From: Aldy Hernandez <aldyh@redhat.com>
To: Andrew MacLeod <amacleod@redhat.com>,
David Malcolm <dmalcolm@redhat.com>,
gcc-patches <gcc-patches@gcc.gnu.org>
Subject: Re: [PATCH] irange_pool class
Date: Fri, 18 Sep 2020 19:03:28 +0200 [thread overview]
Message-ID: <41967a78-56f3-675c-9243-06d07668e8cd@redhat.com> (raw)
In-Reply-To: <40f22a52-03f0-483f-ce84-d80f15a1b705@redhat.com>
[-- Attachment #1: Type: text/plain, Size: 1407 bytes --]
On 9/18/20 6:42 PM, Andrew MacLeod wrote:
> On 9/18/20 8:28 AM, David Malcolm wrote:I think of a "pool allocator" as
> something that makes a small
>>>> number of
>>>> large allocation under the covers, and then uses that to serve
>>>> large
>>>> numbers of fixed sized small allocations and deallocations with
>>>> O(1)
>>>> using a free list.
>>> Ah, I didn't know pool had a different meaning.
>> See e.g. gcc/alloc-pool.h
>
> The name originated when the original v1 version was based on using
> alloc-pool.h. when we went to varying sizes, we switched to and obstack
> implementation and never changed the name.
> <...>
>
>>>> I think it would be clearer to name this "irange_obstack", or
>>>> somesuch.
>>> I'd prefer something more generic. We don't want to tie the name of
>>> the
>>> allocator to the underlying implementation. What if we later change
>>> to
>>> malloc? We'd have to change the name to irange_malloc.
>>> irange_allocator? Or is there something more generically appropriate
>>> here?
>> How about "irange_bump_allocator?" Rather long, but it expresses the
>
>
>
> "irange_allocator" is sufficient . The consumer should not care
> what the implementation is, and we may decide to implement it
> differently down the road. So I don't want to imply something specific
> in the name or we'd have to change it again.
Updated patch attached.
Aldy
[-- Attachment #2: curr.patch --]
[-- Type: text/x-patch, Size: 3026 bytes --]
commit 343463f8887ab510503d9e230268963a299a84ef
Author: Aldy Hernandez <aldyh@redhat.com>
Date: Fri Sep 11 10:15:12 2020 +0200
irange_allocator class
This is the irange storage class. It is used to allocate the
minimum amount of storage needed for a given irange. Storage is
automatically freed at destruction of the storage class.
It is meant for long term storage, as opposed to int_range_max
which is meant for intermediate temporary results on the stack.
The general gist is:
irange_allocator alloc;
// Allocate an irange of 5 sub-ranges.
irange *p = alloc.allocate (5);
// Allocate an irange of 3 sub-ranges.
irange *q = alloc.allocate (3);
// Allocate an irange with as many sub-ranges as are currently
// used in "some_other_range".
irange *r = alloc.allocate (some_other_range);
diff --git a/gcc/value-range.h b/gcc/value-range.h
index 8497791c7b3..c875e713d65 100644
--- a/gcc/value-range.h
+++ b/gcc/value-range.h
@@ -43,6 +43,7 @@ enum value_range_kind
class irange
{
+ friend class irange_allocator;
public:
// In-place setters.
void set (tree, tree, value_range_kind = VR_RANGE);
@@ -619,4 +620,68 @@ vrp_val_min (const_tree type)
return NULL_TREE;
}
+// This is the irange storage class. It is used to allocate the
+// minimum amount of storage needed for a given irange. Storage is
+// automatically freed at destruction of the storage class.
+//
+// It is meant for long term storage, as opposed to int_range_max
+// which is meant for intermediate temporary results on the stack.
+//
+// The newly allocated irange is initialized to the empty set
+// (undefined_p() is true).
+
+class irange_allocator
+{
+public:
+ irange_allocator ();
+ ~irange_allocator ();
+ // Return a new range with NUM_PAIRS.
+ irange *allocate (unsigned num_pairs);
+ // Return a copy of SRC with the minimum amount of sub-ranges needed
+ // to represent it.
+ irange *allocate (const irange &src);
+private:
+ DISABLE_COPY_AND_ASSIGN (irange_allocator);
+ struct obstack m_obstack;
+};
+
+inline
+irange_allocator::irange_allocator ()
+{
+ obstack_init (&m_obstack);
+}
+
+inline
+irange_allocator::~irange_allocator ()
+{
+ obstack_free (&m_obstack, NULL);
+}
+
+// Return a new range with NUM_PAIRS.
+
+inline irange *
+irange_allocator::allocate (unsigned num_pairs)
+{
+ // Never allocate 0 pairs.
+ // Don't allocate 1 either, or we get legacy value_range's.
+ if (num_pairs < 2)
+ num_pairs = 2;
+
+ struct newir {
+ irange range;
+ tree mem[1];
+ };
+ size_t nbytes = (sizeof (newir) + sizeof (tree) * 2 * (num_pairs - 1));
+ struct newir *r = (newir *) obstack_alloc (&m_obstack, nbytes);
+ return new (r) irange (r->mem, num_pairs);
+}
+
+inline irange *
+irange_allocator::allocate (const irange &src)
+{
+ irange *r = allocate (src.num_pairs ());
+ *r = src;
+ return r;
+}
+
#endif // GCC_VALUE_RANGE_H
next prev parent reply other threads:[~2020-09-18 17:03 UTC|newest]
Thread overview: 18+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-09-17 10:36 Aldy Hernandez
2020-09-17 18:02 ` Martin Sebor
2020-09-18 6:17 ` Aldy Hernandez
2020-09-18 1:43 ` David Malcolm
2020-09-18 5:49 ` Aldy Hernandez
2020-09-18 12:28 ` David Malcolm
2020-09-18 14:10 ` Aldy Hernandez
2020-09-18 17:07 ` Martin Sebor
2020-09-18 17:36 ` Andrew MacLeod
2020-09-18 20:35 ` Martin Sebor
2020-09-18 21:09 ` Andrew MacLeod
2020-09-19 20:32 ` Martin Sebor
2020-09-20 0:40 ` Andrew MacLeod
2020-09-20 7:01 ` Aldy Hernandez
2020-09-21 14:14 ` Andrew MacLeod
2020-09-18 16:42 ` Andrew MacLeod
2020-09-18 17:03 ` Aldy Hernandez [this message]
2020-09-28 15:56 ` Andrew MacLeod
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=41967a78-56f3-675c-9243-06d07668e8cd@redhat.com \
--to=aldyh@redhat.com \
--cc=amacleod@redhat.com \
--cc=dmalcolm@redhat.com \
--cc=gcc-patches@gcc.gnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).