<_whitenotifier-5>
[scopehal-apps] azonenberg 702ec7d - Merge pull request #699 from d235j/fix-deep-swapchain Fix for GPUs that have a swapchain shallower or deeper than 2 images
<d1b2>
<johnsel> @ledude0001 you may be interested in joining the Thunderscope discord as well.
<d1b2>
<johnsel> it does seem a like a perfect fit for a proper rendering pass
<d1b2>
<azonenberg> yeah i think i'm going to try multithreading the current code first and use that to lay the groundwork for a full gpu version eventually
<d1b2>
<azonenberg> i think it's going to be similar to how we do waveform rendering
<d1b2>
<azonenberg> preprocessing pass on the CPU to binary search the offset into the data and clock that each GPU thread needs to start from
<d1b2>
<azonenberg> then the GPU will work through each range of the waveform in parallel
<d1b2>
<azonenberg> the tricky part will be the integration since the eye is likely going to be way too big to fit in shared memory
<d1b2>
<azonenberg> and atomics on global memory are slow
<d1b2>
<johnsel> I was more thinking frag/vert shaders
Kerr_ has quit [Ping timeout: 256 seconds]
Kerr_ has joined #scopehal
<d1b2>
<johnsel> I could also work on it in conjunction with the other eye pattern stuff we discussed before if you want. I haven't had time to get back to it as I got a very bad flu but I haven't forgotten about it
<_whitenotifier-5>
[scopehal] azonenberg 1b81696 - Finished initial implementation of 100baseT1 lnk training decode. Works fine for initial training but doesn't yet reset when the link drops.
<_whitenotifier-7>
[scopehal] azonenberg 9596e28 - Initial skeleton of 100BaseT1 link training. No where near finished.