diff options
author | Alan Wu <XrXr@users.noreply.github.com> | 2019-07-30 21:36:05 -0400 |
---|---|---|
committer | 卜部昌平 <shyouhei@ruby-lang.org> | 2019-10-24 18:03:42 +0900 |
commit | 89e7997622038f82115f34dbb4ea382e02bed163 (patch) | |
tree | 993a5f6fb17418381e835be1fd51093dc620148a /mjit_compile.c | |
parent | 38e931fa2ceac6d922f3eabedb8f35f211de0bdb (diff) |
Combine call info and cache to speed up method invocation
To perform a regular method call, the VM needs two structs,
`rb_call_info` and `rb_call_cache`. At the moment, we allocate these two
structures in separate buffers. In the worst case, the CPU needs to read
4 cache lines to complete a method call. Putting the two structures
together reduces the maximum number of cache line reads to 2.
Combining the structures also saves 8 bytes per call site as the current
layout uses separate two pointers for the call info and the call cache.
This saves about 2 MiB on Discourse.
This change improves the Optcarrot benchmark at least 3%. For more
details, see attached bugs.ruby-lang.org ticket.
Complications:
- A new instruction attribute `comptime_sp_inc` is introduced to
calculate SP increase at compile time without using call caches. At
compile time, a `TS_CALLDATA` operand points to a call info struct, but
at runtime, the same operand points to a call data struct. Instruction
that explicitly define `sp_inc` also need to define `comptime_sp_inc`.
- MJIT code for copying call cache becomes slightly more complicated.
- This changes the bytecode format, which might break existing tools.
[Misc #16258]
Notes
Notes:
Merged: https://github.com/ruby/ruby/pull/2564
Diffstat (limited to 'mjit_compile.c')
-rw-r--r-- | mjit_compile.c | 20 |
1 files changed, 18 insertions, 2 deletions
diff --git a/mjit_compile.c b/mjit_compile.c index b6ed984b20..27ea836ef4 100644 --- a/mjit_compile.c +++ b/mjit_compile.c @@ -25,6 +25,21 @@ #define NOT_COMPILED_STACK_SIZE -1 #define ALREADY_COMPILED_P(status, pos) (status->stack_size_for_pos[pos] != NOT_COMPILED_STACK_SIZE) +static size_t +call_data_index(CALL_DATA cd, const struct rb_iseq_constant_body *body) +{ + const struct rb_kwarg_call_data *kw_calls = (const struct rb_kwarg_call_data *)&body->call_data[body->ci_size]; + const struct rb_kwarg_call_data *kw_cd = (const struct rb_kwarg_call_data *)cd; + + VM_ASSERT(cd >= body->call_data && kw_cd < (kw_calls + body->ci_kw_size)); + if (kw_cd < kw_calls) { + return cd - body->call_data; + } + else { + return kw_cd - kw_calls + body->ci_size; + } +} + // For propagating information needed for lazily pushing a frame. struct inlined_call_context { int orig_argc; // ci->orig_argc @@ -383,8 +398,9 @@ precompile_inlinable_iseqs(FILE *f, const rb_iseq_t *iseq, struct compile_status #endif if (insn == BIN(opt_send_without_block)) { // `compile_inlined_cancel_handler` supports only `opt_send_without_block` - CALL_INFO ci = (CALL_INFO)body->iseq_encoded[pos + 1]; - CALL_CACHE cc_copy = status->cc_entries + ((CALL_CACHE)body->iseq_encoded[pos + 2] - body->cc_entries); // use copy to avoid race condition + CALL_DATA cd = (CALL_DATA)body->iseq_encoded[pos + 1]; + CALL_INFO ci = &cd->ci; + CALL_CACHE cc_copy = status->cc_entries + call_data_index(cd, body); // use copy to avoid race condition const rb_iseq_t *child_iseq; if (has_valid_method_type(cc_copy) && |