perfsea

Re: on Sunday October 13, 2024 @09:45AM (#41827985)

After years of repeatedly struggling with vtune installs in cloud environments just to get a few specific hw metrics only to fall back on manually building some perf stat events from Perfmon, I finally just built the functionality I always wanted. It automatically generates the set of hardware counters required for your requested metrics, formats it, and then parses the output to evaluate the metric formulas on the raw events

Re:Flamegraphs, Now What? on Thursday August 01, 2024 @08:41AM (#41128546)

Despite flamegraphs being very intuitive I still have people ask me, what do I look at? Instead of explaining my usual cmd+f's over an over I made a tool which can annotate flamegraphs.

Another downside of flamegraphs is you cannot easily aggregate across multiple stacks without multiple cmd+f's.

The backend is a git repo anyone can add their cmd+f's and optimizations they look for. https://github.com/perfsea/optimizations/blob/main/data/flam...

Flamegraphs, Now What?

Posted by perfsea on Thursday August 01, 2024 @08:41AM
from the perfsea.com dept.

https://perfsea.com/flamegraph

( Read More... | 1 comment | on HN )

Re:Ask HN: What is your go to performance optimization? on Thursday July 25, 2024 @11:33AM (#41069865)

better than transparent?

Re:Ask HN: What is your go to performance optimization? on Thursday July 25, 2024 @10:56AM (#41069447)

I'm conflicted about this one. march native definitely. Have you seen substantial gains from omitting frame pointers?

Re:Ask HN: What is your go to performance optimization? on Thursday July 25, 2024 @10:54AM (#41069424)

lol never knew about this thanks. Has this actually slowed down someone's code?

Re:Ask HN: What is your go to performance optimization? on Thursday July 25, 2024 @09:04AM (#41068204)

also less logging. Seen this pop up way too many times

Ask HN: What is your go to performance optimization?

Posted by perfsea on Thursday July 25, 2024 @08:58AM
from the ask-hn dept.

I'm curious what backpocket optimizations you have from your past experiences which you pull out when a workload costs too much in the cloud. For me its.* disable lower c-states* enable adaptive coalesce in the network driver* mount volumes with noatime

( Read More... | 46 comments | 17 points | on HN )

Re:Show HN: Letterform – Minimalist publishing platform based on Markdown on Wednesday July 24, 2024 @03:26PM (#41060871)

I like the idea. Small nit: the scrolling is a little unintuitive.

Show HN: Perf Sea – performance engineering for everyone

Posted by perfsea on Friday July 12, 2024 @08:39AM
from the show-hn dept.

https://perfsea.com/

I have found that many companies don’t have time for performance engineering and leave random tunables incorrectly configured for their workload. Instead of solving the same problems over and over I wanted an open source place where people could collaboratively share their best practices, knowledge, and methodologies, because in performance engineering no one is really a competitor.It is under...

( Read More... | no comments | 2 points | on HN )

Re:Qualcomm's Oryon Core: A Long Time in the Making on Thursday July 11, 2024 @01:43PM (#40938921)

I wonder how effectively it can utilize all of its execution units for common workloads. Frontend boundedness is often a big issue especially with jit