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]
This reverts commit 43015275b9.
`assert_raise_message` in test-unit is different from
`assert_raise_with_message`. It checks the exception message
only, but not the exception class,
parallel test (`make test-all TESTS=-j8`) runs tests on specified
number of processes. However, some test checkers written in
`runner.rb` are not loaded. This fix enable these checkers on
parallel tests.
See also: https://github.com/ruby/ruby/pull/2508
There are random failures:
> lib/rubygems/core_ext/kernel_require.rb:61:in `require':
> wrong number of arguments (given 1, expected 0) (ArgumentError)
http://ci.rvm.jp/results/trunk-jemalloc@silicon-docker/2275159
To check this failure, I added a small check code.
The original st.c was public domain hash table implementation, but
Ruby's st.c is highly modified, and its data structure is not
compatiblie with the original one.
Therefore, when creating an extension library to wrap C code that uses
the original st.c, the symbols conflict, which leads to segfault.
This changes the prefix `st_*` of st.c functions to `rb_st_*` for
reflecting that they are specific to Ruby's, and avoid symbol conflicts.
The built-in version operates on a buffer of 5 words, much smaller than
the size of jmp_buf defined in libc.
Note, powerpc requires 5 words, while arm and x86_64 just require 3.
Nobuyoshi Nakada
Hiroshi SHIBATA
Yusuke Endoh
卜部昌平
Alan Wu
NARUSE, Yui
Kazuhiro NISHIYAMA
Koichi Sasada
NAKAMURA Usaku
Benoit Daloze
Fangrui Song