munrocket

@msfstef nice solution, thanks! It works on iPhone 6. I am used 2 drawcalls before this too.

Am I right that QFMA care only about performance and become not correctly rounded according to true FMA operation in IEEE 757-2008? There are another way to polyfill FMA with...

With this Quasi FMA proposal you are going to bake new specification that not grantee the same results on different machines and will fail CI tests not only in spec...

Ok, even if we will have QFMA with feature detect https://github.com/WebAssembly/simd/issues/80, we still can create all of the functions. Because ```javascript FMAC(a, b, c) = isFmaAvailable() ? QFMA(a, b, c)...

if `-5.669010335538281e+35 / 2.869936417201091e+38` become zero it is not a problem with atan? But a problem with division? We can try to use Newton's iteration to solve one of the...

`vec4` - for pianists. f32x4 like in wasm or classic vec4 looks better for such basic word. >Consistency, it's the same syntax for vector, matrix, array, etc Then you should...

Agree with @kvark here. Looks like `@vertex` is too visually small for such important thing. `@stage(vertex)` much easier to understand and it's only one line of code in whole shader....

Specify function domains will be practical. For example `pow` function is very popular. But it is undefined if x < 0 or if (x = 0 and y ≤ 0)...

Progressive loading can be discussed with firefox and safari team. Chrome already have a prototype https://github.com/gpuweb/gpuweb/issues/766

@Kangz in chromium: https://chromium-review.googlesource.com/c/chromium/src/+/2166131 in image-decoder-api: https://github.com/dalecurtis/image-decoder-api/issues/3