Ren'Py 6.99.14.1 Prereleased

[PyTom]

6.99.14.1.3194)

I've just updated the prerelease channel. There are three changes:

* The GL Powersave setting is now enabled by default. With some other changes (showing multiple frames before slowing down), there shoudn't be any affect on actually perceived performance, and the new default should be nicer on laptop fans. It can be turned off on the shift+G menu if you'd like, for the purposes of benchmarking Ren'Py's framerate.

* ConditionSwitch prediction has been fixed.

* Imagemap caching has been disabled. This is different from image caching - what imagemap caching was to combine all the button states of an imagemap into a single (smaller) image. This had been interacting poorly with another change, because it required hitting the disk, something that's no longer allowed during image prediction. (Due to it being slow, and causing frame drops.)

And that's it. I feel this is getting close to becoming a point release.

[trooper6]

Hi PyTom!
I'm back from my conference. I downloaded the prerelease and loaded up my game with it...not using the code to increase image cache or anything like that...just the basic code with the new pre-release. It looks great! There is no slowdown or stuttering. Everything is smooth and wonderful. This means that I can get back to finishing this game after sooooo long. Thank you so much!! You are a peach!

What exactly did you do to make it so marvelous?

[Arowana]

Thanks for the updates, PyTom! ConditionSwitch prediction is working much better now. Some of my images are still not being predicted, which I think is related to this. With all the recent improvements to image prediction, do you have any plans to allow prediction for a call followed by a return in the future?

In my particular case, the code I use to show certain backgrounds is repetitive, so I've put it inside a label that I call whenever that background is shown. Something like this:

Code: Select all

label forest_background:
    # code to show all the images that make up the background
    return
    
label scene1:
    call forest_background
    with transition1
    "I'm walking through the forest."
    # other stuff
    
label scene100:
    call forest_background
    with transition2
    "I'm walking through the forest again."
    # other stuff

It sounds like this prevents prediction of the background images unless I put an interaction inside the "forest_background" label. However, I'm not sure what interaction I can put there that won't disrupt the background transition. Does that make sense? Let me know if you have any ideas, and thanks!

[Zetsubou]

With all the recent improvements to image prediction, do you have any plans to allow prediction for a call followed by a return in the future?

In my particular case, the code I use to show certain backgrounds is repetitive, so I've put it inside a label that I call whenever that background is shown. Something like this:

It sounds like this prevents prediction of the background images unless I put an interaction inside the "forest_background" label. However, I'm not sure what interaction I can put there that won't disrupt the background transition. Does that make sense? Let me know if you have any ideas, and thanks!

Do you really need call statements to construct your images? I don't know what you're doing to make the background exactly, so I don't know if this will help, but if you just want to overlap some images, or add animation, you should be able to do something like:

Code: Select all

image forest_background:
  contains:
    "forest" #BG image
  contains:
    "bird" #show a bird on top of the BG up the top right
    xpos 1800
    ypos 100
  contains:
    "mole" #show a mole zoomed out slightly, on the bottom, slowly moving left and right in a loop
    zoom 0.8
    yalign 0.9
    linear 1 xoffset 10
    linear 1 xoffset -10
    repeat

and then use a plain old "show" statement, like with any other image.

[PyTom]

Thanks for the updates, PyTom! ConditionSwitch prediction is working much better now. Some of my images are still not being predicted, which I think is related to this. With all the recent improvements to image prediction, do you have any plans to allow prediction for a call followed by a return in the future?

We've been predicting that for some time, now. I just tried out:

Code: Select all

label showbg:
    scene bg whitehouse
    return

label test:

    "..."

    call showbg
    show eileen happy

    "... done."

label start:
    "..."

    $ renpy.jump("test")

    return

And it predicted both bg whitehouse and eileen happy. Could you be running past the 32-statement prediction window? Maybe try bumping config.predict_statements up and see if that helps? If not, I'd appreciate a small demo showing the problem, so I can work it out. We should be predicting calls.

(The call to renpy.jump intentionally breaks prediction, so I have time to hit f4 to show the screen.)

[Arowana]

Could you be running past the 32-statement prediction window? Maybe try bumping config.predict_statements up and see if that helps?

I think this was the real issue! I doubled config.predict_statements to 64, and now most of my background images are being caught. Thanks so much for the suggestion.

Do you really need call statements to construct your images?

Thanks for trying to help. I actually have to show different parts of the backgrounds on different layers and set up 3D camera stuff, so unfortunately I can't do it in a single image. I wonder if that's why I was running past the default prediction window before - too much stuff going on in those labels.