LCTRL 8 / SCTRL 8 - get/set ASM pointer about samp-plugin-jit HOT 3 CLOSED

Handy function:

stock bool:IsJITASMJumpCapable() {
	#emit lctrl 6
	#emit MOVE.alt
	#emit lctrl 8
	#emit neq
	#emit retn
	return false; // make compiler happy
}

Of course, that will fail if by some remote chance the assembly address of the lctrl 6 instruction is identical to the amx address - but presumably code would then fall back on the generic implementation anyway, so not too much harm done in that unlikely scenario. Relies on the fact that lctrl 8 would be a NO-OP without the JIT (or an older version of the JIT).

from samp-plugin-jit.

This will also help solve #26 since I beleve the problem there is this: When calling a function using SCTRL 6 to jump straight to it, I manually push the return address on to the stack (as CALL would do). However, the return address pushed is one based on the return value of LCTRL 6, so is an AMX address not an x86 address. When the function ends it executes the JIT compiled RETN, which pops an x86 address off the stack and jumps to it. This is why when debugging it was at strange addresses - they are valid AMX addresses being returned to by compiled code.

With this modification, the dynamic call code in YSI and amx_assembly would become:

	#emit lctrl 6
	#emit add.c 0x24
	#emit lctrl 8
	#emit push.pri
	#emit load.s.pri address
	#emit sctrl 6

Again, because LCTRL 8 is a no-op wthout the JIT plugin, this will work in both cases. The add.c must come between then lctrls so that the address is modified in countable cell numbers before being translated to an x86 address.

from samp-plugin-jit.

Another handy function I've put in amx_assembly\amx_base.inc:

stock GetAmxJITBaseAddress() {
	// Only works when the jit is installed, otherwise returns 0.  The code is
	// so short, checking if it had been done before would have been longer than
	// just repeating the calculation.
	#emit zero.pri
	#emit lctrl 8
	#emit retn
	return 0;
}

from samp-plugin-jit.