Klaus Post
Klaus Post
@sigkill What are you thinking off specifically wrt SHA256?
> the AVX512 code doesn't seem like it's sticking to a specific instruction set That seems like a strange expectation. These are the required types: https://github.com/minio/sha256-simd/blob/master/cpuid_other.go#L28 - should match your...
@sigkill If you are able to test it, feel free to send a PR.
The missing instructions sigkill lists AFAICT makes this impossible. So I don't really see any reason to keep this open any more. Feel free to make a PR, if you...
I am always open to things that can reduce the amount of regressions and new bugs. Right now it is a bit hard to commit to a meeting. My main...
@erlapso Sounds great. I will give it a whirl!
Yes. This is by design. Chunks are not parts. Don't use silly sized chunks. Your parts can be any valid size up to 5GB. Reasonable chunk sizes are 64KB, maybe...
I guess this is AWS being silly. This would need a significant rewrite of s3UnsignedChunkedReader to pass through reads and kill the buffer.
It seems when doing *unsigned* chunked uploads the SDK chooses huge chunks. Currently chunks are mapped to buffers and read in full before sending upstream. That of course means that...
Sorry. I didn't have time to go through the PR before now. It seems you are wastly overcomplicating things. There is no need to buffer encoders, since you are only...