Folks, Added fast parser FPU assembly code. Look in the directory: https://www.fractint.org/ftp/experimental/ For: Windows version (32-bit) is: SDL2-fractint-20.05alpha-2019-08-25.zip Pre-compiled 64-bit Linux version is: SDL2-Xfractint-20.05alpha-2019-08-25.tar.xz Jonathan
Added fast parser FPU assembly code. Thanks, Jonathan!
Is my understanding correct that the fast parser assembly code makes floating point *formula* fractals execute faster? I tried the PAR: NYUF004 in: fractint.par that uses the formula: Richard1 in: fractint.frm at 1024x768x256. I kept the MaxIter=150 and tried turning Floating Point off and on. ("no/yes") on the X-screen. "Floating Point=no" (fract001.gif) looks like this: http://tinyurl.com/fract001-No-FP while "Floating Point=yes" (fract002.gif) look like this: http://tinyurl.com/fract002-Yes-FP I thought that possibly the color maps might be different between the two images, but apparently not: http://tinyurl.com/fract001-2-ColrMaps It would seem that some pixels are different values between the two images, then. Is this difference expected? - Hal Lane ######################## # hallane@earthlink.net ######################## -----Original Message----- From: Fractint <fractint-bounces@mailman.xmission.com> On Behalf Of Jonathan Osuch Sent: Sunday, August 25, 2019 2:41 PM To: Fractint and General Fractals Discussion <fractint@mailman.xmission.com> Subject: [Fractint] Update 2019-08-25 to SDL version Folks, Added fast parser FPU assembly code. Look in the directory: https://www.fractint.org/ftp/experimental/ For: Windows version (32-bit) is: SDL2-fractint-20.05alpha-2019-08-25.zip Pre-compiled 64-bit Linux version is: SDL2-Xfractint-20.05alpha-2019-08-25.tar.xz Jonathan _______________________________________________ Fractint mailing list Fractint@mailman.xmission.com https://mailman.xmission.com/cgi-bin/mailman/listinfo/fractint
Hal,
I tried the PAR: NYUF004 in: fractint.par that uses the formula: Richard1 in: fractint.frm at 1024x768x256. I kept the MaxIter=150 and tried turning Floating Point off and on. ("no/yes") on the X-screen.
"Floating Point=no" (fract001.gif) looks like this: http://tinyurl.com/fract001-No-FP
while "Floating Point=yes" (fract002.gif) look like this: http://tinyurl.com/fract002-Yes-FP
This would be a comparison of integer math to floating pointmath. Try setting debug=90 on the <g> screen. This will switch the floating point math back to the C formula parser. Compare the resulting image to the floating point image with one when debug is not set. The floating point image matches the one generated using floating point under Dos Box. The integer math is suspect at this point. Jonathan
Folks, Added assembly language integer math multiply and divide routines so that integer math images should now match what is created using the DOS version. The Linux version may give odd results when division is used. I'll work on it. Look in the directory: https://www.fractint.org/ftp/experimental/ For: Windows version (32-bit) is: SDL2-fractint-20.05alpha-2019-09-08.zip Pre-compiled 64-bit Linux version is: SDL2-Xfractint-20.05alpha-2019-09-08.tar.xz Jonathan
Thank you Jonathan. I'm in awe of your long-term dedication to your work. On Sun, 8 Sep 2019 at 08:03, Jonathan Osuch <osuchj@mediacombb.net> wrote:
Folks,
Added assembly language integer math multiply and divide routines so that integer math images should now match what is created using the DOS version.
The Linux version may give odd results when division is used. I'll work on it.
Look in the directory: https://www.fractint.org/ftp/experimental/
For:
Windows version (32-bit) is:
SDL2-fractint-20.05alpha-2019-09-08.zip
Pre-compiled 64-bit Linux version is:
SDL2-Xfractint-20.05alpha-2019-09-08.tar.xz
Jonathan
_______________________________________________ Fractint mailing list Fractint@mailman.xmission.com https://mailman.xmission.com/cgi-bin/mailman/listinfo/fractint
participants (3)
-
Harold Lane -
John Wilson -
Jonathan Osuch