Skip to content

Benchmarks intel ultracore #2

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Alvicorn merged 4 commits into from
Jun 1, 2026
Merged

Benchmarks intel ultracore #2

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
32a2ecd
add cli benchmark progress tracking
Jun 1, 2026
a670b48
intel64 family 6 model 170 benchmark with python3.14t
Alvicorn Jun 1, 2026
0c308e6
resolve merge conflicts with main
Alvicorn Jun 1, 2026
8f0cd20
fix linting errors
Alvicorn Jun 1, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
126 changes: 126 additions & 0 deletions reports/32a2ecd-amd64-windows-intel-16c22t-1.4ghz-31gb-3.14+freethreaded/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,126 @@
# Benchmark Report — Intel64 Family 6 Model 170 Stepping 4

- **Date:** 2026-06-01
- **Benchmark commit:** `32a2ecd`
- **Geomean speedup:** 1.69×

## Machine

```
Machine: Intel64 Family 6 Model 170 Stepping 4, GenuineIntel
Arch: AMD64 | 16P / 22L cores @ 1.40 GHz | 31.5 GB RAM
OS: Windows 11 (10.0.26200)
Python: CPython 3.14.5 (free-threaded: yes)
Commit: 32a2ecd
```

## Command

```sh
uv run --python 3.14+freethreaded python run_bench.py --runs 10 --mode both
```

## Results

| Workload | Sync | Serial (s) | Parallel (s) | Speedup |
|---|---|---:|---:|---:|
| `adaptive_jacobi` | condition | 0.064 | 0.045 | 1.43x |
| `bfs` | none | 0.106 | 0.117 | 0.90x |
| `bitonic_sort` | barrier | 1.013 | 9.242 | 0.11x |
| `bounded_quadsum` | bounded-semaphore | 0.755 | 0.202 | 3.73x |
| `bounded_workers` | bounded-semaphore | 1.967 | 0.638 | 3.08x |
| `concurrent_hashmap` | striped-lock | 0.022 | 0.019 | 1.12x |
| `cv_bounded_buffer` | condition | 0.230 | 0.167 | 1.38x |
| `dining_philosophers` | fork-locks | 0.084 | 0.041 | 2.07x |
| `early_term_search` | event | 0.083 | 0.041 | 2.03x |
| `factorization_pool` | semaphore | 0.056 | 0.023 | 2.47x |
| `fft` | barrier | 0.277 | 1.455 | 0.19x |
| `floyd_warshall` | barrier | 0.197 | 0.106 | 1.85x |
| `matmul` | none | 0.171 | 0.088 | 1.95x |
| `memo_recursion` | rlock | 6.256 | 1.552 | 4.03x |
| `monte_carlo_pi` | none | 0.193 | 0.049 | 3.98x |
| `nested_counter` | rlock | 2.140 | 0.418 | 5.12x |
| `numerical_integration` | none | 0.817 | 0.184 | 4.43x |
| `page_rank` | barrier | 0.219 | 0.373 | 0.59x |
| `password_crack` | none | 0.176 | 0.081 | 2.18x |
| `permit_pool` | semaphore | 6.052 | 1.197 | 5.05x |
| `pollard_factor` | event | 0.025 | 0.022 | 1.13x |
| `prime_sieve` | none | 0.244 | 0.074 | 3.31x |
| `priority_pipeline` | queue | 0.215 | 0.154 | 1.39x |
| `producer_consumer` | queue | 0.217 | 0.154 | 1.41x |

## Raw output

<details>
<summary>Full <code>run_bench.py</code> stdout</summary>

```
Machine: Intel64 Family 6 Model 170 Stepping 4, GenuineIntel
Arch: AMD64 | 16P / 22L cores @ 1.40 GHz | 31.5 GB RAM
OS: Windows 11 (10.0.26200)
Python: CPython 3.14.5 (free-threaded: yes)
Commit: 32a2ecd

Running 24 workload(s) × 10 run(s) × 2 mode(s) (parallel,serial) = 480 subprocess(es); smoke=False

-> [1/24] adaptive_jacobi parallel wall= 0.045s serial wall= 0.064s speedup= 1.43x par_eff= 17.8%
-> [2/24] bfs parallel wall= 0.117s serial wall= 0.106s speedup= 0.90x par_eff= 11.3%
-> [3/24] bitonic_sort parallel wall= 9.242s serial wall= 1.013s speedup= 0.11x par_eff= 1.4%
-> [4/24] bounded_quadsum parallel wall= 0.202s serial wall= 0.755s speedup= 3.73x par_eff= 46.7%
-> [5/24] bounded_workers parallel wall= 0.638s serial wall= 1.967s speedup= 3.08x par_eff= 38.5%
-> [6/24] concurrent_hashmap parallel wall= 0.019s serial wall= 0.022s speedup= 1.12x par_eff= 14.0%
-> [7/24] cv_bounded_buffer parallel wall= 0.167s serial wall= 0.230s speedup= 1.38x par_eff= 17.3%
-> [8/24] dining_philosophers parallel wall= 0.041s serial wall= 0.084s speedup= 2.07x par_eff= 25.9%
-> [9/24] early_term_search parallel wall= 0.041s serial wall= 0.083s speedup= 2.03x par_eff= 25.3%
-> [10/24] factorization_pool parallel wall= 0.023s serial wall= 0.056s speedup= 2.47x par_eff= 30.8%
-> [11/24] fft parallel wall= 1.455s serial wall= 0.277s speedup= 0.19x par_eff= 2.4%
-> [12/24] floyd_warshall parallel wall= 0.106s serial wall= 0.197s speedup= 1.85x par_eff= 23.1%
-> [13/24] matmul parallel wall= 0.088s serial wall= 0.171s speedup= 1.95x par_eff= 24.3%
-> [14/24] memo_recursion parallel wall= 1.552s serial wall= 6.256s speedup= 4.03x par_eff= 50.4%
-> [15/24] monte_carlo_pi parallel wall= 0.049s serial wall= 0.193s speedup= 3.98x par_eff= 49.7%
-> [16/24] nested_counter parallel wall= 0.418s serial wall= 2.140s speedup= 5.12x par_eff= 64.0%
-> [17/24] numerical_integration parallel wall= 0.184s serial wall= 0.817s speedup= 4.43x par_eff= 55.4%
-> [18/24] page_rank parallel wall= 0.373s serial wall= 0.219s speedup= 0.59x par_eff= 7.3%
-> [19/24] password_crack parallel wall= 0.081s serial wall= 0.176s speedup= 2.18x par_eff= 27.3%
-> [20/24] permit_pool parallel wall= 1.197s serial wall= 6.052s speedup= 5.05x par_eff= 63.2%
-> [21/24] pollard_factor parallel wall= 0.022s serial wall= 0.025s speedup= 1.13x par_eff= 14.2%
-> [22/24] prime_sieve parallel wall= 0.074s serial wall= 0.244s speedup= 3.31x par_eff= 41.4%
-> [23/24] priority_pipeline parallel wall= 0.154s serial wall= 0.215s speedup= 1.39x par_eff= 17.4%
-> [24/24] producer_consumer parallel wall= 0.154s serial wall= 0.217s speedup= 1.41x par_eff= 17.6%

======================================================================
workload runs threads serial wall (s) parallel wall (s) ± stdev speedup par eff peak MB checksum
--------------------- ----- ------- --------------- ----------------- ------- ------- ------- ------- --------
adaptive_jacobi 10/10 8 0.064 0.045 0.005 1.43x 17.8% 27.2 7938836
bfs 10/10 8 0.106 0.117 0.009 0.90x 11.3% 47.0 4973
bitonic_sort 10/10 8 1.013 9.242 0.728 0.11x 1.4% 25.7 4701893
bounded_quadsum 10/10 8 0.755 0.202 0.032 3.73x 46.7% 22.2 8956041
bounded_workers 10/10 8 1.967 0.638 0.051 3.08x 38.5% 23.9 5705378
concurrent_hashmap 10/10 8 0.022 0.019 0.003 1.12x 14.0% 27.9 6297791
cv_bounded_buffer 10/10 8 0.230 0.167 0.015 1.38x 17.3% 24.4 5111357
dining_philosophers 10/10 8 0.084 0.041 0.002 2.07x 25.9% 22.5 960
early_term_search 10/10 8 0.083 0.041 0.005 2.03x 25.3% 24.0 88484
factorization_pool 10/10 8 0.056 0.023 0.005 2.47x 30.8% 22.7 1998560
fft 10/10 8 0.277 1.455 0.207 0.19x 2.4% 36.6 1142602
floyd_warshall 10/10 8 0.197 0.106 0.009 1.85x 23.1% 23.2 8421045
matmul 10/10 8 0.171 0.088 0.007 1.95x 24.3% 24.0 6535330
memo_recursion 10/10 8 6.256 1.552 0.163 4.03x 50.4% 22.5 4881333
monte_carlo_pi 10/10 8 0.193 0.049 0.005 3.98x 49.7% 22.3 1257606
nested_counter 10/10 8 2.140 0.418 0.035 5.12x 64.0% 22.4 4000408
numerical_integration 10/10 8 0.817 0.184 0.007 4.43x 55.4% 22.2 3987766
page_rank 10/10 8 0.219 0.373 0.021 0.59x 7.3% 28.4 9998118
password_crack 10/10 8 0.176 0.081 0.004 2.18x 27.3% 23.8 754
permit_pool 10/10 8 6.052 1.197 0.113 5.05x 63.2% 41.1 327526
pollard_factor 10/10 8 0.025 0.022 0.005 1.13x 14.2% 22.5 9346167
prime_sieve 10/10 8 0.244 0.074 0.013 3.31x 41.4% 23.8 148933
priority_pipeline 10/10 8 0.215 0.154 0.007 1.39x 17.4% 24.6 5111357
producer_consumer 10/10 8 0.217 0.154 0.014 1.41x 17.6% 24.3 5111357
======================================================================

speedup = serial_wall_mean / parallel_wall_mean

par eff = speedup / num_threads
'*' on checksum = unstable across runs; 'DIVERGENT' = serial and parallel disagree.
```

</details>
5 changes: 3 additions & 2 deletions reports/README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -109,5 +109,6 @@ The helper script captures:
## Index of submitted reports

| Report | CPU | RAM | OS | Python | Commit | Geomean speedup |
|---|---|---|---|---|---| ---|
| AMD64 Family 25 Model 68 Stepping 1 | AMD64 | 27.3 GB | Windows 11 | 3.14+freethreaded | `32a2ecd` | 2.28x |
|---|---|---|---|---|---|---|
| AMD64 Family 25 Model 68 Stepping 1 | AMD Ryzen 7 | 27.3 GB | Windows 11 | 3.14+freethreaded | `32a2ecd` | 2.28x |
| Intel64 Family 6 Model 170 Stepping 4 | Intel Core Ultra 7 | 31.5 GB | Windows 11 | 3.14+freethreaded | `32a2ecd` | 1.69x |
Loading