Issues

• can you benchmark the new feature in libtpu==0.0.48 in cross core communication?

  AI-Hypercomputer/accelerator-microbenchmarks#137 · Filter by author pathfinder-pf opened on Sep 17, 2026

• The all_to_all benchmark test results on TPUv7x don't meet expectations.

  AI-Hypercomputer/accelerator-microbenchmarks#73 · Filter by author jimoosciuc opened on Dec 15, 2025
  1

• [Question] I'm confused about what the "_xd" suffix means in the all_to_all benchmark configs

  AI-Hypercomputer/accelerator-microbenchmarks#72 · Filter by author jimoosciuc opened on Dec 15, 2025
  1

• The allreduce benchmark results on TPUv7x do not meet expectations (only achieving 1/10 of the bandwidth)

  AI-Hypercomputer/accelerator-microbenchmarks#71 · Filter by author jimoosciuc opened on Dec 11, 2025