Vendor BQM and add paper benchmarks - #110
Open
klamike wants to merge 21 commits into
Open
Conversation
The CPU baseline now solves the presolved, scaled, standard-form instances the GPU solves, instead of the raw model, and it solves every instance of the largest batch one after the other. For each batch size b the CPU block gives the summed sequential cost of instances 1:b and the max per-instance cost (b CPUs with perfect scaling), the GPU block the cost of the same instances solved as one batch, so both sides always cover the same problems, perturbations and iteration counts included. Both sides report solver construction and solve! as wall-clock times (GPU synchronized) instead of the solvers' internal counters, whose clocks start at different points (after the initial point on the CPU, in the constructor on the GPU). The warmup uses the benchmark's linear solver on the same instance layout, GPU memory is reclaimed between batch sizes, and the DC-OPF per-step timings are taken from a fresh initial point on both sides. Also restore MadNLPHSL for Ma27/Ma57, add the missing sequential CPU helper in the DC-OPF comparison, and fix the --job flag of the DC-OPF script. Result layout: nvar, ncon, nnzj, then per batch size a CPU block (converged, mean iter, summed init, summed solve, max init, max solve) and a GPU block (converged, mean iter, init, solve).
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
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Note the last commit especially. Seems CPU and GPU were timing different things (in particular, unfairly favoring CPU)
Haven't tested on CUDA.