Lynne: > 23 Nov 2021, 12:03 by toq...@outlook.com: > >> Lynne: >> >23 Nov 2021, 10:48 by jianhua...@intel.com: >> >>Lynne: >> >>>>> From: ffmpeg-devel <ffmpeg-devel-boun...@ffmpeg.org> On Behalf Of >>>>> Lynne >>>>> Sent: Tuesday, November 23, 2021 5:23 PM >>>>> To: FFmpeg development discussions and patches <ffmpeg- >>>>> de...@ffmpeg.org> >>>>> Subject: Re: [FFmpeg-devel] [PATCH 4/4] avutil/hwcontext_vulkan: fully >>>>> support customizable validation layers >>>>> >>>>> 23 Nov 2021, 10:01 by jianhua...@intel.com: >>>>> >>>>> > Validation layer is an indispensable part of developing on Vulkan. >>>>> > >>>>> > The following commands is on how to enable validation layers: >>>>> > >>>>> > ffmpeg -init_hw_device >>>>> > >>>>> vulkan=0,debug=1,validation_layers=VK_LAYER_KHRONOS_validation+VK_L >>>>> AYE >>>>> > R_LUNARG_api_dump >>>>> > >>>>> > Signed-off-by: Wu Jianhua <jianhua...@intel.com> >>>>> > --- >>>>> > libavutil/hwcontext_vulkan.c | 110 ++++++++++++++++++++++++++++--- >>>>> ---- >>>>> > libavutil/hwcontext_vulkan.h | 7 +++ >>>>> > 2 files changed, 97 insertions(+), 20 deletions(-) >>>>> > >>>>> > >>>>> > + /** >>>>> > + * Enabled validation layers. >>>>> > + * If no layers are enabled, set these fields to NULL, and 0 >>>>> > respectively. >>>>> > + */ >>>>> > + const char * const *enabled_validation_layers; >>>>> > + int nb_enabled_validation_layers; >>>>> > + >>>>> > >>>>> >>>>> Why are you exposing them? Do API users really need to know this? >>>>> >>>> >>>> It's okay. For it's only integrated in a really small separate function it >>>> could be skipped by the status debug_mode as before. And validation >>>> layers are embed by other specific drivers, platforms(such as those >>>> specific layers in androids) or SDK, the FFmpeg is not need to do anything >>>> more whatever the current is compiled with optimization mode or debug mode. >>>> The use who only want to use filter is simply not able to know how to >>>> enable the debug_mode. For me, as a user also, it is important to me, I >>>> don't want to changed the code then compiling to use the specific >>>> validation >>>> layers again and again. And not only the developer need to use it, those >>>> people who help us test could report a more detailed problem. I think it's >>>> really benefit. >>>> >>> >>> Sorry, I didn't quite understand that. >>> I'm not objecting to being able to use custom debug layers and activating >>> them if they're requested from the user. We already do that for the >>> standard debug layer anyway. I'm just not sure I understand why filters >>> would need to know if a debug layer is ran, and which one it is, by exposing >>> them via the public API. >>> >> >> Oh! My bad. I supposed you mean we may be not necessary to expose >> layer info to the Device Context, right? >> > > Yup, that's right. We'd like to keep the API as small as possible, and > it's already quite large compared to other hwcontexts. In this case, > I don't think it's necessary. API users already know, since they can request > layers, and filters shouldn't have to know. > > So just remove the public API part of the patch. > > As for the rest, could you always include VK_LAYER_KHRONOS_validation > if debug mode is on? This patch removes that. And free the array of > extensions once the instance has been initialized. >
Sure I do. I'll update this patch soon. Thanks, Jianhua _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-requ...@ffmpeg.org with subject "unsubscribe".
