Martin Dimitrov

Results 7 comments of Martin Dimitrov

Splitting the buffer seems to keep performance the same: Performance was tested by: # 1) populate the time-series database with data: get the tool ``` wget -c https://s3.amazonaws.com/benchmarks.redislabs/redistimeseries/tools/tsbs/tsbs_load_redistimeseries_linux_amd64 chmod 755...

> Hi @slice4e and thank you for this. However, you should test this first, at least, before submitting the PR :) Can you try it? Sure - I will.

Hi - I will look into it...

@lrg11 - I believe that this will typically occur is the data you are accessing with a vecrorized load is unaligned. Could you try forcing a 64-byte aligned allocation of...

@lrg11 - after some thought, I belive it will be prudent to first understand what fraction of the execution time we are spending in the "vectorizable" code. This limits the...

@vaclavskala So far any regressions that i observed with kernel 6 were due to the benchmarks running at a lower CPU frequency. Still investigating the root cause... There were use...

@sundb thank you. Submitted related PR for the binary quantization path as well: https://github.com/redis/redis/pull/14492