Robert Muir

also, would be good to compare apples-to-apples here. currently from what i see, benchmark compares `dot8s(MemorySegment..)` vs `BinaryDotProduct(byte[])`. To me this mixes up concerns about memorysegment vs java heap with...

That is the fault of your graviton setup though, you can see it right in your previous `platforms` logic where it is invoking `cCompiler.executable 'gcc10-cc'`. Fix GCC to be in...

sorry @goankur i'm super-behind, vacations and end of summer. i definitely want to try out the VNNI path but just haven't found the cycles yet.

@mikemccand to start with "text mode", just do this: start of benchy: ```console $ nohup vmstat --active --wide --timestamp --unit M 1 > vmstat.log 2>/dev/null & ``` end of benchy:...

also if you don't want output in your local time, set TZ variable: ``` $ TZ=UTC vmstat ... ``` if you don't want `inactive/active` remove the `--active` flag. I prefer...

thanks @mikemccand! can you hook me up with these sample data files? it will be faster to make it look decent.

thank you @mikemccand, that works. crazy how idle your CPU is while doing this.

@mikemccand see https://rmuir.org/vmstat-report/ You can grab the 3 files there to reproduce it: * `vmstat.gpi`: gnuplot file to make several .svg files * `index.html.template`: simple html file to lay them...

This makes sense. These are just a start, we can tweak/beautify later once things are working. I struggled a bit because SVG files were enormous out of box: the program...

> We are not waiting for merges, so I suspect this is just Lucene moving the (big) in-memory segments to disk ... I'll see if I can better tune these...