0	0.8033056259155273	Logs
pass	73dec7f2	1.0	1000.0	0.8033056259155273	Logs
pass	ecd9fbe1	1.0	2000.0	0.9616482257843018	Logs
pass	ecd9fbe1	1.0	2000.0	0.9616482257843018	Logs
pass	97cff20f	1.0	3000.0	1.212583065032959	Logs
pass	97cff20f	1.0	3000.0	1.212583065032959	Logs
pass	79cb2fd1	1.0	4000.0	1.5120949745178223	Logs
pass	79cb2fd1	1.0	4000.0	1.5120949745178223	Logs
pass	64b44271	2.0	1000.0	0.7444722652435303	Logs
pass	64b44271	2.0	1000.0	0.7444722652435303	Logs
pass	48e3c028	2.0	2000.0	0.9612433910369873	Logs
pass	48e3c028	2.0	2000.0	0.9612433910369873	Logs
pass	03f7318f	2.0	3000.0	1.2199804782867432	Logs
pass	03f7318f	2.0	3000.0	1.2199804782867432	Logs
pass	9341436b	2.0	4000.0	1.5131113529205322	Logs
pass	9341436b	2.0	4000.0	1.5131113529205322	Logs
pass	1fe7b1b0	4.0	1000.0	0.7454335689544678	Logs
pass	1fe7b1b0	4.0	1000.0	0.7454335689544678	Logs
pass	b58ae281	4.0	2000.0	0.9570603370666504	Logs
pass	b58ae281	4.0	2000.0	0.9570603370666504	Logs
pass	0fe61403	4.0	3000.0	0.9597663879394531	Logs
pass	0fe61403	4.0	3000.0	0.9597663879394531	Logs
pass	a6450245	4.0	4000.0	1.2114019393920898	Logs
pass	a6450245	4.0	4000.0	1.2114019393920898	Logs

3. Performance Analysis

3.1. Execution Time vs Number of Tasks

This plot shows the execution time breakdown (fill and compute stages) as the number of tasks increases. It highlights parallel scaling behavior for different matrix sizes.

3.2. Execution Time vs Problem Size (N)

This plot illustrates how execution time scales with matrix size N under different task counts. It helps identify compute bottlenecks and memory throughput constraints.

3.3. Speedup Analysis

This speedup plot shows how effectively the computation accelerates when increasing task parallelism. It highlights differences between fill and compute phases and reveals scaling limits for larger matrix sizes.