project-base/server-skynet-source-3rd-jemalloc
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
dev
server-skynet-source-3rd-je.../include/jemalloc/internal/qr.h

141 lines
4.1 KiB
C
Raw Permalink Normal View History

Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
#ifndef JEMALLOC_INTERNAL_QR_H
#define JEMALLOC_INTERNAL_QR_H
QL, QR: Add documentation.
2020-04-11 02:38:33 +08:00
/*
* A ring implementation based on an embedded circular doubly-linked list.
*
* You define your struct like so:
*
* typedef struct my_s my_t;
* struct my_s {
* int data;
* qr(my_t) my_link;
* };
*
* And then pass a my_t * into macros for a_qr arguments, and the token
* "my_link" into a_field fields.
*/
Convert thread-specific caching from magazines, and implement incremental GC. Add the 'G'/'g' and 'H'/'h' MALLOC_OPTIONS flags. Add the malloc_tcache_flush() function. Disable thread-specific caching until the application goes multi-threaded.
2009-12-29 16:09:15 +08:00
/* Ring definitions. */
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define qr(a_type) \
Convert thread-specific caching from magazines, and implement incremental GC. Add the 'G'/'g' and 'H'/'h' MALLOC_OPTIONS flags. Add the malloc_tcache_flush() function. Disable thread-specific caching until the application goes multi-threaded.
2009-12-29 16:09:15 +08:00
struct { \
a_type *qre_next; \
a_type *qre_prev; \
}
QL, QR: Add documentation.
2020-04-11 02:38:33 +08:00
/*
* Initialize a qr link. Every link must be initialized before being used, even
* if that initialization is going to be immediately overwritten (say, by being
* passed into an insertion macro).
*/
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define qr_new(a_qr, a_field) do { \
Convert thread-specific caching from magazines, and implement incremental GC. Add the 'G'/'g' and 'H'/'h' MALLOC_OPTIONS flags. Add the malloc_tcache_flush() function. Disable thread-specific caching until the application goes multi-threaded.
2009-12-29 16:09:15 +08:00
(a_qr)->a_field.qre_next = (a_qr); \
(a_qr)->a_field.qre_prev = (a_qr); \
} while (0)
QL, QR: Add documentation.
2020-04-11 02:38:33 +08:00
/*
* Go forwards or backwards in the ring. Note that (the ring being circular), this
* always succeeds -- you just keep looping around and around the ring if you
* chase pointers without end.
*/
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define qr_next(a_qr, a_field) ((a_qr)->a_field.qre_next)
#define qr_prev(a_qr, a_field) ((a_qr)->a_field.qre_prev)
Convert thread-specific caching from magazines, and implement incremental GC. Add the 'G'/'g' and 'H'/'h' MALLOC_OPTIONS flags. Add the malloc_tcache_flush() function. Disable thread-specific caching until the application goes multi-threaded.
2009-12-29 16:09:15 +08:00
QL, QR: Add documentation.
2020-04-11 02:38:33 +08:00
/*
* Given two rings:
* a -> a_1 -> ... -> a_n --
* ^ |
* |------------------------
*
* b -> b_1 -> ... -> b_n --
* ^ |
* |------------------------
*
* Results in the ring:
* a -> a_1 -> ... -> a_n -> b -> b_1 -> ... -> b_n --
* ^ |
* |-------------------------------------------------|
*
* a_qr_a can directly be a qr_next() macro, but a_qr_b cannot.
*/
Optimize meld in qr module The goal of `qr_meld()` is to change the following four fields `(a->prev, a->prev->next, b->prev, b->prev->next)` from the values `(a->prev, a, b->prev, b)` to `(b->prev, b, a->prev, a)`. This commit changes ``` a->prev->next = b; b->prev->next = a; temp = a->prev; a->prev = b->prev; b->prev = temp; ``` to ``` temp = a->prev; a->prev = b->prev; b->prev = temp; a->prev->next = a; b->prev->next = b; ``` The benefit is that we can use `b->prev->next` for `temp`, and so there's no need to pass in `a_type`. The restriction is that `b` cannot be a `qr_next()` macro, so users of `qr_meld()` must pay attention. (Before this change, neither `a` nor `b` could be a `qr_next()` macro.)
2020-04-02 06:04:24 +08:00
#define qr_meld(a_qr_a, a_qr_b, a_field) do { \
(a_qr_b)->a_field.qre_prev->a_field.qre_next = \
(a_qr_a)->a_field.qre_prev; \
Convert thread-specific caching from magazines, and implement incremental GC. Add the 'G'/'g' and 'H'/'h' MALLOC_OPTIONS flags. Add the malloc_tcache_flush() function. Disable thread-specific caching until the application goes multi-threaded.
2009-12-29 16:09:15 +08:00
(a_qr_a)->a_field.qre_prev = (a_qr_b)->a_field.qre_prev; \
Optimize meld in qr module The goal of `qr_meld()` is to change the following four fields `(a->prev, a->prev->next, b->prev, b->prev->next)` from the values `(a->prev, a, b->prev, b)` to `(b->prev, b, a->prev, a)`. This commit changes ``` a->prev->next = b; b->prev->next = a; temp = a->prev; a->prev = b->prev; b->prev = temp; ``` to ``` temp = a->prev; a->prev = b->prev; b->prev = temp; a->prev->next = a; b->prev->next = b; ``` The benefit is that we can use `b->prev->next` for `temp`, and so there's no need to pass in `a_type`. The restriction is that `b` cannot be a `qr_next()` macro, so users of `qr_meld()` must pay attention. (Before this change, neither `a` nor `b` could be a `qr_next()` macro.)
2020-04-02 06:04:24 +08:00
(a_qr_b)->a_field.qre_prev = \
(a_qr_b)->a_field.qre_prev->a_field.qre_next; \
(a_qr_a)->a_field.qre_prev->a_field.qre_next = (a_qr_a); \
(a_qr_b)->a_field.qre_prev->a_field.qre_next = (a_qr_b); \
Convert thread-specific caching from magazines, and implement incremental GC. Add the 'G'/'g' and 'H'/'h' MALLOC_OPTIONS flags. Add the malloc_tcache_flush() function. Disable thread-specific caching until the application goes multi-threaded.
2009-12-29 16:09:15 +08:00
} while (0)
QL, QR: Add documentation.
2020-04-11 02:38:33 +08:00
/*
* Logically, this is just a meld. The intent, though, is that a_qrelm is a
* single-element ring, so that "before" has a more obvious interpretation than
* meld.
*/
deduplicate insert and delete logic in qr module
2020-04-02 07:13:57 +08:00
#define qr_before_insert(a_qrelm, a_qr, a_field) \
qr_meld((a_qrelm), (a_qr), a_field)
QL, QR: Add documentation.
2020-04-11 02:38:33 +08:00
/* Ditto, but inserting after rather than before. */
deduplicate insert and delete logic in qr module
2020-04-02 07:13:57 +08:00
#define qr_after_insert(a_qrelm, a_qr, a_field) \
qr_before_insert(qr_next(a_qrelm, a_field), (a_qr), a_field)
Style and spelling fixes.
2014-12-09 06:40:14 +08:00
/*
QL, QR: Add documentation.
2020-04-11 02:38:33 +08:00
* Inverts meld; given the ring:
* a -> a_1 -> ... -> a_n -> b -> b_1 -> ... -> b_n --
* ^ |
* |-------------------------------------------------|
*
* Results in two rings:
* a -> a_1 -> ... -> a_n --
* ^ |
* |------------------------
*
* b -> b_1 -> ... -> b_n --
* ^ |
* |------------------------
*
Style and spelling fixes.
2014-12-09 06:40:14 +08:00
* qr_meld() and qr_split() are functionally equivalent, so there's no need to
* have two copies of the code.
*/
Optimize meld in qr module The goal of `qr_meld()` is to change the following four fields `(a->prev, a->prev->next, b->prev, b->prev->next)` from the values `(a->prev, a, b->prev, b)` to `(b->prev, b, a->prev, a)`. This commit changes ``` a->prev->next = b; b->prev->next = a; temp = a->prev; a->prev = b->prev; b->prev = temp; ``` to ``` temp = a->prev; a->prev = b->prev; b->prev = temp; a->prev->next = a; b->prev->next = b; ``` The benefit is that we can use `b->prev->next` for `temp`, and so there's no need to pass in `a_type`. The restriction is that `b` cannot be a `qr_next()` macro, so users of `qr_meld()` must pay attention. (Before this change, neither `a` nor `b` could be a `qr_next()` macro.)
2020-04-02 06:04:24 +08:00
#define qr_split(a_qr_a, a_qr_b, a_field) \
qr_meld((a_qr_a), (a_qr_b), a_field)
Convert thread-specific caching from magazines, and implement incremental GC. Add the 'G'/'g' and 'H'/'h' MALLOC_OPTIONS flags. Add the malloc_tcache_flush() function. Disable thread-specific caching until the application goes multi-threaded.
2009-12-29 16:09:15 +08:00
QL, QR: Add documentation.
2020-04-11 02:38:33 +08:00
/*
* Splits off a_qr from the rest of its ring, so that it becomes a
* single-element ring.
*/
deduplicate insert and delete logic in qr module
2020-04-02 07:13:57 +08:00
#define qr_remove(a_qr, a_field) \
qr_split(qr_next(a_qr, a_field), (a_qr), a_field)
Convert thread-specific caching from magazines, and implement incremental GC. Add the 'G'/'g' and 'H'/'h' MALLOC_OPTIONS flags. Add the malloc_tcache_flush() function. Disable thread-specific caching until the application goes multi-threaded.
2009-12-29 16:09:15 +08:00
QL, QR: Add documentation.
2020-04-11 02:38:33 +08:00
/*
* Helper macro to iterate over each element in a ring exactly once, starting
* with a_qr. The usage is (assuming my_t defined as above):
*
* int sum(my_t *item) {
* int sum = 0;
* my_t *iter;
* qr_foreach(iter, item, link) {
* sum += iter->data;
* }
* return sum;
* }
*/
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define qr_foreach(var, a_qr, a_field) \
Convert thread-specific caching from magazines, and implement incremental GC. Add the 'G'/'g' and 'H'/'h' MALLOC_OPTIONS flags. Add the malloc_tcache_flush() function. Disable thread-specific caching until the application goes multi-threaded.
2009-12-29 16:09:15 +08:00
for ((var) = (a_qr); \
(var) != NULL; \
(var) = (((var)->a_field.qre_next != (a_qr)) \
? (var)->a_field.qre_next : NULL))
QL, QR: Add documentation.
2020-04-11 02:38:33 +08:00
/*
* The same (and with the same usage) as qr_foreach, but in the opposite order,
* ending with a_qr.
*/
Replace tabs following #define with spaces. This resolves #564.
2017-01-20 13:41:41 +08:00
#define qr_reverse_foreach(var, a_qr, a_field) \
Convert thread-specific caching from magazines, and implement incremental GC. Add the 'G'/'g' and 'H'/'h' MALLOC_OPTIONS flags. Add the malloc_tcache_flush() function. Disable thread-specific caching until the application goes multi-threaded.
2009-12-29 16:09:15 +08:00
for ((var) = ((a_qr) != NULL) ? qr_prev(a_qr, a_field) : NULL; \
(var) != NULL; \
(var) = (((var) != (a_qr)) \
? (var)->a_field.qre_prev : NULL))
Break up headers into constituent parts This is part of a broader change to make header files better represent the dependencies between one another (see https://github.com/jemalloc/jemalloc/issues/533). It breaks up component headers into smaller parts that can be made to have a simpler dependency graph. For the autogenerated headers (smoothstep.h and size_classes.h), no splitting was necessary, so I didn't add support to emit multiple headers.
2017-01-11 10:06:31 +08:00
#endif /* JEMALLOC_INTERNAL_QR_H */
Reference in New Issue Copy Permalink