Improved support for "context threading" (not enabled).

This technique uses CPU Register Address Stack for return branch predicion and give 25% improvement on some call intensive micro-becnh

Improved support for "context threading" (not enabled).

This technique uses CPU Register Address Stack for return branch predicion and give 25% improvement on some call intensive micro-becnhmarks. Unfotrunately, it also slow downs other recursive benchmarks (probably because of RAS size). The techniquea is not compatible with VM without register variables.

show more ...

Avoid code duplication

Mark zend_jit_print_regset as unused

This is a debugging function, so it's usually not going to be used
and we don't want to see a warning about that.

Eliminated unreachable jumps

Don't split basic block after RECV, if function checks type hints

Fixed a typo

Avoid code generation for BB with single VM instruction

jump optimization

Fixed crash

Avoid useless code generation for RECV opcodes

JIT code improvement for ocache.jit=1 (IP consistency)

Improve JIT code for smart branches

Avoid generation of dead jumps

Use Zend MM for JIT temporary buffres allocation

Added JIT compiler for x86 and x86_64