100 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
andrew	ec1c1cf43f	Run gc less frequently ... But still run it to account for the version potentially increasing	2024-10-31 15:35:00 -07:00
andrew	eaad0c69a7	Store Result* pointer instead of index	2024-10-31 12:59:36 -07:00
andrew	309e6ab816	Dispatch on pair of types	2024-10-30 22:59:00 -07:00
andrew	12b82c1be5	Fix accidentally-checked-in debug change	2024-10-30 16:44:00 -07:00
andrew	0cce9df8a8	Use constexpr int instead of sizeof	2024-10-30 15:36:46 -07:00
andrew	0df09743da	Don't assign to nextRangeWrite twice	2024-10-30 15:33:40 -07:00
andrew	c4b0aa1085	Skip over point writes in phase 3	2024-10-30 15:08:27 -07:00
andrew	051bfb05fe	Fix up DEBUG_VERBOSE	2024-10-30 14:50:34 -07:00
andrew	7e1bcbf9be	Follow forwarding pointers for end before addWriteRange	2024-10-30 14:26:37 -07:00
andrew	4e685bbc3b	Use deferred release instead of hacky dontInvalidate logic	2024-10-30 14:16:36 -07:00
andrew	b6bfc6f48d	Remove redundant int conversions	2024-10-30 11:48:55 -07:00
andrew	3b858551f3	Switch std::span<const uint8_t> to TrivialSpan	2024-10-30 11:45:56 -07:00
andrew	2c1c26bc88	Enable interleaved range writes	2024-10-30 11:01:23 -07:00
andrew	958ee15cfc	Use forwarding pointers instead of beginIsPrefix	2024-10-29 20:43:58 -07:00
andrew	9015b555de	Allow passing two roots to addWriteRange	2024-10-29 17:24:10 -07:00
andrew	7aac73ee80	Process acceptable subranges interleaved	2024-10-29 14:41:42 -07:00
andrew	c06afeb81e	Add to corpus	2024-10-29 11:59:36 -07:00
andrew	b015711b7c	Fix write metrics for interleaved implementation	2024-10-29 10:42:22 -07:00
andrew	f27ca6d6af	Better support a lack of musttail and preserve_none	2024-10-28 22:40:52 -07:00
andrew	c0bb175b7e	Remove interleaved_insert::begin	2024-10-28 22:17:52 -07:00
andrew	6a6fe5738a	Improve "complete" codegen for interleaved state machines	2024-10-28 22:15:05 -07:00
andrew	dc16eccf06	Use stack allocation for small counts	2024-10-28 21:51:20 -07:00
andrew	3f15db7e82	Avoid longestCommonPrefix if partialKeyLen == 0	2024-10-28 21:44:16 -07:00
andrew	e8a8b5aef1	Implement phase 2	2024-10-28 16:30:06 -07:00
andrew	b8fefff3ba	Remove accidentally-checked-in debug print	2024-10-28 16:20:51 -07:00
andrew	2706b2f65e	Implement "phase 1" of interleaved point writes	2024-10-28 16:02:56 -07:00
andrew	f1292efe41	Add getChildUpdatingMaxVersion	2024-10-28 15:31:00 -07:00
andrew	a2d3d269ec	Add check namespace ... Prepare to add write namespace	2024-10-28 13:57:56 -07:00
andrew	8ff7a112b7	Add forwarding pointers + deferred release	2024-10-28 12:34:20 -07:00
andrew	cf25b8626c	Change server bench to point write heavy workload	2024-10-27 22:28:03 -07:00
andrew	e025f934d8	Fix gc-too-slow issue ... Previously, with a finite set of keys, latest_extant_version could fall all the way behind so that a call to addWrites would have to do O(n) work eventually.	2024-10-27 21:13:27 -07:00
andrew	e5452c0f7d	Only check against sequential if we interleaved	2024-10-17 16:17:21 -07:00
andrew	66fc526a55	Add to corpus	2024-10-17 16:07:15 -07:00
andrew	21f08b9f88	Avoid assigning to continuation where unnecessary	2024-10-17 13:16:29 -07:00
andrew	10c2f06199	Mix type dispatch and interleaving for down left spine	2024-10-17 13:13:16 -07:00
andrew	5cf04e9718	Add to corpus	2024-10-17 10:58:02 -07:00
andrew	707b220fbc	Improve next{Physical,Logical} codegen more	2024-10-17 09:52:03 -07:00
andrew	fd39065498	Add to corpus	2024-10-16 16:38:56 -07:00
andrew	b963d481c9	Improve codegen for nextLogical	2024-10-16 15:54:05 -07:00
andrew	e7ed47e288	Assert that interleaved and sequential report same metrics	2024-10-15 20:58:13 -07:00
andrew	04f138109b	Rename variables from tls	2024-10-15 20:45:11 -07:00
andrew	a0d07dd40c	Remove no-longer-necessary GCOVR_EXCL_LINE ... Now that we exclude noncode lines (why isn't that the default?)	2024-10-15 18:33:40 -07:00
andrew	7fb408b466	Remove tailcall test ... We no longer rely on tail call optimization happening where musttail is not supported	2024-10-15 17:47:36 -07:00
andrew	6d265acfc7	Install llvm-cov	2024-10-15 17:47:01 -07:00
andrew	67a61513b8	Use clang for coverage build	2024-10-15 17:34:30 -07:00
andrew	583f2e7612	Check interleaved and sequential in debug mode	2024-10-15 17:31:06 -07:00
andrew	66e5b033c0	Fix control flow reaches end of non-void function warning	2024-10-15 17:27:36 -07:00
andrew	1d705cd4b7	Add to corpus	2024-10-15 17:12:45 -07:00
andrew	769cf8de9a	Bring back sequential read implementation ... This will be used if there's only one check to perform or if the compiler does not support musttail and preserve_none	2024-10-15 17:12:11 -07:00
andrew	84942a5bf8	Try clang 20	2024-10-15 15:59:58 -07:00
andrew	7ad6872ee8	Use clang-19 in jenkins	2024-10-15 15:55:33 -07:00
andrew	9db5eb960d	Suppress some unhelpful warnings	2024-10-15 15:06:10 -07:00
andrew	5df25a138a	Don't initialize max version if child doesn't exist	2024-10-15 12:55:28 -07:00
andrew	381fbce0c0	Remove duplicate clang build	2024-10-14 22:32:01 -07:00
andrew	87aeb349a3	Use clang for building packages	2024-10-14 22:23:56 -07:00
andrew	28fb0d7faa	Cross-compile to aarch64 with clang	2024-10-14 22:19:50 -07:00
andrew	5013c689a0	Remove dead code	2024-10-14 22:06:48 -07:00
andrew	316bbf679f	Add to corpus	2024-10-14 21:15:45 -07:00
andrew	58aabe83f5	Improve codegen for tracking check metrics	2024-10-14 17:50:38 -07:00
andrew	0c8a051913	Avoid more branches on type	2024-10-14 17:16:56 -07:00
andrew	11e8717da8	Specific node overloads for checkMaxBetweenExclusive	2024-10-14 16:54:44 -07:00
andrew	824037bf32	Remove dead code	2024-10-13 21:28:27 -07:00
andrew	bbe964110e	Disable skip list tests by default	2024-10-13 20:42:14 -07:00
andrew	100449c76c	Add to corpus	2024-10-13 20:39:39 -07:00
andrew	51b5f638a4	Combine terminal down_left_spine paths	2024-10-13 20:39:07 -07:00
andrew	767dacc742	Fully interleaved range read ... Would be great to find a way to make this all more readable	2024-10-13 20:05:06 -07:00
andrew	978a7585b6	Remove backtrack label from checkRangeRightSide	2024-10-13 11:27:36 -07:00
andrew	71b3c7fb7f	Stop tracking first iteration ... We can tell with other means	2024-10-12 23:56:01 -07:00
andrew	420f50c40f	Stop tracking searchPathLen	2024-10-12 23:34:46 -07:00
andrew	69a131df38	Prepare for interleaving checkRangeRightSide	2024-10-12 23:01:47 -07:00
andrew	8a4032e850	Remove dead code	2024-10-12 18:00:19 -07:00
andrew	9c365435ea	Add to corpus	2024-10-12 17:39:27 -07:00
andrew	8eb5e76336	Add short-circuiting back to range checks	2024-10-12 17:20:22 -07:00
andrew	e8982074f2	Add short-circuiting back for range common prefix search	2024-10-12 17:10:19 -07:00
andrew	f60833a57f	Add short-circuiting back in for point and prefix checks	2024-10-12 17:05:47 -07:00
andrew	47fd811efc	Fix new clang warning	2024-10-12 16:42:52 -07:00
andrew	73f93edf49	Remove dead code	2024-10-12 12:56:13 -07:00
andrew	8bac1f66fc	Remove searchPathLen	2024-10-12 11:12:12 -07:00
andrew	352c07cbc9	Use CheckJob::setResult	2024-10-12 00:01:35 -07:00
andrew	2e7e357355	Remove dead code	2024-10-11 23:47:32 -07:00
andrew	147f5af16b	Add to corpus	2024-10-11 23:35:51 -07:00
andrew	323b239411	Stop applying flatten attribute ... This blows up the code size too much. I wonder if having all those distinct possible program counters helps libfuzzer /shrug	2024-10-11 21:31:18 -07:00
andrew	54c7ccb96b	Allow for easily increasing number of range reads ... For worst case for radix tree benchmark	2024-10-11 21:20:45 -07:00
andrew	6a12210866	Make better use of static node type info for check left side	2024-10-11 21:06:57 -07:00
andrew	416504158e	Interleave checkRangeLeftSide	2024-10-11 20:56:40 -07:00
andrew	b0bc68a14e	Move checkRangeStartsWith out of finishCheckRangeRead ... This is incrementally closer to where we want to be I think	2024-10-11 17:52:04 -07:00
andrew	0de85ecda0	Add done_common_prefix_iter	2024-10-11 17:33:31 -07:00
andrew	44afb8be00	Interleave the common prefix search in range check	2024-10-11 17:26:34 -07:00
andrew	ecdbaaf2c1	Add missing MUSTTAIL	2024-10-11 16:51:00 -07:00
andrew	2c253c29b5	Add to corpus	2024-10-11 16:48:13 -07:00
andrew	fe9678787d	Only require coverage of node-specific overloads of getFirstChildExists	2024-10-11 16:47:19 -07:00
andrew	0ac259c782	Interleaved implementation for checking prefix reads	2024-10-11 16:04:55 -07:00
andrew	8b1a0afc58	First continuation for interleaved check range read	2024-10-11 15:45:25 -07:00
andrew	2018fa277c	Use a trampoline if musttail not available ... For the previous approach to work, we would have had to lock in the generated code for every function that needs to make tail calls in assembly.	2024-10-10 17:30:16 -07:00
andrew	1faeb220d5	Implement keepGoing in asm if musttail is not available	2024-10-10 16:52:27 -07:00
andrew	0dc657bfeb	Fix arm build	2024-10-10 16:07:28 -07:00
andrew	b51ef97c71	Fix USE_SIMD_FALLBACK build	2024-10-10 15:56:16 -07:00
andrew	31ad3e8e1c	Add to corpus	2024-10-10 13:21:26 -07:00
andrew	e213237698	Remove some dead code	2024-10-10 13:21:13 -07:00
andrew	a1c61962a1	Workaround lack of musttail in gcc	2024-10-10 13:20:50 -07:00