Cam

I work in both CPU & SoC design at @AmpereComputing. In addition to computers, I enjoy baseball, music, and going up and down mountains

[Cam @RealCamBrown]

Don't try to make assumptions about what is vs. is not a microadjustment. Apply something like k-means per pitch and let the data speak. Cross-correlate with the location of the catcher's mitt to add a validation layer to the clustering (i.e. you should not count up&in when the catcher is set up low&out as a member of an up&in cluster). Spitballing here.

[Cam @RealCamBrown]

@tomdoyo @8_Maxx @MLBONFOX Use aggregate data to infer intent. Look for clusters for a given pitch (maybe a guy throws his curveball for a strike and also to try and elicit chases, in separate situations) and treat some radius around the center of the cluster as accurate execution.

[Cam @RealCamBrown]

@elliotarledge it's a fun post but thinking you're putting 24 LPDDR PHYs on a 144 mm^2 die is a pipe dream

[Cam @RealCamBrown]

@itsclivetime My knee jerk speculation is that because the heavy computational lift in AI/ML Python is all just precompiled C, you’re losing most of the ROI on a pretty considerable design undertaking. Have to make a coprocessor and intf too. Nontrivial integration task

[Cam @RealCamBrown]

@nikolasklimis My college baseball coach had us watch this 5x in a row before a practice

[Cam @RealCamBrown]

@Joe_Dagostinooo @TalkinBaseball_ it was a breaking pitch. it was probably in the zone at the front of the plate.

[Cam @RealCamBrown]

@mreynolds0504 @KutterIsKing that's what happens when it's spinning off the bat and there's high wind above the park

[Cam @RealCamBrown]

@Egon_Unique @MustBeMoose @CodifyBaseball there is a difference between unsuccessful batter challenges and unsuccessful catcher challenges

[Cam @RealCamBrown]

@YourAnonCentral That’s called 9:00 AM Moscow time buddy

[Cam @RealCamBrown]

@oxcrowx an* ls

[Cam @RealCamBrown]

@oxcrowx There are a lot of confounding variables in this. Was this run on a Mac? I.e. a large TAGE predictor that can take more than the execution of and ls to warm up. Notice how many more cycles the parser is. Interesting nonetheless.

[Cam @RealCamBrown]

@rohandevs @bubbleboi eh, I feel like 12 year CEO of half the EDA duopoly is a decent credential

[Cam @RealCamBrown]

@dioscuri I mean yeah if you want something built on the latest node. But that's just if you need the bleeding edge. Plenty of fabrication done on older nodes for things like consumer appliances. I don't need my smoke detector fabbed on N3P.

[Cam @RealCamBrown]

@schelldave @SteveHusseyMusi @barstoolsports what if CF wants to throw behind a lazy round of first base on a single to right-center where baserunner puts his head down and walks back to first? i'd routinely ask my first basemen to be ready for a throw behind on balls hit to right center and they'd usually just post up

[Cam @RealCamBrown]

@elusivecurrent Tiger at Breck. Solid choice.

[Cam @RealCamBrown]

@itsclivetime @skibidblackhole Our Altra (80 core) and Altra Max (128 cores) are based on ARM Neoverse, similar to Google Axion and NVIDIA Grace

[Cam @RealCamBrown]

@UmptyMaTwo @SwiftOnSecurity Physically closer and integrated into the pipeline as part of hardware prefetching as well!

[Cam @RealCamBrown]

@HowDoIPlayGT @liquid_buddha @bmix012 Work in CPU logic design.
Google “Tiny Tapeout GDS web viewer” if you want an interactive 3D model of an IC that you can play around with. Highly recommend @matthewvenn and his work on @tinytapeout if this interests you

[Cam @RealCamBrown]

@liquid_buddha @bmix012 No