On 4/17/2025 10:11 PM, Lynne wrote:
> On 18/04/2025 01:55, IndecisiveTurtle wrote:
>> From: IndecisiveTurtle <geoster3d@gmail.com>
>>
>> Performance wise, encoding a 1080p 1-minute video is performed in 
>> about 2.5 minutes with the cpu encoder running on my Ryzen 5 4600H, 
>> while it takes about 30 seconds on my NVIDIA GTX 1650
>>
>> Haar shader has a subgroup optimized variant that applies when 
>> configured wavelet depth allows it
>> ---
>>   configure                                    |   1 +
>>   libavcodec/Makefile                          |   3 +
>>   libavcodec/allcodecs.c                       |   1 +
>>   libavcodec/vc2enc_vulkan.c                   | 959 +++++++++++++++++++
>>   libavcodec/vulkan/vc2_dwt_haar.comp          |  82 ++
>>   libavcodec/vulkan/vc2_dwt_haar_subgroup.comp |  75 ++
>>   libavcodec/vulkan/vc2_dwt_hor_legall.comp    |  82 ++
>>   libavcodec/vulkan/vc2_dwt_upload.comp        |  96 ++
>>   libavcodec/vulkan/vc2_dwt_ver_legall.comp    |  78 ++
>>   libavcodec/vulkan/vc2_encode.comp            | 169 ++++
>>   libavcodec/vulkan/vc2_slice_sizes.comp       | 170 ++++
>>   11 files changed, 1716 insertions(+)
>>   create mode 100644 libavcodec/vc2enc_vulkan.c
>>   create mode 100644 libavcodec/vulkan/vc2_dwt_haar.comp
>>   create mode 100644 libavcodec/vulkan/vc2_dwt_haar_subgroup.comp
>>   create mode 100644 libavcodec/vulkan/vc2_dwt_hor_legall.comp
>>   create mode 100644 libavcodec/vulkan/vc2_dwt_upload.comp
>>   create mode 100644 libavcodec/vulkan/vc2_dwt_ver_legall.comp
>>   create mode 100644 libavcodec/vulkan/vc2_encode.comp
>>   create mode 100644 libavcodec/vulkan/vc2_slice_sizes.comp
> 
> LGTM.
> Planning to push this tomorrow unless there are objections.

This set was sent today, so please wait a couple days before pushing it.