KarpelesLab / gowebp / 24428524917 / 1
93%
main: 93%

DEFAULT BRANCH: main
Ran
Files 12
Run time 0s
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 95.577%. Remained the same
24428524917.1

push

github

MagicalTux 
benchmark: add public API Encode benchmarks + rewrite README

Adds BenchmarkEncodeGallery (gowebp lossless + 5 lossy tiers) and
BenchmarkEncodePNGBaseline against the 5 Google WebP Gallery photos
that serve as the library's realistic regression corpus. Benchmarks
skip cleanly when testdata/ isn't populated; populate with
NATIVEWEBP_FETCH=1 go test -run TestGalleryPSNR.

Rewrites README to reflect the current state of the library:

- New top-of-file description: pure Go, no libwebp/cgo, both lossless
  and lossy, with animation + alpha.
- Consolidated usage section: three short examples (lossless, lossy,
  animation) instead of scattered snippets.
- Method-tier table refined to focus on what changes between
  tiers rather than abstract "speed/size" labels.
- Benchmarks replaced with fresh measurements from the new
  BenchmarkEncode* suites on an i9-14900K. All five gallery photos
  now have head-to-head numbers for: PNG, gowebp lossless, and four
  lossy tiers (Q=50 M=4, Q=75 M=1/3/4, Q=90 M=4).
- Y-PSNR table for Q=75 M=3, showing the Y-PSNR floor assertions
  baked into TestGalleryPSNR.
- BT.601-range note preserved because downstream users decoding
  through x/image/webp hit this every time.

Measurements (reproducible; see README for commands):

  Lossless vs PNG (image/png BestCompression):
    13-23 % smaller files, 1.5-2.8× faster encode time.

  Lossy Q=75 Method=3 vs lossless:
    1.8×-3.3× smaller files, Y-PSNR 36.7-43.5 dB on natural photos.

  Lossy Q=50 Method=4:
    up to 7× smaller than PNG, still 32+ dB Y-PSNR on natural content.

2939 of 3075 relevant lines covered (95.58%)

112570.73 hits per line

Source Files on job 24428524917.1