# Rewriting the CUDA cache memory allocator

**URL:** <https://discuss.pytorch.org/t/rewriting-the-cuda-cache-memory-allocator/192696>\
**Category:** Memory Format\
**Created:** [November 28, 2023, 4:48am UTC](https://discuss.pytorch.org/t/rewriting-the-cuda-cache-memory-allocator/192696 "2023-11-28T04:48:06Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![Chi\_Zhang](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/chi_zhang/32/65834_2.png) [@Chi\_Zhang](https://discuss.pytorch.org/u/Chi_Zhang)\
**Post date:** [November 28, 2023, 4:48am UTC](https://discuss.pytorch.org/t/rewriting-the-cuda-cache-memory-allocator/192696/1 "2023-11-28T04:48:06Z")

</div>

Hi,  
I am trying to rewrite the CUDA cache memory allocator based on the CUDAPluggableAllocator, since I think the current CUDA memory allocator may not fit for all the cases. I want to confirm if the CUDA memory allocator is really appropriate or not for all the hardware settings.