This is an automated email from the git hooks/post-receive script.
Git pushed a commit to branch master
in repository ffmpeg.
The following commit(s) were added to refs/heads/master by this push:
new bcbc183264 doc/swscale-v2: update file names
bcbc183264 is described below
commit bcbc1832649770fbcedee069bbaae1b5622f9f9a
Author: Sungjoon Moon <[email protected]>
AuthorDate: Sat Nov 14 15:47:50 2026 +0900
Commit: Niklas Haas <[email protected]>
CommitDate: Sat Nov 14 17:39:41 2026 +0000
doc/swscale-v2: update file names
Signed-off-by: Sungjoon Moon <[email protected]>
---
doc/swscale-v2.txt | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/doc/swscale-v2.txt b/doc/swscale-v2.txt
index a8bc71f426..285b229f9f 100644
--- a/doc/swscale-v2.txt
+++ b/doc/swscale-v2.txt
@@ -46,13 +46,13 @@ The basic idea is to split logic between three major
components:
the resulting output is always identical, no matter how many layers of
optimization we add.
- This code lives in libswscale/ops.c
+ This code lives in libswscale/ops_optimizer.c
3. The compiler. Once we have a sequence of operations as output by the
optimizer, we "compile" this down to a callable function. This is then
applied by the dispatch wrapper by striping it over the input image.
- See libswscale/ops_backend.c for the reference backend, or
+ See libswscale/uops_backend.c for the reference backend, or
libswscale/x86/ops.c for a more complex SIMD example.
This overall approach has a considerable number of benefits:
@@ -198,8 +198,8 @@ As a result, we arrive at a design where we first divide
the image into small
chunks, or "blocks", and then dispatch the "chain" of kernels on each chunk in
sequence. Each kernel processes a fixed number of pixels, with the overall
entry point taking care of looping. Remaining pixels (the "tail") are handled
-generically by the backend-invariant dispatch code (located in ops.c), using a
-partial memcpy into a suitably sized temporary buffer.
+generically by the backend-invariant dispatch code (located in ops_dispatch.c),
+using a partial memcpy into a suitably sized temporary buffer.
To minimize the per-kernel function call overhead, we use a "continuation
passing style" for chaining kernels. Each operation computes its result and
--
To stop receiving notification emails like this one, please contact
[email protected].
_______________________________________________
ffmpeg-cvslog mailing list -- [email protected]
To unsubscribe send an email to [email protected]