Added a new notifyer, NC_SPACE_CHANGED, to signal an editor that

replaces another so it can do updates (e.g. dopesheet editor can
sync channel selection).

Also coded a simple optimization for allocating small objects,
based on mempools.  It's #ifdef'd out, you can enabled it by 
defining OPTIMIZE_SMALL_BLOCKS (e.g. adding -DDOPTIMIZE_SMALL_BLOCKS to
your compiler flags).

We suffer from a great deal of performance loss from the system allocator
(vgroups, ghash, edgehash, the singly-linked list implementation in blenlib,
editmesh, and likely a great many areas I'm forgetting), and this is the 
common solution for handling the many-small-objects problem.  It's not
really production-ready yet (it's long-term memory consequencers need to
be profiled first, and the implementation tweaked as necassary), but for
people on systems with slow system allocators it's worth trying.

Note that since this creates a guardedalloc<->blenlib link, the build systems
need to be updated accordingly (I've already done this for scons, though I'm
not sure if the player builds).

1 files changed, 57 insertions, 0 deletions

diff --git a/intern/guardedalloc/intern/mallocn.c b/intern/guardedalloc/intern/mallocn.c
index d1114af2437..15236a2bee4 100644
--- a/intern/guardedalloc/intern/mallocn.c
+++ b/intern/guardedalloc/intern/mallocn.c
@@ -37,6 +37,8 @@
 #include <string.h>	/* memcpy */
 #include <stdarg.h>
 
+#include "../../../source/blender/blenlib/BLI_mempool.h"
+
 /* mmap exception */
 #if defined(WIN32)
 #include <sys/types.h>
@@ -138,6 +140,40 @@ static int malloc_debug_memset= 0;
 /* implementation                                                        */
 /* --------------------------------------------------------------------- */
 
+#ifdef OPTIMIZE_SMALL_BLOCKS
+
+/*allocator for small objects
+
+  basic strategy: most slowdowns from overuse of the system allocator tends
+  to be from relatively small allocations.
+*/
+
+/*this is a little bit bigger then it perhaps needs to be, to accomodate vgroup
+  allocations (which are one system alloc per vertex of a mesh, and is a major
+  source of performance loss)*/
+#define SMALL_BLOCK_LIMIT	1024
+#define POOL_CHUNK_SIZE		512 /*size in number of elements, not bytes*/
+
+static BLI_mempool *alloc_pools[SMALL_BLOCK_LIMIT+sizeof(MemHead)+sizeof(MemTail)] = {NULL,};
+
+static void *mempool_alloc_mem(int size) {
+	size += sizeof(MemHead)+sizeof(MemTail);
+
+	if (!alloc_pools[size]) {
+		alloc_pools[size] = BLI_mempool_create(size, 1, POOL_CHUNK_SIZE, 1);
+	}
+	
+	return BLI_mempool_alloc(alloc_pools[size]);
+}
+
+static void mempool_free(void *mem, int size) {
+	size += sizeof(MemHead)+sizeof(MemTail);
+
+	BLI_mempool_free(alloc_pools[size], mem);
+}
+
+#endif
+
 static void print_error(const char *str, ...)
 {
 	char buf[1024];
@@ -254,7 +290,14 @@ void *MEM_mallocN(unsigned int len, const char *str)
 
 	len = (len + 3 ) & ~3; 	/* allocate in units of 4 */
 	
+#ifdef OPTIMIZE_SMALL_BLOCKS
+	if (len < SMALL_BLOCK_LIMIT)
+		memh= mempool_alloc_mem(len);
+	else
+		memh= (MemHead *)malloc(len+sizeof(MemHead)+sizeof(MemTail));
+#else
 	memh= (MemHead *)malloc(len+sizeof(MemHead)+sizeof(MemTail));
+#endif
 
 	if(memh) {
 		make_memhead_header(memh, len, str);
@@ -276,7 +319,16 @@ void *MEM_callocN(unsigned int len, const char *str)
 
 	len = (len + 3 ) & ~3; 	/* allocate in units of 4 */
 
+#ifdef OPTIMIZE_SMALL_BLOCKS
+	if (len < SMALL_BLOCK_LIMIT) {
+		memh= mempool_alloc_mem(len);
+		memset(memh, 0, len+sizeof(MemHead)+sizeof(MemTail));
+	} else {
+		memh= (MemHead *)calloc(len+sizeof(MemHead)+sizeof(MemTail),1);
+	}
+#else
 	memh= (MemHead *)calloc(len+sizeof(MemHead)+sizeof(MemTail),1);
+#endif
 
 	if(memh) {
 		make_memhead_header(memh, len, str);
@@ -629,6 +681,11 @@ static void rem_memblock(MemHead *memh)
         if (munmap(memh, memh->len + sizeof(MemHead) + sizeof(MemTail)))
             printf("Couldn't unmap memory %s\n", memh->name);
     }
+#ifdef OPTIMIZE_SMALL_BLOCKS
+	else if (memh->len < SMALL_BLOCK_LIMIT) {
+		mempool_free(memh, memh->len);
+	}
+#endif
     else {
 		if(malloc_debug_memset && memh->len)
 			memset(memh+1, 255, memh->len);