Improve element-wise tensor operation performance - #454

Merged

Dobiasd merged 7 commits into

optimize-inference-performance

Mar 13, 2026

Commits on Mar 13, 2026

Add fast path for the common standard ReLU (max(0, x))
Dobiasd
committed
cb18805
Optimize convolution
Dobiasd
committed
a7bab79
Revert conv change which did not actually help
Dobiasd
committed
03e2227
it was just the inlining
Dobiasd
committed
5f26c03
relu special case with simple lambda
Dobiasd
committed
1a404d1
Use std::transform with pre-allocated buffer in transform_tensor
Dobiasd
committed
7b813c5
auto-format
Dobiasd
committed
98902c9