Sepehr Taghdisian Profile
Sepehr Taghdisian

@septagh

Followers
546
Following
7K
Media
140
Statuses
775

Programmer SandboxInteractive (Ex. @UbisoftBerlin)

Joined May 2009
Don't wanna be here? Send us removal request.
@septagh
Sepehr Taghdisian
1 year
Staying on a healthy diet for compile times :)
0
0
5
@septagh
Sepehr Taghdisian
1 year
Nice! With the new improved Vulkan backend, separate transfer queue and batch upload submission, the performance of this drastically improved. Now for 1.4GB of unbaked asset data, it takes 1.5 secs. For baked assets, it takes 270ms (5.2 GB/s). Very close to my SSD drive speed.
@septagh
Sepehr Taghdisian
1 year
Asset manager test. First working implementation. It loads stuff in chunks. so here, each chunk is around 60MB of data (texture+model). It's pretty much a snap (~30-40ms for each chunk). In total, 1.4 GB of asset data loads in 700 ms. baking all data takes 2.2 secs (512 assets).
0
0
9
@septagh
Sepehr Taghdisian
1 year
But one major drawback of having such control and power is that it's so easy to over-engineer, over-abstract and come up with generalized abstraction API designs that could make your backend perform worse than D3d11 drivers. (5/5)
0
0
2
@septagh
Sepehr Taghdisian
1 year
It's obviously not perfect at all, although it took me about 4~5kloc to do basic imgui rendering, but the control freak in me is satisfied with how much control you have over memory, queues and multi-threaded rendering. (4/5)
1
0
1
@septagh
Sepehr Taghdisian
1 year
The latter confused me a lot on how frame and swapchain syncing and syncing in general. The former starts with 1.3 and is better organised for beginners. (3/5)
1
0
1
@septagh
Sepehr Taghdisian
1 year
My first try was about 2 yrs ago, but then recently I had to revisit all the stuff again and rewrite my toy rendering backend to learn it more properly. I think newcomers, better start with https://t.co/cNfGOA7wpO tuts instead of https://t.co/hjhVhWOHHc (2/5)
vkguide.dev
Guide to the Vulkan API
1
0
3
@septagh
Sepehr Taghdisian
1 year
After 6-7 yrs of wanting to learn a modern graphics API, I finally managed to put enough time into learning Vulkan. I think I finally got comfortable with it so that validation errors don't freak me out and I usually have a good idea what's going on. (1/5)
2
0
7
@PhysInHistory
Physics In History
1 year
There are times when I feel like I'm in a big forest and don't know where I'm going. But then somehow I come to the top of a hill and can see everything more clearly. When that happens, it's really exciting. - Maryam Mirzakhani
22
128
1K
@septagh
Sepehr Taghdisian
1 year
Graphic objects creation code is still messy. no specific optimizations, fancy IO (DirectStoragee/etc) or compression algorithms is yet used.
0
0
1
@septagh
Sepehr Taghdisian
1 year
Asset manager test. First working implementation. It loads stuff in chunks. so here, each chunk is around 60MB of data (texture+model). It's pretty much a snap (~30-40ms for each chunk). In total, 1.4 GB of asset data loads in 700 ms. baking all data takes 2.2 secs (512 assets).
2
0
11
@septagh
Sepehr Taghdisian
1 year
0
0
0
@septagh
Sepehr Taghdisian
1 year
I rewrote the asset system in junkyard. The system now loads groups of assets instead of single loads. Much cleaner data/code, faster and vastly simplified new asset type impl. Supports all prev feats, multi-threaded, hot-reload and remote loads. Time to test/dbg and a blog post
0
0
11
@Dachsjaeger
Alexander Battaglia
1 year
Star Wars Outlaws on PC is one of the best looking games this gen with a PC version that is not artificially held back by consoles. What does RTXDI do? How is perf? What about the hidden max settings or DLSS Ray Reconstruction? I cover that and more! https://t.co/WFkvn2pe17
88
106
1K
@rfleury
Ryan Fleury
1 year
You can build a custom standard library on top of C that’s much better than both libc, and the STL/C++ runtime, as I’ve done many times. The C++ standards are run by overly-bureaucratic, overly-abstract, and non-auteur-driven organizations; they have no chance of success.
@seanbax
Sean Baxter
1 year
Something evident with Rust is that its library was developed as a co-design process with the language. C++ stdlib on the other hand is overflowing with workarounds. It creates a big technical debt right out of the gates. It's not a language problem, it's a process problem.
10
14
312
@_plop_
Jeremy Laumon
2 years
Here it is! Let me present you Asset Cooker. It's a build system aimed at game assets, with a UI. It is using the USN journals from Windows to robustly keep track of file changes, and only cook what needs to be cooked. You can leave it running in the background. It is fast.
@_plop_
Jeremy Laumon
2 years
In the next few days, I'm going to open source my hobby project of the last few months. It's a build system aimed at game assets (and with a UI). It will be potentially useful to anyone making/using a custom engine and hopefully a great upgrade for many small engines/studios.
12
75
309
@IntNostalgia
Internet Nostalgia
2 years
Do you remember Visual Basic? 🥹
373
280
3K
@tsoding
Тsфdiиg
2 years
A lot of memory management models assume that there are many objects and many lifetimes (a lifetime per each object). The reality is that there are many object and very few lifetimes (usually like 3-5). So a lot of languages are solving pretty much non-existing problems.
33
50
770
@akoshlo
Oleksandr Koshlo
2 years
My GDC talk “Ray Tracing in Snowdrop: Scene Representation and Custom BVH” is available on Vault https://t.co/NSlFyKtaTI. My DMs are open so feel free to ask any questions.
0
12
52
@septagh
Sepehr Taghdisian
2 years
Looks like some sort of driver thread pool ? the count is actually 1 more than my logical processor count. I kind of had the impression that with vulkan, drivers wouldn't need to do much extra work in background
0
0
1
@septagh
Sepehr Taghdisian
2 years
Does someone have any idea why as soon as vkQueuePresentKHR gets called for the first time, we get a zillion new threads by the nvidia driver ? does this happen on other drivers as well ? (not using any validation layers) #vulkan #windows #NVIDIA
2
0
3