Valentin Deleplace
Valentin Deleplace
That would save me precious minutes
Thanks Michael and Michael, I wasn't aware yet that (1) Darwin was special here and (2) inlining could have bad effects on performance, besides increasing the binary size.
@mknyszek I relaunched my original scripts on Linux amd64, and observed the same ~5% perf loss. My best guess is that the cause is not specific to Darwin.
@prattmic I can confirm your awesome findings on my Linux amd64: - a large part of the extra seconds are spent in [`flate.(*compressor).deflate`](https://cs.opensource.google/go/go/+/master:src/compress/flate/deflate.go;l=380;drc=40a0986959167e7b82ae16a35a5dd2a347609eff) - inlining is kicking in on `findMatch`,...
Thank you Michael
As everyone mentioned, making channels properly work over the network is complex. There was a real attempt to have "netchan" in the stdlib but afaik this didn't land : [video](https://www.youtube.com/watch?v=sln-gJaURzk&t=2010s),...
Yes. IMO it's fine to have code duplicated in the modules of each samples. Factoring would imply putting image+video in the same sample folder, which is fine too.
In this page we want to show a distinct snippet in the two section "Image" and "Video" https://cloud.google.com/vertex-ai/docs/generative-ai/multimodal/send-multimodal-prompts#gemini-send-multimodal-samples-go
> OK, fair enough, then please make it as identical as possible, except perhaps a comment and the link itself. I think there are some extra differences like whitespace right...
Merged from main, and fixed conflicts