Heinz N. Gies
Heinz N. Gies
super quick first hacky test: ``` Running benches/read_rustdoc.rs (target/release/deps/read_rustdoc-cecf84416ef1155c) Benchmarking aws_sdk_ec2: Warming up for 3.0000 s Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target...
> Am I reading the data correctly, that the EC2 crate parses half a second slower with SIMD than without? 👀 Yes, the context here is important. Those tests are...
Here the results from a Digital Oceans (2 core, 4GB memory) server [results-server.txt](https://github.com/user-attachments/files/16851675/results-server.txt)
ja copying the 500MB file to the server takes time 😂 , it's a WIP
but FWIW results from my local machine are already in for the aws crate ``` aws_sdk_ec2/SIMD/parse time: [1.3610 s 1.3683 s 1.3761 s] Found 3 outliers among 100 measurements (3.00%)...
> I'm a bit surprised the SIMD benefit is only about ~12%. Is there something we might be doing that is pessimizing for SIMD? Could we adapt our approach in...
If there is a chance, it's worth doing the research :) if @dmatos2012 we can go through that experiment together.
Just thoughts out loud before I forget what I was thinking :D Could we get rid of batched as a concept altogether and move this to userland? This sure would...
Manuall is totally fine, the problem I ran into is that in the rust-side of things I couldn't express "this is a pointer to unknown" as any approach I tried...
@dependabot rebase