@BSzili

I see. Hope he didn't suffer to much post-trauma.

I'd like to get hold of him. Maybe I could get the source to have something to build upon.

@broadblues

I would be very surprised (looking at the changes on that link) if the changes were so sever that it would be impossible to adapt to the current jpeg.library API. I mean, it doesn't expose too much detail about the internals of libjpeg. I mean, jpeg.library could theoretically house any implementation, not just libjpeg (or libjpeg-turbo, which is what I will put in there).

But I understand a simple recompile wouldn't work. It wouldn't even compile. Still, I'd rather much get the source to build upon then reinventing the wheel and not being able to use the new library anyway (because no apps would use it).

I mean, making a new library is meaningless for me, then I could just link towards libjpeg directly instead, if it was only a personal project. Much easier.

@BSzili

Some renamed and added symbols, fine
The code could be adapted to this in a few hours.

@thomas

Darius' website does not exist anymore and presumably not his email either then. The original author didn't want to make it open source a million years ago. That is void now (read: I don't care) if we get the source and he doesn't respond to an e-mail address he probably doesn't even have anymore.

What is not compatible, and which Amiga software would require those special features that are no longer available in v9?

Does jpeg.library (or datatypes) even export such functionality that is so obscure that libjpeg could just decide to drop them while still being able to call itself libjpeg?

I feel like such application would have compile in their own libjpeg instead of using the shared library...

Does anyone have, or know where the source of jpeg.library can be found?

It is old and needs to be updated. I can't find the source anywhere, only the headers, and I don't want to waste time reinventing the wheel. The links in the readme are dead and I assume the emails are too, although I tried to send.

The source of jpeg.library should be open-source.

Why I ask is because the current version is based on ijg v6a. Current version is v9a and it is 10-50% faster. I also have an Altivec build at hand that is 200-300% faster!

@all

So, conclusion is that feanor's work is void because noone wants to build mplayer (at a reasonable price in Amiga-land). That's too bad, I was looking forward to this new version...

@kas1e

So how much do I have to pay you to give us a new Odyssey with the latest webkit?? The current engine is showing its age with an increasing number of webpages starting to fail. We need a new build soon...

@samo79

I'll donate to a bounty for that.

@feanor

I recently found libjpeg-turbo which is a simd-accelerated version of libjpeg. I don't know, but maybe you could use some of its Altivec code for further j/mpeg acceleration?

It is tremendously fast! Between 2-3 times (200-300%) faster than the original libjpeg for both jpeg encoding and decoding. I'm currently trying to make a jpeg.library out of it, and I emailed Oliver Roberts to make a new WarpDT.jpeg out of it. No response yet though... Seems jpeg.library is not open-source, which is kind of stupid.

On a swedish layout it is using AltGr (or just Alt) and the ?+ key which is to the right of 0(zero) (on a swedish layout). Worth trying... Try Alt and all keys in the upper-right corner, and the key between LShift and Z as ppl mentioned. I have |(pipe) and arrows on that key though...

@kas1e

What was the bug? How could Dopus5 affect MUI in that way?

Thanks!

Hi

I'm trying to configure a generic tool but I get this:
......
config.status: executing depfiles commands
./config.status[2061]: cannot create ././.deps/jmemnobs.Plo: No such file or directory
./config.status[2061]: cannot create ././.deps/cdjpeg.Po: No such file or directory
...and many of these...

Seems like sh (abc-shell) isn't handling "./" correctly or something..?
The directory ".deps" is created but empty.

Help!

Using OS4-FE.

I have always use DOpus on its own screen, than way I can keep a set of listers and a button bar constantly open without obstructing WB where I have all other programs running.

I use the 5th mouse button to switch screens (configured by Allkeys). Other system wide hotkeys I use ToolsDaemon or AllKeys for. In DOpus I only have local hotkeys for copy, move, edit, view, etc.

@Severin

Sorry for hijacking the thread, but is it possible to use a USB device with RDB/SFS on WinUAE?

I mean, I have tried to mix MBR/FAT with RDB/SFS to some extent, just to be able to read the data in case all my Amigas die. But if I can do this on WinUAE there is no need to have MBR/FAT on the device at all.

@LiveForIt

Quote:

I would find it strange if these were both broken. They use unrelated but both use a fairly high bitrate though, 10-12Mbit/s. They are my main source of HD material, but waaay to slow to watch, even with comp_yuv

Quote:

Peek? I use the docky that shows a rainbow-colored graph over time. All previous versions of MPlayer shows 100% CPU when the video isn't playing at full speed. Maybe this version waits for the DMA transfer while the old versions were busy using the CPU to copy data to screen.

Quote:

Switching overhead is very small. The problem is that it seems to skip a lot of frames and be less smooth than comp_yuv, even if the overall play speed seems a little higher.

Quote:

Noo. Any program could use 100% CPU without affecting WB or USB if its priority is 0 or lower. All previous versions of MPlayer did that when a video was too heavy.

Is MPlayer able to use two cores at the moment, if AmigaOS suddenly got SMP tomorrow?

@LiveForIt

I noticed that I could not play 720p (mp4 saved by my phone, or MOV saved by my action camera) at full speed even with your latest version and comp_yuv, but still the CPU was not at 100%, more like 80-90%. Is that because of vblank sync? Also tried comp_yuv2 which seemed a tiny bit faster (used like 95% CPU), but less smooth.

I wonder if feanor's improvements are giong to be able to fix these 720 clips...

Is there anyway to switch off vblank timing for maximum speed and 100% CPU usage? Care more about speed than tearing here... It's mostly about previewing action clips rather than watching a movie.

@samo79

I noticed that muimplayer closes bsdsocket.library twice on exit even if it was only opened once. It seems it only happens when no streaming from the net, because then it is only opened once, but it is still closed twice.

When you have used muimplayer a few times, the bsdsocket opencount variable goes to zero and bsdsocket is expunged from memory, but other programs are still using it and we have a crash.

@Raziel

Quote:

I tested a heatpipe once (designed for a gfx-card) by holding it in my hand, expecting the other end to reach hand-temperatur pretty quickly, but that never happend. The heat transfer was obviously very poor at 35degC.

To make a pipe most efficient, it would have to cool something that is hotter than the boiling point, and cooled by something that is less that the boiling point. Only then will you get the maximum effect through the evaporation and condensation of the liquid/gas and fast transfer of the heat in the form of gasified medium.

Maybe water still is efficient with a diff between 40degC (CPU) and 20degC (ambient), I don't know, but it doesn't feel like you would get enough flow without a mechanical pump in this case. Having vacuum in the tube would lower the boiling point from 100degC to something lower, but how low?

"something with a semblance of accuracy"

Like what?

The PassMark seems like a fairly good indicator when buying a PC. But the program is probably written especially for x86 and would not compare even if it was ported to PPC I suppose.

We got memtester which is universal. It would be nice with a CPU tester (maybe just single-core) that could be made somewhat universal in ANSI-C ((un?)fortunately then not testing vector units).

Maybe also try to include the new altivec code from this thread?

I was wondering if there was any program with which one could compare performance between different PPCs and x86?

Currently I look at https://www.cpubenchmark.net/ when I'm getting a PC to see how fast it really is, since CPU model, GHz and number of cores has little to say.

My crap laptop at home (AMD C50) scores 460 passmark, my old work laptop (i5-something) scored 2800, my new work laptop (i7-4810MQ) scores 8500.

But where on this list does a G4/800 or PA6T/1.800 or ARM Cortex-A7 fit in?

Finally an Amiga keyboard with (a very few) multimedia-keys

I've been so disappointed with earlier Amiga keyboards that did not have one single multimedia-key on then. This one has at least four :)