florian 3d9daba4c7 * some fixes to get the new cg compiling again 26 years ago
..
alpha 1c638f2952 * small fix to operator overloading when in MMX mode 26 years ago
i386 1c638f2952 * small fix to operator overloading when in MMX mode 26 years ago
powerpc b3276fa59e + a_jmp_cond() 26 years ago
symtable 157bc635ce * Fixed bugs. 26 years ago
Makefile 8b9633266b * new codegenerator compiles again 26 years ago
README fa19f0aa09 - removed the hcodegen use in the new cg, use cgbase instead 26 years ago
TODO a184e5ff43 * well, more changes, especially parts of secondload ported 26 years ago
agatt.pas 4abc13fd78 * defined some necessary virtual helper methods 26 years ago
aopt.pas 56524f33df + compilable! 26 years ago
aoptbase.pas 28ddb689cb * some small changes so the noew optimizer is again compilable 26 years ago
aoptcs.pas 5fc1435036 * changed pai386, paippc and paiapha (same for tai*) to paicpu (taicpu) 26 years ago
aoptda.pas 56524f33df + compilable! 26 years ago
aoptobj.pas 167f8d654f * fixes from daopt386.pas integrated 26 years ago
cgbase.pas 3d9daba4c7 * some fixes to get the new cg compiling again 26 years ago
cgflw.pas e10deeca37 + initial implementation 26 years ago
cgobj.pas 3d9daba4c7 * some fixes to get the new cg compiling again 26 years ago
convtree.pas 134030a529 * only small updates to work with the current compiler 26 years ago
ncon.pas 8abdd311fd * implemented some stuff for assignments 26 years ago
nmem.pas 8b9633266b * new codegenerator compiles again 26 years ago
nstatmnt.pas fdc1e9792c * some fixes for the floating point registers 26 years ago
pass_1.pas 8b9633266b * new codegenerator compiles again 26 years ago
pass_2.pas 8b9633266b * new codegenerator compiles again 26 years ago
pp.pas 8b9633266b * new codegenerator compiles again 26 years ago
tainst.pas a9d8bfa1f7 * some bug fixes (e.g. must_be_valid and procinfo.funcret_is_valid) 26 years ago
temp_gen.pas 98ad42cb87 + dummy implementation 26 years ago
tempgen.inc 1cf1665d73 * the alpha compiler can be compiled now 26 years ago
tgeni386.pas eac806034d * the main branch psub.pas is now used for 26 years ago
tgobj.pas fa19f0aa09 - removed the hcodegen use in the new cg, use cgbase instead 26 years ago
transn.pas 8b9633266b * new codegenerator compiles again 26 years ago
tree.pas 3d9daba4c7 * some fixes to get the new cg compiling again 26 years ago

README

$Id$

History
-------

13th oct 1999 remark about alignment added (FK)

CVS Log see at the end of that file

frame_pointer contains the register used as frame pointer
stack_pointer contains the register used as stack pointer
self_pointer contains the register used as self pointer


unusedregsint set of Currently available integer registers
unusedregsfpu set of Currently available fpu registers
unusedregsmm set of Currently available mm registers

availabletempregsint set of maximally available integer registers
availabletempregsfpu set of maximally available fpu registers
availabletempregsmm set of maximally available mm registers

countusableregsint count of currently available integer registers
countusableregsfpu count of currently available fpu registers
countusableregsmm count of currently available mm registers

c_countusableregsint count of max. available int registers (in the current procedure)
c_countusableregsfpu count of max. available fpu registers (in the current procedure)
c_countusableregsmm count of max. available mm registers (in the current procedure)

intregs all!! available integer register
fpuregs all!! available fpu register
mmregs all!! available multimedia register

Intel specific
--------------
unusedregssse
availabletempregssse
countusableregssse

Jonas Maebe schrieb:
>
> Hello,
>
> Is there any difference between the localsize parameter of
> g_stackframe_entry and the parasize parameter of g_return_from_proc, or
> are they both the same value?

They are different, I think the value of g_return_from_proc doesn't matter
for the PowerPC. It's the size of parameters passed on the stack
and only important for the i386/m68k probably.

>
> And for the PowerPC, what will they contain? Just the size of the local
> variables and parameters, or also the maximum needed size for parameters
> of any procedure called by the current one (the caller must reserve space
> for the callee's parameters on it's own stack because you can't push
> values on the stack in the middle of a procedure (no frame pointer))
>
> Jonas

the parameter passed to g_stackframe_entry contains the size of the all local space which is
needed
except
that one for saving registers: the set procinfo.registerstosave (not yet implemented,
I'll commit it soon) will contain
all registers which must be saved by the entry and restored by the exit code of a procedure
and you have to add extra space to do that.



Alignment
---------

The alignment is handled very easily: treference contains a field
alignment which describes the ensured alignment for the node, possible
values: 1,2,4,8,16 (1 means unligned). The code generator must update
that field at the appropriate places and take care of it when
generating the code


CVS Log
-------

$Log$
Revision 1.4 1999-10-14 14:57:54 florian
- removed the hcodegen use in the new cg, use cgbase instead