~cnx/roux - Alternative QBE compiler

Age	Commit message (Collapse)	Author
2016-02-24	use default use/def counts in newtmp()	Quentin Carbonneaux
	This is paliative, before I check that all use counts in isel.c are correct.
2016-02-24	do not print null offsets	Quentin Carbonneaux

2016-02-24	fix swapped operands on emitcopy() call	Quentin Carbonneaux

2016-02-24	prepare for big structs passing code	Quentin Carbonneaux

2016-02-24	support memory class arguemnts	Quentin Carbonneaux

2016-02-24	oh oh, subs are not folded!	Quentin Carbonneaux

2016-02-24	simply use memset to 0 stuff	Quentin Carbonneaux

2016-02-24	oops, phi nodes rewrite for fast locals was trashed	Quentin Carbonneaux
	The phi fixing mechanism can use emit(), so we need to set curi before performing the rewrite. Otherwise, we are writing at random places in the instruction buffer (not so bad because it is bounds checked), but then we loose the instructions written (bad)!
2016-02-23	add fun example, does not compile yet	Quentin Carbonneaux

2016-02-23	cosmetics in isel.c	Quentin Carbonneaux

2016-02-23	print new jump instructions	Quentin Carbonneaux

2016-02-23	fix uninitialized variable in selpar()	Quentin Carbonneaux

2016-02-23	patch return, might not work	Quentin Carbonneaux

2016-02-22	fix buggy name changes in isel	Quentin Carbonneaux

2016-02-22	simplify emit table	Quentin Carbonneaux

2016-02-22	cosmetics	Quentin Carbonneaux

2016-02-22	use isstore() in parser	Quentin Carbonneaux

2016-02-21	do not spill dead phis	Quentin Carbonneaux
	Regalloc will be able to handle these spurious phis, however, some other spurious dead instructions can be emitted. It would be better to get rid of them upfront; maybe by modifying isel, or by inserting a proper dce pass. An example of that undesirable behavior is exposed below. @l1 %foo =w ... @l2 %dead =w phi @l1 %foo, ...
2016-02-21	complete fp support for small structs	Quentin Carbonneaux
	There is an oddity/bug though, we use OStorel to store possibly fp registers. Gas does not complain, but this is wrong. The fix is probably to have a simple OStore, like in the OLoad case.
2016-02-21	genius or idiot?	Quentin Carbonneaux

2016-02-19	add lexing sugar for backward compatibility	Quentin Carbonneaux

2016-02-18	stop using OXxx1 and use new OLoad	Quentin Carbonneaux

2016-02-18	use classes in arg classification	Quentin Carbonneaux

2016-02-18	complete argcls (pretty ugly...)	Quentin Carbonneaux

2016-02-18	start completing the fp abi	Quentin Carbonneaux

2016-02-16	fix dumb bug found by Andrew Chambers	Quentin Carbonneaux

2016-02-15	better variable name in selcall()	Quentin Carbonneaux

2016-02-15	always print RACall with 3 digits	Quentin Carbonneaux

2016-02-15	use arguments in mandelbrot test	Quentin Carbonneaux

2016-02-15	more fp calling conventions	Quentin Carbonneaux

2016-02-15	be future proof in usage message :)	Quentin Carbonneaux

2016-02-15	fix comments in emit	Quentin Carbonneaux

2016-02-15	uniformize notations in mandelbrot test	Quentin Carbonneaux

2016-02-15	scale up the mandelbrot test	Quentin Carbonneaux

2016-02-15	quickly patch emit (fp load and store)	Quentin Carbonneaux

2016-02-15	initialize class correctly in pmgen()	Quentin Carbonneaux

2016-02-15	patch isel for store{s,d}	Quentin Carbonneaux

2016-02-15	completely hide xmm15	Quentin Carbonneaux

2016-02-15	more stressful test for fp	Quentin Carbonneaux

2016-02-15	partial fix argcls() in isel	Quentin Carbonneaux

2016-02-15	collect and emit fp constants	Quentin Carbonneaux

2016-02-13	add new test by Andrew Chambers	Quentin Carbonneaux

2016-02-12	test int syntax for one fp constant	Quentin Carbonneaux

2016-02-12	cut code, use scanf a little more	Quentin Carbonneaux

2016-02-12	new syntax for float literals	Quentin Carbonneaux

2016-02-12	scan ints using scanf	Quentin Carbonneaux

2016-02-12	fix harmless typo in isel	Quentin Carbonneaux

2016-02-12	use ICX{np,p} to clarify enum definitions	Quentin Carbonneaux

2016-02-11	patch minic for new comparisons	Quentin Carbonneaux
	This solved one pending bug: comparisons of long variables are now compiled properly. A bug for comparisons < and <= of pointers remain, it is related to signedness, not width. This can be easily fixed by the reader!
2016-02-11	fixed the comparison issue	Quentin Carbonneaux