Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion.

Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules.

  

Go Back   Doom9's Forum > Search Forums
Reload this Page Search Results
Register FAQ Calendar Search Today's Posts Mark Forums Read

Showing results 1 to 18 of 18
Search took 0.00 seconds. Search: Posts Made By: DTL
Forum: Avisynth Development 13th May 2023, 03:11
Replies: 45
Views: 10,976
Posted By DTL
Sample of simple colour-space converting plugin...

Sample of simple colour-space converting plugin (YV12 to RGB32 decoding) using AVX2 optimization for both memory transfer and SIMD computing:...
Forum: Avisynth Development 1st May 2023, 10:30
Replies: 45
Views: 10,976
Posted By DTL
Some longread about superscalar programming for...

Some longread about superscalar programming for SIMD:

Many new CPU chips (may be after 199x years already) have some limited capability of superscalar computing. It mean in some cases more...
Forum: Avisynth Development 12th March 2023, 02:31
Replies: 45
Views: 10,976
Posted By DTL
Some interesting way of SIMD processing of rows...

Some interesting way of SIMD processing of rows without adding special processing for last columns if they are not integer divide to number of columns in SIMD 'workunit' size -...
Forum: Avisynth Development 29th January 2023, 21:46
Replies: 45
Views: 10,976
Posted By DTL
It looks intel SDE even latest 9.14.0 only...

It looks intel SDE even latest 9.14.0 only support full debugging integration only with VS2017. As readme says no support for VS2019 yet.

So if running in VS2019 as debug application it run...
Forum: Avisynth Development 29th January 2023, 20:06
Replies: 45
Views: 10,976
Posted By DTL
SDE typically install as standalone. And may have...

SDE typically install as standalone. And may have addon to VisualStudio. After installing addon you configure it with the path to SDE installation (or may be addon try to do it in auto way). After...
Forum: Avisynth Development 29th January 2023, 17:25
Replies: 45
Views: 10,976
Posted By DTL
I have only experience with DirectX12 and compute...

I have only experience with DirectX12 and compute shader to process. The compute shader itself for inverting is very simple. But DX12 resources init and data upload/download to/from HWAcc is not very...
Forum: Avisynth Development 29th January 2023, 07:15
Replies: 45
Views: 10,976
Posted By DTL
It is very good to add at least 'single register...

It is very good to add at least 'single register SIMD' immediate processing after 'massive full-registerfile SIMD' and before pure C scalar ending. As described in...
Forum: Avisynth Development 29th January 2023, 05:43
Replies: 45
Views: 10,976
Posted By DTL
For up to VisualStudio 2017 you can install Intel...

For up to VisualStudio 2017 you can install Intel SDE and Visual Studio addon ...
Forum: Avisynth Development 28th January 2023, 22:59
Replies: 45
Views: 10,976
Posted By DTL
"do you advice even more?" Most slowdown...

"do you advice even more?"

Most slowdown happens on something like UHDTV sized frames. As you see even small 4K frame width can not be loaded in the 2KBytes register file of AVX512 in single load...
Forum: Avisynth Development 28th January 2023, 22:36
Replies: 45
Views: 10,976
Posted By DTL
For integers: for (auto...

For integers:


for (auto column = 0; column < row_size_mod64; column += 64)
{
__m512i vector_src_00 = _mm512_loadu_si512(local_srcp);
...
Forum: Avisynth Development 28th January 2023, 21:41
Replies: 45
Views: 10,976
Posted By DTL
"some functions have not a _m256i, _m128i one" ...

"some functions have not a _m256i, _m128i one"

What are the examples ? If you do not have single op instruction or macro you make a sequence to workaround (and may make/define your own macro to...
Forum: Avisynth Development 28th January 2023, 19:48
Replies: 45
Views: 10,976
Posted By DTL
float* line_dstp = dstp + y * dst_pitch; ...

float* line_dstp = dstp + y * dst_pitch;
float* line_srcp = srcp + y * src_pitch;
float* src_float_ptr = (float*)line_srcp;
float* dst_float_ptr =...
Forum: Avisynth Development 28th January 2023, 12:14
Replies: 45
Views: 10,976
Posted By DTL
This is example of OpenMP internal MT ('threads'...

This is example of OpenMP internal MT ('threads' user control param with default 1) and SIMD up to AVX512 in float32 processing only (to make example shorter).

#include <windows.h>
#include...
Forum: Avisynth Development 27th January 2023, 22:32
Replies: 45
Views: 10,976
Posted By DTL
height = src->GetHeight(planes[p]) *...

height = src->GetHeight(planes[p]) * vi.ComponentSize();


this looks like error. Number of lines in a frame and rows in a storage buffer do not depend on bitdepth. Only measured in 8bit bytes row...
Forum: Avisynth Development 27th January 2023, 09:07
Replies: 45
Views: 10,976
Posted By DTL
To add HBD you add copy of your processing...

To add HBD you add copy of your processing functions with 16bit and float32 input and output for source and target planes. The copy over templating may be recommended because of later adding...
Forum: Avisynth Development 26th January 2023, 23:48
Replies: 45
Views: 10,976
Posted By DTL
In dstp[x] = int(i * 0.0625f + 0.5f); ...

In

dstp[x] = int(i * 0.0625f + 0.5f);

your 'i' variable declared as int, so to make division to 16 integer you can use

dstp[x] = i >> 4;

to avoid slow conversion integer to float and...
Forum: Avisynth Development 26th January 2023, 18:05
Replies: 45
Views: 10,976
Posted By DTL
Also for integer samples i = (int)(i /...

Also for integer samples

i = (int)(i / 16.0f + 0.5f);


may be replaced with integer shift without converting to float (and slow float division that any good compiler will replace with...
Forum: Avisynth Development 26th January 2023, 12:21
Replies: 45
Views: 10,976
Posted By DTL
To make it faster onCPU you need to use SIMD...

To make it faster onCPU you need to use SIMD co-processor.

So in the plugin init you put some selector of the main processing function for each combination of input params and SIMD co-processor...
Showing results 1 to 18 of 18

 
Forum Jump

All times are GMT +1. The time now is 08:40.

Doom9.org - Archive - Top

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2023, vBulletin Solutions Inc.