Performance optimizations for the GBC code

Description

Now that the CGB code has been merged !16 (merged) we need to focus on the optimization process to be able to have CGB support without to much of a performance impact.

Analyze the performance impact of this commit f61e3f9e because of the whole let cycles_n = cycles / self.mmu().speed.multiplier();.

The main areas that should be the focus of optimizations are:

• The PPU conditional code for CGB and DMG with color compatibility
• The variable clock speed code of the main GB cycle https://gitlab.stage.hive.pt/joamag/boytacean/-/blob/master/src/gb.rs#L396
• The usage of the get_flipped() method seems to be one of the culprits

️ The performance impact of the CGB changes is measured to be in the order of 30% less performance. And most of it should be recoverable.

Version 0.7.4 of Boytacean should be used as a reference for benchmarking.

changed title from Performance optimizations for the GBC code to Performance optimizations for the GBC code

added quick-win label

assigned to @joamag

mentioned in merge request !16 (merged)

changed the description

changed title from Performance optimizations for the GBC code to Performance optimizations for the GBC code

changed the description

@joamag I must run some base benchmark operations in the native SDL port and post here together with the machine specs

changed the description

Since cdcfc326 there's now support for benchmark included in the GitHub Actions. This should be used to control performance evolution of the system.