Felix Handte

Software Engineer working on compression @facebook.
Cinematographer at

[Meta Open Source @MetaOpenSource]

🚀 Introducing OpenZL, a new #opensource, format-aware compression framework!

It uses a unique “graph model” for compression, enabling modular, high-performance codecs and a universal decoder - perfect for structured data at scale

Link: https://t.co/XynPVXXw3I

#DataCompression

[Yann Collet @Cyan4973]

We are open-sourcing OpenZL, a new data compression library and training tools to generator specialized compressors for structured data, achieving performance levels inaccessible to classic generic algorithms:
https://t.co/cJJVpqGBK6

[Terror Films @Terror_Films]

Can't cram in a movie on your busy schedule??

Settle in for an unsettling 45 minutes with CRAM! 😈

#CRAM is available today on @Tubi and more!

TUBI: https://t.co/YMIPm6zfNi

[Felix Handte @FelixHandte]

@absidell @coopercooperco E.g., the fix for this would be to slide the bottom left control point of the green channel curve to the right a bit.

[Felix Handte @FelixHandte]

@absidell @coopercooperco In response, my first adjustment in Lightroom is now always to adjust the channel scales myself to fix these mismatch errors, by manipulating the endpoints of the channel curves.

[Abie Sidell @absidell]

Wow if I had grades like this in school CRAM wouldn't exist

[Abie Sidell @absidell]

With 10 DAYS left to study like your life depends on it before CRAM drops on-demand & streaming, I’m excited to share more rad behind-the-scenes highlights🤘

And there’s nobody radder than one of the original Rad Rhinos himself: @felixhandte 🦏 — CRAM’s cinematographer! 🧵

[Felix Handte @FelixHandte]

@dhueuskndskjsm @danluu There were many hard parts.

[Dan Luu @danluu]

I wonder how much waste has been eliminated by Yann Collect creating zstd.

When I ran the numbers at Twitter, which is tiny compared to the huge tech companies, switching to HDFS to zstd was ~ mid 8 figs/yr.

Across the world (not annualized), it seems like must be >= 9 figs?

[Felix Handte @FelixHandte]

@mattroll_ Don't worry, C's got you covered: statically allocate all your memory.

[Felix Handte @FelixHandte]

@mattroll_ There is no god but the one true god, whose name is ISO C.

[Felix Handte @FelixHandte]

@steveyedlin Man, I’m sorry. I’m really trying to resist going further down the https://t.co/IxQn2ryrq4 rabbit-hole.

But I think it comes down to this: I’m talking about cd/m^2 and you’re talking about lm/m^2.

Energy (lm) = intensity (cd) * area (aperture).

[Felix Handte @FelixHandte]

@steveyedlin At any rate, thanks for chatting! I've really enjoyed your work, both on screen and #NerdyFilmTechStuff.

[Felix Handte @FelixHandte]

@steveyedlin (Obviously the total energy incident on the film plane changes, but that's a different thing...)

[Felix Handte @FelixHandte]

@steveyedlin Is it that intensity (cd/m^2) matches between the object and the image? Or that total flux (lm/m^2) matches between the object and the image? Or something else?

[Felix Handte @FelixHandte]

@steveyedlin Fair. I tried! But it's not clear to me what the fundamental conclusion you're reaching is. Apparently, the statement I thought you were arguing for is one we both don't agree with. So what is true of shooting at t/1.0 that you're saying isn't true of other apertures?