2.7 regression in GL performance on Mac

[dweymouth]

Checklist

• I have searched the issue tracker for open issues that relate to the same problem, before opening a new one.
• This issue only relates to a single bug. I will open new issues for any other problems.

Describe the bug

The issue is pretty well-described here fyne-io/terminal#133. I am just opening an issue in fyne itself rather than the terminal project since this is where the root cause lies.

How to reproduce

see linked issue

Screenshots

No response

Example code

see linked issue

Fyne version

2.7.0

Go compiler version

n/a

Operating system and version

Mac M series (unsure if it impacts Intel mac or other OSes)

Additional Information

No response

[andydotxyz]

Thanks @dweymouth - I guess we can switch out the use of the sub-data optimisation for macOS path? (but not others / iOS as they all seemed to benefit)

[dweymouth]

Have we confirmed this isn't an issue on other desktop platforms? I don't have low-ish spec windows or linux hardware to test with

[redawl]

I don't see an issue on my laptop, which is definitely lower spec than an m1 mac

[redawl]

I suppose we could only use BufferSubData on mobile platforms, where the benefit is noticeable

[andydotxyz]

That is OK with me too.
Though on Linux it does seem to be absolutely flying. However it's happily fast without.

[dweymouth]

It might very well be a Mac metal/GL thing. I'll try to see if I notice any responsiveness difference on my Linux/Win dual-boot desktop, but that is a pretty beefy machine so it may obscure any differences

[drunlade]

Please see comment on fyne-io/terminal#133 (comment) but TL;DR; it has the same issue on Windows according to my testing.

[dweymouth]

Fixed on develop