https://git.ruby-lang.org/ruby.git/commit/?id=b06a4dc6f1

From b06a4dc6f1fbef12ff7b4d57e7b5f12fd1f6cb5b Mon Sep 17 00:00:00 2001
From: Aaron Patterson <tenderlove@r...>
Date: Thu, 25 Jun 2020 15:20:25 -0700
Subject: Expand heap pages to be exactly 16kb

This commit expands heap pages to be exactly 16KiB and eliminates the
`REQUIRED_SIZE_BY_MALLOC` constant.

I believe the goal of `REQUIRED_SIZE_BY_MALLOC` was to make the heap
pages consume some multiple of OS page size.  16KiB is convenient because
OS page size is typically 4KiB, so one Ruby page is four OS pages.

Do not guess how malloc works
=============================

We should not try to guess how `malloc` works and instead request (and
use) four OS pages.

Here is my reasoning:

1. Not all mallocs will store metadata in the same region as user requested
memory.  jemalloc specifically states[1]:

> Information about the states of the runs is stored as a page map at the beginning of each chunk.

2. We're using `posix_memalign` to request memory.  This means that the
   first address must be divisible by the alignment.  Our allocation is
   page aligned, so if malloc is storing metadata *before* the page,
   then we've already crossed page boundaries.

3. Some allocators like glibc will use the memory at the end of the
   page.  I am able to demonstrate that glibc will return pointers
   within the page boundary that contains `heap_page_body`[2].  We
   *expected* the allocation to look like this:

![Expected alignment](https://user-images.githubusercontent.com/3124/85803661-8a81d600-b6fc-11ea-8cb6-7dbdb434a43b.png)

   But since `heap_page` is allocated immediately after
   `heap_page_body`[3], instead the layout looks like this:

![Actual alignment](https://user-images.githubusercontent.com/3124/85803714-a1c0c380-b6fc-11ea-8c17-8b37369e17ee.png)

   This is not optimal because `heap_page` gets allocated immediately
   after `heap_page_body`.  We frequently write to `heap_page`, so the
   bottom OS page of `heap_page_body` is very likely to be copied.

One more object per page
========================

In jemalloc, allocation requests are rounded to the nearest boundary,
which in this case is 16KiB[4], so `REQUIRED_SIZE_BY_MALLOC` space is
just wasted on jemalloc.

On glibc, the space is not wasted, but instead it is very likely to
cause page faults.

Instead of wasting space or causing page faults, lets just use the space
to store one more Ruby object.  Using the space to store one more Ruby
object will prevent page faults, stop wasting space, decrease memory
usage, decrease GC time, etc.

1. https://people.freebsd.org/~jasone/jemalloc/bsdcan2006/jemalloc.pdf
2. https://github.com/ruby/ruby/commit/33390d15e7a6f803823efcb41205167c8b126fbb
3  https://github.com/ruby/ruby/blob/289a28e68f30e879760fd000833b512d506a0805/gc.c#L1757-L1763
4. https://people.freebsd.org/~jasone/jemalloc/bsdcan2006/jemalloc.pdf page 4

Co-authored-by: John Hawthorn <john@h...>

diff --git a/gc.c b/gc.c
index a2a7997..d1af8ba 100644
--- a/gc.c
+++ b/gc.c
@@ -805,8 +805,7 @@ typedef struct rb_objspace { https://github.com/ruby/ruby/blob/trunk/gc.c#L805
 enum {
     HEAP_PAGE_ALIGN = (1UL << HEAP_PAGE_ALIGN_LOG),
     HEAP_PAGE_ALIGN_MASK = (~(~0UL << HEAP_PAGE_ALIGN_LOG)),
-    REQUIRED_SIZE_BY_MALLOC = (sizeof(size_t) * 5),
-    HEAP_PAGE_SIZE = (HEAP_PAGE_ALIGN - REQUIRED_SIZE_BY_MALLOC),
+    HEAP_PAGE_SIZE = HEAP_PAGE_ALIGN,
     HEAP_PAGE_OBJ_LIMIT = (unsigned int)((HEAP_PAGE_SIZE - sizeof(struct heap_page_header))/sizeof(struct RVALUE)),
     HEAP_PAGE_BITMAP_LIMIT = CEILDIV(CEILDIV(HEAP_PAGE_SIZE, sizeof(struct RVALUE)), BITS_BITLENGTH),
     HEAP_PAGE_BITMAP_SIZE = (BITS_SIZE * HEAP_PAGE_BITMAP_LIMIT),
@@ -4187,20 +4186,20 @@ gc_page_sweep(rb_objspace_t *objspace, rb_heap_t *heap, struct heap_page *sweep_ https://github.com/ruby/ruby/blob/trunk/gc.c#L4186
 {
     int i;
     int empty_slots = 0, freed_slots = 0, final_slots = 0;
-    RVALUE *p, *pend,*offset;
+    RVALUE *p, *offset;
     bits_t *bits, bitset;
 
     gc_report(2, objspace, "page_sweep: start.\n");
 
     sweep_page->flags.before_sweep = FALSE;
 
-    p = sweep_page->start; pend = p + sweep_page->total_slots;
+    p = sweep_page->start;
     offset = p - NUM_IN_PAGE(p);
     bits = sweep_page->mark_bits;
 
     /* create guard : fill 1 out-of-range */
     bits[BITMAP_INDEX(p)] |= BITMAP_BIT(p)-1;
-    bits[BITMAP_INDEX(pend)] |= ~(BITMAP_BIT(pend) - 1);
+    bits[BITMAP_INDEX(p) + sweep_page->total_slots / BITS_BITLENGTH] |= ~((1 << ((NUM_IN_PAGE(p) + sweep_page->total_slots) % BITS_BITLENGTH)) - 1);
 
     for (i=0; i < HEAP_PAGE_BITMAP_LIMIT; i++) {
 	bitset = ~bits[i];
-- 
cgit v0.10.2


--
ML: ruby-changes@q...
Info: http://www.atdot.net/~ko1/quickml/