2c2c36d3da
add rocm-7.2.1-pr21344 toolbox (gfx1151 MMQ/MMVQ tile + nwarp tuning)
...
Adds a new toolbox variant based on PR #21344 (pedapudi/llama.cpp@gfx1151-opt)
which tunes MMQ tile sizes (x_max=48, y=64) and warp counts (nwarps=4) for
RDNA3_5 gfx1151, yielding up to +100% prefill throughput at small batch sizes.
Also adds BMI2/FMA/F16C CPU SIMD flags and GGML_CUDA_FA_ALL_QUANTS=ON to match
the benchmark build used in the PR. Wire up CI (build matrix + prune), the
refresh script, and run_benchmarks.sh so results land alongside rocm-7.2.1.
2026-04-15 09:23:58 +01:00
4ac481e7d1
chore: upgrade ROCm version from 7.2 to 7.2.1 across configuration and documentation
2026-04-09 18:33:52 +01:00
06fc789eba
chore: deprecate and remove ROCm 7.1.1 toolbox and all associated references.
2026-02-04 17:56:41 +00:00
1807e8cff2
Adding ROCm 7.2 backend
2026-01-23 08:07:40 +00:00
783998589e
neclean up of legacy toolboxes, removal of rocwmma and renamed rocm7-alpha to rocm-7nightlies. Added new benchmarks
2026-01-10 10:31:04 +00:00
9ba6812003
feat: upgrade ROCm to 7.1.1 and update associated tooling and documentation
2025-12-07 09:30:14 +00:00
df54882433
remove manual application of RPC performance PR (this is merged into master now)
2025-11-28 14:20:03 +00:00
1b5ced1255
make PR-15405 application explicit in logs
2025-11-25 10:02:32 +00:00
5105f6cf10
add flag to remvoe RPC PR (for testing)
2025-11-24 17:16:06 +00:00
40a47116a9
Merge remote-tracking branch 'origin/main' into pr-20
2025-11-12 13:19:56 +00:00
abeec3526c
fix: rocm-7alpha pipelines and refresh-toolboxes.sh
2025-11-12 08:35:11 +01:00
8b2cc6a0cc
feat: rocm-7alpha
2025-11-09 15:57:19 +01:00
81a5743bfb
Add basic support for ROCm 7.1 release
2025-11-08 18:39:05 +00:00
f62739d243
update build scripts
2025-10-11 18:52:16 +01:00
a5f27f609d
remove all_quants toolbox
2025-10-11 18:34:47 +01:00
b02ffb55f7
adding rocm-6.4.4
2025-09-27 18:57:03 +01:00
e86bf4c418
Adding bilding steps for FA ALL QUANTS
2025-09-16 11:07:46 +01:00
b71a37647f
Updated benchmakrs, removed old toolboxes and results
2025-08-17 12:32:08 +01:00
5a417bc3b5
removed rocm7beta from builds and fixed 7rc build
2025-08-16 10:58:10 +01:00
8628db44d3
Adding rocm-6.4.3 toolboxes
2025-08-11 16:13:38 +01:00
a9618d881b
- Corrected typo in WMMA (was spelt wrong as waam)
...
- Included rocm-7rc-rocwmma toolbox
- Included updated results from benchmarks including rocm 7rc with ROMWMMA and hipBLASLt
2025-08-10 13:21:06 +01:00
0d6b2dc731
Addin rocm7-rocwaam toolbox
2025-08-09 19:21:52 +01:00
9b7ddaea4a
Added rocm-6.4.2 with ROCWAAM for faster Flash attention kernels
2025-08-09 10:10:29 +01:00
70460248f7
different strategy to free up disk space
2025-08-06 10:39:09 +01:00
7d94b17c07
Free up disk space
2025-08-06 10:35:54 +01:00
de05eb2318
Consolidating build actions
2025-08-06 10:11:15 +01:00
82efce49a1
Reverted action configuration
2025-08-06 09:49:08 +01:00
55d96f4d93
Add manual‐only, parameterized build workflow
2025-08-06 09:46:49 +01:00
c8019bdc0e
Attempt to fix build pipeline
2025-08-06 09:43:56 +01:00
dff443a0e5
Parameterize build project
2025-08-06 09:18:31 +01:00
f2006194dc
Updating build scripts
2025-08-06 09:06:18 +01:00
2c90eac378
Add build-and-push workflow
2025-08-06 08:56:58 +01:00
Donato Capitella