Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
3 changes: 3 additions & 0 deletions CHANGELOG.md
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,9 @@ project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

### Added

- `scripts/solver_factorize.py --cudss` times NVIDIA's cuDSS GPU solver on the real ecoinvent matrix: a
first factorization and one that reuses its analysis, as a Monte Carlo draw does. `--pardiso` now
times PARDISO's reuse too ([method](docs/method.md)).
- When a `floors`, `pre`, `conda-forge` or `suites` compat job fails, its summary and artifact now
hold a draft of an upstream issue: the failed test IDs (or the resolve error), the versions, the
pins that changed since the job's last green set, and the run URL. Posting it is left to a person.
Expand Down
3 changes: 2 additions & 1 deletion ROADMAP.md
Original file line number Diff line number Diff line change
Expand Up @@ -17,7 +17,8 @@ here. Ideas and bugs go in [GitHub issues](https://github.com/simonvanlierde/bri
- **The synthetic inputs as a dataset** on Zenodo, one version per generator version, so the
matrices and ecoSpold2 files can be used without running the generator.
- **More solver columns for bw2calc:** NVIDIA cuDSS behind a GPU opt-in, kept out of the default
run, and Apple's Accelerate LU on macOS.
run (`scripts/solver_factorize.py --cudss` already times it on real ecoinvent), and Apple's
Accelerate LU on macOS.
- **Scaling cells for bw_processing:** how the time to read a datapackage grows with the number of
resource groups, from 1,000 to 10,000.
- **bw2data's write path:** the SQLite database against the Parquet tables proposed for a
Expand Down
2 changes: 1 addition & 1 deletion docs/method.md
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,7 @@ The profile is harder on a direct solver than real ecoinvent: SuperLU's factors
as many entries as the matrix, against 175 for ecoinvent. So SuperLU times overstate real
ecoinvent ([#54](https://github.com/simonvanlierde/brightway-benchmarks/issues/54)). With a
licence of your own, [`scripts/solver_factorize.py`](../scripts/solver_factorize.py) times each
direct solver on the real matrix.
direct solver on the real matrix, and with `--cudss` NVIDIA's cuDSS on a GPU.

`just bw2io-import synthetic smoke,ecoinvent` times bw2io's ecoSpold2 import without a licence:
it writes a profile's system as ecoSpold2 files with bw2io's own `biosphere3` flows, so every
Expand Down
Loading