Research from your subscriptions

Paper

When a larger cache makes a GPU slower

Review

Separating cache capacity from replacement policy

Replication

Reproducing a memory-bound inference result

Negative result

A scheduling heuristic that did not generalize

Paper

Warm-up is part of the measurement

Review

What belongs in a minimal reproduction package?

Paper·Recommended

The hidden cost of a little more precision

Related to Computer architecture, a topic you follow.

Paper

One command to reproduce the central experiment

Paper

The latency hidden inside a larger batch

Paper

A random seed is not a complete experimental input

Review

What an average memory latency leaves out

Replication

A smaller test of the same claim