Repository navigation
Conversation
…0a nightly (hipBLASLt 1.5.0) tools/hip/gfx1100-hipblaslt-100500.txt: the same 16 dense GEMM geometries as the other gfx1100 tables (bf16 and f16, T=4096 and T=8192, 32 rows), calibrated on an RX 7900 XTX with tune_hipblaslt (relative L2 <= 1e-4, max abs <= 1e-2, no padding writes) against a ROCm 10.2.0a20261003 nightly SDK (libamdhip64.so.7.17.26392), hipBLASLt 1.5.0 (version number 100500). The engine refuses a table of another hipBLASLt version, so on this ROCm the shipped gfx1100 tables (100100, 100200) are refused and prefill runs on plain hipBLAS - the same gap the gfx1201 1.5.0 table (Niko1221#481) closed for the R9700. setup picks this one by name once the installed hipBLASLt reports 100500. Measured on a 0.1.38-lineage nightly build of this tree, Qwen3.8-Flash-Next IQ3_S greedy, arms differing only in STRATA_HIPBLASLT_TUNING (median of 3 clean runs each): | prompt tokens | with table | plain hipBLAS | gain | | 131071 | 1687 tok/s | 926 tok/s | 1.82x | decode unchanged (62.3 vs 60.7 tok/s, overlapping ranges). Per-run data, the calibration command and the tuner log: Niko1221#745 (bench/results/2026-10-04-community-rx7900xtx-hipblaslt-100500/). docs/AMD_HIP.md: the shipped-tables list carries the entry.
xyzzing
added a commit
to xyzzing/Strata
that referenced
this pull request
Oct 4, 2026
…d opt-in switches, and the status of the two open performance PRs (Niko1221#755 table, Niko1221#744 autotune)
xyzzing
added a commit
to xyzzing/Strata
that referenced
this pull request
Oct 4, 2026
This was referenced Oct 5, 2026
Owner
|
Merged for 0.1.40 (gfx1100 hipBLASLt table 100500). |
Niko1221
pushed a commit
that referenced
this pull request
Oct 7, 2026
…5 (PR #917) Interactive Q2_0 and UD-IQ4_XS chats against llama.cpp Vulkan, then controlled A-E runs of the gfx1151 hipBLASLt tuning table (250.9 to 528.9 prompt tok/s), dense MMQ (+0.2%) and the shared-expert decode stream (+2.5%, 8 requests). The build is 0.1.39 plus then-unmerged PRs (#895, #820, a tuning table adapted from #755), not the 0.1.40 release; the index line in COMMUNITY_BENCHMARKS.md says so and notes that 0.1.40 turned the shared-expert stream off on HIP (#826).
0xrjman
pushed a commit
to 0xrjman/Strata
that referenced
this pull request
Oct 7, 2026
…5 (PR Niko1221#917) Interactive Q2_0 and UD-IQ4_XS chats against llama.cpp Vulkan, then controlled A-E runs of the gfx1151 hipBLASLt tuning table (250.9 to 528.9 prompt tok/s), dense MMQ (+0.2%) and the shared-expert decode stream (+2.5%, 8 requests). The build is 0.1.39 plus then-unmerged PRs (Niko1221#895, Niko1221#820, a tuning table adapted from Niko1221#755), not the 0.1.40 release; the index line in COMMUNITY_BENCHMARKS.md says so and notes that 0.1.40 turned the shared-expert stream off on HIP (Niko1221#826).
xyzzing
added a commit
to xyzzing/Strata
that referenced
this pull request
Oct 13, 2026
… — median of 3 clean cells per tier, the shipped config Engine 0.1.43.1 (refresh of the 0.1.38-era row), ROCm 10.2 nightly, the tuned gfx1100 hipBLASLt-100500 table (Niko1221#755), Ryzen 9 7900X / 96 GB, one-shot greedy runs, 256 generated tokens, median of 3 clean cells per tier (the recorded methodology). Decode now carries STRATA_ROUTE_TAIL_SKIP=7 (adopted after a task-battery-gated A/B; answers differ slightly, =0 restores) — the decode row moves 59.2-65.1 -> 71.4-81.3 tok/s; prefill 32K-128K moves 1,494-1,641 -> 2,117-2,149 tok/s.
xyzzing
added a commit
to xyzzing/Strata
that referenced
this pull request
Oct 13, 2026
… — median of 3 clean cells per tier, the shipped config Engine 0.1.43.1 (refresh of the 0.1.38-era row), ROCm 10.2 nightly, the tuned gfx1100 hipBLASLt-100500 table (Niko1221#755), Ryzen 9 7900X / 96 GB, one-shot greedy runs, 256 generated tokens, median of 3 clean cells per tier (the recorded methodology). Decode now carries STRATA_ROUTE_TAIL_SKIP=7 (adopted after a task-battery-gated A/B; answers differ slightly, =0 restores) — the decode row moves 59.2-65.1 -> 71.4-81.3 tok/s; prefill 32K-128K moves 1,494-1,641 -> 2,117-2,149 tok/s.
xyzzing
added a commit
to xyzzing/Strata
that referenced
this pull request
Oct 13, 2026
… — median of 3 clean cells per tier, the shipped config Engine 0.1.43.1 (refresh of the 0.1.38-era row), ROCm 10.2 nightly, the tuned gfx1100 hipBLASLt-100500 table (Niko1221#755), Ryzen 9 7900X / 96 GB, one-shot greedy runs, 256 generated tokens, median of 3 clean cells per tier (the recorded methodology). Decode now carries STRATA_ROUTE_TAIL_SKIP=7 (adopted after a task-battery-gated A/B; answers differ slightly, =0 restores) — the decode row moves 59.2-65.1 -> 71.4-81.3 tok/s; prefill 32K-128K moves 1,494-1,641 -> 2,117-2,149 tok/s.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Ship
tools/hip/gfx1100-hipblaslt-100500.txt— a hipBLASLt tuning table for the RX 7900 XTX calibrated against hipBLASLt 1.5.0 (version number 100500), the same gap #481 closed for gfx1201. This is an engine-tree companion to the results-only #745 (its second commit measures exactly this table; per-run data lives there).Why: the engine refuses a table whose version doesn't match the installed hipBLASLt, so on a ROCm 10.2 nightly (hipBLASLt 1.5.0) the shipped
gfx1100-hipblaslt-100100/100200.txtare refused and prefill silently runs on plain hipBLAS. Every gfx1100 nightly user is on the untuned path today.What: the same 16 dense GEMM geometries as the other gfx1100 tables (bf16 + f16, T = 4096 and 8192 → 32 rows), calibrated with
tune_hipblasltrun against that exact nightly library (ROCm 10.2.0a20261003 SDK,libamdhip64.so.7.17.26392; every candidate gated at rel-L2 ≤ 1e-4 / max-abs ≤ 1e-2 / no padding writes; each row is its case's measured best).setuppicks it up by name once the installed hipBLASLt reports 100500 — no config change needed.Measured (0.1.38-lineage nightly binary of this tree, Qwen3.8-Flash-Next IQ3_S greedy, arms differing only in
STRATA_HIPBLASLT_TUNING, median of 3 clean runs each):Decode unchanged (62.3 vs 60.7 tok/s, overlapping ranges — the table is prefill-only, as designed). Per-run data, calibration command and tuner log: bench/results/2026-10-04-community-rx7900xtx-hipblaslt-100500/ in #745.
docs/AMD_HIP.md: one entry in the shipped-tables list, mirroring the gfx1201 1.5.0 entry (including the caveat that the version number is the only thing the engine can check — recalibrate with
tune_hipblasltfor a different 1.5.0 build).Checklist
docs/COMMUNITY_BENCHMARKS.md.