In this patch series, clock scaling and core selection methods are
updated. Current clock scaling and core selection methods are same
for vpu4 and previous versions. Introducing load calculations using
vpp cycles, which indicates the cycles required by video hardware to
process each macroblock. Also adding vsp cycles, cycles require by
stream processor. Clock scaling is now done more precisely using vpp
and vsp cycles. Instance is assigned to core with minimum load, instead
of static assignment.
Changes since v5:
- Corrected load_per_core calculations.
Changes since v4:
- Added call to load_scale_clocks from venus_helper_vb2_buf_queue.
- Modified check to match core_id in core_selection.
Changes since v3:
- vsp_cycles and vpp_cyles are now unsigned long.
- Core number counting aligned with VIDC_CORE_ID_.
- Aligned hardware overload handling of scale_clocks_v4 with scale_clocks.
- Added bitrate based clock scaling patch in this patch series.
- Instance state check is now moved from scale_clocks to load_scale_clocks.
Aniket Masule (4):
media: venus: Add codec data table
media: venus: Update clock scaling
media: venus: Update to bitrate based clock scaling
media: venus: Update core selection
drivers/media/platform/qcom/venus/core.c | 13 ++
drivers/media/platform/qcom/venus/core.h | 16 ++
drivers/media/platform/qcom/venus/helpers.c | 215 +++++++++++++++++++++++--
drivers/media/platform/qcom/venus/helpers.h | 3 +-
drivers/media/platform/qcom/venus/hfi_helper.h | 1 +
drivers/media/platform/qcom/venus/hfi_parser.h | 5 +
drivers/media/platform/qcom/venus/vdec.c | 6 +-
drivers/media/platform/qcom/venus/venc.c | 6 +-
8 files changed, 252 insertions(+), 13 deletions(-)
--
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project