Repository Issue Activity (beta)

#1587 Unbounded memory growth in libvmaf 3.2.0 when the decoder outpaces feature extraction (regression from 3.1.0)

#1109 Wrong PSNR and VMAF values in FFmpeg 5.1.1

#818 libvmaf API does not support `perc*` pool methods

#1242 Dictionary ownership model seems error prone

#743 Garbled Progress Spinner on Windows

#766 escaping of special characters in --model/--feature cli params

#930 Possible unwanted float to double promotion

#1455 Libvmaf: missing support for SHSTK (Intel CET)

#1555 x86 AVX512 build enables VBMI, but runtime dispatch appears to gate only on AVX512 (can cause illegal instruction on non-VBMI AVX512 CPUs)

#1580 mirror()-style boundary reflection reads out-of-bounds for width/height < 3 (motion feature, all backends)

#1576 Consider adding a minimal CUDA compile check to CI

#1541 Build fails due to race condition: vcs_version.h: No such file or directory during parallel ninja build

#1301 libvmaf with CUDA -- how to build for Windows

#1356 Models are the same as those 4 years ago?

#1415 Fails to build when CUDA is enabled

#1098 ffmpeg libvmaf filter: Segmentation fault with very small dimensions

#1547 `Speed_chroma_feature_speed_chroma_uv_score` error with `libvmaf` and `ffmpeg` from homebrew, with v1 models

#1394 How does VMAF decides how many threads to use

#1549 v1 CUDA extractors

#1568 Windows: FFmpeg `libvmaf` filter creates a mojibake filename for non-ASCII `log_path`

#1567 Feature request: CUDA feature extractor for CAMBI

#1566 CUDA 16bpc motion kernel applies byte stride as uint16_t element stride — motion2 ~-58%, VMAF ~1.5 points low for all >8-bit input

#1564 integer ADM: CUDA cm diverges from CPU (border mirror + rounding order) and x86 SIMD dwt corrupts the mirrored last output column — fixes attached

#1562 cuda: motion score diverges from CPU — edge mirror off-by-one in calculate_motion_score_kernel_{8,16}bpc

#1561 Which (V1) Model ?