Commit Graph

78 Commits (3dabdcbdc3b343302ab594d2b7e748f929a64fd9)

Author SHA1 Message Date
Slava Pestov e5e51c40a1 vm: implement frame-based SEH for 64-bit Windows 2010-04-08 10:32:14 -07:00
Slava Pestov 26cae00eeb Add code heap introspection primitive to VM, and make a code-blocks word in tools.memory that wraps a useful interface around this. Add a lookup-return-address word and use this in tools.disassembler.utils 2010-02-03 23:11:32 +13:00
Slava Pestov 3259ef4bec vm: modify-code-heap primitive now takes a pair of parameters, update-existing? and reset-pics?. If no generic words were changed, PICs do not need to be reset. If no existing words were redefined, the code heap doesn't have to be traced at all. Speeds up compilation of PEGs since those create lots of compilation units 2010-02-03 23:11:28 +13:00
Slava Pestov 637a77c018 Stop wearing monocle and use the term "entry point" instead of "XT" throughout VM and compiler; also remove two unused relocation types 2010-01-18 20:54:00 +13:00
Slava Pestov 63edd20a55 vm: eliminating register variables work in progress. Works on x86-32 with non-optimizing compiler 2009-12-19 10:59:56 +13:00
Doug Coleman 1454e534fe fix a couple of warnings on win64 2009-12-13 20:00:02 -06:00
Slava Pestov 066d2836ca Get modify-code-heap working if stack traces are stripped (but then it won't update existing code blocks), and rename code_heap::code_heap_free() to code_heap::free() 2009-12-02 05:27:48 -06:00
Slava Pestov 68c09f0e93 Split literal table into literal and parameter tables, literal table is discarded after code block initialization 2009-12-02 04:28:15 -06:00
Slava Pestov c6602bda27 Big cleanup of literal table and relocation-related code
- new way to iterate over instruction_operands cleans up relocation code
- move some methods out of the VM class to clean up code
- if debugging is on, die earlier if attempting to allocate inside GC
- callback heap entries are now code_blocks
- replace block_granularity with data_alignment
2009-12-02 00:48:41 -06:00
Slava Pestov bd02eef38e Fix non-tail recursive inline recursive compilation 2009-11-25 19:12:09 -06:00
Slava Pestov 4ebaf956c3 vm: non-optimizing compiler now compiles word definition quotations with the owner set to the word object 2009-11-24 22:44:04 -06:00
Slava Pestov bfd4cce364 vm: compaction now updates the needs_fixup set 2009-11-24 19:29:59 -06:00
Slava Pestov 001a3f2847 vm: mark sweep now traces code block references; rename embedded_pointer to instruction_operand 2009-11-23 19:17:12 -06:00
Slava Pestov 18a2ce1f8c vm: remove some dead code 2009-11-05 19:32:26 -06:00
Slava Pestov 72ab6ec548 vm: rewrite 'become' primitive so that it uses a slot visitor instead of GC 2009-11-05 19:03:51 -06:00
Slava Pestov 6c2c87758a vm: rename gc_root to data_root, add code_root to fix a problem where code blocks would move underneath the PIC compiler if PIC compiler allocated enough 2009-11-02 18:10:34 -06:00
Slava Pestov 5536003cf8 vm: new GC policy: mark sweep after N megabytes promoted, mark compact if tenured space cannot fit nursery+aging in one contiguous block. N = 32 on 32-bit and 64 on 64-bit 2009-10-30 02:26:57 -05:00
Slava Pestov a1aac42786 vm: clean up gc events, remove -verbosegc switch, fix compaction bug 2009-10-27 21:31:28 -05:00
Slava Pestov d95a98eb9c vm: add primitives for getting at GC events, data-room and code-room primitives now return structs instead of arrays 2009-10-27 03:32:28 -05:00
Slava Pestov 91cec17e52 vm: get GC events working, -verbosegc switch now produces more info 2009-10-26 23:57:26 -05:00
Slava Pestov e793a72060 vm: remove crummy old GC stats, split off free list code, clean up various other things 2009-10-26 22:08:35 -05:00
Slava Pestov cf247c23a2 vm: room. now prints mark stack size, and total/contiguous free space 2009-10-25 13:18:06 -05:00
Slava Pestov fae27fb361 vm: make some more ctors explicit just for kicks 2009-10-24 23:02:58 -05:00
Slava Pestov 62e718eaa9 vm: combine heap_block and code_block structs, eliminates some boilerplate 2009-10-24 21:33:31 -05:00
Slava Pestov 29a27cfde4 vm: data heap compaction work in progress 2009-10-24 04:18:33 -05:00
Slava Pestov fd1e992e7d vm: factor out code that visits object slots and code heap blocks into slot_visitor and code_block_visitor 2009-10-24 03:54:53 -05:00
Slava Pestov 814f6371d6 vm: mark sweep gc for tenured space work in progress 2009-10-20 22:20:49 -05:00
Slava Pestov f0816d72f1 vm: split off free_list_allocator from heap class, rename zone to bump_allocator 2009-10-20 15:26:11 -05:00
Slava Pestov a9dbbd1efb vm: simplify code heap by eliminating HeapLayout template parameter 2009-10-20 14:01:46 -05:00
Slava Pestov acdcb181e0 vm: working on making heap more generic 2009-10-20 10:22:06 -05:00
Slava Pestov 838a44e901 vm: change code heap layout somewhat, remove unused allocation bitmap from mark_bits 2009-10-20 09:37:24 -05:00
Slava Pestov 4ddd63d83e vm: move compaction algorithm to mark_bits.hpp since it doesn't rely on properties of heaps per se 2009-10-19 03:05:20 -05:00
Slava Pestov ac25b8ebbb vm: new mark_bits data structure replaces hashtable when compacting code heap 2009-10-19 02:21:11 -05:00
Slava Pestov 464aac14cf vm: working on new compaction implementation 2009-10-19 00:07:43 -05:00
Slava Pestov 2e65366c6b vm: rename myvm and parent_vm instance variables to parent, clean up casts in primitive definitions 2009-10-18 20:31:59 -05:00
Slava Pestov 236588208b vm: rename F to false_object, and rename T to true_object 2009-10-18 20:26:21 -05:00
Slava Pestov 107c96f642 vm: code heap compaction at runtime using compact-gc primitive 2009-10-16 11:39:35 -05:00
Slava Pestov c04412219b vm: allocate stacks and data heap segments without execute permissions 2009-10-16 05:16:12 -05:00
Slava Pestov 85b746e7b5 vm: simpler and more efficient code heap compaction; restructure a few things to get ready for runtime code heap compaction 2009-10-16 04:33:35 -05:00
Slava Pestov 0a3d08bc52 vm: make the code nice and pretty 2009-10-16 02:55:02 -05:00
Slava Pestov 45eb68fa38 vm: during tenuring stage of aging collection, if tenured space fills up, it would attempt a to_tenured collection. this will succeed if all roots were tenured. however, this is unsound, because there's now an untraced segment of tenured space. fix: if tenuring fails, go on to do a full collection instead 2009-10-15 05:51:11 -05:00
Slava Pestov 21f55ab1a3 vm: more efficient code heap remembered set 2009-10-08 23:10:32 -05:00
Slava Pestov 651a7fb6fa vm: replace generations array with individual instance variables referencing zone objects 2009-10-07 11:59:59 -05:00
Slava Pestov 5c774d8c84 vm: more code cleanups 2009-10-07 08:33:54 -05:00
Slava Pestov 697e2342d0 vm: put code block owner directly in the header, instead of as the first entry in the literal table. Reduces x86-64 image size by ~700kb, also eliminates separate 'strip' set of staging images from deploy tool 2009-10-06 06:25:07 -05:00
Slava Pestov 16c3251072 vm: free up a cell in compiled code block headers 2009-10-06 05:52:45 -05:00
Slava Pestov ed1ee19ce1 vm: add a remembered set for code blocks which may reference young literals. Improves loading time 2009-10-06 04:36:34 -05:00
Slava Pestov d10e27149c vm: cleanup 2009-10-06 01:42:17 -05:00
Slava Pestov bae2240f62 vm: refactoring garbage collector 2009-10-05 22:16:08 -05:00
Slava Pestov c4ef640f4d Big VM cleanup
- Move forward declarations of 'struct factor_vm' to one place
- Rename template parameters from T and TYPE to descriptive names. New convention: CamelCase for template parameters
- Change some higher-order functions taking function pointers into templates, and define classes overriding operator(). There's a bit of new boilerplate here but its more consistent than the old mish-mash approaches
- Put GC state into a gc_state struct
- Use exceptions instead of longjmp for non-local control transfer in GC
- In code GC, instead of interleaving code block tracing with copying, add code blocks which need to be revisited to an std::set stored in the gc_state
2009-10-03 08:47:05 -05:00