Nicolas Silvagni

[Nicolas Silvagni @galop1n]

@Force1studios @Variety The old arnold running man movie is more akin to how AI today is a big problem with picture/video generation that can completely fabricate lie on demand and used as missinformation and propaganda!

[Nicolas Silvagni @galop1n]

@pATjako @OlexGameDev The best ideal way remains to draw the vector data directly by triangulating the interior, have encompassing triangles at the boundaries and solve the spline in the shader. Sadly I think there is some patents that prevents anyone to do so if I recall

[Nicolas Silvagni @galop1n]

@pickover Only if he is a bird...

[Nicolas Silvagni @galop1n]

@TheJimbothy @charlieINTEL The downfall to me was post BO cold war, when Treyarch had to drop their engine and all studios now had to work hand in hand on the same IW engine codebase.

[Nicolas Silvagni @galop1n]

@blokswrd @ThatNgonGuy Except you can tesselate analyticaly all you want regular quad ( center quad of a 9 quad patch ) but have to iterate recursively on irregular quads, tris and ngon. So much more cpu/gpu intensive and inneficient

[Nicolas Silvagni @galop1n]

@mazeincoding Weekly update

[Nicolas Silvagni @galop1n]

@DeveloperHarris It is only insane if you do not learn the physics that drove the power line design.

[Nicolas Silvagni @galop1n]

@zuhaitz_dev A test like that should also run on variable matrix sizes and plot. To confirm if we are O(NNN) in cost or if something come in play at certain thresholds. Plus what methodology on priming the data without accidentally having it load in cache before hand too.

[Nicolas Silvagni @galop1n]

@zuhaitz_dev This looks like it mostly test memory bandwidth given these 1000sq matrix does not fit caches plus basic repeated memory access to read+add+write that often won't be optimized. It could at least gen a cell value in a local tmp and write once.

[Nicolas Silvagni @galop1n]

@krastavichok @ccallac7 Yeah, all geo data available in an unified way is the way for visbuf and other advanced rendering technique that reduce draw counts. Just barrier things properly, and use the copy queue, you will be fine.

[Nicolas Silvagni @galop1n]

@ChShersh How many allocations and intermediate sub string copies during that code ?

[Nicolas Silvagni @galop1n]

@ThePrimeagen @thoughtlesslabs When your max value was 65535 this was not too hard to read without it

[Nicolas Silvagni @galop1n]

@bmcnett @SebAaltonen The reason was to avoid data thiefs. Imagine a world you wouldstore data in a brain rather than to use robust proven encryption on conventional storage

[Nicolas Silvagni @galop1n]

@chefsevenn Flavor, this is what is missing !

[Nicolas Silvagni @galop1n]

@_sanjeev_singh_ @Adarsh_Web3 @LewisCTech does the code need to ref any of these enum by name ? Can't you merge anything into a single variant with an inside sub state ? Is it code gen ? large enum in msvc/clang c++ are also surprisingly slow to compile fyi! the lsp should handle it yes, but was it really the best way?

[Nicolas Silvagni @galop1n]

@johnsontradez @_devJNS The 3 nested if can be collapse into a single one now, would be a bit better

[Nicolas Silvagni @galop1n]

@_devJNS 2 spaces indentation is the only sane settings ! There is all you need to know

[Nicolas Silvagni @galop1n]

@bmcnett We do not really have an alternative at the moment in most game renderer scenario sadly.

[Nicolas Silvagni @galop1n]

@JoeWilsonEQ1 @I_F32_I sure a couple thousands caps on screen won't hurt. but topology matter if you have millions polys to draw and do use raytracing. Artists having to go back and optimize geo is sure a thing they have to do

[Nicolas Silvagni @galop1n]

@JoeWilsonEQ1 @I_F32_I The problem is not the vertex, it is worsen pixel occupancy has more fragments (2x2 pixels) contains helper lanes. raytraced data structure also worsen with slant triangles.