Reimplement the Canvas transformation functions

[m-sasha]

via the corresponding Skia functions directly, rather than by concatenating with a transformation matrix.

This appears to be significantly faster. Specifically in Compose, the canvas is translated for every node twice (by its position, and then back). Changing the translate function to perform a direct call to Skia transform makes drawing simple nodes much faster.

[MatkovIvan]

Could you please post your benchmark / measurements for both JVM and native?

[m-sasha]

Could you please post your benchmark / measurements for both JVM and native?

The improvement I saw was in the "Color Sparks" benchmark from the Fleet team. It's just a huge amount of small nodes that redraw each frame.

I don't think it's particularly important to "prove" the performance improvement in this case because this particular change doesn't really have a chance to make it worse. Plus, it makes sense regardless of performance.

[MatkovIvan]

this particular change doesn't really have a chance to make it worse

It does. Native calls might be expensive 1 2

Compared to a pure Java method call, calling a user-written native method usually has a significant overhead. The reasons for this are as much to do with optimisations that the JVM can't make compared to regular Java methods:

• the JVM can't inline the native method;
• the JVM doesn't know enough about the method to make optimisations that it could make when compiling a regular Java method (for example, it has to assume that all of the parameters passed in are always used);
• the JVM can't make other optimisations that it could make if it were dynamically compiling the code (e.g. compiling a constant parameter is a constant operand to a machine instruction rather than placing it on the stack and reading it off again);
  in order to make the call into the DLL or library, the JVM may have to perform extra work, such as rearranging items on the stack.

[m-sasha]

@MatkovIvan Indeed, but this PR replaces a call to toInterop + a native call with just one native call.