56 commits

Author	SHA1	Message	Date
Dale Weiler	7ffda37513	s/atanhf/atanh/	2014-02-08 02:50:06 -05:00
Dale Weiler	24f9313952	Screw clang	2014-01-07 13:13:25 -05:00
Dale Weiler	a140b749ff	Fix lshift/rshift for runtime and const-fold consistency.	2014-01-07 12:58:53 -05:00
Dale Weiler	a934e0fe4b	Happy new year!	2014-01-01 06:24:16 -05:00
Dale Weiler	9b92cb0897	Simplify	2013-11-29 13:50:06 -05:00
Dale Weiler	43e9885a08	Fix fold_op_cmp	2013-11-29 13:48:03 -05:00
Dale Weiler	ac7e1a557d	Fixes	2013-11-29 13:36:35 -05:00
Dale Weiler	dc48af195d	Pregenerate 2.0 for const folder since it's used often	2013-11-25 13:25:29 -05:00
Dale Weiler	63c0917d24	More compiler intrinsics for math functions	2013-11-23 21:14:13 -05:00
Dale Weiler	e2bfaf8109	Implemented exp2 intrinsic	2013-11-23 07:37:26 -05:00
Wolfgang Bumiller	49bb172a09	folder: change ~ for floats too	2013-11-14 11:41:11 +01:00
Wolfgang Bumiller	f43106017f	bit more concise	2013-11-12 16:28:11 +01:00
Wolfgang Bumiller	eb4486a7ac	working fold_op_cmp implementation	2013-11-12 16:15:02 +01:00
Wolfgang Bumiller	1d5229ee8c	vec3_not: be consistent with compile-time behavior here	2013-11-12 16:05:54 +01:00
Wolfgang Bumiller	eee7d0aac1	vec3_pbool: or, not and	2013-11-12 15:56:04 +01:00
Wolfgang Bumiller	de12a24bc9	folder: fix -fperl-logic	2013-11-12 14:39:25 +01:00
Wolfgang Bumiller	6ea88c9a3f	subtracting != adding	2013-11-12 14:31:11 +01:00
Wolfgang Bumiller	de14d514f3	fix 0-x being turned into x	2013-11-06 16:57:04 +01:00
Wolfgang Bumiller	b1016c7f48	fold_binary now used instead of ast_binary_new, which calls fold_superfluous	2013-10-25 13:40:31 +02:00
Dale Weiler	8699053887	Fix handling on intrinsic folding, this closes #118.	2013-10-17 00:14:42 -04:00
Dale Weiler	eca82511c6	divVerent suggestion to use memcmp here	2013-10-14 22:40:36 -04:00
Dale Weiler	6d8d7ee923	Make divVerent happy about handling divison by zero/inf/nan and negitive versions. This code now assumes IEEE 754.	2013-10-14 22:31:37 -04:00
Dale Weiler	300fb9905b	Fix fold_superfluous	2013-10-04 07:10:58 -04:00
Dale Weiler	cc69370575	Another peephole optimization which removes superfluous expressions such as (A + 0), (A - 0), (A * 1) and (A / 1).	2013-10-04 06:46:54 -04:00
Dale Weiler	15b0555546	Implement constant folding on ternary operations via fold_cond.	2013-09-26 06:51:49 -04:00
Dale Weiler	f25fff1e3d	Remove debug printfs	2013-09-18 19:18:10 -04:00
Dale Weiler	e18849fa42	Generate (non_const_float * (1.0 / constant_float)) for (non_const_float / constant_float) expressions.	2013-09-18 19:15:24 -04:00
Dale Weiler	64661f54ea	Implemented >< (vector cross product operator). Currently support for constants only.	2013-08-31 13:41:25 -04:00
Dale Weiler	69252071ba	Prevent divide by zero for / and % operations in constant folding (previously caused compiler SIGFPE), instead "inf" is generated for both cases. This closes #124	2013-08-31 12:57:24 -04:00
Dale Weiler	a50635bcd7	intrinsic folding cleanups (and improvements.)	2013-08-30 07:12:16 -04:00
Dale Weiler	51ef277e21	Fix folding logic for conditions.	2013-08-29 00:18:48 -04:00
Dale Weiler	3b4a5667ea	Constant fold intrinsics if their arguments are constant. TODO: reference count intrinsics such that they're not generated unless they're used, currently when an intrinsic can be folded-away it's marked for generation and makes it to the final output binary even though it isn't used.	2013-08-28 12:46:22 -04:00
Dale Weiler	ee428b9081	Fix fold-dce for if(0)	2013-08-28 10:39:48 -04:00
Rudolf Polzer	e50b7a2719	Support vector bitor/bitand/bitxor. ... Signed-off-by: Rudolf Polzer <divverent@xonotic.org>	2013-08-26 18:32:28 +02:00
Dale Weiler	c64005966f	Remove some trailing whitespace	2013-08-17 23:43:41 +00:00
Dale Weiler	a622d5163b	Rework some build stuff for better output and to enable strict prototypes	2013-08-16 03:28:02 +00:00
Dale Weiler	91c7209146	Track constant folds in opts_optimization list .. this could be handled better I assume.	2013-08-15 03:45:50 +00:00
Dale Weiler	f82097b6b8	Update deps	2013-08-14 06:12:43 +00:00
Dale Weiler	90e2e5a4ad	Some typos	2013-08-14 02:07:12 +00:00
Dale Weiler	7669a99c7f	increment the optimization counter for DCE'd folds, also enable -O3 for compilation.	2013-08-01 14:24:10 +00:00
Dale Weiler	db69d14995	Disable the macros to keep clang happy (just uncomment them for when more constant folding optimizations need them)	2013-08-01 07:12:21 +00:00
Dale Weiler	6f749d61b1	Added -Oconst-fold-dce (dead code elimination optimization for when constant expressions form the basis of the dead code, i.e if else with constant expression). Fixed deps and added documentation. Cleaned up folds for ir_value (can now use the same macros as the ast_value ones).	2013-08-01 07:07:59 +00:00
Dale Weiler	10b75fd8b9	Move const-branch-elision into fold.c	2013-07-31 19:34:38 +00:00
Dale Weiler	960cb7034a	The problem was so subtle	2013-07-31 18:53:00 +00:00
Dale Weiler	3fa74da2b5	Fixes	2013-07-31 17:24:32 +00:00
Dale Weiler	5f2b7e3d57	fixed vector ops constant folding.	2013-07-31 17:05:43 +00:00
Dale Weiler	d0ee56f25f	more fixes	2013-07-31 16:31:45 +00:00
Dale Weiler	1538e69f93	Fixes	2013-07-31 15:56:56 +00:00
Dale Weiler	fa5ad1212e	Operator constant folding rewrite almost complete, just need to track down why two tests are failing.	2013-07-31 15:49:45 +00:00
Dale Weiler	86adb94d7d	folding for lteqgt (less than equal to or greater than) operator a.k.a <=> which maps values to -1,0,1 depending on the result (think of the result as a troolan) .. it's a perl thing :P	2013-07-31 12:59:34 +00:00