Radzivon Bartoshyk
Radzivon Bartoshyk
We implemented with @Shnatsel more or less reasonable solution for alpha premultiplication in #2223. However, I don't think there is any agreement on its integration into the existing pipelines.
SIMD will be much better here on x86 since Rust make a lot efforts on f32 -> u8, with NEON conversion is done in one instruction and there will be...
Who wants to check: https://godbolt.org/z/YYGqjzasf https://developer.arm.com/architectures/instruction-sets/intrinsics/vcvtaq_u32_f32
It's close to not reducible further with current android packaging system. Approximate size of libraries: AVIF encoder 5mb, AVIF decoder 1mb, any of HEIF requires libheif which is 1mb, HEIF...
I tried actually some time ago ( about a year ago ) SVT instead AOM and kvazaar instead of x265. SVT is using threading model not compatible with the JVM...
As I noted here https://github.com/awxkee/avif-coder/issues/35#issuecomment-2839518633: decoding and encoding using DataSpace/ColorSpace is not implemented. Any image is tonemapped to sRGB. I created the library before HDR displays became common ( and...
No, this is actually a really bad workaround. It will eliminate linking issues but introduce OOM and memory management problems on the JVM (or KVM) side, because it requires to...
From my perspective, it seems that this is more often done by creating two separate APIs: one that accepts a slice, and another that accepts a dynamic stream (```Box```). When...
Calls on dynamic method usually costs ( with context saving & restoring ) 15 - 30 CPU cycles on some the worst cases. If method needs to be loaded from...
How about do the following: 1. Since there seems to be agreement that nobody likes the current BufReader, we can copy its implementation and change the defaults to ones we...