Split/Second - ULUS10513

[DragonNeos]

This game was tested with revision 5c381c50. There are several issues with running this game on the emulator:

1) Performance during gameplay is around 5 - 10 FPS on either graphics backend. Is there anything to analyze that can possibly help with performance?

2) When the opening video is being played, the following message appears:

Code:

17:48:35  WARN atrac3plus - ScePsmfPlayerDecodeThread - GHA Phase shifting

I added a log that included the following but I don't see any sceAtrac3plus being played during the opening video:

Code:

<logger name='hle.scePsmf'> <level value='debug' /> </logger>
<logger name='hle.sceAtrac3plus'> <level value='debug' /> </logger>

3) On OpenGL, the following messages appear endlessly during gameplay:

Code:

16:25:05  WARN       ge - GUI - Unimplemented Blue color mask 0x0F
...
16:25:06  WARN       ge - GUI - UNSUPPORTED: Both different SFIX (0x1C1C1C) and DFIX (0x404040) are not supported (blend equation=0), approximating with 0x000000/0x404040

See the 1st, 2nd, and 3rd Screenshot (OpenGL without "Enable the shader implementation for the "Stencil Test" (more accurate but slower)" and "Enable the shader implementation for the "Color Mask" (more accurate but slower)") at how the foreground/background looks:
           

Once both options are enabled, the warnings don't appear anymore but everything looks extremely worse off (parts of text is missing and certain things can't be seen with OpenGL with "Enable the shader implementation for the "Stencil Test" (more accurate but slower)" and "Enable the shader implementation for the "Color Mask" (more accurate but slower)"):
           

The game displays properly on External Software Renderer in races for the most part but the performance is similar to OpenGL (around 5 - 10 FPS).
           

[gid15]

1) Performance during gameplay is around 5 - 10 FPS on either graphics backend. Is there anything to analyze that can possibly help with performance?

How was the attached profiler information being taken?
Have you followed the recommendations from the README about profiling:
https://github.com/jpcsp/jpcsp/blob/mast...E.txt#L753

Thank you!

[gid15]

1) Performance during gameplay is around 5 - 10 FPS on either graphics backend. Is there anything to analyze that can possibly help with performance?

How was the attached profiler information being taken?
Have you followed the recommendations from the README about profiling:
https://github.com/jpcsp/jpcsp/blob/mast...E.txt#L753

Thank you!

I've added a new video option "Hide some video effects (experimental, might improve performance)" in 0d59542. Could you please try if it is somehow improving the performance on your computer? When this option is enabled, I've tried to disable some video effects which might be time consuming. This is just meant to try to identify which OpenGL operations are time consuming, the graphics will probably be (partially) broken when this option is enabled.

It would be helpful to have a profiler information with and without this option enabled, and in each case to reset the profiler information when reaching the area where there is a performance issue, like recommended into the README.

[DragonNeos]

1) Performance during gameplay is around 5 - 10 FPS on either graphics backend. Is there anything to analyze that can possibly help with performance?

How was the attached profiler information being taken?
Have you followed the recommendations from the README about profiling:
https://github.com/jpcsp/jpcsp/blob/mast...E.txt#L753

Thank you!

I've added a new video option "Hide some video effects (experimental, might improve performance)" in 0d59542. Could you please try if it is somehow improving the performance on your computer? When this option is enabled, I've tried to disable some video effects which might be time consuming. This is just meant to try to identify which OpenGL operations are time consuming, the graphics will probably be (partially) broken when this option is enabled.

It would be helpful to have a profiler information with and without this option enabled, and in each case to reset the profiler information when reaching the area where there is a performance issue, like recommended into the README.

On revision da6886e0, using the video option "Hide some video effects (experimental, might improve performance)" improves the performance of this game (around 24 - 30 FPS) compared to not using this option (around 12 - 15 FPS).

I clicked on "Reset profiler information" to capture the area with the performance issue for about 1 minute (one with the video option "Hide some video effects (experimental, might improve performance)" and the other without the video option "Hide some video effects (experimental, might improve performance)"). I also attached a capture of the profiler data with the External Software Renderer (revision 3d1bbb0) in the area with performance issues (around 7 - 9 FPS).

I think the video option "Enable the shader implementation for the "Stencil Test" (more accurate but slower)" needs to be fixed. As I mention in my last post, the text appears cut off and the graphics don't display properly on the race course (white screen). I have also attached profiler logs with the Stencil Test enabled and video option "Hide some video effects (experimental, might improve performance)" disabled (around 4 - 7 FPS) and enabled (around 8 - 11 FPS) as a comparison.
           

I have attached screenshots showing the video option "Hide some video effects (experimental, might improve performance)" being disabled and enabled without the Stencil Test video option.
       

[gid15]

On revision da6886e0, using the video option "Hide some video effects (experimental, might improve performance)" improves the performance of this game (around 24 - 30 FPS) compared to not using this option (around 12 - 15 FPS).

I clicked on "Reset profiler information" to capture the area with the performance issue for about 1 minute (one with the video option "Hide some video effects (experimental, might improve performance)" and the other without the video option "Hide some video effects (experimental, might improve performance)"). I also attached a capture of the profiler data with the External Software Renderer in the area with performance issues.

Interesting, thank you.
The performance issue seems to be coming from the "Copy GE to memory" operations:
with hiding of video effects:

Code:

Texture load count: 43, average 0.0 per GE list
Copy GE to memory: 0, average 0.0 per GE list
GE list duration: 28996ms, average 23.7ms per GE list, max FPS is 42.2

without hiding of video effects:

Code:

Texture load count: 12812, average 10.3 per GE list
Copy GE to memory: 46513, average 37.2 per GE list
GE list duration: 74206ms, average 59.4ms per GE list, max FPS is 16.8

The other elements have slightly changed, but do not seem to be significant.

It seems you are using a laptop integrated video card

Code:

07:30:47  INFO       ge - GUI - OpenGL version: 3.3.11672 Compatibility Profile Context, vender: ATI Technologies Inc., renderer: ATI Radeon HD 4200

The Radeon HD 4200 is quite old (more than 10 years old), right?

[DragonNeos]

It seems you are using a laptop integrated video card

Code:

07:30:47  INFO       ge - GUI - OpenGL version: 3.3.11672 Compatibility Profile Context, vender: ATI Technologies Inc., renderer: ATI Radeon HD 4200

The Radeon HD 4200 is quite old (more than 10 years old), right?

Yes, but I can only make do with what I have on my computer. I'm hoping some performance improvements can be made on this game.