Issues

[enc] An L1T4 config with only short-term references runs out of reconstruction buffers

rockchip-linux/mpp#1022 · opened

on Oct 24, 2026

[h265e] With temporal layers, each frame gets the next frame's temporal id

rockchip-linux/mpp#1021 · opened

on Oct 24, 2026

[h264e] Intra refresh does nothing on VEPU510: the refresh band is never programmed

rockchip-linux/mpp#1020 · opened

on Oct 24, 2026

[enc] VEPU510/511/540C keep the ROI of an earlier frame on frames without KEY_ROI_DATA

rockchip-linux/mpp#1019 · opened

on Oct 24, 2026

[vdpp] VEP refuses a 1920-wide frame whose pitch is 1984, the pitch the H.264 decoder uses

rockchip-linux/mpp#1018 · opened

on Oct 24, 2026

[vdpp] hwpq contexts share one frame counter, so a later context can skip its config

rockchip-linux/mpp#1016 · opened

on Oct 24, 2026

[iep2] IEP_CMD_RUN_SYNC returns MPP_OK when the frame is refused

rockchip-linux/mpp#1014 · opened

on Oct 24, 2026

[hal_av1d] Reference CDFs are lost when the AV1 frame size grows (spatial layers, resize)

rockchip-linux/mpp#1012 · opened

on Oct 23, 2026

[av1d] current_operating_point uses the operating point idc as an array index

rockchip-linux/mpp#1010 · opened

on Oct 23, 2026

[h265e] The VPS level ignores the frame rate: 1080p60 is coded as level 4, which is too low

rockchip-linux/mpp#1008 · opened

on Oct 23, 2026

[h264e] The SPS level ignores the frame rate: 1080p60 is coded as level 4.0, which is too low

rockchip-linux/mpp#1006 · opened

on Oct 23, 2026

[vp9d] Intra-only frames in profile 2 skip color_config, so the rest of the uncompressed header is misparsed

rockchip-linux/mpp#1004 · opened

on Oct 23, 2026

[h264d] First frame of an interlaced stream is reported as progressive when the picture timing SEI comes before the SPS

rockchip-linux/mpp#1002 · opened

on Oct 23, 2026

[jpegd] Requesting an output format the JPEG post-processor cannot produce makes the decoder write past the output buffer

rockchip-linux/mpp#1000 · opened

on Oct 23, 2026

[dma_heap] Cacheable DMA32 buffers come from the uncached heap when the board has no DMA32 heap

rockchip-linux/mpp#998 · opened

on Oct 23, 2026

[vp9d] 12-bit VP9 profile 2 is accepted and output as MPP_FMT_YUV420SP_10BIT since the parser rewrite

rockchip-linux/mpp#996 · opened

on Oct 23, 2026

[h265d] 10-bit 4:0:0 HEVC frames are labelled 8-bit MPP_FMT_YUV400 with a 10-bit stride

rockchip-linux/mpp#994 · opened

on Oct 23, 2026

[av1d] 8-bit 4:2:2 AV1 is accepted, labelled planar MPP_FMT_YUV422P, and decodes with reference errors

rockchip-linux/mpp#992 · opened

on Oct 23, 2026

[h265d] HDR10+ / HDR Vivid dynamic metadata keeps emulation prevention bytes and loses its tail

rockchip-linux/mpp#990 · opened

on Oct 23, 2026

[mpp_enc] One invalid colour value in MPP_ENC_SET_CFG discards the valid colour range, and reserved primaries reach the VUI

rockchip-linux/mpp#988 · opened

on Oct 23, 2026

[h264e] YUV400 input with Baseline or Main profile produces an H.264 stream that decodes with errors

rockchip-linux/mpp#986 · opened

on Oct 23, 2026

[hal_jpege] VPU720 JPEG encoder writes a corrupt JPEG for RGB input instead of failing

rockchip-linux/mpp#984 · opened

on Oct 23, 2026

[jpegd] Grayscale JPEG is returned tagged as NV12 when NV12 output is requested

rockchip-linux/mpp#982 · opened

on Oct 23, 2026

[jpegd] Decoded JPEG frames carry no colour range, so full-range JFIF is treated as limited range

rockchip-linux/mpp#980 · opened

on Oct 23, 2026

[mpp_dec] JPEG decode into a too-small caller buffer still runs the hardware and causes an IOMMU write fault

rockchip-linux/mpp#978 · opened

on Oct 23, 2026

Navigation Menu

Repository navigation

Issues

[enc] An L1T4 config with only short-term references runs out of reconstruction buffers

[h265e] With temporal layers, each frame gets the next frame's temporal id

[h264e] Intra refresh does nothing on VEPU510: the refresh band is never programmed

[enc] VEPU510/511/540C keep the ROI of an earlier frame on frames without KEY_ROI_DATA

[vdpp] VEP refuses a 1920-wide frame whose pitch is 1984, the pitch the H.264 decoder uses

[vdpp] hwpq contexts share one frame counter, so a later context can skip its config

[iep2] IEP_CMD_RUN_SYNC returns MPP_OK when the frame is refused

[hal_av1d] Reference CDFs are lost when the AV1 frame size grows (spatial layers, resize)

[av1d] current_operating_point uses the operating point idc as an array index

[h265e] The VPS level ignores the frame rate: 1080p60 is coded as level 4, which is too low

[h264e] The SPS level ignores the frame rate: 1080p60 is coded as level 4.0, which is too low

[vp9d] Intra-only frames in profile 2 skip color_config, so the rest of the uncompressed header is misparsed

[h264d] First frame of an interlaced stream is reported as progressive when the picture timing SEI comes before the SPS

[jpegd] Requesting an output format the JPEG post-processor cannot produce makes the decoder write past the output buffer

[dma_heap] Cacheable DMA32 buffers come from the uncached heap when the board has no DMA32 heap

[vp9d] 12-bit VP9 profile 2 is accepted and output as MPP_FMT_YUV420SP_10BIT since the parser rewrite

[h265d] 10-bit 4:0:0 HEVC frames are labelled 8-bit MPP_FMT_YUV400 with a 10-bit stride

[av1d] 8-bit 4:2:2 AV1 is accepted, labelled planar MPP_FMT_YUV422P, and decodes with reference errors

[h265d] HDR10+ / HDR Vivid dynamic metadata keeps emulation prevention bytes and loses its tail

[mpp_enc] One invalid colour value in MPP_ENC_SET_CFG discards the valid colour range, and reserved primaries reach the VUI

[h264e] YUV400 input with Baseline or Main profile produces an H.264 stream that decodes with errors

[hal_jpege] VPU720 JPEG encoder writes a corrupt JPEG for RGB input instead of failing

[jpegd] Grayscale JPEG is returned tagged as NV12 when NV12 output is requested

[jpegd] Decoded JPEG frames carry no colour range, so full-range JFIF is treated as limited range

[mpp_dec] JPEG decode into a too-small caller buffer still runs the hardware and causes an IOMMU write fault