Bug 197979

Summary:

[JSC] Implement op_wide16 / op_wide32 and introduce 16bit version bytecode

Product:

WebKit

Reporter:

Yusuke Suzuki <ysuzuki>

Component:

JavaScriptCore

Assignee:

Yusuke Suzuki <ysuzuki>

Status:

RESOLVED FIXED

Severity:

Normal

CC:

benjamin, cdumez, chris.reid, cmarcelo, dbates, ews-watchlist, fpizlo, keith_miller, mark.lam, msaboff, rniwa, saam, tzagallo, webkit-bug-importer

Priority:

Keywords:

InRadar

Version:

WebKit Nightly Build

Hardware:

Unspecified

OS:

Unspecified

Bug Depends on:

Bug Blocks:

198173, 197940

Attachments:

Description	Flags
wip	none
Patch	none
Patch	none
Patch	none
Patch	none
Archive of layout-test-results from ews101 for mac-highsierra	none
Archive of layout-test-results from ews116 for mac-highsierra	none
Archive of layout-test-results from ews106 for mac-highsierra-wk2	none
Archive of layout-test-results from ews125 for ios-simulator-wk2	none
Patch	none
Patch	none
Archive of layout-test-results from ews116 for mac-highsierra	none
Patch	none
Archive of layout-test-results from ews213 for win-future	none
Patch	none
Patch	none
Patch	none
Patch	none
Archive of layout-test-results from ews214 for win-future	none
Patch	none
Patch	none
Archive of layout-test-results from ews211 for win-future	none
Archive of layout-test-results from ews210 for win-future	none
Patch	none
Archive of layout-test-results from ews214 for win-future	none
Patch	none
Patch	none
Patch	none
Patch	none
Patch	none
Patch	none
Patch	none
Patch	fpizlo: review+
Patch	none
Patch	none

Yusuke Suzuki

Reported 2019-05-17 01:46:03 PDT

The data in Gmail indicates that it is profitable.

Attachments
wip (62.39 KB, patch) 2019-05-17 11:26 PDT, Tadeu Zagallo	no flags	Details Formatted Diff Diff
Patch (66.81 KB, patch) 2019-05-18 19:20 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (66.81 KB, patch) 2019-05-18 19:36 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (72.69 KB, patch) 2019-05-19 20:52 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (76.61 KB, patch) 2019-05-21 03:04 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Archive of layout-test-results from ews101 for mac-highsierra (3.12 MB, application/zip) 2019-05-21 04:14 PDT, EWS Watchlist	no flags	Details
Archive of layout-test-results from ews116 for mac-highsierra (524.78 KB, application/zip) 2019-05-21 04:18 PDT, EWS Watchlist	no flags	Details
Archive of layout-test-results from ews106 for mac-highsierra-wk2 (2.69 MB, application/zip) 2019-05-21 04:30 PDT, EWS Watchlist	no flags	Details
Archive of layout-test-results from ews125 for ios-simulator-wk2 (8.05 MB, application/zip) 2019-05-21 05:02 PDT, EWS Watchlist	no flags	Details
Patch (79.63 KB, patch) 2019-05-22 19:17 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (79.47 KB, patch) 2019-05-22 19:20 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Archive of layout-test-results from ews116 for mac-highsierra (531.23 KB, application/zip) 2019-05-22 20:33 PDT, EWS Watchlist	no flags	Details
Patch (83.16 KB, patch) 2019-05-22 22:17 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Archive of layout-test-results from ews213 for win-future (13.59 MB, application/zip) 2019-05-23 01:22 PDT, EWS Watchlist	no flags	Details
Patch (83.16 KB, patch) 2019-05-23 13:14 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (84.94 KB, patch) 2019-05-23 15:05 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (85.77 KB, patch) 2019-05-23 15:52 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (85.81 KB, patch) 2019-05-23 17:27 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Archive of layout-test-results from ews214 for win-future (13.41 MB, application/zip) 2019-05-23 20:14 PDT, EWS Watchlist	no flags	Details
Patch (85.83 KB, patch) 2019-05-24 17:06 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (83.97 KB, patch) 2019-05-24 18:45 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Archive of layout-test-results from ews211 for win-future (13.19 MB, application/zip) 2019-05-24 23:40 PDT, EWS Watchlist	no flags	Details
Archive of layout-test-results from ews210 for win-future (13.57 MB, application/zip) 2019-05-25 00:41 PDT, EWS Watchlist	no flags	Details
Patch (85.30 KB, patch) 2019-05-27 05:05 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Archive of layout-test-results from ews214 for win-future (13.66 MB, application/zip) 2019-05-27 07:50 PDT, EWS Watchlist	no flags	Details
Patch (99.90 KB, patch) 2019-05-28 00:18 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (100.07 KB, patch) 2019-05-28 00:25 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (100.37 KB, patch) 2019-05-28 01:52 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (106.01 KB, patch) 2019-05-28 14:17 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (105.95 KB, patch) 2019-05-28 15:24 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (109.43 KB, patch) 2019-05-28 16:10 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (105.99 KB, patch) 2019-05-28 16:11 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (105.99 KB, patch) 2019-05-28 18:39 PDT, Yusuke Suzuki	fpizlo: review+	Details Formatted Diff Diff
Patch (285.20 KB, patch) 2019-05-29 22:21 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Patch (288.98 KB, patch) 2019-05-30 02:24 PDT, Yusuke Suzuki	no flags	Details Formatted Diff Diff
Show Obsolete (24) View All Add attachment proposed patch, testcase, etc.

Tadeu Zagallo

Comment 1 2019-05-17 11:26:05 PDT

Created attachment 370128 [details] wip Here's the code I had, in case it helps. It passed tests on x64 and built for i386 and arm, but I didn't test any further.

Yusuke Suzuki

Comment 2 2019-05-18 19:20:58 PDT

Created attachment 370215 [details] Patch WIP, update 32bit wide16 things, add loadhsp

Yusuke Suzuki

Comment 3 2019-05-18 19:36:49 PDT

Created attachment 370216 [details] Patch WIP, add code for 32bit

Tadeu Zagallo

Comment 4 2019-05-18 22:52:36 PDT

(In reply to Yusuke Suzuki from comment #2) > Created attachment 370215 [details] > Patch > > WIP, update 32bit wide16 things, add loadhsp Did you see crashes with `loadhs`? I had updated it to use :quad for the destination operand on 64-bit, which is the same that `loadis` does. In fact, I don't think I needed to add `loadbsp` in the past, just fix `loadbs` for 64-bit as well.

Yusuke Suzuki

Comment 5 2019-05-19 20:27:06 PDT

(In reply to Tadeu Zagallo from comment #4) > (In reply to Yusuke Suzuki from comment #2) > > Created attachment 370215 [details] > > Patch > > > > WIP, update 32bit wide16 things, add loadhsp > > Did you see crashes with `loadhs`? I had updated it to use :quad for the > destination operand on 64-bit, which is the same that `loadis` does. In > fact, I don't think I needed to add `loadbsp` in the past, just fix `loadbs` > for 64-bit as well. No, I don't see the crash, loadhsp change is just aligning to loadbsp. But now I wonder if both loadbsp and loadhsp have a bug in ARM32_64 environment. They will load 32bit value into the register (b/c pointer is 32bit). But later, if we use this register in addressing form in ARM32_64, it would be incorrect. Like, loadbsp xxx, t1 loadi [t0, t1, 8], t2 I'll remove loadbsp and loadhsp. Extending loadbsp to quad and using loadbs / loadhs.

Yusuke Suzuki

Comment 6 2019-05-19 20:52:39 PDT

Created attachment 370235 [details] Patch WIP

Yusuke Suzuki

Comment 7 2019-05-20 23:24:51 PDT

Gmail, after full GC, OP distribution. Almost all wide opcodes become wide16. 1. OPCODE[op_jmp] has some wide32. It is because of large jump. We could add some instruction like, op_jmp32 epecially to remove op_wide32 dispatch given that only op_jmp has wide32 versions in this data. 2. OPCODE[op_put_to_scope] has small amount of wide32. Not sure what it is for now. 3. OPCODE[op_get_by_val] has some wide32. I believe this is https://bugs.webkit.org/show_bug.cgi?id=190929, and we can remove it easily. 4. Instruction Stream reduction is very large. It seems 7MB. I'll add (3) part too and ensure wide32 will be gone for get_by_val in this workload. METADATA TABLES SIZE 1863752 METADATA[op_call] = 6687 METADATA[op_call_eval] = 0 METADATA[op_call_varargs] = 39 METADATA[op_tail_call_varargs] = 12 METADATA[op_tail_call_forward_arguments] = 3 METADATA[op_construct] = 396 METADATA[op_construct_varargs] = 0 METADATA[op_get_argument] = 13 METADATA[op_jneq_ptr] = 407 METADATA[op_new_object] = 145 METADATA[op_resolve_scope] = 6101 METADATA[op_get_from_scope] = 6547 METADATA[op_put_to_scope] = 390 METADATA[op_get_from_arguments] = 53 METADATA[op_in_by_val] = 38 METADATA[op_new_array] = 242 METADATA[op_get_by_id] = 9541 METADATA[op_get_by_id_with_this] = 0 METADATA[op_get_by_val_with_this] = 0 METADATA[op_catch] = 115 METADATA[op_get_by_id_direct] = 14 METADATA[op_try_get_by_id] = 19 METADATA[op_put_by_id] = 2285 METADATA[op_tail_call] = 72 METADATA[op_profile_type] = 0 METADATA[op_profile_control_flow] = 0 METADATA[op_mul] = 43 METADATA[op_has_indexed_property] = 30 METADATA[op_get_by_val] = 873 METADATA[op_put_by_val] = 241 METADATA[op_get_direct_pname] = 73 METADATA[op_div] = 13 METADATA[op_put_by_val_direct] = 35 METADATA[op_sub] = 148 METADATA[op_bitand] = 77 METADATA[op_bitor] = 78 METADATA[op_bitxor] = 16 METADATA[op_to_object] = 14 METADATA[op_to_number] = 105 METADATA[op_negate] = 15 METADATA[op_bitnot] = 4 METADATA[op_new_array_with_size] = 24 METADATA[op_new_array_buffer] = 14 METADATA[op_create_this] = 188 METADATA[op_to_this] = 1074 METADATA[op_add] = 318 UnlinkedCodeBlocks 24386, InstructionStream bytes 13009050 OPCODE[op_call] = narrow: 110375, wide16: 32469, wide32: 0 OPCODE[op_call_eval] = narrow: 3, wide16: 0, wide32: 0 OPCODE[op_call_varargs] = narrow: 251, wide16: 0, wide32: 0 OPCODE[op_tail_call_varargs] = narrow: 9, wide16: 0, wide32: 0 OPCODE[op_tail_call_forward_arguments] = narrow: 1, wide16: 0, wide32: 0 OPCODE[op_construct] = narrow: 8066, wide16: 717, wide32: 0 OPCODE[op_construct_varargs] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_get_argument] = narrow: 15, wide16: 0, wide32: 0 OPCODE[op_jneq_ptr] = narrow: 2833, wide16: 0, wide32: 0 OPCODE[op_new_object] = narrow: 5267, wide16: 4766, wide32: 0 OPCODE[op_resolve_scope] = narrow: 116431, wide16: 154418, wide32: 0 OPCODE[op_get_from_scope] = narrow: 120640, wide16: 119466, wide32: 0 OPCODE[op_put_to_scope] = narrow: 15572, wide16: 49584, wide32: 23 OPCODE[op_get_from_arguments] = narrow: 305, wide16: 0, wide32: 0 OPCODE[op_in_by_val] = narrow: 174, wide16: 0, wide32: 0 OPCODE[op_new_array] = narrow: 5052, wide16: 6021, wide32: 0 OPCODE[op_get_by_id] = narrow: 112581, wide16: 36042, wide32: 0 OPCODE[op_get_by_id_with_this] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_get_by_val_with_this] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_catch] = narrow: 2744, wide16: 489, wide32: 0 OPCODE[op_get_by_id_direct] = narrow: 16, wide16: 0, wide32: 0 OPCODE[op_try_get_by_id] = narrow: 16, wide16: 0, wide32: 0 OPCODE[op_put_by_id] = narrow: 40755, wide16: 61557, wide32: 0 OPCODE[op_tail_call] = narrow: 290, wide16: 0, wide32: 0 OPCODE[op_profile_type] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_profile_control_flow] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_mul] = narrow: 308, wide16: 14, wide32: 0 OPCODE[op_has_indexed_property] = narrow: 258, wide16: 0, wide32: 0 OPCODE[op_get_by_val] = narrow: 6720, wide16: 66, wide32: 335 OPCODE[op_put_by_val] = narrow: 2349, wide16: 2413, wide32: 0 OPCODE[op_get_direct_pname] = narrow: 0, wide16: 0, wide32: 323 OPCODE[op_div] = narrow: 158, wide16: 2, wide32: 0 OPCODE[op_put_by_val_direct] = narrow: 442, wide16: 14787, wide32: 0 OPCODE[op_sub] = narrow: 973, wide16: 6, wide32: 0 OPCODE[op_bitand] = narrow: 458, wide16: 0, wide32: 0 OPCODE[op_bitor] = narrow: 349, wide16: 3, wide32: 0 OPCODE[op_bitxor] = narrow: 196, wide16: 1, wide32: 0 OPCODE[op_to_object] = narrow: 13, wide16: 0, wide32: 0 OPCODE[op_to_number] = narrow: 479, wide16: 0, wide32: 0 OPCODE[op_negate] = narrow: 54, wide16: 0, wide32: 0 OPCODE[op_bitnot] = narrow: 32, wide16: 0, wide32: 0 OPCODE[op_new_array_with_size] = narrow: 69, wide16: 0, wide32: 0 OPCODE[op_new_array_buffer] = narrow: 656, wide16: 6213, wide32: 0 OPCODE[op_create_this] = narrow: 3466, wide16: 0, wide32: 0 OPCODE[op_to_this] = narrow: 6876, wide16: 0, wide32: 0 OPCODE[op_add] = narrow: 3869, wide16: 24, wide32: 0 OPCODE[op_instanceof_custom] = narrow: 593, wide16: 0, wide32: 0 OPCODE[op_typeof] = narrow: 146, wide16: 0, wide32: 0 OPCODE[op_is_cell_with_type] = narrow: 45, wide16: 0, wide32: 0 OPCODE[op_in_by_id] = narrow: 56, wide16: 11, wide32: 0 OPCODE[op_put_by_id_with_this] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_del_by_id] = narrow: 77, wide16: 0, wide32: 0 OPCODE[op_put_by_val_with_this] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_del_by_val] = narrow: 132, wide16: 0, wide32: 0 OPCODE[op_put_getter_by_id] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_put_setter_by_id] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_put_getter_setter_by_id] = narrow: 6, wide16: 2, wide32: 0 OPCODE[op_put_getter_by_val] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_put_setter_by_val] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_define_data_property] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_define_accessor_property] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_jmp] = narrow: 18297, wide16: 737, wide32: 2304 OPCODE[op_jtrue] = narrow: 11422, wide16: 279, wide32: 0 OPCODE[op_jfalse] = narrow: 23152, wide16: 163, wide32: 0 OPCODE[op_jeq_null] = narrow: 558, wide16: 124, wide32: 0 OPCODE[op_jneq_null] = narrow: 237, wide16: 1, wide32: 0 OPCODE[op_jeq] = narrow: 801, wide16: 12, wide32: 0 OPCODE[op_jstricteq] = narrow: 492, wide16: 0, wide32: 0 OPCODE[op_jneq] = narrow: 1872, wide16: 7, wide32: 0 OPCODE[op_jnstricteq] = narrow: 683, wide16: 9, wide32: 0 OPCODE[op_jless] = narrow: 1110, wide16: 376, wide32: 0 OPCODE[op_jlesseq] = narrow: 108, wide16: 29, wide32: 0 OPCODE[op_jgreater] = narrow: 130, wide16: 18, wide32: 0 OPCODE[op_jgreatereq] = narrow: 29, wide16: 2, wide32: 0 OPCODE[op_jnless] = narrow: 2028, wide16: 0, wide32: 0 OPCODE[op_jnlesseq] = narrow: 285, wide16: 5, wide32: 0 OPCODE[op_wide16] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_jngreatereq] = narrow: 148, wide16: 0, wide32: 0 OPCODE[op_jbelow] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_jbeloweq] = narrow: 2, wide16: 0, wide32: 0 OPCODE[op_loop_hint] = narrow: 3297, wide16: 0, wide32: 0 OPCODE[op_switch_imm] = narrow: 202, wide16: 0, wide32: 0 OPCODE[op_switch_char] = narrow: 10, wide16: 0, wide32: 0 OPCODE[op_switch_string] = narrow: 32, wide16: 0, wide32: 0 OPCODE[op_new_func] = narrow: 943, wide16: 0, wide32: 0 OPCODE[op_new_func_exp] = narrow: 16711, wide16: 67018, wide32: 0 OPCODE[op_new_generator_func] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_new_generator_func_exp] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_new_async_func] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_new_async_func_exp] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_new_async_generator_func] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_new_async_generator_func_exp] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_set_function_name] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_ret] = narrow: 28033, wide16: 5, wide32: 0 OPCODE[op_strcat] = narrow: 711, wide16: 0, wide32: 0 OPCODE[op_to_primitive] = narrow: 2027, wide16: 0, wide32: 0 OPCODE[op_put_to_arguments] = narrow: 8, wide16: 0, wide32: 0 OPCODE[op_push_with_scope] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_create_lexical_environment] = narrow: 2204, wide16: 0, wide32: 0 OPCODE[op_create_generator_frame_environment] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_get_parent_scope] = narrow: 5, wide16: 0, wide32: 0 OPCODE[op_throw] = narrow: 1137, wide16: 1, wide32: 0 OPCODE[op_throw_static_error] = narrow: 660, wide16: 0, wide32: 0 OPCODE[op_debug] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_end] = narrow: 209, wide16: 0, wide32: 0 OPCODE[op_get_enumerable_length] = narrow: 258, wide16: 0, wide32: 0 OPCODE[op_has_structure_property] = narrow: 258, wide16: 0, wide32: 0 OPCODE[op_has_generic_property] = narrow: 258, wide16: 0, wide32: 0 OPCODE[op_get_property_enumerator] = narrow: 258, wide16: 0, wide32: 0 OPCODE[op_enumerator_structure_pname] = narrow: 516, wide16: 0, wide32: 0 OPCODE[op_enumerator_generic_pname] = narrow: 516, wide16: 0, wide32: 0 OPCODE[op_to_index_string] = narrow: 258, wide16: 0, wide32: 0 OPCODE[op_unreachable] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_create_rest] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_get_rest_length] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_yield] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_check_traps] = narrow: 27683, wide16: 0, wide32: 0 OPCODE[op_log_shadow_chicken_prologue] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_log_shadow_chicken_tail] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_resolve_scope_for_hoisting_func_decl_in_eval] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_nop] = narrow: 48, wide16: 0, wide32: 0 OPCODE[op_super_sampler_begin] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_super_sampler_end] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_jngreater] = narrow: 402, wide16: 4, wide32: 0 OPCODE[op_wide32] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_enter] = narrow: 24386, wide16: 0, wide32: 0 OPCODE[op_get_scope] = narrow: 24386, wide16: 0, wide32: 0 OPCODE[op_create_direct_arguments] = narrow: 221, wide16: 0, wide32: 0 OPCODE[op_create_scoped_arguments] = narrow: 13, wide16: 0, wide32: 0 OPCODE[op_create_cloned_arguments] = narrow: 24, wide16: 0, wide32: 0 OPCODE[op_argument_count] = narrow: 2, wide16: 0, wide32: 0 OPCODE[op_check_tdz] = narrow: 26, wide16: 0, wide32: 0 OPCODE[op_new_array_with_spread] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_spread] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_new_regexp] = narrow: 975, wide16: 634, wide32: 0 OPCODE[op_mov] = narrow: 299918, wide16: 16419, wide32: 0 OPCODE[op_eq] = narrow: 1418, wide16: 0, wide32: 0 OPCODE[op_neq] = narrow: 722, wide16: 0, wide32: 0 OPCODE[op_stricteq] = narrow: 1152, wide16: 1, wide32: 0 OPCODE[op_nstricteq] = narrow: 356, wide16: 4, wide32: 0 OPCODE[op_less] = narrow: 688, wide16: 1, wide32: 0 OPCODE[op_lesseq] = narrow: 180, wide16: 4, wide32: 0 OPCODE[op_greater] = narrow: 191, wide16: 1, wide32: 0 OPCODE[op_greatereq] = narrow: 74, wide16: 0, wide32: 0 OPCODE[op_below] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_beloweq] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_mod] = narrow: 80, wide16: 0, wide32: 0 OPCODE[op_pow] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_lshift] = narrow: 167, wide16: 0, wide32: 0 OPCODE[op_rshift] = narrow: 87, wide16: 0, wide32: 0 OPCODE[op_urshift] = narrow: 157, wide16: 2, wide32: 0 OPCODE[op_eq_null] = narrow: 637, wide16: 0, wide32: 0 OPCODE[op_neq_null] = narrow: 720, wide16: 0, wide32: 0 OPCODE[op_to_string] = narrow: 14, wide16: 0, wide32: 0 OPCODE[op_unsigned] = narrow: 155, wide16: 0, wide32: 0 OPCODE[op_is_empty] = narrow: 6, wide16: 0, wide32: 0 OPCODE[op_is_undefined] = narrow: 277, wide16: 0, wide32: 0 OPCODE[op_is_undefined_or_null] = narrow: 10, wide16: 0, wide32: 0 OPCODE[op_is_boolean] = narrow: 5, wide16: 0, wide32: 0 OPCODE[op_is_number] = narrow: 48, wide16: 0, wide32: 0 OPCODE[op_is_object] = narrow: 624, wide16: 0, wide32: 0 OPCODE[op_is_object_or_null] = narrow: 167, wide16: 0, wide32: 0 OPCODE[op_is_function] = narrow: 228, wide16: 0, wide32: 0 OPCODE[op_inc] = narrow: 2908, wide16: 0, wide32: 0 OPCODE[op_dec] = narrow: 279, wide16: 0, wide32: 0 OPCODE[op_not] = narrow: 3085, wide16: 3, wide32: 0 OPCODE[op_identity_with_profile] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_overrides_has_instance] = narrow: 593, wide16: 0, wide32: 0 OPCODE[op_instanceof] = narrow: 593, wide16: 0, wide32: 0 OPCODE[handleUncaughtException] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_eval_prologue] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_module_program_prologue] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_function_for_call_prologue] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_function_for_construct_prologue] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_function_for_call_arity_check] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_function_for_construct_arity_check] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_generic_return_point] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_throw_from_slow_path_trampoline] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_throw_during_call_trampoline] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_native_call_trampoline] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_native_construct_trampoline] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_internal_function_call_trampoline] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_internal_function_construct_trampoline] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_program_prologue] = narrow: 0, wide16: 0, wide32: 0

Yusuke Suzuki

Comment 8 2019-05-20 23:26:11 PDT

op_get_direct_pname also exists. And this is almost the same to op_get_by_val, and I think this "wide32" thing is the same reason.

Yusuke Suzuki

Comment 9 2019-05-21 02:13:13 PDT

Another thing I wonder is how many out-of-line jumps exist.

Tadeu Zagallo

Comment 10 2019-05-21 02:14:42 PDT

(In reply to Yusuke Suzuki from comment #9) > Another thing I wonder is how many out-of-line jumps exist. Yes, that'd be very interesting to know. Depending on the numbers, it might be worth generating forward jumps as wide16 opcodes, or even some simple heuristic to decide whether to emit as narrow or wide16.

Yusuke Suzuki

Comment 11 2019-05-21 02:20:10 PDT

(In reply to Tadeu Zagallo from comment #10) > (In reply to Yusuke Suzuki from comment #9) > > Another thing I wonder is how many out-of-line jumps exist. > > Yes, that'd be very interesting to know. Depending on the numbers, it might > be worth generating forward jumps as wide16 opcodes, or even some simple > heuristic to decide whether to emit as narrow or wide16. Yeah, right! It would be cool if we can choose appropriate narrow/wide16/wide32 based on various information in BytecodeGenerator.

Yusuke Suzuki

Comment 12 2019-05-21 02:20:28 PDT

Add OpGetByVal shrinking for for-in-context. Wide32 OpGetByVal decreases. 6 exists, but I think this is converted OpGetByval from OpGetDirectPname. UnlinkedCodeBlocks 25488, InstructionStream bytes 13306296 OPCODE[op_call] = narrow: 118474, wide16: 33166, wide32: 0 OPCODE[op_call_eval] = narrow: 3, wide16: 0, wide32: 0 OPCODE[op_call_varargs] = narrow: 252, wide16: 0, wide32: 0 OPCODE[op_tail_call_varargs] = narrow: 9, wide16: 0, wide32: 0 OPCODE[op_tail_call_forward_arguments] = narrow: 1, wide16: 0, wide32: 0 OPCODE[op_construct] = narrow: 8256, wide16: 717, wide32: 0 OPCODE[op_construct_varargs] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_get_argument] = narrow: 15, wide16: 0, wide32: 0 OPCODE[op_jneq_ptr] = narrow: 2848, wide16: 0, wide32: 0 OPCODE[op_new_object] = narrow: 5301, wide16: 4769, wide32: 0 OPCODE[op_resolve_scope] = narrow: 124695, wide16: 155198, wide32: 0 OPCODE[op_get_from_scope] = narrow: 129305, wide16: 120256, wide32: 0 OPCODE[op_put_to_scope] = narrow: 16232, wide16: 49588, wide32: 23 OPCODE[op_get_from_arguments] = narrow: 305, wide16: 0, wide32: 0 OPCODE[op_in_by_val] = narrow: 173, wide16: 0, wide32: 0 OPCODE[op_new_array] = narrow: 5144, wide16: 6021, wide32: 0 OPCODE[op_get_by_id] = narrow: 117291, wide16: 36107, wide32: 0 OPCODE[op_get_by_id_with_this] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_get_by_val_with_this] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_catch] = narrow: 2751, wide16: 489, wide32: 0 OPCODE[op_get_by_id_direct] = narrow: 16, wide16: 0, wide32: 0 OPCODE[op_try_get_by_id] = narrow: 16, wide16: 0, wide32: 0 OPCODE[op_put_by_id] = narrow: 41597, wide16: 61596, wide32: 0 OPCODE[op_tail_call] = narrow: 290, wide16: 0, wide32: 0 OPCODE[op_profile_type] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_profile_control_flow] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_mul] = narrow: 312, wide16: 14, wide32: 0 OPCODE[op_has_indexed_property] = narrow: 260, wide16: 0, wide32: 0 OPCODE[op_get_by_val] = narrow: 7239, wide16: 102, wide32: 6 OPCODE[op_put_by_val] = narrow: 2364, wide16: 2413, wide32: 0 OPCODE[op_get_direct_pname] = narrow: 0, wide16: 0, wide32: 326 OPCODE[op_div] = narrow: 158, wide16: 2, wide32: 0 OPCODE[op_put_by_val_direct] = narrow: 442, wide16: 14787, wide32: 0 OPCODE[op_sub] = narrow: 994, wide16: 6, wide32: 0 OPCODE[op_bitand] = narrow: 478, wide16: 0, wide32: 0 OPCODE[op_bitor] = narrow: 351, wide16: 3, wide32: 0 OPCODE[op_bitxor] = narrow: 202, wide16: 1, wide32: 0 OPCODE[op_to_object] = narrow: 13, wide16: 0, wide32: 0 OPCODE[op_to_number] = narrow: 495, wide16: 0, wide32: 0 OPCODE[op_negate] = narrow: 56, wide16: 0, wide32: 0 OPCODE[op_bitnot] = narrow: 32, wide16: 0, wide32: 0 OPCODE[op_new_array_with_size] = narrow: 70, wide16: 0, wide32: 0 OPCODE[op_new_array_buffer] = narrow: 666, wide16: 6213, wide32: 0 OPCODE[op_create_this] = narrow: 3524, wide16: 0, wide32: 0 OPCODE[op_to_this] = narrow: 7050, wide16: 0, wide32: 0 OPCODE[op_add] = narrow: 3925, wide16: 30, wide32: 0 OPCODE[op_instanceof_custom] = narrow: 601, wide16: 0, wide32: 0 OPCODE[op_typeof] = narrow: 147, wide16: 0, wide32: 0 OPCODE[op_is_cell_with_type] = narrow: 45, wide16: 0, wide32: 0 OPCODE[op_in_by_id] = narrow: 56, wide16: 11, wide32: 0 OPCODE[op_put_by_id_with_this] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_del_by_id] = narrow: 77, wide16: 0, wide32: 0 OPCODE[op_put_by_val_with_this] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_del_by_val] = narrow: 132, wide16: 0, wide32: 0 OPCODE[op_put_getter_by_id] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_put_setter_by_id] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_put_getter_setter_by_id] = narrow: 6, wide16: 2, wide32: 0 OPCODE[op_put_getter_by_val] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_put_setter_by_val] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_define_data_property] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_define_accessor_property] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_jmp] = narrow: 18777, wide16: 719, wide32: 2304 OPCODE[op_jtrue] = narrow: 11703, wide16: 356, wide32: 0 OPCODE[op_jfalse] = narrow: 24080, wide16: 187, wide32: 0 OPCODE[op_jeq_null] = narrow: 619, wide16: 130, wide32: 0 OPCODE[op_jneq_null] = narrow: 250, wide16: 1, wide32: 0 OPCODE[op_jeq] = narrow: 805, wide16: 12, wide32: 0 OPCODE[op_jstricteq] = narrow: 493, wide16: 0, wide32: 0 OPCODE[op_jneq] = narrow: 1897, wide16: 7, wide32: 0 OPCODE[op_jnstricteq] = narrow: 685, wide16: 11, wide32: 0 OPCODE[op_jless] = narrow: 1127, wide16: 379, wide32: 0 OPCODE[op_jlesseq] = narrow: 111, wide16: 31, wide32: 0 OPCODE[op_jgreater] = narrow: 138, wide16: 18, wide32: 0 OPCODE[op_jgreatereq] = narrow: 30, wide16: 2, wide32: 0 OPCODE[op_jnless] = narrow: 2060, wide16: 1, wide32: 0 OPCODE[op_jnlesseq] = narrow: 295, wide16: 5, wide32: 0 OPCODE[op_wide16] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_jngreatereq] = narrow: 162, wide16: 0, wide32: 0 OPCODE[op_jbelow] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_jbeloweq] = narrow: 2, wide16: 0, wide32: 0 OPCODE[op_loop_hint] = narrow: 3467, wide16: 0, wide32: 0 OPCODE[op_switch_imm] = narrow: 225, wide16: 5, wide32: 0 OPCODE[op_switch_char] = narrow: 10, wide16: 0, wide32: 0 OPCODE[op_switch_string] = narrow: 32, wide16: 0, wide32: 0 OPCODE[op_new_func] = narrow: 944, wide16: 0, wide32: 0 OPCODE[op_new_func_exp] = narrow: 17012, wide16: 67018, wide32: 0 OPCODE[op_new_generator_func] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_new_generator_func_exp] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_new_async_func] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_new_async_func_exp] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_new_async_generator_func] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_new_async_generator_func_exp] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_set_function_name] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_ret] = narrow: 29410, wide16: 13, wide32: 0 OPCODE[op_strcat] = narrow: 713, wide16: 0, wide32: 0 OPCODE[op_to_primitive] = narrow: 2032, wide16: 0, wide32: 0 OPCODE[op_put_to_arguments] = narrow: 8, wide16: 0, wide32: 0 OPCODE[op_push_with_scope] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_create_lexical_environment] = narrow: 2357, wide16: 0, wide32: 0 OPCODE[op_create_generator_frame_environment] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_get_parent_scope] = narrow: 5, wide16: 0, wide32: 0 OPCODE[op_throw] = narrow: 1181, wide16: 9, wide32: 0 OPCODE[op_throw_static_error] = narrow: 668, wide16: 0, wide32: 0 OPCODE[op_debug] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_end] = narrow: 208, wide16: 0, wide32: 0 OPCODE[op_get_enumerable_length] = narrow: 260, wide16: 0, wide32: 0 OPCODE[op_has_structure_property] = narrow: 260, wide16: 0, wide32: 0 OPCODE[op_has_generic_property] = narrow: 260, wide16: 0, wide32: 0 OPCODE[op_get_property_enumerator] = narrow: 260, wide16: 0, wide32: 0 OPCODE[op_enumerator_structure_pname] = narrow: 520, wide16: 0, wide32: 0 OPCODE[op_enumerator_generic_pname] = narrow: 520, wide16: 0, wide32: 0 OPCODE[op_to_index_string] = narrow: 260, wide16: 0, wide32: 0 OPCODE[op_unreachable] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_create_rest] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_get_rest_length] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_yield] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_check_traps] = narrow: 28955, wide16: 0, wide32: 0 OPCODE[op_log_shadow_chicken_prologue] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_log_shadow_chicken_tail] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_resolve_scope_for_hoisting_func_decl_in_eval] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_nop] = narrow: 48, wide16: 0, wide32: 0 OPCODE[op_super_sampler_begin] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_super_sampler_end] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_jngreater] = narrow: 410, wide16: 4, wide32: 0 OPCODE[op_wide32] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_enter] = narrow: 25488, wide16: 0, wide32: 0 OPCODE[op_get_scope] = narrow: 25488, wide16: 0, wide32: 0 OPCODE[op_create_direct_arguments] = narrow: 222, wide16: 0, wide32: 0 OPCODE[op_create_scoped_arguments] = narrow: 13, wide16: 0, wide32: 0 OPCODE[op_create_cloned_arguments] = narrow: 24, wide16: 0, wide32: 0 OPCODE[op_argument_count] = narrow: 2, wide16: 0, wide32: 0 OPCODE[op_check_tdz] = narrow: 26, wide16: 0, wide32: 0 OPCODE[op_new_array_with_spread] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_spread] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_new_regexp] = narrow: 975, wide16: 634, wide32: 0 OPCODE[op_mov] = narrow: 310261, wide16: 16968, wide32: 0 OPCODE[op_eq] = narrow: 1441, wide16: 1, wide32: 0 OPCODE[op_neq] = narrow: 727, wide16: 0, wide32: 0 OPCODE[op_stricteq] = narrow: 1181, wide16: 5, wide32: 0 OPCODE[op_nstricteq] = narrow: 364, wide16: 4, wide32: 0 OPCODE[op_less] = narrow: 702, wide16: 1, wide32: 0 OPCODE[op_lesseq] = narrow: 185, wide16: 4, wide32: 0 OPCODE[op_greater] = narrow: 195, wide16: 1, wide32: 0 OPCODE[op_greatereq] = narrow: 80, wide16: 0, wide32: 0 OPCODE[op_below] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_beloweq] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_mod] = narrow: 81, wide16: 0, wide32: 0 OPCODE[op_pow] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_lshift] = narrow: 166, wide16: 0, wide32: 0 OPCODE[op_rshift] = narrow: 88, wide16: 0, wide32: 0 OPCODE[op_urshift] = narrow: 158, wide16: 2, wide32: 0 OPCODE[op_eq_null] = narrow: 649, wide16: 2, wide32: 0 OPCODE[op_neq_null] = narrow: 803, wide16: 28, wide32: 0 OPCODE[op_to_string] = narrow: 14, wide16: 0, wide32: 0 OPCODE[op_unsigned] = narrow: 156, wide16: 0, wide32: 0 OPCODE[op_is_empty] = narrow: 6, wide16: 0, wide32: 0 OPCODE[op_is_undefined] = narrow: 277, wide16: 0, wide32: 0 OPCODE[op_is_undefined_or_null] = narrow: 10, wide16: 0, wide32: 0 OPCODE[op_is_boolean] = narrow: 5, wide16: 0, wide32: 0 OPCODE[op_is_number] = narrow: 48, wide16: 0, wide32: 0 OPCODE[op_is_object] = narrow: 632, wide16: 0, wide32: 0 OPCODE[op_is_object_or_null] = narrow: 167, wide16: 0, wide32: 0 OPCODE[op_is_function] = narrow: 229, wide16: 0, wide32: 0 OPCODE[op_inc] = narrow: 2957, wide16: 0, wide32: 0 OPCODE[op_dec] = narrow: 286, wide16: 0, wide32: 0 OPCODE[op_not] = narrow: 3172, wide16: 6, wide32: 0 OPCODE[op_identity_with_profile] = narrow: 0, wide16: 0, wide32: 0 OPCODE[op_overrides_has_instance] = narrow: 601, wide16: 0, wide32: 0 OPCODE[op_instanceof] = narrow: 601, wide16: 0, wide32: 0 OPCODE[handleUncaughtException] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_eval_prologue] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_module_program_prologue] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_function_for_call_prologue] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_function_for_construct_prologue] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_function_for_call_arity_check] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_function_for_construct_arity_check] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_generic_return_point] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_throw_from_slow_path_trampoline] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_throw_during_call_trampoline] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_native_call_trampoline] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_native_construct_trampoline] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_internal_function_call_trampoline] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_internal_function_construct_trampoline] = narrow: 0, wide16: 0, wide32: 0 OPCODE[llint_program_prologue] = narrow: 0, wide16: 0, wide32: 0

Yusuke Suzuki

Comment 13 2019-05-21 03:04:29 PDT

Created attachment 370308 [details] Patch WIP, pick small GetByVal in for-in-context by using OpXXX::{checkWithoutMetadataID,emitWithSmallestSizeRequirement}, drop Fits<Special::Pointer> due to enum specialization,

EWS Watchlist

Comment 14 2019-05-21 03:06:39 PDT

Attachment 370308 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 33 files If any of these errors are false positives, please file a bug against check-webkit-style.

Yusuke Suzuki

Comment 15 2019-05-21 03:54:24 PDT

*** Bug 190929 has been marked as a duplicate of this bug. ***

EWS Watchlist

Comment 16 2019-05-21 04:14:43 PDT

Comment on attachment 370308 [details] Patch Attachment 370308 [details] did not pass mac-ews (mac): Output: https://webkit-queues.webkit.org/results/12245355 New failing tests: webgl/1.0.2/conformance/typedarrays/typed-arrays-in-workers.html js/dom/typed-array-access.html webgl/2.0.0/conformance/typedarrays/typed-arrays-in-workers.html fast/canvas/webgl/array-set-with-offset.html fast/canvas/webgl/array-setters.html fast/canvas/webgl/typed-arrays-in-workers.html

EWS Watchlist

Comment 17 2019-05-21 04:14:45 PDT

Created attachment 370311 [details] Archive of layout-test-results from ews101 for mac-highsierra The attached test failures were seen while running run-webkit-tests on the mac-ews. Bot: ews101 Port: mac-highsierra Platform: Mac OS X 10.13.6

EWS Watchlist

Comment 18 2019-05-21 04:18:06 PDT

Comment on attachment 370308 [details] Patch Attachment 370308 [details] did not pass mac-debug-ews (mac): Output: https://webkit-queues.webkit.org/results/12245338 Number of test failures exceeded the failure limit.

EWS Watchlist

Comment 19 2019-05-21 04:18:08 PDT

Created attachment 370312 [details] Archive of layout-test-results from ews116 for mac-highsierra The attached test failures were seen while running run-webkit-tests on the mac-debug-ews. Bot: ews116 Port: mac-highsierra Platform: Mac OS X 10.13.6

EWS Watchlist

Comment 20 2019-05-21 04:30:06 PDT

Comment on attachment 370308 [details] Patch Attachment 370308 [details] did not pass mac-wk2-ews (mac-wk2): Output: https://webkit-queues.webkit.org/results/12245381 New failing tests: webgl/1.0.2/conformance/typedarrays/typed-arrays-in-workers.html js/dom/typed-array-access.html webgl/2.0.0/conformance/typedarrays/typed-arrays-in-workers.html fast/canvas/webgl/array-set-with-offset.html fast/canvas/webgl/array-setters.html fast/canvas/webgl/typed-arrays-in-workers.html

EWS Watchlist

Comment 21 2019-05-21 04:30:08 PDT

Created attachment 370313 [details] Archive of layout-test-results from ews106 for mac-highsierra-wk2 The attached test failures were seen while running run-webkit-tests on the mac-wk2-ews. Bot: ews106 Port: mac-highsierra-wk2 Platform: Mac OS X 10.13.6

Yusuke Suzuki

Comment 22 2019-05-21 04:38:13 PDT

Ah, loadbs :quad breaks Int8Array GetByVal (it assumes that upper 32bit register part is zeroed). I'll fix things tomorrow.

Yusuke Suzuki

Comment 23 2019-05-21 04:41:42 PDT

(In reply to Yusuke Suzuki from comment #22) > Ah, loadbs :quad breaks Int8Array GetByVal (it assumes that upper 32bit > register part is zeroed). I'll fix things tomorrow. I'll align things to MacroAssembler. Maybe, signed load things, having extension width in instruction would be better.

Yusuke Suzuki

Comment 24 2019-05-21 04:46:53 PDT

(In reply to Yusuke Suzuki from comment #23) > (In reply to Yusuke Suzuki from comment #22) > > Ah, loadbs :quad breaks Int8Array GetByVal (it assumes that upper 32bit > > register part is zeroed). I'll fix things tomorrow. > > I'll align things to MacroAssembler. Maybe, signed load things, having > extension width in instruction would be better. I'll add loadbsq / loadhsq explicitly to use it for bytecode loading. Interesting! :)

EWS Watchlist

Comment 25 2019-05-21 04:48:23 PDT

Comment on attachment 370308 [details] Patch Attachment 370308 [details] did not pass jsc-ews (mac): Output: https://webkit-queues.webkit.org/results/12245382 New failing tests: stress/dataview-jit-set.js.dfg-eager stress/dataview-jit-set.js.no-ftl wasm.yaml/wasm/js-api/Module.customSection.js.default-wasm stress/dataview-jit-set.js.bytecode-cache stress/dataview-jit-set.js.dfg-eager-no-cjit-validate wasm.yaml/wasm/js-api/Module.customSection.js.wasm-no-call-ic wasm.yaml/wasm/js-api/Module.customSection.js.wasm-eager-jettison stress/dataview-jit-set.js.ftl-no-cjit-b3o0 stress/dataview-jit-set.js.ftl-no-cjit-validate-sampling-profiler wasm.yaml/wasm/js-api/Module.customSection.js.wasm-slow-memory stress/dataview-jit-set.js.ftl-no-cjit-no-put-stack-validate stress/dataview-jit-set.js.no-cjit-collect-continuously stress/dataview-jit-set.js.no-cjit-validate-phases wasm.yaml/wasm/js-api/Module.customSection.js.wasm-no-tls-context es6.yaml/es6/typed_arrays_Int8Array.js.default stress/dataview-jit-set.js.ftl-eager-no-cjit stress/dataview-jit-set.js.dfg-maximal-flush-validate-no-cjit stress/dataview-jit-set.js.ftl-eager-no-cjit-b3o1 stress/dataview-jit-set.js.ftl-no-cjit-small-pool es6.yaml/es6/typed_arrays_Int16Array.js.default stress/dataview-jit-set.js.default wasm.yaml/wasm/js-api/Module.customSection.js.wasm-no-air stress/dataview-jit-set.js.ftl-no-cjit-no-inline-validate wasm.yaml/wasm/js-api/Module.customSection.js.wasm-no-cjit-yes-tls-context stress/dataview-jit-set.js.ftl-eager apiTests

EWS Watchlist

Comment 26 2019-05-21 05:02:35 PDT

Comment on attachment 370308 [details] Patch Attachment 370308 [details] did not pass ios-sim-ews (ios-simulator-wk2): Output: https://webkit-queues.webkit.org/results/12245419 New failing tests: webgl/1.0.2/conformance/typedarrays/typed-arrays-in-workers.html js/dom/typed-array-access.html webgl/2.0.0/conformance/typedarrays/typed-arrays-in-workers.html fast/canvas/webgl/array-set-with-offset.html fast/canvas/webgl/array-setters.html fast/canvas/webgl/typed-arrays-in-workers.html

EWS Watchlist

Comment 27 2019-05-21 05:02:37 PDT

Created attachment 370314 [details] Archive of layout-test-results from ews125 for ios-simulator-wk2 The attached test failures were seen while running run-webkit-tests on the ios-sim-ews. Bot: ews125 Port: ios-simulator-wk2 Platform: Mac OS X 10.14.5

Tadeu Zagallo

Comment 28 2019-05-21 05:21:52 PDT

(In reply to Yusuke Suzuki from comment #22) > Ah, loadbs :quad breaks Int8Array GetByVal (it assumes that upper 32bit > register part is zeroed). I'll fix things tomorrow. That'd explain why I had to add loadbsp... I did not remember it. My bad.

Yusuke Suzuki

Comment 29 2019-05-22 17:18:44 PDT

Another thing I'm considering now is the policy in BytecodeGenerator::newRegister(). It tends to allocate more and more locals if some locals having high number are live. This register allocation policy works perfectly for allocating contiguous register sequence for call/construct. But I think it tends to allocate registers having large number index, and it leads us to using 16bit operations in many bytecode. I think we could reduce # of 16bit bytecodes further by introducing a small tweak.

Yusuke Suzuki

Comment 30 2019-05-22 19:17:36 PDT

Created attachment 370475 [details] Patch

EWS Watchlist

Comment 31 2019-05-22 19:19:53 PDT

Attachment 370475 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 35 files If any of these errors are false positives, please file a bug against check-webkit-style.

Yusuke Suzuki

Comment 32 2019-05-22 19:20:04 PDT

Created attachment 370476 [details] Patch

EWS Watchlist

Comment 33 2019-05-22 19:22:57 PDT

Attachment 370476 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 35 files If any of these errors are false positives, please file a bug against check-webkit-style.

EWS Watchlist

Comment 34 2019-05-22 20:33:00 PDT

Comment on attachment 370476 [details] Patch Attachment 370476 [details] did not pass mac-debug-ews (mac): Output: https://webkit-queues.webkit.org/results/12264918 Number of test failures exceeded the failure limit.

EWS Watchlist

Comment 35 2019-05-22 20:33:01 PDT

Created attachment 370481 [details] Archive of layout-test-results from ews116 for mac-highsierra The attached test failures were seen while running run-webkit-tests on the mac-debug-ews. Bot: ews116 Port: mac-highsierra Platform: Mac OS X 10.13.6

Yusuke Suzuki

Comment 36 2019-05-22 22:17:36 PDT

Created attachment 370486 [details] Patch WIP, fix & compact OperandTypes

EWS Watchlist

Comment 37 2019-05-22 22:20:30 PDT

Attachment 370486 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 36 files If any of these errors are false positives, please file a bug against check-webkit-style.

Tadeu Zagallo

Comment 38 2019-05-22 22:58:20 PDT

Comment on attachment 370486 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=370486&action=review > Source/JavaScriptCore/bytecode/Fits.h:250 > + if constexpr (size == OpcodeSize::Narrow) { Nice, TIL!

Radar WebKit Bug Importer

Comment 39 2019-05-23 00:19:22 PDT

<rdar://problem/51058295>

EWS Watchlist

Comment 40 2019-05-23 01:21:58 PDT

Comment on attachment 370486 [details] Patch Attachment 370486 [details] did not pass win-ews (win): Output: https://webkit-queues.webkit.org/results/12266720 New failing tests: js/dom/modules/import-from-javascript-url.html editing/style/apply-style-iframe-crash.html editing/style/iframe-onload-crash-mac.html js/dom/modules/import-from-handler.html js/array-string-recursion.html

EWS Watchlist

Comment 41 2019-05-23 01:22:01 PDT

Created attachment 370496 [details] Archive of layout-test-results from ews213 for win-future The attached test failures were seen while running run-webkit-tests on the win-ews. Bot: ews213 Port: win-future Platform: CYGWIN_NT-10.0-17763-3.0.5-338.x86_64-x86_64-64bit

Yusuke Suzuki

Comment 42 2019-05-23 13:14:28 PDT

Created attachment 370516 [details] Patch WIP

EWS Watchlist

Comment 43 2019-05-23 13:15:41 PDT

Attachment 370516 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 36 files If any of these errors are false positives, please file a bug against check-webkit-style.

Yusuke Suzuki

Comment 44 2019-05-23 15:05:20 PDT

Created attachment 370524 [details] Patch WIP

EWS Watchlist

Comment 45 2019-05-23 15:07:45 PDT

Attachment 370524 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/bytecode/Fits.h:250: An else statement can be removed when the prior "if" concludes with a return, break, continue or goto statement. [readability/control_flow] [4] ERROR: Source/JavaScriptCore/bytecode/Fits.h:264: An else statement can be removed when the prior "if" concludes with a return, break, continue or goto statement. [readability/control_flow] [4] ERROR: Source/JavaScriptCore/bytecode/Fits.h:279: An else statement can be removed when the prior "if" concludes with a return, break, continue or goto statement. [readability/control_flow] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 5 in 39 files If any of these errors are false positives, please file a bug against check-webkit-style.

Yusuke Suzuki

Comment 46 2019-05-23 15:52:31 PDT

Created attachment 370529 [details] Patch WIP

EWS Watchlist

Comment 47 2019-05-23 15:55:20 PDT

Attachment 370529 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/bytecode/Fits.h:250: An else statement can be removed when the prior "if" concludes with a return, break, continue or goto statement. [readability/control_flow] [4] ERROR: Source/JavaScriptCore/bytecode/Fits.h:264: An else statement can be removed when the prior "if" concludes with a return, break, continue or goto statement. [readability/control_flow] [4] ERROR: Source/JavaScriptCore/bytecode/Fits.h:279: An else statement can be removed when the prior "if" concludes with a return, break, continue or goto statement. [readability/control_flow] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 5 in 40 files If any of these errors are false positives, please file a bug against check-webkit-style.

Yusuke Suzuki

Comment 48 2019-05-23 17:27:57 PDT

Created attachment 370534 [details] Patch WIP

EWS Watchlist

Comment 49 2019-05-23 17:29:56 PDT

Attachment 370534 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/bytecode/Fits.h:250: An else statement can be removed when the prior "if" concludes with a return, break, continue or goto statement. [readability/control_flow] [4] ERROR: Source/JavaScriptCore/bytecode/Fits.h:264: An else statement can be removed when the prior "if" concludes with a return, break, continue or goto statement. [readability/control_flow] [4] ERROR: Source/JavaScriptCore/bytecode/Fits.h:279: An else statement can be removed when the prior "if" concludes with a return, break, continue or goto statement. [readability/control_flow] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 5 in 40 files If any of these errors are false positives, please file a bug against check-webkit-style.

EWS Watchlist

Comment 50 2019-05-23 20:14:42 PDT

Comment on attachment 370534 [details] Patch Attachment 370534 [details] did not pass win-ews (win): Output: https://webkit-queues.webkit.org/results/12274231 New failing tests: js/dom/modules/import-from-javascript-url.html editing/style/apply-style-iframe-crash.html editing/style/iframe-onload-crash-mac.html js/dom/modules/import-from-handler.html js/array-string-recursion.html

EWS Watchlist

Comment 51 2019-05-23 20:14:45 PDT

Created attachment 370556 [details] Archive of layout-test-results from ews214 for win-future The attached test failures were seen while running run-webkit-tests on the win-ews. Bot: ews214 Port: win-future Platform: CYGWIN_NT-10.0-17763-3.0.5-338.x86_64-x86_64-64bit

Yusuke Suzuki

Comment 52 2019-05-24 17:06:22 PDT

Created attachment 370607 [details] Patch

EWS Watchlist

Comment 53 2019-05-24 17:09:03 PDT

Attachment 370607 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 40 files If any of these errors are false positives, please file a bug against check-webkit-style.

Yusuke Suzuki

Comment 54 2019-05-24 18:45:07 PDT

Created attachment 370616 [details] Patch

EWS Watchlist

Comment 55 2019-05-24 18:47:43 PDT

Attachment 370616 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 37 files If any of these errors are false positives, please file a bug against check-webkit-style.

EWS Watchlist

Comment 56 2019-05-24 23:40:35 PDT

Comment on attachment 370616 [details] Patch Attachment 370616 [details] did not pass win-ews (win): Output: https://webkit-queues.webkit.org/results/12285231 Number of test failures exceeded the failure limit.

EWS Watchlist

Comment 57 2019-05-24 23:40:38 PDT

Created attachment 370623 [details] Archive of layout-test-results from ews211 for win-future The attached test failures were seen while running run-webkit-tests on the win-ews. Bot: ews211 Port: win-future Platform: CYGWIN_NT-10.0-17763-3.0.5-338.x86_64-x86_64-64bit

EWS Watchlist

Comment 58 2019-05-25 00:41:42 PDT

Comment on attachment 370616 [details] Patch Attachment 370616 [details] did not pass win-ews (win): Output: https://webkit-queues.webkit.org/results/12285759 New failing tests: js/dom/modules/import-from-javascript-url.html editing/style/iframe-onload-crash-mac.html editing/style/apply-style-iframe-crash.html js/dom/modules/import-from-handler.html js/array-string-recursion.html

EWS Watchlist

Comment 59 2019-05-25 00:41:48 PDT

Created attachment 370626 [details] Archive of layout-test-results from ews210 for win-future The attached test failures were seen while running run-webkit-tests on the win-ews. Bot: ews210 Port: win-future Platform: CYGWIN_NT-10.0-17763-3.0.5-338.x86_64-x86_64-64bit

Yusuke Suzuki

Comment 60 2019-05-27 05:05:30 PDT

Created attachment 370690 [details] Patch WIP

EWS Watchlist

Comment 61 2019-05-27 05:09:04 PDT

Attachment 370690 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 39 files If any of these errors are false positives, please file a bug against check-webkit-style.

EWS Watchlist

Comment 62 2019-05-27 07:50:40 PDT

Comment on attachment 370690 [details] Patch Attachment 370690 [details] did not pass win-ews (win): Output: https://webkit-queues.webkit.org/results/12300831 New failing tests: js/dom/modules/import-from-javascript-url.html editing/style/apply-style-iframe-crash.html editing/style/iframe-onload-crash-mac.html js/dom/modules/import-from-handler.html js/array-string-recursion.html

EWS Watchlist

Comment 63 2019-05-27 07:50:43 PDT

Created attachment 370691 [details] Archive of layout-test-results from ews214 for win-future The attached test failures were seen while running run-webkit-tests on the win-ews. Bot: ews214 Port: win-future Platform: CYGWIN_NT-10.0-17763-3.0.5-338.x86_64-x86_64-64bit

Yusuke Suzuki

Comment 64 2019-05-27 15:05:40 PDT

In Windows, we exhaust the stack by reentering CLoop::execute only 4 times. My guess is that VC++2017 compiler accidentally compiles CLoop::execute with very large stack requirement. Not sure why they do such a thing.

Yusuke Suzuki

Comment 65 2019-05-27 17:07:16 PDT

(In reply to Yusuke Suzuki from comment #64) > In Windows, we exhaust the stack by reentering CLoop::execute only 4 times. > My guess is that VC++2017 compiler accidentally compiles CLoop::execute with > very large stack requirement. Not sure why they do such a thing. Yes, the problem seems that CLoop::execute starts consuming super large stack height.

Yusuke Suzuki

Comment 66 2019-05-27 19:42:01 PDT

Not sure, but it seems that MSVC generates code which consumes so much stack space as CLoop::execute code size increases. Is this a bug in MSVC? Not sure. But one way to avoid the problem is disabling op_wide16 feature for MSVC CLoop combination.

Yusuke Suzuki

Comment 67 2019-05-28 00:18:24 PDT

Created attachment 370714 [details] Patch WIP, disable wide16 optimization in Windows CLoop due to MSVC miscompilation. Add C_LOOP_WIN backend for this mode

EWS Watchlist

Comment 68 2019-05-28 00:21:13 PDT

Attachment 370714 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 41 files If any of these errors are false positives, please file a bug against check-webkit-style.

Yusuke Suzuki

Comment 69 2019-05-28 00:25:30 PDT

Created attachment 370715 [details] Patch WIP

EWS Watchlist

Comment 70 2019-05-28 00:27:23 PDT

Attachment 370715 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 41 files If any of these errors are false positives, please file a bug against check-webkit-style.

Yusuke Suzuki

Comment 71 2019-05-28 01:52:31 PDT

Created attachment 370718 [details] Patch WIP

EWS Watchlist

Comment 72 2019-05-28 01:54:54 PDT

Attachment 370718 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 41 files If any of these errors are false positives, please file a bug against check-webkit-style.

Yusuke Suzuki

Comment 73 2019-05-28 14:17:47 PDT

Created attachment 370783 [details] Patch

EWS Watchlist

Comment 74 2019-05-28 14:19:47 PDT

Attachment 370783 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 44 files If any of these errors are false positives, please file a bug against check-webkit-style.

Tadeu Zagallo

Comment 75 2019-05-28 14:49:08 PDT

Comment on attachment 370783 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=370783&action=review > Source/JavaScriptCore/bytecode/BytecodeConventions.h:35 > +static constexpr int FirstConstantRegisterIndex16 = 64; FYI I did not test other sizes for this constants, but originally I never saw any instructions going wide for arg > 16, so I think it should be fine. > Source/JavaScriptCore/bytecode/BytecodeDumper.cpp:212 > + ": %lu instructions (%lu 16-byte instructions, %lu 32-byte instructions, %lu instructions with metadata); %lu bytes (%lu metadata bytes); %d parameter(s); %d callee register(s); %d variable(s)", oops, this hsould be 16- and 32-*bit*, or maybe just say wide16 and wide32. > Source/JavaScriptCore/generator/Opcode.rb:230 > + # extraneous # > Source/JavaScriptCore/llint/LowLevelInterpreter.asm:377 > +if not C_LOOP_WIN This is the only place where you actually check for C_LOOP_WIN. Wouldn't it be better to export a macro `WIN` check for `C_LOOP and WIN` here rather than `C_LOOP or C_LOOP_WIN` everywhere else? > Source/JavaScriptCore/llint/LowLevelInterpreter.cpp:291 > + static_assert(sizeof(CLoopRegister) == sizeof(intptr_t)); Does static_assert without a message work everywhere? > Source/JavaScriptCore/parser/ResultType.h:217 > + static_assert(sizeof(OperandTypes) == sizeof(uint16_t)); > + return bitwise_cast<uint16_t>(*this); nice! > JSTests/stress/32bit-code.js:6 > +function f(i) { return i }; this seems unused.

Yusuke Suzuki

Comment 76 2019-05-28 15:16:22 PDT

Comment on attachment 370783 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=370783&action=review Thanks! >> Source/JavaScriptCore/bytecode/BytecodeDumper.cpp:212 >> + ": %lu instructions (%lu 16-byte instructions, %lu 32-byte instructions, %lu instructions with metadata); %lu bytes (%lu metadata bytes); %d parameter(s); %d callee register(s); %d variable(s)", > > oops, this hsould be 16- and 32-*bit*, or maybe just say wide16 and wide32. Fixed. >> Source/JavaScriptCore/generator/Opcode.rb:230 >> + # > > extraneous # Fixed. >> Source/JavaScriptCore/llint/LowLevelInterpreter.asm:377 >> +if not C_LOOP_WIN > > This is the only place where you actually check for C_LOOP_WIN. Wouldn't it be better to export a macro `WIN` check for `C_LOOP and WIN` here rather than `C_LOOP or C_LOOP_WIN` everywhere else? I think the style C_LOOP_WIN is fine here because we already have X86_WIN / X86_64_WIN (this is CPU(X86) + OS(WINDOWS)). Since they will generate different LLInt backend code, having C_LOOP_WIN is cleaner. >> Source/JavaScriptCore/llint/LowLevelInterpreter.cpp:291 >> + static_assert(sizeof(CLoopRegister) == sizeof(intptr_t)); > > Does static_assert without a message work everywhere? Yeah, this is C++17 feature. >> JSTests/stress/32bit-code.js:6 >> +function f(i) { return i }; > > this seems unused. Oops, right. Dropped.

Yusuke Suzuki

Comment 77 2019-05-28 15:21:21 PDT

Comment on attachment 370783 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=370783&action=review >> Source/JavaScriptCore/bytecode/BytecodeConventions.h:35 >> +static constexpr int FirstConstantRegisterIndex16 = 64; > > FYI I did not test other sizes for this constants, but originally I never saw any instructions going wide for arg > 16, so I think it should be fine. In Gmail case, I didn't see the case like exceeding 64 arguments. So I think this is fine too.

Yusuke Suzuki

Comment 78 2019-05-28 15:24:21 PDT

Created attachment 370791 [details] Patch

EWS Watchlist

Comment 79 2019-05-28 15:27:58 PDT

Attachment 370791 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 44 files If any of these errors are false positives, please file a bug against check-webkit-style.

Yusuke Suzuki

Comment 80 2019-05-28 16:10:08 PDT

Created attachment 370797 [details] Patch

Yusuke Suzuki

Comment 81 2019-05-28 16:11:25 PDT

Created attachment 370798 [details] Patch

EWS Watchlist

Comment 82 2019-05-28 16:14:41 PDT

Attachment 370798 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 44 files If any of these errors are false positives, please file a bug against check-webkit-style.

EWS Watchlist

Comment 83 2019-05-28 18:27:23 PDT

Comment on attachment 370798 [details] Patch Attachment 370798 [details] did not pass jsc-ews (mac): Output: https://webkit-queues.webkit.org/results/12312784 New failing tests: stress/16bit-code.js.no-llint stress/16bit-code.js.ftl-eager-no-cjit stress/16bit-code.js.default stress/16bit-code.js.bytecode-cache stress/16bit-code.js.ftl-eager stress/16bit-code.js.ftl-no-cjit-validate-sampling-profiler stress/16bit-code.js.dfg-eager stress/16bit-code.js.dfg-maximal-flush-validate-no-cjit stress/16bit-code.js.ftl-no-cjit-no-put-stack-validate stress/16bit-code.js.dfg-eager-no-cjit-validate stress/16bit-code.js.ftl-no-cjit-no-inline-validate stress/16bit-code.js.no-cjit-collect-continuously stress/16bit-code.js.ftl-eager-no-cjit-b3o1 stress/16bit-code.js.no-ftl stress/16bit-code.js.ftl-no-cjit-b3o0 stress/16bit-code.js.no-cjit-validate-phases stress/16bit-code.js.ftl-no-cjit-small-pool apiTests

Yusuke Suzuki

Comment 84 2019-05-28 18:30:24 PDT

(In reply to Build Bot from comment #83) > Comment on attachment 370798 [details] > Patch > > Attachment 370798 [details] did not pass jsc-ews (mac): > Output: https://webkit-queues.webkit.org/results/12312784 > > New failing tests: > stress/16bit-code.js.no-llint > stress/16bit-code.js.ftl-eager-no-cjit > stress/16bit-code.js.default > stress/16bit-code.js.bytecode-cache > stress/16bit-code.js.ftl-eager > stress/16bit-code.js.ftl-no-cjit-validate-sampling-profiler > stress/16bit-code.js.dfg-eager > stress/16bit-code.js.dfg-maximal-flush-validate-no-cjit > stress/16bit-code.js.ftl-no-cjit-no-put-stack-validate > stress/16bit-code.js.dfg-eager-no-cjit-validate > stress/16bit-code.js.ftl-no-cjit-no-inline-validate > stress/16bit-code.js.no-cjit-collect-continuously > stress/16bit-code.js.ftl-eager-no-cjit-b3o1 > stress/16bit-code.js.no-ftl > stress/16bit-code.js.ftl-no-cjit-b3o0 > stress/16bit-code.js.no-cjit-validate-phases > stress/16bit-code.js.ftl-no-cjit-small-pool > apiTests It's OutOfMemory error b/c we run them concurrently. I'll reduce the counter of repeat.

Yusuke Suzuki

Comment 85 2019-05-28 18:39:32 PDT

Created attachment 370819 [details] Patch

EWS Watchlist

Comment 86 2019-05-28 18:42:04 PDT

Attachment 370819 [details] did not pass style-queue: ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:52: g_opcodeMapWide16 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] ERROR: Source/JavaScriptCore/llint/LLIntData.cpp:53: g_opcodeMapWide32 is incorrectly named. Don't use underscores in your identifier names. [readability/naming/underscores] [4] Total errors found: 2 in 44 files If any of these errors are false positives, please file a bug against check-webkit-style.

Yusuke Suzuki

Comment 87 2019-05-29 22:21:05 PDT

Created attachment 370922 [details] Patch Less ifdef version :)

Yusuke Suzuki

Comment 88 2019-05-29 22:21:41 PDT

(In reply to Yusuke Suzuki from comment #87) > Created attachment 370922 [details] > Patch > > Less ifdef version :) Oops, uploaded the patch to the incorrect bug :P

Yusuke Suzuki

Comment 89 2019-05-30 02:24:19 PDT

Created attachment 370929 [details] Patch

Yusuke Suzuki

Comment 90 2019-05-30 14:24:47 PDT

Thanks!!

Yusuke Suzuki

Comment 91 2019-05-30 14:40:41 PDT

Committed r245906: <https://trac.webkit.org/changeset/245906>

Yusuke Suzuki

Comment 92 2019-05-30 16:45:38 PDT

InstructionStream becomes from 23.7 to 15.3. 8.4MB reduction. https://bugs.webkit.org/show_bug.cgi?id=186422#c35 VIRTUAL RESIDENT DIRTY SWAPPED ALLOCATION BYTES DIRTY+SWAP REGION MALLOC ZONE SIZE SIZE SIZE SIZE COUNT ALLOCATED FRAG SIZE % FRAG COUNT =========== ======= ========= ========= ========= ========= ========= ========= ====== ====== Vector_0x10bbac000 213.0M 107.9M 64.9M 0K 297105 41.1M 23.8M 37% 43 WebKit Using System Malloc_0x10598f000 188.0M 58.1M 41.7M 0K 199543 32.8M 9088K 22% 31 HashTable_0x10a382000 174.0M 57.7M 45.1M 0K 20644 33.7M 11.5M 26% 26 StringImpl_0x10a38e000 163.0M 37.2M 34.5M 0K 148428 25.3M 9475K 27% 26 DefaultMallocZone_0x1040e6000 160.0M 23.2M 22.9M 0K 247114 17.9M 5130K 22% 30 MetadataTable_0x10ce98000 156.0M 36.5M 26.8M 0K 25067 4819K 22.1M 83% 29 WebKit IsoAlignedMemoryAllocator_0x10c922000 153.0M 77.5M 68.0M 0K 4062 63.5M 4632K 7% 18 AssemblerData_0x10c995000 150.0M 11.6M 8256K 0K 2 3600 8252K 100% 25 StringBuffer_0x10cc75000 150.0M 3980K 3980K 0K 1 16 3980K 100% 24 InstructionStream_0x10cec6000 150.0M 25.6M 21.0M 0K 25359 15.3M 5831K 28% 24 WebKit FastMallocAlignedMemoryAllocator_0x10c8f4000 145.0M 69.3M 56.4M 0K 2812 43.9M 12.5M 23% 16 RefCountedArray_0x10c898000 118.0M 2360K 2360K 0K 5478 346K 2014K 86% 20 SmallPtrSet_0x10cedd000 118.0M 2032K 2032K 0K 1 16 2032K 100% 19 PropertyTable_0x10cce3000 102.0M 6544K 6544K 0K 6837 2608K 3936K 61% 18 UniqueArray_0x105991000 86.0M 22.1M 22.0M 0K 203870 16.2M 5916K 27% 30 SegmentedVector_0x10cd11000 59.0M 7716K 7716K 0K 12417 4711K 3005K 39% 15 WebKit XMLHttpRequest_0x10cf67000 49.0M 140K 140K 0K 4 5K 135K 97% 7 WebKit FrameView_0x10c9f1000 49.0M 80K 80K 0K 15 21K 59K 74% 7 GCSegmentedArray_0x10c8dd000 49.0M 1700K 1700K 0K 20 76K 1624K 96% 5 IdentifierArena_0x10ce25000 49.0M 60K 60K 0K 1 16 60K 100% 7 SpeculativeJIT_0x10d1c5000 49.0M 60K 60K 0K 1 16 60K 100% 7 HTMLDocumentParser_0x10cb78000 49.0M 76K 76K 0K 1 32 76K 100% 7 ParserArena_0x10ce53000 49.0M 20.3M 20.3M 0K 1 16 20.3M 100% 7 CachedResource_0x10cc47000 49.0M 356K 356K 0K 203 303K 53K 15% 7 WebKit HTMLDocument_0x10ca08000 49.0M 124K 124K 0K 14 46K 78K 64% 7 TreeResolverScope_0x10cbd4000 49.0M 104K 104K 0K 1 32 104K 100% 7 ResourceLoader_0x10cc5e000 49.0M 132K 132K 0K 2 1552 130K 99% 7 SourceProviderCacheItem_0x10ce6a000 46.0M 2380K 2380K 0K 1 32 2380K 100% 11 DocumentLoader_0x10c9da000 41.0M 128K 128K 0K 15 49K 79K 62% 6 WebKit HTMLAudioElement_0x10e3ac000 25.0M 40K 40K 0K 4 8K 32K 82% 2 CStringBuffer_0x10a38c000 22.0M 120K 120K 0K 67 2976 117K 98% 8 UniqueArrayElement_0x10c825000 22.0M 1680K 1680K 0K 38283 1600K 80K 5% 8 QuartzCore_0x10c9c3000 22.0M 292K 292K 0K 1875 139K 153K 53% 8 BagNode_0x10d124000 18.0M 15.8M 11.7M 0K 75875 4973K 6979K 59% 14 VM_0x10d169000 14.0M 80K 80K 0K 6 3920 76K 96% 7 ConcurrentBuffer_0x10ccb5000 13.0M 68K 68K 0K 40 7K 61K 90% 6 CSSValue_0x10c86a000 10.0M 5028K 5028K 0K 142700 4882K 146K 3% 10 WebKit SVGDocument_0x10e166000 9216K 24K 24K 0K 3 7K 17K 71% 2 StyleRareNonInheritedData_0x10ca7b000 7168K 3232K 2228K 0K 1688 817K 1411K 64% 7 ImmutableStyleProperties_0x10c83c000 7168K 3828K 3828K 0K 37643 3732K 96K 3% 7 DFGNode_0x10d197000 7168K 5116K 4112K 0K 1 16 4112K 100% 7 StringSourceProvider_0x10c90b000 6144K 96K 96K 0K 273 30K 66K 69% 6 FastBitVector_0x10c939000 6144K 748K 748K 0K 41123 653K 95K 13% 6 StyleTransformData_0x10ca92000 6144K 72K 72K 0K 32 1520 71K 98% 6 MarkedBlockHandle_0x10c950000 6144K 1484K 1484K 0K 7880 1108K 376K 26% 6 WebKit DOMWindow_0x10ca36000 6144K 72K 72K 0K 16 5K 67K 93% 6 WebKit HTMLImageElement_0x10df80000 6144K 168K 168K 0K 235 73K 95K 57% 6 WebKit RenderInline_0x10de6c000 6144K 200K 200K 0K 374 70K 130K 66% 6 MetaAllocatorFreeSpace_0x10a384000 6144K 460K 460K 0K 4683 220K 240K 53% 6 WebKit RenderTextFragment_0x10d0c5000 6144K 268K 268K 0K 712 100K 168K 63% 6 WebKit InlineTextBox_0x10d121000 6144K 300K 300K 0K 584 55K 245K 82% 5 WebKit StaticElementList_0x10d292000 6144K 72K 72K 0K 2 80 72K 100% 6 WebKit HTMLUListElement_0x10dfc5000 6144K 92K 92K 0K 205 22K 70K 76% 6 SourceProviderCache_0x10ce3c000 6144K 72K 72K 0K 3 96 72K 100% 6 WebKit EllipsisBox_0x10e297000 6144K 132K 132K 0K 181 14K 118K 90% 4 WebKit HTMLAnchorElement_0x10d008000 6144K 80K 80K 0K 174 24K 56K 70% 6 WebKit Performance_0x10d07b000 6144K 72K 72K 0K 13 3680 68K 96% 6 WebKit HTMLScriptElement_0x10ce0e000 6144K 76K 76K 0K 99 17K 59K 78% 6 MutableStyleProperties_0x10c8c6000 6144K 188K 188K 0K 934 88K 100K 54% 6 WebKit RenderImage_0x10e038000 6144K 72K 72K 0K 15 3616 68K 96% 6 WatchpointSet_0x10cc9e000 6144K 3020K 3020K 0K 90355 2824K 196K 7% 6 MessageEvent_0x10d10a000 6144K 124K 124K 0K 5 784 123K 100% 6 PerformanceEntry_0x10d0f3000 6144K 136K 136K 0K 200 75K 61K 46% 6 StyleInheritedData_0x10cac0000 6144K 1180K 1180K 0K 1791 280K 900K 77% 6 WebKit RootInlineBox_0x10d208000 6144K 508K 508K 0K 606 104K 404K 80% 6 WebKit RenderFlexibleBox_0x10de55000 6144K 1244K 1244K 0K 1616 580K 664K 54% 6 SymbolTableEntryFatEntry_0x10cd28000 6144K 1148K 1148K 0K 65515 1024K 124K 11% 6 CodeBlockRareData_0x10cef4000 6144K 136K 136K 0K 197 18K 118K 87% 6 WebKit CSSComputedStyleDeclaration_0x10def6000 6144K 108K 108K 0K 33 1072 107K 100% 6 QualifiedNameQualifiedNameImpl_0x10bbdf000 6144K 116K 116K 0K 1079 51K 65K 57% 6 NodeListsNodeData_0x10d264000 6144K 88K 88K 0K 155 14K 74K 84% 6 CSSSelectorRareData_0x10c881000 6144K 116K 116K 0K 681 53K 63K 55% 6 ParserArenaRoot_0x10ce81000 6144K 72K 72K 0K 1 16 72K 100% 6 WebKit UndoManager_0x10ca1f000 6144K 72K 72K 0K 16 752 71K 99% 6 AccessCase_0x10d13b000 6144K 1256K 1256K 0K 12955 614K 642K 52% 6 WebKit RenderView_0x10cb05000 6144K 72K 72K 0K 15 9K 63K 88% 6 RenderLayer_0x10cb4a000 6144K 1500K 1500K 0K 2138 735K 765K 52% 6 Font_0x10cc02000 6144K 124K 124K 0K 60 35K 89K 72% 5 WebKit HTMLBodyElement_0x10cbbd000 6144K 72K 72K 0K 14 1488 71K 98% 6 StyleRuleBase_0x10c80e000 6144K 100K 100K 0K 1046 37K 63K 64% 6 WebKit DOMRectReadOnly_0x10de10000 6144K 92K 92K 0K 1 32 92K 100% 6 WebKit HTMLHeadElement_0x10cba6000 6144K 72K 72K 0K 14 1488 71K 98% 6 WebKit HTMLSpanElement_0x10de3e000 6144K 516K 516K 0K 2881 315K 201K 39% 6 SerializedScriptValue_0x10d221000 6144K 96K 96K 0K 6 512 96K 100% 6 NodeRareData_0x10cdc9000 6144K 552K 552K 0K 2747 385K 167K 31% 6 WebKit HTMLTableRowElement_0x10deb1000 6144K 92K 92K 0K 205 22K 70K 76% 6 WebKit HTMLTableCellElement_0x10dec8000 6144K 388K 388K 0K 2011 220K 168K 44% 6 WebKit Location_0x10cf39000 6144K 72K 72K 0K 12 368 72K 100% 6 WebKit ChildNodeList_0x10de27000 6144K 72K 72K 0K 1 32 72K 100% 6 WebKit InlineCSSStyleDeclaration_0x10ddcb000 6144K 92K 92K 0K 308 14K 78K 85% 6 MetaAllocatorHandle_0x10c9ac000 6144K 1812K 1812K 0K 12935 808K 1004K 56% 6 BitVector_0x10d1dc000 6144K 72K 72K 0K 1 16 72K 100% 6 WebKit InlineElementBox_0x10d24d000 6144K 72K 72K 0K 43 2720 69K 97% 6 StyleRule_0x10c853000 6144K 1224K 1224K 0K 36856 1152K 72K 6% 6 BasicBlock_0x10d180000 6144K 332K 332K 0K 1 16 332K 100% 6 RuleSet_0x10bbf6000 6144K 120K 120K 0K 74 49K 71K 60% 6 DirectJITCode_0x10ccfa000 6144K 1484K 1484K 0K 4185 1068K 416K 28% 6 StringWrapperCFAllocator_0x10bbc3000 6144K 72K 72K 0K 87 5K 67K 93% 5 WebKit PseudoElement_0x10cc9c000 6144K 240K 240K 0K 713 89K 151K 63% 6 PolymorphicAccess_0x10d152000 6144K 884K 884K 0K 5997 375K 509K 58% 5 WebKit HTMLHtmlElement_0x10cb8f000 6144K 72K 72K 0K 14 1488 71K 98% 6 WebKit RenderScrollbarPart_0x10e11e000 6144K 72K 72K 0K 27 6K 66K 92% 6 WebKit ClassCollection_0x10dfdc000 6144K 108K 108K 0K 216 50K 58K 54% 6 WebKit RenderBlockFlow_0x10cc19000 6144K 1056K 1056K 0K 1914 508K 548K 52% 5 Watchpoint_0x10cf0b000 6144K 240K 240K 0K 3853 120K 120K 50% 6 BytecodeGenerator_0x10ceaf000 6144K 72K 72K 0K 1 32 72K 100% 6 WebKit HTMLDivElement_0x10cf7e000 6144K 500K 500K 0K 3316 363K 137K 28% 6 WebKit RenderText_0x10d0dc000 6144K 204K 204K 0K 616 67K 137K 68% 6 WebKit DocumentFragment_0x10d2a9000 6144K 72K 72K 0K 1 32 72K 100% 6 StyleRareInheritedData_0x10caa9000 6144K 308K 308K 0K 376 100K 208K 68% 6 ShareableElementData_0x10cd56000 6144K 252K 252K 0K 1772 133K 119K 48% 6 ElementData_0x10cdb2000 6144K 260K 260K 0K 1039 130K 130K 51% 6 BlockDirectory_0x10c967000 6144K 84K 84K 0K 83 22K 62K 75% 6 WebKit HTMLTagCollection_0x10e14f000 6144K 72K 72K 0K 17 3872 68K 95% 6 StyleBoxData_0x10ca4d000 6144K 620K 620K 0K 1856 145K 475K 77% 6 StyleResolver_0x10cb61000 6144K 76K 76K 0K 16 14K 62K 82% 6 BytecodeBasicBlock_0x10d1ae000 6144K 1648K 1648K 0K 20099 1570K 78K 5% 6 Region_0x10cb1c000 6144K 72K 72K 0K 1 16 72K 100% 6 WebKit RenderImageResource_0x10e04f000 6144K 72K 72K 0K 13 416 72K 100% 6 WeakBlock_0x10cccc000 6144K 3892K 3892K 0K 15085 3771K 121K 4% 6 WebKit Text_0x10cd9b000 6144K 404K 404K 0K 3194 249K 155K 39% 6 RegisterAtOffsetList_0x10c97e000 6144K 96K 96K 0K 1944 30K 66K 69% 6 StyleSurroundData_0x10ca64000 6144K 1328K 1328K 0K 1346 315K 1013K 77% 6 FontCascadeFonts_0x10cb33000 6144K 92K 92K 0K 72 16K 76K 84% 6 RuleDataVector_0x10c8af000 6144K 1348K 1348K 0K 19865 931K 417K 31% 6 RenderStyle_0x10cbeb000 6144K 1704K 1704K 0K 1380 129K 1575K 93% 6 WebKit HTMLLIElement_0x10d036000 6144K 196K 196K 0K 842 92K 104K 54% 6 WebKit HTMLFormControlsCollection_0x10e135000 5120K 60K 60K 0K 1 48 60K 100% 5 WebKit RenderTable_0x10e0c2000 5120K 60K 60K 0K 7 2432 58K 97% 5 WebKit VisualViewport_0x10cc30000 5120K 60K 60K 0K 13 1952 58K 97% 5 WebKit AllDescendantsCollection_0x10df24000 5120K 60K 60K 0K 10 2048 58K 97% 5 WebKit Navigator_0x10cf22000 5120K 60K 60K 0K 9 928 59K 99% 5 WebKit RenderIFrame_0x10ddf9000 5120K 60K 60K 0K 7 1664 58K 98% 5 StyleFillData_0x10caee000 5120K 60K 60K 0K 21 976 59K 99% 5 WebKit Storage_0x10cf50000 5120K 60K 60K 0K 8 240 60K 100% 5 WebKit RenderTableSection_0x10e0f0000 5120K 60K 60K 0K 6 1712 58K 98% 5 WebKit HTMLIFrameElement_0x10dde2000 5120K 60K 60K 0K 12 3024 57K 96% 5 WebKit HTMLStyleElement_0x10cdf7000 5120K 60K 60K 0K 16 2432 58K 97% 5 WebKit_GenericChildrenOnlyCachedHTMLCollection_0x10e138000 5120K 60K 60K 0K 13 1200 59K 99% 5 WebKit RenderTableRow_0x10e107000 5120K 60K 60K 0K 12 2672 57K 96% 5 SVGRenderStyle_0x10cad7000 5120K 60K 60K 0K 30 2800 57K 96% 5 WebKit RenderTableCell_0x10e0ab000 5120K 60K 60K 0K 12 3200 57K 95% 5 TableLayout_0x10e0d9000 5120K 60K 60K 0K 7 400 60K 100% 5 WebKit HTMLTableSectionElement_0x10de9a000 4096K 48K 48K 0K 7 704 47K 99% 4 WebKit SVGSVGElement_0x10df3b000 4096K 48K 48K 0K 18 7K 41K 87% 4 WebKit DocumentType_0x10cd3f000 4096K 48K 48K 0K 11 992 47K 98% 4 WebKit HTMLElement_0x10cfda000 4096K 48K 48K 0K 8 816 47K 99% 4 WebKit HTMLMetaElement_0x10cd6d000 4096K 48K 48K 0K 15 1600 46K 97% 4 WebKit MessagePort_0x10e205000 4096K 48K 48K 0K 13 1952 46K 97% 4 WebKit RenderSVGRoot_0x10e00a000 4096K 48K 48K 0K 11 4K 44K 91% 4 WebKit HTMLTableElement_0x10de83000 4096K 48K 48K 0K 8 928 47K 99% 4 WebKit RenderSVGPath_0x10e021000 4096K 48K 48K 0K 21 6K 42K 87% 4 WebKit HTMLHeadingElement_0x10e17d000 4096K 48K 48K 0K 7 704 47K 99% 4 WebKit SVGPathElement_0x10df69000 4096K 52K 52K 0K 32 10K 42K 82% 4 WebKit_GenericDescendantsCachedHTMLCollection_0x10d27b000 4096K 48K 48K 0K 6 1168 47K 98% 4 SMILTimeContainer_0x10df52000 4096K 48K 48K 0K 18 2480 46K 95% 4 WebKit HTMLBRElement_0x10dedf000 4096K 48K 48K 0K 7 704 47K 99% 4 WebKit HTMLInputElement_0x10cfac000 3072K 36K 36K 0K 10 2192 34K 95% 3 WebKit InlineFlowBox_0x10e2c8000 3072K 36K 36K 0K 3 256 36K 100% 3 CSSFontFace_0x10cc8c000 3072K 52K 52K 0K 78 16K 36K 70% 2 WebKit TextControlInnerTextElement_0x10d04d000 3072K 36K 36K 0K 4 384 36K 99% 3 WebKit HTMLTableRowsCollection_0x10e2d5000 3072K 36K 36K 0K 5 352 36K 100% 3 WebKit RenderButton_0x10d0ae000 3072K 36K 36K 0K 8 2720 33K 93% 3 WebKit HTMLLinkElement_0x10cde0000 3072K 36K 36K 0K 11 2752 33K 93% 3 WebKit RenderImageResourceStyleImage_0x10e094000 3072K 36K 36K 0K 3 144 36K 100% 3 WebKit Screen_0x10e1ee000 3072K 36K 36K 0K 5 144 36K 100% 3 WebKit DOMMimeTypeArray_0x10e1ad000 3072K 36K 36K 0K 4 128 36K 100% 3 WebKit ShadowRoot_0x10cfc3000 3072K 36K 36K 0K 10 1904 34K 95% 3 WebKit HTMLTableColElement_0x10e269000 3072K 36K 36K 0K 23 2496 34K 94% 3 WebKit DOMPluginArray_0x10e196000 3072K 36K 36K 0K 4 128 36K 100% 3 WebKit RenderTableCol_0x10e280000 2048K 24K 24K 0K 12 2672 21K 90% 2 WebKit HTMLTitleElement_0x10cd84000 2048K 24K 24K 0K 4 416 24K 99% 2 WebKit IDBDatabase_0x10e311000 2048K 24K 24K 0K 1 32 24K 100% 2 WebKit HTMLButtonElement_0x10df97000 2048K 24K 24K 0K 11 1792 22K 93% 2 WebKit IDBOpenDBRequest_0x10e2ae000 2048K 24K 24K 0K 1 32 24K 100% 2 WebKit HTMLBaseElement_0x10d064000 2048K 24K 24K 0K 4 368 24K 99% 2 WebKit HTMLFormElement_0x10cf95000 2048K 24K 24K 0K 5 992 23K 96% 2 WebKit TrailingFloatsRootInlineBox_0x10e2fa000 2048K 24K 24K 0K 2 224 24K 100% 2 WebKit SVGRectElement_0x10e367000 1024K 12K 12K 0K 5 1504 11K 88% 1 WebKit TextControlPlaceholderElement_0x10dfae000 1024K 12K 12K 0K 2 160 12K 99% 1 WebKit History_0x10e23a000 1024K 12K 12K 0K 3 176 12K 99% 1 WebKit_SimulatedMouseEvent_0x10e23d000 1024K 12K 12K 0K 2 224 12K 99% 1 WebKit IDBRequest_0x10e2d1000 1024K 12K 12K 0K 1 32 12K 100% 1 WebKit IDBTransaction_0x10e2c5000 1024K 12K 12K 0K 1 32 12K 100% 1 WebKit RenderTextControlSingleLine_0x10e066000 1024K 12K 12K 0K 2 320 12K 98% 1 WebKit ServiceWorkerContainer_0x10d20a000 1024K 12K 12K 0K 3 544 11K 96% 1 WebKit HTMLHRElement_0x10df0d000 1024K 12K 12K 0K 2 144 12K 99% 1 WebKit HTMLCanvasElement_0x10e40f000 1024K 12K 12K 0K 3 480 12K 97% 1 WebKit RenderLineBreak_0x10dff3000 1024K 12K 12K 0K 4 608 11K 96% 1 WebKit CanvasRenderingContext2D_0x10e426000 1024K 12K 12K 0K 3 960 11K 93% 1 WebKit HTMLOListElement_0x10d01f000 1024K 12K 12K 0K 2 160 12K 99% 1 WebKit RenderTextControlInnerBlock_0x10e07d000 1024K 12K 12K 0K 2 320 12K 98% 1 WebKit TrackListBase_0x10e3c3000 1024K 12K 12K 0K 4 752 11K 94% 1 WebKit SVGGElement_0x10e2d3000 1024K 12K 12K 0K 3 640 11K 95% 1 WebKit MutationObserver_0x10e17f000 1024K 12K 12K 0K 2 128 12K 99% 1 WebKit AbortSignal_0x10e37e000 1024K 12K 12K 0K 1 32 12K 100% 1 WebKit HTMLParagraphElement_0x10cff1000 1024K 12K 12K 0K 2 144 12K 99% 1 WebKit HTMLUnknownElement_0x10e395000 1024K 12K 12K 0K 1 32 12K 100% 1 DefaultPurgeableMallocZone_0x10e21c000 116K 116K 116K 0K 2 112K 4K 4% 2 =========== ======= ========= ========= ========= ========= ========= ========= ====== ====== TOTAL 3.7G 650.1M 528.8M 0K 1881880 340.1M 188.7M 36% 1293

Yusuke Suzuki

Comment 93 2019-05-31 07:17:51 PDT

1% improvement on iOS RAMification

Christopher Reid

Comment 94 2019-05-31 17:21:35 PDT

WinCairo is crashing in llint_entry after this change. https://build.webkit.org/builders/WinCairo%2064-bit%20WKL%20Release%20%28Tests%29?numbuilds=50 It hits an access violation writing to null inside llint_entry. llint_entry asm section up to the point of the crash: > llint_entry PROC PUBLIC > ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:876 > ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:878 > push rbp > mov rbp, rsp ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:885 > call _relativePCBase ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:1319 > > _relativePCBase: > pop r9 ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:1321 > lea r10, [_llint_op_has_indexed_property - _relativePCBase + r9] ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:1350 > xor rcx, rcx ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:1351 > mov qword ptr [0 + rcx + rcx * 8], r10 ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:1352

Yusuke Suzuki

Comment 95 2019-05-31 17:43:01 PDT

(In reply to Christopher Reid from comment #94) > WinCairo is crashing in llint_entry after this change. > https://build.webkit.org/builders/WinCairo%2064- > bit%20WKL%20Release%20%28Tests%29?numbuilds=50 > > It hits an access violation writing to null inside llint_entry. > llint_entry asm section up to the point of the crash: > > > llint_entry PROC PUBLIC > > ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:876 > > ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:878 > > push rbp > > mov rbp, rsp ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:885 > > call _relativePCBase ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:1319 > > > > _relativePCBase: > > pop r9 ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:1321 > > lea r10, [_llint_op_has_indexed_property - _relativePCBase + r9] ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:1350 > > xor rcx, rcx ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:1351 > > mov qword ptr [0 + rcx + rcx * 8], r10 ; C:/git/webkit/Source/JavaScriptCore/llint/LowLevelInterpreter.asm:1352 Maybe I have a fix now. I'll land it as an unreviewed fix and see the results in the bots. But why is EWS green?

Yusuke Suzuki

Comment 96 2019-05-31 18:00:06 PDT

Committed r246006: <https://trac.webkit.org/changeset/246006>

Yusuke Suzuki

Comment 97 2019-05-31 18:00:56 PDT

I see, WinCairo EWS is just a build testing.....

Note You need to log in before you can comment on or make changes to this bug.