Re: [FFmpeg-devel] [PATCH 2/2] lavc/dxvenc: migrate DXT1 encoder to lavu hashtable and improve Resolume compatibility

2025-04-20 Thread James Almer
On 4/20/2025 10:08 AM, Emma Worley wrote: Offers a modest performance gain due to the switch from naive linear probling to robin hood, as well as improved compatibility with Resolume. A few numbers would be nice since you're adding new public API and a lot of code changes. OpenPGP_signatur

Re: [FFmpeg-devel] [PATCH 2/2] lavc/dxvenc: migrate DXT1 encoder to lavu hashtable and improve Resolume compatibility

2025-04-20 Thread Nicolas George
Emma Worley (HE12025-04-20): > Subject: Re: [FFmpeg-devel] [PATCH 2/2] lavc/dxvenc: migrate DXT1 encoder to > lavu hashtable and improve Resolume compatibility Thanks for the patch. Are the two indissociably linked? I have trouble imagining it. Unless there is a strong reason mak

[FFmpeg-devel] [PATCH 2/2] lavc/dxvenc: migrate DXT1 encoder to lavu hashtable and improve Resolume compatibility

2025-04-20 Thread Emma Worley
Offers a modest performance gain due to the switch from naive linear probling to robin hood, as well as improved compatibility with Resolume. Signed-off-by: Emma Worley --- libavcodec/dxvenc.c | 190 +--- tests/ref/fate/dxv3enc-dxt1 | 2 +- 2 files chang

Re: [FFmpeg-devel] [PATCH 2/2] lavc/dxvenc: migrate DXT1 encoder to lavu hashtable

2024-02-05 Thread Connor Worley
On Mon, Feb 5, 2024 at 12:06 PM Andreas Rheinhardt < andreas.rheinha...@outlook.com> wrote: > Connor Worley: > >> > >> How much would one gain if the hash function knew that key_size and > >> val_size are four? > >> > > > > That yields a nice 10-20% speedup on my machine. Are you thinking of > mac

Re: [FFmpeg-devel] [PATCH 2/2] lavc/dxvenc: migrate DXT1 encoder to lavu hashtable

2024-02-05 Thread Andreas Rheinhardt
Connor Worley: >> >> How much would one gain if the hash function knew that key_size and >> val_size are four? >> > > That yields a nice 10-20% speedup on my machine. Are you thinking of macros > to parametrize key/val size, or possibly optimized versions for common > sizes? I am thinking about i

Re: [FFmpeg-devel] [PATCH 2/2] lavc/dxvenc: migrate DXT1 encoder to lavu hashtable

2024-02-05 Thread Connor Worley
> > How much would one gain if the hash function knew that key_size and > val_size are four? > That yields a nice 10-20% speedup on my machine. Are you thinking of macros to parametrize key/val size, or possibly optimized versions for common sizes? -- Connor Worley ___

Re: [FFmpeg-devel] [PATCH 2/2] lavc/dxvenc: migrate DXT1 encoder to lavu hashtable

2024-02-05 Thread Andreas Rheinhardt
Connor Worley: > Offers a modest performance gain due to the switch from naive linear > probling to robin hood. How much would one gain if the hash function knew that key_size and val_size are four? > > Signed-off-by: Connor Worley > --- > libavcodec/dxvenc.c | 121 +---

[FFmpeg-devel] [PATCH 2/2] lavc/dxvenc: migrate DXT1 encoder to lavu hashtable

2024-02-04 Thread Connor Worley
Offers a modest performance gain due to the switch from naive linear probling to robin hood. Signed-off-by: Connor Worley --- libavcodec/dxvenc.c | 121 +--- 1 file changed, 35 insertions(+), 86 deletions(-) diff --git a/libavcodec/dxvenc.c b/libavcodec/d