-
### 🚀 The feature, motivation and pitch
There is a growing need to mix-and-match cuda allocators in PyTorch for new NVIDIA architectures. For instance:
- Reductions on NVIDIA Switches ([NVLS Red…
-
### Version
1
### DataCap Applicant
Bond
### Project ID
1
### Data Owner Name
Bond
### Data Owner Country/Region
China
### Data Owner Industry
Information, Media & Telecommunications
### W…
-
Function `malloc` and friends are handled specially in `ConstraintResolver::analyzeAllocExpr` and in addition are treated specially when inferring bounds. They almost work properly without special tre…
-
Hi litehtml team,
For the use case that I am interested in (game UI) litehtml seems quite promising. There is a slight problem though. Most game engines (including mine) use custom memory allocation …
-
We currently support in the Java side some combinations of allocators and wrappers that make it easy to setup pools (default, arena, and now async), backing allocators (cuda, managed memory). But it i…
-
This repeats the second from last line a few 100 times and ends with the last line
Logs:
2024-05-11 09:21:27.764059-07:00"memorysetup-temp-allocator-size-gfx=262144"
2024-05-11 09:21:29.014183-07…
-
n-dub updated
7 months ago
-
At the moment, the parsers heavily rely on boxing to provide the indirection needed for our recursive data types. However this puts a lot of pressure on the memory allocator for probably relatively sh…
-
CreateComittedResource is fine,
However is good to add interface to CreateTexture, CreateTextureEx
Particularly on d3d12 this allocator is very often used
https://github.com/GPUOpen-Libraries…
-
It was a safety requirement that they *don't* for GlobalAlloc.
Should unsafe code need to deal with allocators unwinding (as opposed to aborting or otherwise halting the program)?