-
Notifications
You must be signed in to change notification settings - Fork 6
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Benchmarking 'Bansheequeen' and 'Overwatch' drawings #1
Comments
By the way, could you make a command line (.BAT) file to compile with GCC (Windows)?
|
Hi Georgi, The en-/decoding speed can be heavily dependent on the target machine, so in order to get a comparable benchmark I recommend running it on your machine. :) I've disabled OpenCL for builds that don't use premake so that you don't get any linker errors because of OpenCL + some other small portability improvements. In it's updated form I was able to compile it through both, GCC on Windows (both WSL & MSYS2) and the Intel Compiler. Intel Compiler
GCC via MSYS2
Please let me know if your problem is resolved. |
Thanks, will run the benchmark as soon as I have the binaries, tried the new .ZIP:
And the GCC attempt:
Don't know enough to make it work, tried adding -std=c99 but it complained again. |
I'm a bit confused about the Could you please provide the output of this:
|
I have two laptops, the old one (with Core 2) running Windows XP 32bit has installed Intel v15 and GCC v4.7.2:
And the newer laptop i5-7200U running Windows 10 64bit has GCC v7.3.0 64bit:
|
I think the gcc build on your i5-7200U should have produced an executable. |
Indeed, please wait some 30 minutes, already running successfully it, along with it for comparison running Oodle 'Selkie', Lizard, LzTurbo and trle ... on both .BMPs .... |
Here we go... First, Overwatch-Cast-4K-Wallpaper.bmp:
Second, The_Undead_Lady_Sylvanas_Windrunner_(The_Banshee_Queen)_Warchief_of_the_Horde_3840x1600.bmp:
And mixed with TurboBench results:
Wow, congratulations for setting Pareto with rle8 Extreme on ... WoW! Note: In the hurry forgot to start the shell in Administrative mode, results should be slightly better in it. Filetest-wise, if you have some good one in mind I will test it as well, I find the 'girl.bmp' too "untelling", if such a file has to represent scanned old FAX dumps at least it has to feature text and logos. Oh, and it would be interesting to run the negative modes of latest Zstd, as a complementary. |
And for good measure latest Zstd:
Zstd has become much faster, it sets Pareto frontier, the FASTEST compressor? |
Wow, that's impressive! A 24 bit run length encoding might produce much better compression ratios with BMPs, potentially I can add one to rle8. |
Right, that's one of the hidden aspects/properties of benchmarking - it reveals hidden potential/scenarios.
Very telling is this 'Overwatch' testfile. |
I have added 24 Bit and 48 Bit en-/decoders to rle8 extreme. |
This latest .ZIP package compiles without any warnings with GCC v7.3.0. The CAST:
The Banshee:
Congratulations Christoph, your 24bit RLE sets Pareto for Decompression, on both beauties.
Wot! 3x faster compression and 2x faster decompression than Lizard 11, CRAAAZYY! Regarding the fluctuating speed rates, allow me to suggest start using the line that made Nakamichi's speed stable:
Above timings were gotten with Administrator rights. Didn't look up the source, have you entered HIGHEST PRIORITY? |
I’ve added a new benchmarking tool to rle8 in the previous release. That will start with a dry run which eliminates most of the fluctuations. Just use If compiled for Windows the application runs in high priority mode and restricts execution to CPU core 0. |
Hi Christoph,
wanted to see your superb RLE [de]crunching two random drawings from my 'Deathship' compression corpus:
Previously (the old trle) did scream on 'overwatch' setting a Pareto:
Sanmayce/Nakamichi#1 (comment)
The 'Banshee Queen' is not so RLEish, wonder how your RLE8 extreme fares against the trle...
The .BMPs were created out by these .JPGs:
The text was updated successfully, but these errors were encountered: