David Michael Barr

Results 43 comments of David Michael Barr

[AWCY results at the default speed](https://beta.arewecompressedyet.com/?job=master-87c0297c631cf40bc1e9b03e8130928bd8e223b7&job=segment-opt%402021-03-17T09%3A28%3A36.965Z) indicate more work is needed from this initial version: | PSNR Y | PSNR Cb | PSNR Cr | CIEDE2000 | SSIM | MS-SSIM...

Update: Rebased on top of recent changes. AWCY results with number of [segments fixed at 3 and dynamic offsets on objective-fast-1 at default speed](https://beta.arewecompressedyet.com/?job=master-c0b1f562452c2ec8caf6396ff23e4612dfc7c0a1&job=segment-opt-3-seg%402022-08-19T18%3A31%3A42.751Z): | PSNR Y | PSNR Cb...

The goal is to retain a dynamic number of segments. The current testing is to measure the performance of an 'oracle' that picks the best number of segments for each...

AWCY results for [objective-1-fast at default speed](https://beta.arewecompressedyet.com/?job=master-1816a1c625b9279369d868328f6a2e801637ea43&job=segment-opt%402022-08-21T08%3A41%3A58.294Z) | PSNR Y | PSNR Cb | PSNR Cr | CIEDE2000 | SSIM | MS-SSIM | PSNR-HVS Y | PSNR-HVS Cb | PSNR-HVS...

The following dependencies need to be packaged before rav1e for Debian: ``` dpkg-checkbuilddeps: error: Unmet build dependencies: librust-arbitrary-0.2+default-dev librust-arg-enum-proc-macro-0.3+default-dev librust-arrayvec-0.5+default-dev librust-better-panic-0.2+default-dev librust-bitstream-io-0.8+default-dev librust-err-derive-0.2+default-dev (>= 0.2.1-~~) librust-interpolate-name-0.2+default-dev (>= 0.2.2-~~) librust-ivf-0.1+default-dev librust-nasm-rs-0.1+default-dev...

We are waiting on https://github.com/Homebrew/homebrew-core/pull/46975 so that we can include the dev artifacts in the formula. This will enable the bottle to be used for building against the C-ABI. **Update:**...

I was able to find desyncs by adapting the fuzzer to run 8-in-16. Here is one small such config: ```rust DecodeTestParameters { w: 16, h: 31, speed: 6, q: 0,...

FYI, I adapted https://github.com/Kagami/av1-bench to try to match encoding time and bit-rate for **still images**. Just one luma-only metric does not tell the whole story but this gives an indication...

This is still a major opportunity for improvement. With the current structure of luma-chroma RDO, there should be negligible loss in compression efficiency from moving to: best luma mode (luma-only...