~cnx/roux - Alternative QBE compiler

Age	Commit message (Collapse)	Author
2016-04-12	check invalid instruction types	Quentin Carbonneaux

2016-04-12	both comparison arguments need to be fixed	Quentin Carbonneaux
	The second argument will not be a constant because of the assertion. But it could be a fast local, and this situation needs to be handled by the last case of fixarg().
2016-04-12	frendlier error message in emit	Quentin Carbonneaux

2016-04-12	add simple il validation	Quentin Carbonneaux

2016-04-12	fix typo in emit	Quentin Carbonneaux

2016-04-12	handle dumb conditional jumps in isel	Quentin Carbonneaux
	I also removed the code to handle jumps with a constant argument since those should be eliminated by the folding pass.
2016-04-12	fix bug in predecessors filling code	Quentin Carbonneaux

2016-04-12	fix type bug in abi3 test	Quentin Carbonneaux

2016-04-12	simplify fillpreds() code	Quentin Carbonneaux

2016-04-12	diagnose some undefined uses	Quentin Carbonneaux

2016-04-12	oops, wrong test in abi classify()	Quentin Carbonneaux

2016-04-11	improve help message slightly	Quentin Carbonneaux

2016-04-09	this can be false	Quentin Carbonneaux
	I think I should check for Top when rewriting. Because for sure, we don't want to replace some temporary by Top, first, I can't represent it, and second, it'd mean that it is a temporary used undefined. The example that triggered the assertion was like that: @b0 jnz 1, @b1, @b3 @b1 %x =w phi @b0 10, @b2 %x1 jnz %x, @b2, @b3 @b2 %x1 =w sub %x, 1 jmp @b1 @b3 %x2 =w phi @b1 %x, @b0 42 So SCCP optimistically assumes %x is 10 when it goes through @b1, then at the branch, @b1->@b3 is left for dead. After that, @b2 is processed and the flow worklist is empty. Now uses are processed, %x2 will get processed regardless if its block is dead or not (it is at that time), then all the back-edges are dead so the phi is set to the lattice merge of tops only. BOOM! This leads to the question, should phis of dead blocks be processed? I'll check that later.
2016-04-09	cosmetic fixes in llvm comparison	Quentin Carbonneaux

2016-04-09	rebuild rpo after fold	Quentin Carbonneaux

2016-04-09	did I loose my c?	Quentin Carbonneaux

2016-04-09	enable constant folding	Quentin Carbonneaux

2016-04-09	fix wrong assertion in fold	Quentin Carbonneaux

2016-04-09	oops, forgot to patch phi arguments	Quentin Carbonneaux

2016-04-09	more debug tweaks in fold	Quentin Carbonneaux

2016-04-09	add a proper block deletion routine	Quentin Carbonneaux

2016-04-09	nicer debug info	Quentin Carbonneaux

2016-04-09	quickly hack fold rewriting	Quentin Carbonneaux

2016-04-08	avoid gcc warning in emit	Quentin Carbonneaux

2016-04-08	simplify a buggy test	Quentin Carbonneaux
	I found it by compiling -O2 and seeing the ABI code fail. Further investigation revealed GCC trimmed away the last iteration of the loop because I was accessing the third element of an array of size two. This is undefined behavior, so GCC "proved" that the last iteration was never run.
2016-04-08	prevent gcc warning in rega	Quentin Carbonneaux

2016-04-08	fix loop header detection bug in spill	Quentin Carbonneaux

2016-04-08	use union for punning in emit	Quentin Carbonneaux

2016-04-07	use cast in czero()	Quentin Carbonneaux

2016-04-07	inline latmerge() (cross fingers)	Quentin Carbonneaux

2016-04-07	adjustments in sccp	Quentin Carbonneaux

2016-04-07	add a canfold field to opdesc	Quentin Carbonneaux

2016-04-07	add boring folding code	Quentin Carbonneaux

2016-04-06	start work on constant propagation	Quentin Carbonneaux

2016-04-05	default to gnu gas format	Quentin Carbonneaux

2016-04-05	use bsiter() for better performance	Quentin Carbonneaux

2016-04-05	speedup bscount()	Quentin Carbonneaux

2016-04-05	yay, support freebsd	Quentin Carbonneaux

2016-04-04	fat il!	Quentin Carbonneaux

2016-04-04	reorder instructions	Quentin Carbonneaux

2016-04-04	cosmetics in sysv abi	Quentin Carbonneaux
	A struct of size 0 is now marked as passed in memory. All the ABI code assumes structs passed in registers have size at least 8. This could have an impact on the alignment in the stack, but eh, I guess they are rare.
2016-04-04	remove old selpar() code	Quentin Carbonneaux

2016-04-03	fix alignment code in selpar()	Quentin Carbonneaux

2016-04-03	rewrite of selpar() for factoring	Quentin Carbonneaux

2016-04-01	typo in readme	Quentin Carbonneaux

2016-04-01	don't try to keep use counts in abi()	Quentin Carbonneaux
	Abi lowering does not need use counts, but they are needed for instruction selection. I changed main to call filluse() between these two passes.
2016-04-01	cheap massive performance gain on brainfuck	Quentin Carbonneaux

2016-04-01	tradeoff the type of bsiter()	Quentin Carbonneaux
	int is used all over the place for temporaries, maybe this should be changed, I don't know. Another thing to consider is that temporaries are currently on 12 bits (and will be on 29 or 30 bits in the future), so int will always be safe to store them. We just loose the free invariant of non-negativity.
2016-04-01	use bsiter in critical loop	Quentin Carbonneaux

2016-04-01	add huge mandelbrot brainfuck example	Quentin Carbonneaux