From: samuel@...
Date: 2019-07-16T02:34:41+00:00
Subject: [ruby-core:93801] [Ruby master Feature#15997] Improve performance of fiber creation by using pool allocation strategy.

Issue #15997 has been updated by ioquatix (Samuel Williams).


I've updated the implementation for fiber pool, which now has some functionality controlled by `#define FIBER_POOL_ALLOCATION_FREE`.

The normal (CPU efficient, memory expensive) implementation creates and reuses `fiber_pool_allocation` indefinitely, and never returns the resources back to the system, so address space is not released.

If you defined `FIBER_POOL_ALLOCATION_FREE`, during `fiber_pool_stack_release`:

1/ We use `madvise(free)` to clear dirty bit on stack memory (to avoid swapping to disk under memory pressure) and,
2/ We use `munmap` on the `fiber_pool_allocation` when it���s usage drops to 0 (extra book keeping required), and:
    - remove it from allocation list (double linked list required)
    - remove all `fiber_pool_vacancy` from vacancy list (double linked list required).

The consequence of `#define FIBER_POOL_ALLOCATION_FREE` is that `fiber_pool_stack_acquire` and `fiber_pool_stack_release` becomes more CPU expensive, but address space is released back to the system when possible and dirty pages are cleared so that swap space is not consumed.

1/ `fiber_pool_stack_release` will always call `madvise(free)` and occasionally `munmap`.
2/ `fiber_pool_stack_acquire` is more likely to call `mmap` and `mprotect` if more stacks are required.

We could merge this PR, and decide whether we want to be conservative or not, or maybe do it on a arch basis (e.g. 32-bit could be conservative vs 64-bit since address space is less of a concern).

----------------------------------------
Feature #15997: Improve performance of fiber creation by using pool allocation strategy.
https://bugs.ruby-lang.org/issues/15997#change-79667

* Author: ioquatix (Samuel Williams)
* Status: Open
* Priority: Normal
* Assignee: ko1 (Koichi Sasada)
* Target version: 
----------------------------------------
https://github.com/ruby/ruby/pull/2224

This PR improves the performance of fiber allocation and reuse by implementing a better stack cache.

The fiber pool manages a singly linked list of fiber pool allocations. The fiber pool allocation contains 1 or more stack (typically more, e.g. 512). It uses N^2 allocation strategy, starting at 8 initial stacks, next is 8, 16, 32, etc.

```
//
// base = +-------------------------------+-----------------------+  +
//        |VM Stack       |VM Stack       |                       |  |
//        |               |               |                       |  |
//        |               |               |                       |  |
//        +-------------------------------+                       |  |
//        |Machine Stack  |Machine Stack  |                       |  |
//        |               |               |                       |  |
//        |               |               |                       |  |
//        |               |               | .  .  .  .            |  |  size
//        |               |               |                       |  |
//        |               |               |                       |  |
//        |               |               |                       |  |
//        |               |               |                       |  |
//        |               |               |                       |  |
//        +-------------------------------+                       |  |
//        |Guard Page     |Guard Page     |                       |  |
//        +-------------------------------+-----------------------+  v
//
//        +------------------------------------------------------->
//
//                                  count
//
```

The performance improvement depends on usage:

```
Calculating -------------------------------------
                     compare-ruby  built-ruby 
  vm2_fiber_allocate     132.900k    180.852k i/s -    100.000k times in 0.752447s 0.552939s
     vm2_fiber_count       5.317k    110.724k i/s -    100.000k times in 18.806479s 0.903145s
     vm2_fiber_reuse      160.128     347.663 i/s -     200.000 times in 1.249003s 0.575269s
    vm2_fiber_switch      13.429M     13.490M i/s -     20.000M times in 1.489303s 1.482549s

Comparison:
               vm2_fiber_allocate
          built-ruby:    180851.6 i/s 
        compare-ruby:    132899.7 i/s - 1.36x  slower

                  vm2_fiber_count
          built-ruby:    110724.3 i/s 
        compare-ruby:      5317.3 i/s - 20.82x  slower

                  vm2_fiber_reuse
          built-ruby:       347.7 i/s 
        compare-ruby:       160.1 i/s - 2.17x  slower

                 vm2_fiber_switch
          built-ruby:  13490282.4 i/s 
        compare-ruby:  13429100.0 i/s - 1.00x  slower
```

This test is run on Linux server with 64GB memory and 4-core Xeon (Intel Xeon CPU E3-1240 v6 @ 3.70GHz). "compare-ruby" is `master`, and "built-ruby" is `master+fiber-pool`.

Additionally, we conservatively use `madvise(free)` to avoid swap space usage for unused fiber stacks. However, if you remove this requirement, we can get 6x - 10x performance improvement in `vm2_fiber_reuse` benchmark. There are some options to deal with this (e.g. moving it to `GC.compact`) but as this is still a net win, I'd like to merge this PR as is.





-- 
https://bugs.ruby-lang.org/

Unsubscribe: <mailto:ruby-core-request@ruby-lang.org?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>