News for hackers. Stuff that matters.

A comment on XLA: linear algebra library for TensorFlow.

Re:XLA: linear algebra library for TensorFlow by probdist on Tuesday March 07, 2017 @11:21PM (#13817868)
Why does this support JIT but not AOT for NVIDIA GPUs?
[ Parent | on HN ]
Re:XLA: linear algebra library for TensorFlow by learyg on Wednesday March 08, 2017 @12:04AM (#13818045)
AOT for GPUs is doable. Do you have a killer use case?

For CPU, mobile code footprint reduction was the driving force.

[ Parent | on HN ]
Re:XLA: linear algebra library for TensorFlow by puzzle on Wednesday March 08, 2017 @07:52AM (#13819848)
Not having to ship a toolchain (nvcc, gpucc or whatever equivalent linked as a library)?
[ Parent | on HN ]
Re:XLA: linear algebra library for TensorFlow by learyg on Wednesday March 08, 2017 @07:07PM (#13825258)
It's gpucc, so it builds from LLVM when you enable XLA in the TF configure step.

Trying to understand: you don't want to ship a compiler library on principle, or is it some kind of product requirement, or ? There's lots of cool work to be done in the compiler space so use cases help to prioritize. :-) Thanks!

[ Parent | on HN ]
Re:XLA: linear algebra library for TensorFlow by posterboy on Wednesday March 08, 2017 @12:53PM (#13822011)
which from dev POV means having to ship a tool(chain) (for the AoT stuff)!
[ Parent | on HN ]