benchmark: compare3.jl — FFTW vs registry FFTA vs checkouts, side by side - #1
Open
pankgeorg wants to merge 21 commits into
Open
benchmark: compare3.jl — FFTW vs registry FFTA vs checkouts, side by side#1pankgeorg wants to merge 21 commits into
pankgeorg wants to merge 21 commits into
Conversation
…ate processes Runs FFTW, a registry FFTA version and any number of FFTA checkouts on the suite.jl case list (moved to cases.jl), one process and environment each, and renders one table with ratios to FFTW, speedups over a reference column, per-class geometric means and thread scaling.
…to reuse finished columns
Re-runs the power-of-two, composite and Bluestein kernels with stages removed (leaves / butterfly passes / twiddle pass / sub-transforms) so the stage times add up to the total, next to a memory-pass floor and FFTW.
…ch64, 1/4/16 threads)
…aarch64, 1 thread)
…odelets (aarch64, 1 thread)
… experiment columns (aarch64, 1 thread)
…on/experiments), FFTW MEASURE sweeps, DSP.jl workloads (aarch64)
…g branch (aarch64, 1 thread, rfft cases)
…ng reuse only columns that cover the current run
…SURE sweeps on integration/experiments
… pencils (aarch64, 1+16 threads)
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds
benchmark/compare3.jl(+compare3_worker.jl,cases.jlshared withsuite.jl): each--impl NAME=SPECcolumn (fftw,@0.3.1, or a checkout path) runs in its own process/environment on the suite's case list, per--threadscount; one markdown table gives times, ratio to FFTW, speedup over--ref, per-class geomeans, thread scaling and a checksum check;--classes/--skip-existingrestrict classes and reuse finished columns. Alsokernel_stages.jl+KERNEL_STAGES.md(per-stage time breakdown inside the kernels).Baseline, aarch64 Neoverse-N1, 1 thread (
benchmark/compare3/aarch64/baseline/COMPARE3.md): FFTA 0.3.1 / FFTW 6.49× and integration/all / FFTW 2.45× (geomean, 353 matched cases; 0.3.1's prime/awkward classes capped at 2^12 — it is O(n²) there, 8.8 s per call at n = 49757). By class, 0.3.1 → integration/all: pow2 2.07→1.23×, smooth 4.11→2.92×, prime 7.30→2.00×, 2D 8.59→1.95×, rfft 2D 19.1→2.38× (ComplexF64/Float64).Tested:
--quicksmoke run,suite.jl --quickafter thecases.jlsplit, the full three-way run.