Introduction to Computer Graphics. School of Computing, University of Utah. Full playlist: • Introduction to Comput... Course website: graphics.cs.ut...
I really appreciate the efforts put in the lecture series , this is full university level education for free, accessible to everyone . Your enthusiasm related to computer graphics , is really contagious ! Hope to learn more from you in upcoming lectures.
Wow, this whole series was very helpful while studying for my exams. I really appreciate the effort you put into this lecture. Greetings from Germany :)
Why don't we approach this problem in two stages, first rendering all the opaque geometry and only after that rendering transparent objects? I feel like it would address the sorting problem but there must be something I'm missing.
5 min read, TL;DR why not just use an FPGA for the HWRT?: I've gone through all these videos up to this point. Really great material, I've learned a lot. However there seems to be a lot of "hacks" going around in the background. Such as the perspective matrix, using the W=Z to get X/Z and Y/Z in the canonical space. At 1:02:41 you mention that you're in the HWRT group for designing GPU's that more or less will only be raytracing to render the scene. How come the HWRT group focus on making dedicated hardware just for this instead of using an FPGA? It doesn't even have to be fast, 50 MHz would definitely suffice. Then the VHDL/Verilog code you upload would become the shaders.. well... all the settings actually, and it would be extremely fast. My bets are that it would outperform a GPU easily since you can define your own multipliers and adders in your own custom pipeline. Because throughout the video series it looks like people more or less want to use an FPGA but they don't seem to know what an FPGA is and are forced to use the awkward GPU pipeline architecture. My reasoning for saying this is related to how you set the texture sampler (among many other things), it's very clunky. First you attach a texture to the texture unit and then the texture unit to a program. Overall the way you use the GPU is very awkward and what you're actually trying to do is VHDL/Verilog but in some GLSL instead. Had everything been on an FPGA (attached to the PCIE) there wouldn't have been any problems doing anything really, even the pipeline I assume you're trying to design with an absurd amount of raytracing. The price range I'm talking about would be what a typical GPU costs on the market today for your average Joe, around $300.
Unfortunately, FPGA is much slower than dedicated hardware and it is not nearly fast enough. It is not even a good solution for replacing the shader software on the GPU and it is certainly not an alternative to actual hardware. FPGA has its use cases, but real-time graphics does not seem to be one, at least not today.
@@cem_yuksel Huh interesting, and here I thought an FPGA was like a Swiss Army knife when it came to parallel problems. Thanks for the response. I really appreciate it.
Thank you Cem! Love the raw creative energy these videos have! 🙂
I really appreciate the efforts put in the lecture series , this is full university level education for free, accessible to everyone . Your enthusiasm related to computer graphics , is really contagious ! Hope to learn more from you in upcoming lectures.
Cem has mastered the art of explaining concepts so beautifully. Love this playlist, hope it never ends :)
The 'GoodBye' gesture combined with 'Ok I will see you then' at the end of each video is iconic! I kept learning because of that, lol.
Wow, this whole series was very helpful while studying for my exams. I really appreciate the effort you put into this lecture.
Greetings from Germany :)
Thank you so much for uploading new lessons. Very much appreciated.
A fan simulating a mild breeze through your hair would really bring this to life
Excellent explanation about Ray Tracing!
Why don't we approach this problem in two stages, first rendering all the opaque geometry and only after that rendering transparent objects? I feel like it would address the sorting problem but there must be something I'm missing.
That is exactly how transparent objects are handled with rasterization. This way, you just need to sort the transparent objects.
Super video! Thank you
Need books related Rasterization. without using OpenGL or other Api's
Read fundamentals of computer graphics by steve marschener it explains rasterization and ray tracing it is good read in my opinion.
5 min read, TL;DR why not just use an FPGA for the HWRT?:
I've gone through all these videos up to this point. Really great material, I've learned a lot. However there seems to be a lot of "hacks" going around in the background. Such as the perspective matrix, using the W=Z to get X/Z and Y/Z in the canonical space.
At 1:02:41 you mention that you're in the HWRT group for designing GPU's that more or less will only be raytracing to render the scene. How come the HWRT group focus on making dedicated hardware just for this instead of using an FPGA? It doesn't even have to be fast, 50 MHz would definitely suffice. Then the VHDL/Verilog code you upload would become the shaders.. well... all the settings actually, and it would be extremely fast. My bets are that it would outperform a GPU easily since you can define your own multipliers and adders in your own custom pipeline.
Because throughout the video series it looks like people more or less want to use an FPGA but they don't seem to know what an FPGA is and are forced to use the awkward GPU pipeline architecture. My reasoning for saying this is related to how you set the texture sampler (among many other things), it's very clunky. First you attach a texture to the texture unit and then the texture unit to a program. Overall the way you use the GPU is very awkward and what you're actually trying to do is VHDL/Verilog but in some GLSL instead. Had everything been on an FPGA (attached to the PCIE) there wouldn't have been any problems doing anything really, even the pipeline I assume you're trying to design with an absurd amount of raytracing. The price range I'm talking about would be what a typical GPU costs on the market today for your average Joe, around $300.
Unfortunately, FPGA is much slower than dedicated hardware and it is not nearly fast enough. It is not even a good solution for replacing the shader software on the GPU and it is certainly not an alternative to actual hardware. FPGA has its use cases, but real-time graphics does not seem to be one, at least not today.
@@cem_yuksel Huh interesting, and here I thought an FPGA was like a Swiss Army knife when it came to parallel problems.
Thanks for the response. I really appreciate it.
thanks for sharing the content!
Thank you so much 🙏🏻☺️
Fun lecture.
this is gold
why someone dislike
Thank you ..!!!!!!!
raster of puppets