Issues

• Upstream draft for review: an XLA feature request to let float64 dots use cuBLAS's float64 emulation

  QuantEcon/benchmarks#30 · Filter by author mmcky opened on Nov 3, 2026

• Upstream draft for review: RTX PRO 6000 Blackwell results for JAX's float64 column-scatter issue

  QuantEcon/benchmarks#29 · Filter by author mmcky opened on Nov 3, 2026

• TPUs: float64 without float64 hardware, and QuantEcon's JAX workloads on Google Cloud

  QuantEcon/benchmarks#27 · Filter by author mmcky opened on Nov 2, 2026

• Float32 fidelity: which float64 lectures need float64, and can the heaviest step drop to float32?

  QuantEcon/benchmarks#11 · Filter by author mmcky opened on Oct 25, 2026
  2

• Problem types: NumPy, Numba and JAX on CPU and GPU, to guide the choice of stack

  QuantEcon/benchmarks#5 · Filter by author mmcky opened on Oct 25, 2026
  1

• Whole lectures, as used: every JAX lecture on an Apple CPU, a Turing GPU and an x86 CPU

  QuantEcon/benchmarks#4 · Filter by author mmcky opened on Oct 25, 2026
  1

• Precision and platform: float64 against float32 on CPU, CUDA and Apple Silicon

  QuantEcon/benchmarks#1 · Filter by author mmcky opened on Oct 24, 2026
  2