The existing linear mapping between allocation size and pool index
does not scale when moving to a 64 KB granularity or beyond. With
a granularity of 64 KB, 2048 (!) bins will be created for each
memory type, each differing 32 bytes in size with the next one.
Instead, introduce an exponential scheme where each bin size is
the sum of the two previous ones.
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Ard Biesheuvel <ard.biesheu...@linaro.org>
---
MdeModulePkg/Core/Dxe/Mem/Pool.c | 36 ++++++++++++++++++++++++++++--------
1 file changed, 28 insertions(+), 8 deletions(-)
diff --git a/MdeModulePkg/Core/Dxe/Mem/Pool.c b/MdeModulePkg/Core/Dxe/Mem/Pool.c
index f99293da5c51..c7998e80cce6 100644
--- a/MdeModulePkg/Core/Dxe/Mem/Pool.c
+++ b/MdeModulePkg/Core/Dxe/Mem/Pool.c
@@ -41,19 +41,24 @@ typedef struct {
UINTN Size;
} POOL_TAIL;
-
-#define POOL_SHIFT 7
-
#define POOL_OVERHEAD (SIZE_OF_POOL_HEAD + sizeof(POOL_TAIL))
#define HEAD_TO_TAIL(a) \
((POOL_TAIL *) (((CHAR8 *) (a)) + (a)->Size - sizeof(POOL_TAIL)));
+//
+// Each element is the sum of the 2 previous ones: this allows us to migrate
+// blocks between bins by splitting them up, while not wasting too much memory
+// as we would in a strict power-of-2 sequence
+//
+STATIC CONST UINT16 mPoolSizeTable[] = {
+ 64, 128, 192, 320, 512, 832, 1344, 2176, 3520, 5696, 9216, 14912, 24128
+};
-#define SIZE_TO_LIST(a) ((a) >> POOL_SHIFT)
-#define LIST_TO_SIZE(a) ((a+1) << POOL_SHIFT)
+#define SIZE_TO_LIST(a) (GetPoolIndexFromSize (a))
+#define LIST_TO_SIZE(a) (mPoolSizeTable [a])
-#define MAX_POOL_LIST SIZE_TO_LIST(DEFAULT_PAGE_ALLOCATION)
+#define MAX_POOL_LIST (sizeof (mPoolSizeTable) / sizeof
(mPoolSizeTable[0]))
#define MAX_POOL_SIZE (MAX_ADDRESS - POOL_OVERHEAD)
@@ -80,6 +85,21 @@ POOL mPoolHead[EfiMaxMemoryType];
//
LIST_ENTRY mPoolHeadList = INITIALIZE_LIST_HEAD_VARIABLE (mPoolHeadList);
+STATIC
+UINTN
+GetPoolIndexFromSize (
+ UINTN Size
+ )
+{
+ UINTN Index;
+
+ for (Index = 0; Index < MAX_POOL_LIST; Index++) {
+ if (mPoolSizeTable [Index] >= Size) {
+ return Index;
+ }
+ }
+ return MAX_POOL_LIST;
+}
/**
Called to initialize the pool.
@@ -311,7 +331,7 @@ CoreAllocatePoolI (
// If allocation is over max size, just allocate pages for the request
// (slow)
//
- if (Index >= MAX_POOL_LIST) {
+ if (Index >= SIZE_TO_LIST (Granularity)) {
NoPages = EFI_SIZE_TO_PAGES(Size) + EFI_SIZE_TO_PAGES (Granularity) - 1;
NoPages &= ~(UINTN)(EFI_SIZE_TO_PAGES (Granularity) - 1);
Head = CoreAllocatePoolPages (PoolType, NoPages, Granularity);
@@ -539,7 +559,7 @@ CoreFreePoolI (
//
// If it's not on the list, it must be pool pages
//
- if (Index >= MAX_POOL_LIST) {
+ if (Index >= SIZE_TO_LIST (Granularity)) {
//
// Return the memory pages back to free memory
--
1.8.3.2
------------------------------------------------------------------------------
Dive into the World of Parallel Programming. The Go Parallel Website,
sponsored by Intel and developed in partnership with Slashdot Media, is your
hub for all things parallel software development, from weekly thought
leadership blogs to news, videos, case studies, tutorials and more. Take a
look and join the conversation now. http://goparallel.sourceforge.net/
_______________________________________________
edk2-devel mailing list
edk2-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/edk2-devel
