|
1 | | -OpenSolvers explores how open-source scientific software runs on real hardware — starting with **RISC-V** boards and the tools that make that practical (EESSI, OpenBLAS, HPL, and friends). This site documents what we learn along the way. |
| 1 | +OpenSolvers explores how open-source scientific software runs on real hardware — starting with **RISC-V** boards and the tools that make that practical (EESSI, OpenBLAS, and friends). This site documents what we learn along the way. |
2 | 2 |
|
3 | 3 | ## What we're working on |
4 | 4 |
|
5 | | -We're benchmarking **High Performance Linpack (HPL)** across several consumer RISC-V boards through the [EESSI](https://www.eessi.io/) stack. On the VisionFive 2, U74-tuned OpenBLAS lifts HPL **3.13 → 5.28 GFLOP/s** (**1.69×**). On the Orange Pi RV2, fixing an OpenBLAS RVV `gemv_n` bug turns a failing stock run into **10.53 GFLOP/s** (PASSED). |
| 5 | +We benchmark **scientific libraries** and **applications** on consumer RISC-V boards through the [EESSI](https://www.eessi.io/) stack — from BLAS kernels up to full app runs — swapping fixed OpenBLAS builds via FlexiBLAS without rebuilding downstream code. |
6 | 6 |
|
7 | | -See the [HPL results overview](apps/hpl.html) for a cross-board summary. |
| 7 | +Recent highlights on the Orange Pi RV2 (SpaceMiT X60, RVV): fixing an OpenBLAS `gemv_n` bug restores correctness across [BLAS](scientific-libs/blas.html), [LAPACK](scientific-libs/lapack.html), [ELPA](scientific-libs/elpa.html), and [HPL](apps/hpl.html) — with patched RVV reaching **10.53 GFLOP/s** on Linpack and **1.58×** on a dense eigensolve. |
| 8 | + |
| 9 | +## Scientific libs |
| 10 | + |
| 11 | +Library-level probes — performance *and* numerical correctness: |
| 12 | + |
| 13 | +- **[BLAS](scientific-libs/blas.html)** — OpenBLAS improvements (U74 tuned kernel, X60 RVV `gemv_n` fix) and `difftest` / `bench_dgemm` verification |
| 14 | +- **[LAPACK](scientific-libs/lapack.html)** — NumPy `dgemm` and `eigvalsh` as a real SciPy-stack probe |
| 15 | +- **[ELPA](scientific-libs/elpa.html)** — dense eigensolver mixing BLAS-2 and BLAS-3 (CP2K / VASP class workloads) |
| 16 | + |
| 17 | +## Apps |
| 18 | + |
| 19 | +End-to-end application benchmarks on the same boards and EESSI toolchain: |
| 20 | + |
| 21 | +- **[HPL](apps/hpl.html)** — High Performance Linpack; cross-board summary and A/B configs from [opensolvers/benchmarks](https://github.com/opensolvers/benchmarks) |
8 | 22 |
|
9 | 23 | ## Boards |
10 | 24 |
|
11 | | -- **[StarFive VisionFive 2](boards/VisionFive2.html)** — JH7110 SoC, 4× SiFive U74 (`rv64gc`). EESSI HPL **3.13 → 5.28 GFLOP/s** with U74 OpenBLAS tuning. |
12 | | -- **[Orange Pi RV2](boards/RV2.html)** — SpaceMiT K1 SoC, 8× SpacemiT X60 (RVV). EESSI HPL **FAILED (`nan`) → 10.53 GFLOP/s** with fixed OpenBLAS. |
13 | | -- **[Banana Pi F3](boards/F3.html)** — SpacemiT K1 SoC, 8× SpacemiT X60. Same RVV fix as RV2; peak HPL not yet recorded. |
| 25 | +- **[StarFive VisionFive 2](boards/VisionFive2.html)** — JH7110 SoC, 4× SiFive U74 (`rv64gc`). U74 OpenBLAS tuning: HPL **3.13 → 5.28 GFLOP/s**. |
| 26 | +- **[Orange Pi RV2](boards/RV2.html)** — SpaceMiT K1, 8× X60 (RVV). Fixed OpenBLAS: HPL **FAILED (`nan`) → 10.53 GFLOP/s**; ELPA **34.81 s** (vs 54.92 s scalar). |
| 27 | +- **[Banana Pi F3](boards/F3.html)** — same K1 / X60 SoC; same fixes apply, peak results not yet recorded. |
14 | 28 |
|
15 | 29 | Use the menu above to jump to a board, app, or scientific lib page. |
0 commit comments