28419 commits

This branch

This branch

All branches

Author	SHA1	Message	Date
Douglas Katzman	da7d90ca9a	Optimize SSET-INTERSECTION ... Impose a stricter upper bound on consing as explained in the comments. Test cases by Gemini	2026-07-19 11:33:20 -04:00
Stas Boukarev	0843efc927	x86-64, utf8-sap-to-string: remove redundant moves	2026-07-19 07:49:35 +03:00
Douglas Katzman	26a27bc8d8	Optimize SSET-DIFFERENCE ... Iterating over the second set is preferred as it avoids an intermediate list, and is worse only if the second set is much larger than the first. Test cases by Gemini	2026-07-18 23:06:28 -04:00
Stas Boukarev	246af00412	Don't signal errors for (log qnan) ... Fixes lp#2160268	2026-07-19 02:10:15 +03:00
Douglas Katzman	3751ecd356	Restore the ability to use SSETs as CONSETs ... The bit-vector choice is still the default, however now we can run experiments to see if an adaptive approach would make sense. First of all most consets are small (< 64 bits) so they really could use just 1 word in the CONSET struct instead of a separately allocated bit-vector. Over 64 bits the bit-vector is ok but only up to a certain size. Eventually the vector waste so much more space compared to an SSET because most CONSETs in large universes are not dense.	2026-07-18 17:10:26 -04:00
Stas Boukarev	3b5740a297	x86-64, utf8-sap-to-string: use a single ymm register for two xmm	2026-07-18 13:58:34 +03:00
Stas Boukarev	9939d74866	Use a memory operand instead of loading into a register, 2	2026-07-18 13:44:16 +03:00
Stas Boukarev	88b11e8094	Use a memory operand instead of loading into a register	2026-07-18 13:34:21 +03:00
Stas Boukarev	121f13c48b	Correct #-	2026-07-18 06:09:49 +03:00
Stas Boukarev	eb3e4d50f8	x86-64: simd encoding of the whole utf8 range for C strings	2026-07-18 05:28:37 +03:00
Stas Boukarev	094a492c82	Add utf8 strlen test	2026-07-18 05:28:05 +03:00
Stas Boukarev	d6129e1920	Add tests for simd-copy-ascii	2026-07-18 05:28:05 +03:00
Douglas Katzman	d3fa62156f	Limit the scope of warning suppressions ... * Rev 906a406a quieted all compiler notes coming from make-host-1, rendering moot the muffle-conditions code-deletion-note in shared.lisp, not to mention (satisfies unable-to-optimize-note-p) there as well. * Prevent introduction of new occurrences of the optional+key warning by confining the suppression to precisely where there is old code. We need to fix whatever is wrong with cross-typep so that make-host-2 correctly utilizes the &OPTIONAL-AND-&KEY-IN-LAMBDA-LIST condition.	2026-07-17 23:59:27 +00:00
Douglas Katzman	9c82bf8bdc	Add missing feature to arm64 crossbuild test	2026-07-17 22:18:51 +00:00
Douglas Katzman	9bc6ee5903	Silence a bunch of gcc warnings ... of this variety warning: suggest parentheses around ‘-’ in operand of ‘&’	2026-07-17 21:43:12 +00:00
Douglas Katzman	940bb873d1	build-all-cores	2026-07-17 21:06:42 +00:00
Douglas Katzman	81c43250dc	Heed the optional+key style warning	2026-07-17 20:58:00 +00:00
Douglas Katzman	573b76f069	Revert do-conset-elements to older logic on 3 backends	2026-07-17 20:17:33 +00:00
Douglas Katzman	0aeedbcaf6	mips: Implement integer-length vop without looping	2026-07-17 19:51:05 +00:00
Douglas Katzman	e4cd2e2548	Improve DO-CONSET-ELEMENTS a little more	2026-07-17 17:37:25 +00:00
Stas Boukarev	13527b1f60	x86-64: simd decoding of the whole unicode range for C strings	2026-07-17 07:05:36 +03:00
Stas Boukarev	3f1f44a4be	x86-64: support byte-array inline constants	2026-07-17 04:55:59 +03:00
Stas Boukarev	b30453bed8	arm64: speed up length calculation in copy-character-string-to-utf8	2026-07-17 04:29:33 +03:00
Stas Boukarev	0c48a0eec1	arm64, utf8-sap-to-character-string: reduce instruction dependencies	2026-07-17 03:06:41 +03:00
Stas Boukarev	a6183f3c0d	arm64: better bit packing in copy-utf8-sap-to-character-string	2026-07-16 21:59:16 +03:00
Douglas Katzman	8796281182	Fix arena-related crash	2026-07-16 15:46:05 +00:00
Stas Boukarev	e3c2db021f	Don't fmakunbound functions present in *handled-conditions* ... Otherwise they themselves can't be compiled.	2026-07-16 18:01:32 +03:00
Stas Boukarev	eda10e3e57	Fix a compile-time type error on bad arguments to concatenate ... Fixes lp#2160747	2026-07-16 17:24:54 +03:00
Douglas Katzman	b1c1f00f62	Add yet another way to run the parallelized build of cold-sbcl.core ... Using a Bazel rule - which we can certainly commit to our git repo though probably nobody will use it - I was able to get the end-to-end build time excluding contribs down to 30 seconds using 300 execution machines. The usual caveat pertains about *DERIVE-FUNCTION-TYPES* not informing subsequent files of learned information. But it's great for making small algorithmic changes and running some basic smoke tests.	2026-07-16 03:06:24 +00:00
Douglas Katzman	6bd776dee6	Remove a line of output from genesis ... Also fix some stype warnings if muffler gets reloaded	2026-07-16 02:56:38 +00:00
Stas Boukarev	3025a181b0	Replace a MUL with USHL	2026-07-16 04:45:53 +03:00
Stas Boukarev	0e34c1c4a8	More optimizations in copy-utf8-sap-to-character-string	2026-07-16 04:16:54 +03:00
Stas Boukarev	ab04e16dd5	Micro-optimizations in copy-utf8-sap-to-character-string	2026-07-15 06:49:44 +03:00
Stas Boukarev	280e13f233	arm64: simd decoding of the whole unicode range for C strings	2026-07-15 04:12:18 +03:00
Stas Boukarev	4d6772fd1b	arm64: support byte-array inline constants ... Fewer instructions to load. Saves registers on not having to pin in a descriptor register.	2026-07-15 00:32:12 +03:00
Douglas Katzman	2d6b8ff746	Move assembly step first in make-host-2 ... In my attempt to infer a somewhat graph-based compilation order rather than simply a linear order, my build was telling me that the assembly routines depended on output/ucd/* because they follow (by a lot) target-char linearly and therefore must depend on anything that target-char does. Obviously that's wrong. Putting the asm files first proves that they depend on no other file. I would not have been surprised if this failed the crossbuild test because some asm file depends on a constant that isn't defined until later, but it seems they're actually all fine this way.	2026-07-14 17:40:47 +00:00
Douglas Katzman	83c7aafa19	Delete work files without pre-testing for existence	2026-07-14 15:18:42 +00:00
Stas Boukarev	6927ec50b9	arm64: encode more vector shift instructions	2026-07-14 00:43:39 +03:00
Douglas Katzman	809276aedd	Avoid zillons of lstat() calls from %sbcl-homedir-pathname	2026-07-13 19:58:13 +00:00
Douglas Katzman	cfbeebabb3	Set *derive-function-types* to T in make-contrib	2026-07-13 16:08:32 +00:00
Hayley Patton	90ce44aad5	Get the encodable ranges right for SAP-REF-{8,16,32}	2026-07-12 22:51:38 +03:00
Stas Boukarev	99af116141	arm64: encode prfm	2026-07-12 22:51:38 +03:00
Stas Boukarev	3bd658d129	arm64: better copy-character-string-to-utf8-byte-array ... Use a single TBL instead of multiple BSL and multiple different TBLs.	2026-07-12 22:51:38 +03:00
Douglas Katzman	8aa85ec581	Enable *derive-function-types* in PCL build ... Trade one hack for another basically	2026-07-12 19:13:27 +00:00
Stas Boukarev	ba08d271de	source-transform-union-typep: eql complexes are not numeric-types ... Fixes lp#2160429	2026-07-11 21:18:44 +03:00
Stas Boukarev	14d60e26b1	Avoid using avx512 registers in generic routines ... Which forces avx512-capable allocation tramps. Fixes lp#2160413	2026-07-11 18:03:41 +03:00
Douglas Katzman	173185836f	Improve SSET capacity tracking and SSET-ADJOIN ... * Since tombstones do not exist, there is no reason to separately record the available cell count and in-use element count. * SSET-ADJOIN need not eagerly grow the SSET except when the vector length is 0. Only if it discovers that it must insert and the capacity is reached should it call SSET-GROW. * SSET-GROW can avoid some conditional branching now that it won't be called with a vector length of 0. Test cases and some assistance by Gemini	2026-07-10 21:11:40 +00:00
Douglas Katzman	eba8865cfd	Add test manifest entry	2026-07-10 15:23:54 +00:00
Douglas Katzman	1ce38f9567	Change SSET to not use tombstones ... Avoids 1 comparison and branch in the probe loop	2026-07-10 01:15:48 +00:00
Stas Boukarev	81229e243f	Optimize instructions in copy-character-string-to-utf8-byte-array	2026-07-10 02:47:17 +03:00