Jim Drygiannakis

Work in progress

@sparsevoxel Happy to help! I've updated the gist to include the USBDescBuilder functions in case you don't have the missing descriptors at hand :)

@sparsevoxel This is used in an HID + CDC device. Enumerate's correctly on Windows and (afair) works :) The repo isn't public. If you need something else referenced in the file but not shown, please tell me and I'll update the gist.

@sparsevoxel I don't remember much but here is my code for a CDC device. Afair it was based on NXP's iMX SDK. Check usbSerialClassRequest() gist.github.com/jdryg/5af306c2…

DLS in now free (name your own price) on itch.io makingartstudios.itch.io/dls

Today I’m releasing the source code of my new experiment - multi-threaded, CPU-based vector graphics rasterizer - github.com/aurimasg/blaze

Triangle Rasterization - Part 2: More optimizations cohost.org/jdryg/post/625… TLDR: Managed to save ~0.4ms (1.34ms/frame -> 0.94ms/frame) using only SSE2 and ~0.5ms (0.82ms/frame) using SSSE3/SSE4.1

Trying to optimize a simple triangle rasterization function I found on the internet. cohost.org/jdryg/post/603…

@sparsevoxel I remember doing something similar for my RISCVEmu project a long time ago: #L162" target="_blank" rel="nofollow noopener">github.com/jdryg/RISCVEmu… My code is a bit more complicated than yours but it does basically the same things (load program headers into memory and jump to the entry point).

@sparsevoxel @ChristerEricson Unfortunately he uses Altium which isn't free but apparently makes things a lot easier for such designs. IIRC he demostrates how to do the same calcs for KiCad using external tools in older videos.

@sparsevoxel @ChristerEricson You might also be interested in this video youtube.com/watch?v=wQ37Nx… if you have 15mins more to spare, as well as other videos from Phil's channel.

@ChristerEricson @sparsevoxel AFAIK when dealing with digital signals, rise time is more important than the frequency (rate of change) of the signal. Regarding diff pairs, this might have some answers: youtube.com/watch?v=xwrdvh…

@sparsevoxel At what frequency do you have to worry about same trace lengths, and to what precision? Do you also have to worry about trace width/spacing/impedance?

@bkaradzic You are probably referring to MsgWaitForMultipleObjects() docs.microsoft.com/en-us/windows/…

@jdryg Hm, check there is similar API where you can specify wait timeout for messages instead peeking, that would block if there are no messages.

What's the proper way to handle rendering on minimized window? Render to a 1x1 framebuffer and swap/vsync as you normally would? Skip rendering completely and Sleep(16)? Something else?

@bkaradzic I thought all WaitXXX Win32 functions use the same timer as Sleep(), so is there really any benefit over just Sleep(1)? Only waitable timers can be high-res so they are a bit different in this regard.

@bkaradzic Didn't know about WaitForInputIdle(). I'll try it. Looking at the docs, it doesn't seem to do what I'm looking for, unless I'm missing something.

@jdryg You can switch your windows message pump from looping with PeekMessageW to WaitForInputIdle with timeout when app is minimized, and rendering is skipped.

@kinjalkishor That's exactly what I currently have. I guess it should be enough. If my update() func depends on the rate it's called, I should fix that instead of complicating things.

@jdryg github.com/kinjalkishor/s…

@bkaradzic Only really possible with a high resolution waitable timer because Sleep()'s resolution is fixed to ~15ms.

@bkaradzic But I just read the post I linked on Win32 timer resolutions, and I was wondering if it's better to measure the amount of time update() takes and only "sleep" for the rest of the 60Hz frame.

And a semi-related question. Are high resolution waitable timers a good choice for a custom frame limiter (e.g. no vsync, limit FPS to 120Hz)? Anyone tried it? randomascii.wordpress.com/2020/10/04/win…