Aug 3, 2023, 22:39 by [email protected]:
> Michael Niedermayer:
>
>> On Thu, Aug 03, 2023 at 06:26:16PM +0200, Lynne wrote:
>>
>>> These are in-place transforms, required for DCT-I and DST-I.
>>>
>>> Templated as the mod2 variant requires minor modifications, and is
>>> required specifically for DCT-I/DST-I.
>>>
>>> Quite optimized, as there's no need for any additional buffer storage.
>>>
>>> doc/APIchanges | 3
>>> libavutil/tx.c | 18 ++++
>>> libavutil/tx.h | 10 ++
>>> libavutil/tx_template.c | 175
>>> +++++++++++++++++++++++++++++++++++++-----------
>>> libavutil/version.h | 2
>>> 5 files changed, 167 insertions(+), 41 deletions(-)
>>> 6e6308365cd78a84c7db5800207e6b5977945079
>>> 0001-lavu-tx-add-real-to-real-and-real-to-imaginary-RDFT-.patch
>>> From 2ea5e2541c2551bf1b56e967d35946289a85aa49 Mon Sep 17 00:00:00 2001
>>> From: Lynne <[email protected]>
>>> Date: Thu, 3 Aug 2023 18:21:23 +0200
>>> Subject: [PATCH 1/2] lavu/tx: add real to real and real to imaginary RDFT
>>> transforms
>>>
>>> These are in-place transforms, required for DCT-I and DST-I.
>>>
>>> Templated as the mod2 variant requires minor modifications, and is
>>> required specifically for DCT-I/DST-I.
>>>
>>
>> not sure if i forgot some patch but
>> this seems to break fate-binkaudio-dct
>>
>
> It's not just you; patchwork reported four failing tests:
> make: *** [fate-filter-firequalizer] Error 218
> make: *** [fate-binkaudio-rdft] Error 218
> make: *** [fate-binkaudio-dct] Error 218
> make: *** [fate-qdm2] Error 218
>
> - Andreas
>
Thanks, I had a typo on the standard, non-half RDFTs. One line change:
@@ -1711,7 +1703,7 @@ static const FFTXCodelet TX_NAME(ff_tx_rdft_ ##n## _def)
= { \
.function = TX_NAME(ff_tx_rdft_ ##n),
\
.type = TX_TYPE(RDFT),
\
.flags = AV_TX_UNALIGNED | AV_TX_INPLACE | FF_TX_OUT_OF_PLACE |
\
- inv ? FF_TX_INVERSE_ONLY : FF_TX_FORWARD_ONLY,
\
+ (inv ? FF_TX_INVERSE_ONLY : FF_TX_FORWARD_ONLY),
\
.factors = { 4, TX_FACTOR_ANY },
\
.nb_factors = 2,
\
.min_len = 4,
\
V2 sent.
_______________________________________________
ffmpeg-devel mailing list
[email protected]
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
[email protected] with subject "unsubscribe".
