Commit Graph

8401 Commits

Author SHA1 Message Date
Tomasz Dąbrowski c8341dbf7c fix fmt_fp.c(178) : warning C4244: 'return' : conversion from 'ptrdiff_t' to 'int', possible loss of data 2017-08-25 14:10:43 +02:00
Tomasz Dąbrowski 5d099e40e6 fix fmt_fp.c(123) : warning C4244: 'return' : conversion from 'ptrdiff_t' to 'int', possible loss of data 2017-08-25 14:10:24 +02:00
Yukihiro "Matz" Matsumoto 63f10a1ee5 Merge pull request #3788 from christopheraue/fix_codegen_op_send_nregs
Fixed register windows of OP_SENDs generated during codegen
2017-08-24 14:38:03 +09:00
Christopher Aue 3366894a40 Fixed register windows of OP_SENDs generated by NODE_{DREGX,REGX}; ref #3783 2017-08-23 18:16:05 +02:00
Christopher Aue 499498a6a7 Tested register windows of OP_SENDs generated by NODE_{DXSTR,XSTR}; ref #3783 2017-08-23 18:15:54 +02:00
Christopher Aue 84d8ecc1d7 Fixed register windows of OP_SENDs generated by NODE_OP_ASGN; ref #3783 2017-08-23 18:15:44 +02:00
Christopher Aue 4aca9ef98e Refactored code around generation of OP_SEND in NODE_OP_ASGN 2017-08-23 18:15:38 +02:00
Christopher Aue 1221f7676e Fixed register windows of OP_SENDs generated by NODE_{FOR,SYMBOLS}; ref #3783 2017-08-23 18:15:27 +02:00
Christopher Aue c040c8d92b Tested register windows of OP_SENDs generated by NODE_{RESCUE,HASH,ALIAS}; ref #3783 2017-08-23 18:15:08 +02:00
Christopher Aue c6596cf26a Fixed register windows of OP_SENDs generated by NODE_{SCALL,CASE,YIELD,UNDEF}; ref #3783 2017-08-23 18:12:35 +02:00
Christopher Aue 64e5208404 Asserted bidx < ci->nregs for OP_SEND and OP_SUPER 2017-08-23 15:18:06 +02:00
Yukihiro "Matz" Matsumoto 4c9a604877 Added method cache.
To enable method cache, define `MRB_METHOD_CACHE` or
`MRB_METHOD_CACHE_SIZE`. The cache size must be power of 2.
The default cache size is 128.

The measurement:
I measured simple benchmarks found in benchmark/ directory. With
method cache enabled, we gained 6-8% performance improvement, with
97-99% cache hit rate.
2017-08-22 22:17:01 +09:00
Yukihiro "Matz" Matsumoto a65f4dbb32 Remove possible path that leaves a local variable uninitialized. 2017-08-22 15:50:08 +09:00
Yukihiro "Matz" Matsumoto db437b04a4 mrb_obj_respond_to to use mrb_method_search_vm. 2017-08-22 14:23:46 +09:00
Yukihiro "Matz" Matsumoto 0d865a77d5 c (mrbc_context) may be NULL; fix #3787 2017-08-22 09:47:10 +09:00
Yukihiro "Matz" Matsumoto ab3b6dbd2a strlen returns size_t; need to cast before assigning to int. 2017-08-22 08:14:09 +09:00
Yukihiro "Matz" Matsumoto 3f90af6b42 (Try to) fix mixture of int and size_t in UTF-8 conversion.
This only effects VC.
2017-08-22 07:35:51 +09:00
Yukihiro "Matz" Matsumoto a7d611fb23 Reduce signed/unsigned warnings in dump.c 2017-08-19 21:50:36 +09:00
Yukihiro "Matz" Matsumoto 5d0fc7e57a Reduce signed/unsigned warnings in fmt_fp.c 2017-08-19 17:09:30 +09:00
Yukihiro "Matz" Matsumoto 8cea1f8907 Use ptrdiff_t instead of int. 2017-08-19 15:51:08 +09:00
Yukihiro "Matz" Matsumoto a117ec771b Remove mixed signed/unsigned comparison in debug.c. 2017-08-19 14:42:30 +09:00
Yukihiro "Matz" Matsumoto 834523945e Replace stack pop() by push_n(2); pop_n(3); fix #3783
To calculate correct register windows size. The fix was suggested
by Christopher Aue.
2017-08-19 11:56:04 +09:00
Yukihiro "Matz" Matsumoto db098d3e0e Type check before traversing irep->outer; fix #3782 2017-08-19 11:39:42 +09:00
Yukihiro "Matz" Matsumoto 2294ebdd44 Call mrb_full_gc before gc_each_objects; ref #3782
Otherwise dead object may be seen from `#each_object`.
2017-08-19 11:38:11 +09:00
Yukihiro "Matz" Matsumoto 9146d5c34f Zero width unshift should not touch memory; ref #3780 2017-08-19 11:11:26 +09:00
Yukihiro "Matz" Matsumoto 8b3298b2d6 Unshift pointer move size was wrong (not len but alen); fix #3780 2017-08-19 11:08:46 +09:00
Yukihiro "Matz" Matsumoto 29ad0a7964 capacify' for mrb_str_new_capa should not be 0`. 2017-08-18 23:42:34 +09:00
Yukihiro "Matz" Matsumoto a8bd06904d mrb_str_cat: capa should not be zero to avoid infinite loops. 2017-08-18 23:42:34 +09:00
Yukihiro "Matz" Matsumoto a06adb176f mrb_str_cat: capa should be bigger than total. 2017-08-18 23:42:33 +09:00
Yukihiro "Matz" Matsumoto 8b5be554c7 ARY_CAPA handles capacity for embedded arrays by itself. 2017-08-18 23:42:33 +09:00
Yukihiro "Matz" Matsumoto 8fbc4bef55 Reduce signed/unsigned comparison warnings; ref #3785 2017-08-18 22:54:59 +09:00
Yukihiro "Matz" Matsumoto f1767fd079 Separate mrb_str_buf_new and mrb_str_new_capa.
`mrb_str_buf_new` is an old function that ensures capacity size of
`MRB_STR_BUF_MIN_SIZE` minimum. Usually one need to use
`mrb_str_new_capa` instead.
2017-08-18 22:17:48 +09:00
Yukihiro "Matz" Matsumoto 722d123b87 Remove code duplication in mrb_str_concat. 2017-08-18 22:17:48 +09:00
Yukihiro "Matz" Matsumoto 198c898817 Rename mrb_str_concat2 to mrb_str_concat_m.
According to the naming convention a function that implements
a method should be suffixed by `_m`.
2017-08-18 22:17:47 +09:00
Yukihiro "Matz" Matsumoto 4c25738ac7 Merge str_buf_cat and mrb_str_cat. 2017-08-18 22:17:47 +09:00
Yukihiro "Matz" Matsumoto 2adb0c201e Merge pull request #3786 from christopheraue/3784_fix
Reset ci in OP_SUPER after potential realloc
2017-08-18 22:17:33 +09:00
Yukihiro "Matz" Matsumoto 11c3ad999e Merge pull request #3781 from clayton-shopify/fix-hash-compact-bang
Check whether internal khash is initialized in Hash#compact!
2017-08-18 22:13:33 +09:00
Yukihiro "Matz" Matsumoto f229ed917e Merge pull request #3785 from miura1729/original
Improve Array structure
2017-08-18 20:37:31 +09:00
Miura Hideki 75ec86ed00 Improve Array structure 2017-08-18 20:12:00 +09:00
Christopher Aue 6b0860ce61 Reset ci in OP_SUPER after potential realloc 2017-08-18 13:11:08 +02:00
Clayton Smith 5dc688fb12 Check whether internal khash is initialized in Hash#compact! 2017-08-17 12:13:41 -04:00
Yukihiro "Matz" Matsumoto 0789ec71e0 Merge pull request #3779 from christopheraue/vm_refactoring
Some more vm refactorings
2017-08-13 07:11:21 +09:00
Christopher Aue d5ac785b31 Reintroduced not storing converted proc directly in the stack 2017-08-12 23:59:40 +02:00
Christopher Aue 198df61039 Removed unneeded ci->nregs checks in OP_SEND and OP_SUPER
Because of #3504 `ci->nregs = bidx+1` was introduced in b64f08784c.
This led to the follow up error #3551 whose fix introduced the `if (bidx >= ci->nregs)`
check in 071164b799 and the `stack_extend(mrb, ci->nregs)`
in 93d802987e.

Then, the code causing #3504 reappeared again in #3590. The fix for it moved the code
dealing with the block in OP_SUPER from below the `cipush`  to above the `cipush`
in d9fb8b69b0. The `if (bidx >= ci->nregs) { ... }` from
then on works with the original callinfo and not the pushed one. `ci->nregs` needed to
be modified for the pushed one because it is initialized to 0. But for the original ci
it is propertly set and a check is not needed.
2017-08-12 23:27:17 +02:00
Christopher Aue f2d4640998 Extended stack always based on ci->nregs 2017-08-12 14:47:22 +02:00
Christopher Aue 3f591d71c6 Refactored variable usage in OP_SEND and OP_SUPER 2017-08-12 14:31:18 +02:00
Yukihiro "Matz" Matsumoto 8bf492f127 Reduce integer type mismatch warnings in VC. 2017-08-12 09:35:35 +09:00
Yukihiro "Matz" Matsumoto baa5d2e3f7 Remove some empty lines; ref #3778 2017-08-11 22:39:31 +09:00
Yukihiro "Matz" Matsumoto 0828a962c0 Merge pull request #3778 from christopheraue/vm_refactoring
Refactored OP_SEND and OP_SUPER and calculate argc right at the top
2017-08-11 22:38:32 +09:00
Yukihiro "Matz" Matsumoto 42f9af41fb Merge pull request #3777 from christopheraue/super_method_missing_test
Added basic test for calling a missing method through super
2017-08-11 22:35:40 +09:00