Stream: brlcad

Topic: interactive ray tracing

Sean (Jan 25 2018 at 03:48):

For those interested, here's a sneak peak at a new interactive ray tracing interface I've been playing with: ort_havoc.png

Sean (Jan 25 2018 at 03:48):

@Daniel Rossberg you might find that interesting ;)

Sean (Jan 25 2018 at 03:51):

havoc is a considerably more impressive model when you take away the colors.. the attention to detail is fantastic

Sean (Jul 11 2026 at 16:27):

Finally figured out what was wrong with the IBRT build integration!

Was working well on reviewing and integrating our collab with TAMU last fall, then hit a snag when it was converted over to using BEXT for its dependencies. Visual artifacts galore and long debugging iterations over the past few weeks trying to figure out what was wrong.

Turns out it was an ISPC ABI mismatch... (!)

ISPC is the glue between OSPRay, Embree, and LIBRT. Now that's fixed, we're finally good to go, ready to get CI producing releases, and finally publish it!

For anyone thinking "what is IBRT??" it's the real-time interactive BRL-CAD ray tracing geometry visualization prototype I showed a few months back. Intent is for this to eventually become a modern visualization mode available to most of our graphical apps.

Sean (Jul 11 2026 at 17:44):

Sean (Jul 11 2026 at 20:25):

This is what ISPC ABI incompatibility looks like:

Sean (Jul 12 2026 at 21:37):

And now with wireframes working!

Sean (Jul 12 2026 at 21:41):

ibrt_demo.mp4

Waleed Ahmed (Jul 30 2026 at 13:00):

I wanted to ask about OpenCL GPGPU Raytracing effort, wanting to study it for a while. Also why don't we use a vulkan compute as an alternative for the ray tracing path since OpenCL is a bit old and slowly starting to lose the support.

Sean (Aug 03 2026 at 05:34):

@Waleed Ahmed The OpenCL code is in the checkout -- just need to pass a cmake flag to turn it on, then it's the -z option to rt to test it.

We could use vulcan compute, but that's a portability headache to get working everywhere. OpenCL low-level direct is no longer the only option for us, though, like it was some 10+ years ago when that effort started. Today, SYCL is the way to go for how we're using it. Would greatly simplify the code and a lot easier to migrate and get working everywhere.

Waleed Ahmed (Aug 03 2026 at 06:39):

Makes sense, didn't know SYCL was the current path. I'll get the OpenCL build running first, then dig into SYCL and follow up with what I find.

Last updated: Aug 17 2026 at 00:38 UTC