Experiments / expE_partial_gemm

expE_partial_gemm

Partial matrix multiplication: quality vs blocks evaluated, incl. a Metal kernel bandwidth check

View benchmark implementation (benchmark.py) →

Hypothesis

documentexpE_partial_gemm/hypothesisauthorSimon-Pierre BouchercreatedMon Aug 10 2026 20:00:00 GMT-0400 (heure avancée de l’Est)statusdraft

Hypothesis — expE_partial_gemm#

Hypothesis
  <what we believe and why>

Falsification criterion
  <the concrete measurable outcome that would prove this wrong>

Method
  <exact procedure, model(s), data, seeds, measurement points>

Baseline
  <what this is compared against — no straw men>

Result
  <filled after the run: numbers, with mean/median/std and run count>

Interpretation
  <what the numbers mean; alternative explanations considered>

Next experiment
  <the most informative follow-up given this result>

Analysis

documentexpE_partial_gemm/analysisauthorSimon-Pierre BouchercreatedMon Aug 10 2026 20:00:00 GMT-0400 (heure avancée de l’Est)statusdraft

Analysis — expE_partial_gemm#

To be written after results exist. Must include the seven-field block and the evidence standard of CLAUDE.md §10.

README

documentexpE_partial_gemm/READMEauthorSimon-Pierre BouchercreatedMon Aug 10 2026 20:00:00 GMT-0400 (heure avancée de l’Est)statusdraft

expE_partial_gemm#

Partial matrix multiplication: quality vs blocks evaluated, incl. a Metal kernel bandwidth check

Status: scaffolded 2026-08-11, not yet run.