South Park Rally crashes using interpreter CPU core

[GoogleCodeExporter]

Describe your system:
 - Operating System: Linux
 - Machine type: x86 32-bit
 - Mupen64Plus version: 1.5
 - Plugins used: Any
 - Game name and CRC/MD5 hash: South Park Rally (1C494719032FF99382B167C43FB11762)

Describe the problem:

This game is extremely sensitive to the floating point rounding mode and 
particularly the correct 
operation of the FLOOR.W.D instruction.  The behavior of the interpreter and 
dynamic recompiler do 
not match.  If you select interpreter, the game will crash during the intro.

Original issue reported on code.google.com by a8033...@gmail.com on 8 Feb 2010 at 1:07

[GoogleCodeExporter]

The reason this occurs is because the interpreter is relying on a hack to set 
the rounding mode, using the set_floor macro.  This does not work because gcc 
sets the rounding mode back to truncate.

Don't rely on hacks like this.  Either use the libm floor() function, or use 
__asm__ to output exactly the instructions that you want.

Original comment by a8033...@gmail.com on 27 Aug 2010 at 3:39

[GoogleCodeExporter]

Very interesting, thanks for report. Maybe it could be easely change in the 
repo.

If you have time, it could be good that you look at the project:

http://bitbucket.org/richard42/mupen64plus-core/overview

And prupose a patch.

Thanks in advance. =)

Original comment by dorian.f...@gmail.com on 27 Aug 2010 at 9:57

[GoogleCodeExporter]

It's more than just the floor instruction, many of the floating point 
conversions are incorrect.

There is a similar problem in the Goldeneye intro, where the in-game sequences 
do not finish.  It doesn't crash, just restarts.

Original comment by a8033...@gmail.com on 8 Sep 2010 at 11:44

[GoogleCodeExporter]

Thanks! This prob is important indeed... What do you suggest to do with it?

You seem to be right about the conversion prob (even if it doesn't solve the 
prob).

When you said that "many of the floating point conversions are incorrect", are 
they easely identifiable?

Don't hesitate to go on the #mupen64plus IRC on freenode, maybe there will be 
people that will understand the problem and help you. :)

Original comment by dorian.f...@gmail.com on 14 Sep 2010 at 2:44

[GoogleCodeExporter]

Original comment by richard...@gmail.com on 22 May 2011 at 4:00

• Added labels: Component-Core, Milestone-2.0, Type-Defect

[GoogleCodeExporter]

Just pushed fd0d64765ec6 (currently only in 
https://bitbucket.org/lazhur/mupen64plus-core/ ). Can you please test if it 
works for you?

Original comment by sven@narfation.org on 10 Jul 2011 at 8:29

[GoogleCodeExporter]

It is now merged in https://bitbucket.org/richard42

Original comment by s...@narfation.org on 15 Jul 2011 at 12:13

[GoogleCodeExporter]

Original comment by s...@narfation.org on 29 Aug 2011 at 2:43

• Changed state: Fixed