86 commits

Author	SHA1	Message	Date
Douglas Katzman	3d4fe67e29	build-all-cores	2026-09-09 16:14:36 -04:00
Douglas Katzman	e5daa56dc3	build-all-cores	2026-08-30 12:28:02 -04:00
Douglas Katzman	210ad59740	build-all-cores	2026-08-29 13:25:17 +00:00
Douglas Katzman	18f8a0c8ea	build-all-cores	2026-08-21 22:19:29 +00:00
Douglas Katzman	d1d85bbaa1	build-all-cores	2026-08-12 22:45:49 -04:00
Douglas Katzman	100c9c2bdc	Perform local common subexpression elimination for some memory loads ... The technique is to find "equivalent" loads in between which there is no computation that affects the result of the load. Also it needs a surrounding LET, which is augmented with a new temp variable as if the user did that. This is slightly deficient for various reasons: 1. it won't do anything without the LET 2. the more kinds of common subexpressions we allow (such as math), the slower IR1-OPTIMIZE-COMBINATION is going to run 3. there are possibly other node types that should be allowed to intervene On the plus side, it's not all that hard to extend the logic to accept other functions as participants in common subexpressions. All test cases plus a little bit of assistance from Gemini	2026-08-04 00:58:21 +00:00
Douglas Katzman	629a0df182	Fix failure to choose a jump table for a key type of (MOD 4) ... Ironically it was working when the key's derived type was _not_ known to be exactly (MOD 4) with a CASE over 4 keys.	2026-07-29 20:56:46 +00:00
Douglas Katzman	74f07eb8f4	Fix crossbuild-runner ... It doesn't work to omit avx512 from the darwin config which may have been inadvertent. It's suspicious that a new non-working Makefile was checked in. And I'm not sure why there are 5 features keywords to do one thing if all of them are in fact mandatory as of now.	2026-07-04 20:42:56 +00:00
Douglas Katzman	35332792be	build-all-cores	2026-03-29 23:59:15 -04:00
Douglas Katzman	223369311a	build-all-cores	2026-03-11 22:55:23 -04:00
Douglas Katzman	57ce36ce66	build-all-cores	2026-03-06 11:11:58 -05:00
Stas Boukarev	2cfca0fa32	ppc32, ppc64: remove LRA ... TODO: put return-pc-passing-location into the LR special register. flag-based multiple value returns.	2026-02-28 02:22:29 +03:00
Douglas Katzman	2baf0b0a27	build-all-cores	2026-02-24 22:09:02 -05:00
Douglas Katzman	d71bb5f2ba	build-all-cores	2026-02-21 12:11:01 -05:00
Douglas Katzman	51aa6af904	build-all-cores	2026-02-11 08:48:39 -05:00
Douglas Katzman	9ffabf0a4d	build-all-cores	2026-01-20 14:14:37 -05:00
Douglas Katzman	b6f186ac9b	build-all-cores	2026-01-14 11:09:15 -05:00
Douglas Katzman	91d76e968a	build-all-cores	2026-01-02 21:01:13 -05:00
Douglas Katzman	2df0ca31ea	build-all-cores	2025-12-04 21:33:43 -05:00
Christophe Rhodes	a8450880fd	xperfecthash updates	2025-11-30 14:51:07 +00:00
Douglas Katzman	ffb8ede1dd	build-all-cores	2025-11-06 12:36:27 -05:00
Douglas Katzman	68e9440d17	build-all-cores	2025-09-08 08:12:59 -04:00
Douglas Katzman	3b5041d5a5	build-all-cores	2025-09-01 15:41:30 -04:00
Douglas Katzman	eb7d126e85	build-all-cores	2025-08-23 01:58:42 +00:00
Douglas Katzman	84543e0050	Rebuild xperfecthash files	2025-08-15 16:46:28 +00:00
Douglas Katzman	43af8c9253	Optimize out a few instructions in perfect hash calculation ... For sets of only character or only fixnum, there is no need to guard against an object having the wrong type. So instead of doing (if (fixnump x) (ldb ... x)) just use (ldb ... (get-lisp-obj-address x)) eliminating an initial branch. Also the AND operation - at least on x86-64 - avoids use of a raw constant. This just changes some shifts in the perfect hash expression. It drops a bit (or more) of precision, but we're already doing that in descriptor-hash32 for the situations where the keyset contains a mixture of immediates and symbols.	2025-07-07 18:28:54 +00:00
Douglas Katzman	7b3816c6b1	Transform typep on frozen structure unions	2025-06-30 20:47:29 +00:00
Douglas Katzman	3d5139bef7	Use descriptor-hash32 only for a mixture of CASE key types ... Otherwise, just CHAR-CODE or just SYMBOL-NAME-HASH may be enough.	2025-06-17 00:42:43 +00:00
Douglas Katzman	21a23d809a	Use perfect hash more often in sequence transforms ... The ASSOC,MEMBER,etc functions were only handling symbols with perfect hashing. This change extends them to fixnum and character keys.	2025-06-16 20:07:45 +00:00
Douglas Katzman	ba509cda22	Implement DESCRIPTOR-HASH32 ... which is just like POINTER-HASH (removed at rev 1857b789) but with at most 32 significant bits of output. Fewer for some platforms. The vop reduces a multiway COND to an IF when computing the input to the perfect hash function in a CASE expression whose keys include {symbol|character|fixnum}. There's no need to test CHARACTERP or FIXNUMP, and 0 and #\null can both be keys of the same CASE because they hash differently now.	2025-06-16 19:52:00 +00:00
Douglas Katzman	3655d025c9	Remove extraneous whitespace in xperfecthash files	2025-06-14 01:38:39 +00:00
Douglas Katzman	3cc36fad5a	Regenerate xoerfecthash files	2025-05-02 13:47:00 +00:00
Douglas Katzman	5c9841249d	Remove :symbol-value fixup flavor	2025-04-16 12:48:50 -04:00
Stas Boukarev	7f469f58a5	Faster make-sequence on simple types. ... Don't call specifier-type on simple type specifiers.	2025-03-30 20:05:52 +03:00
Christophe Rhodes	c3168e3ebc	xperfecthash regeneration	2025-03-22 17:33:18 +00:00
Douglas Katzman	09294a674c	build-all-cores	2025-03-14 23:51:47 -04:00
Douglas Katzman	fc501b66ec	build-all-cores	2025-03-13 02:16:40 +00:00
Douglas Katzman	5d4f0ccd0c	Rebuild xperfecthash files from scratch	2025-03-10 13:50:53 +00:00
Douglas Katzman	31b0c0d03e	build-all-cores	2025-02-17 15:23:50 -05:00
Douglas Katzman	c4dcdc659a	build-all-cores	2025-01-21 14:15:27 -05:00
Douglas Katzman	f9c1797aac	build-all-cores	2025-01-08 13:18:52 +00:00
Stas Boukarev	1aef34f7ef	Redo numeric-union-type. ... Use a single type for numbers, numeric-union-type, even if it only contains one range. Makes the methods simpler. A single-range type could be optimized later. Use a single type for rationals. It's easier to normalize unions of rationals, ratios and integers. Fixes lp#308937 lp#1734959 lp#2073544	2024-12-29 17:59:38 +03:00
Stas Boukarev	fcc83f20fa	Reduce node.lisp constructor lambda lists. ... Smaller code, faster.	2024-12-19 08:47:17 +03:00
Stas Boukarev	f07ab4574d	Revert numeric-range-type ... It don't work right.	2024-12-17 07:19:22 +03:00
Stas Boukarev	41a6964128	numeric-range-type: more thorough canonicalization. ... Fixes lp#2091767	2024-12-14 14:44:56 +03:00
Stas Boukarev	21d84823f0	Add a dedicated type for unions of numeric-types. ... Intersection, union, subtypep, difference are implemented more efficiently on a vector of sorted ranges. Fixes lp#2073544	2024-12-13 03:28:23 +03:00
Stas Boukarev	554360c906	xperfecthash	2024-12-02 01:58:28 +03:00
Christophe Rhodes	5fe3559bb2	build-all-cores from empty xperfecthash files	2024-11-29 12:18:31 +00:00
Douglas Katzman	741581124f	build-all-cores	2024-11-13 23:45:01 +00:00
Douglas Katzman	d6f5d95a8f	Regenerate xperfecthash files	2024-10-18 23:31:37 +00:00