heap: Quality updates to heap management

- Allow reuse of unused sections that fit exactly to selected allocation size. Decreases fragmentation dramatically.
- Always allocate and align mapped memory to selected alignment. Avoids having fragmented unused maps that are not aligned.
- Use a static alignment based on BPMP and generally average cache line size. Boosts performance when MMU is used.

bootloader/mem/heap.c

@@ -17,7 +17,8 @@
 
 #include <string.h>
 #include "heap.h"
-#include "../../common/common_heap.h"
+#include "../gfx/gfx.h"
+#include "../../../common/common_heap.h"
 
 static void _heap_create(heap_t *heap, u32 start)
 {
@@ -25,12 +26,13 @@ static void _heap_create(heap_t *heap, u32 start)
 	heap->first = NULL;
 }
 
-static u32 _heap_alloc(heap_t *heap, u32 size, u32 alignment)
+// Node info is before node address.
+static u32 _heap_alloc(heap_t *heap, u32 size)
 {
 	hnode_t *node, *new;
-	int search = 1;
 
-	size = ALIGN(size, alignment);
+	// Align to cache line size.
+	size = ALIGN(size, sizeof(hnode_t));
 
 	if (!heap->first)
 	{
@@ -45,27 +47,35 @@ static u32 _heap_alloc(heap_t *heap, u32 size, u32 alignment)
 	}
 
 	node = heap->first;
-	while (search)
+	while (true)
 	{
-		if (!node->used && size + sizeof(hnode_t) < node->size)
+		if (!node->used && (size <= node->size))
 		{
+			u32 new_size = node->size - size;
 			new = (hnode_t *)((u32)node + sizeof(hnode_t) + size);
 
-			new->size = node->size - sizeof(hnode_t) - size;
+			// If there's aligned leftover space, create a new node.
+			if (new_size >= (sizeof(hnode_t) << 2))
+			{
+				new->size = new_size - sizeof(hnode_t);
+				new->used = 0;
+				new->next = node->next;
+				new->next->prev = new;
+				new->prev = node;
+				node->next = new;
+			}
+			else
+				size += new_size;
+
 			node->size = size;
 			node->used = 1;
-			new->used = 0;
-			new->next = node->next;
-			new->next->prev = new;
-			new->prev = node;
-			node->next = new;
 
 			return (u32)node + sizeof(hnode_t);
 		}
 		if (node->next)
 			node = node->next;
 		else
-			search = 0;
+			break;
 	}
 
 	new = (hnode_t *)((u32)node + sizeof(hnode_t) + node->size);
@@ -108,12 +118,12 @@ void heap_init(u32 base)
 
 void *malloc(u32 size)
 {
-	return (void *)_heap_alloc(&_heap, size, sizeof(hnode_t));
+	return (void *)_heap_alloc(&_heap, size);
 }
 
 void *calloc(u32 num, u32 size)
 {
-	void *res = (void *)_heap_alloc(&_heap, num * size, sizeof(hnode_t));
+	void *res = (void *)_heap_alloc(&_heap, num * size);
 	memset(res, 0, num * size);
 	return res;
 }
@@ -123,3 +133,30 @@ void free(void *buf)
 	if ((u32)buf >= _heap.start)
 		_heap_free(&_heap, (u32)buf);
 }
+
+void heap_monitor(heap_monitor_t *mon, bool print_node_stats)
+{
+	u32 count = 0;
+	memset(mon, 0, sizeof(heap_monitor));
+
+	hnode_t *node = _heap.first;
+	while (true)
+	{
+		if (node->used)
+			mon->used += node->size + sizeof(hnode_t);
+		else
+			mon->total += node->size + sizeof(hnode_t);
+
+		if (print_node_stats)
+			gfx_printf("%3d - %d, addr: 0x%08X, size: 0x%X\n",
+				count, node->used, (u32)node + sizeof(hnode_t), node->size);
+
+		count++;
+		
+		if (node->next)
+			node = node->next;
+		else
+			break;
+	}
+	mon->total += mon->used;
+}

bootloader/mem/heap.h

@@ -18,10 +18,12 @@
 #define _HEAP_H_
 
 #include "../utils/types.h"
+#include "../../../common/common_heap.h"
 
 void heap_init(u32 base);
 void *malloc(u32 size);
 void *calloc(u32 num, u32 size);
 void free(void *buf);
+void heap_monitor(heap_monitor_t *mon, bool print_node_stats);
 
 #endif

common/common_heap.h

@@ -25,6 +25,7 @@ typedef struct _hnode
 	u32 size;
 	struct _hnode *prev;
 	struct _hnode *next;
+	u32 align[4]; // Align to arch cache line size.
 } hnode_t;
 
 typedef struct _heap
@@ -32,3 +33,9 @@ typedef struct _heap
 	u32 start;
 	hnode_t *first;
 } heap_t;
+
+typedef struct
+{
+    u32 total;
+    u32 used;
+} heap_monitor_t;

nyx/nyx_gui/mem/heap.c

@@ -17,6 +17,7 @@
 
 #include <string.h>
 #include "heap.h"
+#include "../gfx/gfx.h"
 #include "../../../common/common_heap.h"
 
 static void _heap_create(heap_t *heap, u32 start)
@@ -25,12 +26,13 @@ static void _heap_create(heap_t *heap, u32 start)
 	heap->first = NULL;
 }
 
-static u32 _heap_alloc(heap_t *heap, u32 size, u32 alignment)
+// Node info is before node address.
+static u32 _heap_alloc(heap_t *heap, u32 size)
 {
 	hnode_t *node, *new;
-	int search = 1;
 
-	size = ALIGN(size, alignment);
+	// Align to cache line size.
+	size = ALIGN(size, sizeof(hnode_t));
 
 	if (!heap->first)
 	{
@@ -45,27 +47,35 @@ static u32 _heap_alloc(heap_t *heap, u32 size, u32 alignment)
 	}
 
 	node = heap->first;
-	while (search)
+	while (true)
 	{
-		if (!node->used && size + sizeof(hnode_t) < node->size)
+		if (!node->used && (size <= node->size))
 		{
+			u32 new_size = node->size - size;
 			new = (hnode_t *)((u32)node + sizeof(hnode_t) + size);
 
-			new->size = node->size - sizeof(hnode_t) - size;
+			// If there's aligned leftover space, create a new node.
+			if (new_size >= (sizeof(hnode_t) << 2))
+			{
+				new->size = new_size - sizeof(hnode_t);
+				new->used = 0;
+				new->next = node->next;
+				new->next->prev = new;
+				new->prev = node;
+				node->next = new;
+			}
+			else
+				size += new_size;
+
 			node->size = size;
 			node->used = 1;
-			new->used = 0;
-			new->next = node->next;
-			new->next->prev = new;
-			new->prev = node;
-			node->next = new;
 
 			return (u32)node + sizeof(hnode_t);
 		}
 		if (node->next)
 			node = node->next;
 		else
-			search = 0;
+			break;
 	}
 
 	new = (hnode_t *)((u32)node + sizeof(hnode_t) + node->size);
@@ -108,12 +118,12 @@ void heap_init(u32 base)
 
 void *malloc(u32 size)
 {
-	return (void *)_heap_alloc(&_heap, size, sizeof(hnode_t));
+	return (void *)_heap_alloc(&_heap, size);
 }
 
 void *calloc(u32 num, u32 size)
 {
-	void *res = (void *)_heap_alloc(&_heap, num * size, sizeof(hnode_t));
+	void *res = (void *)_heap_alloc(&_heap, num * size);
 	memset(res, 0, num * size);
 	return res;
 }
@@ -123,3 +133,30 @@ void free(void *buf)
 	if ((u32)buf >= _heap.start)
 		_heap_free(&_heap, (u32)buf);
 }
+
+void heap_monitor(heap_monitor_t *mon, bool print_node_stats)
+{
+	u32 count = 0;
+	memset(mon, 0, sizeof(heap_monitor));
+
+	hnode_t *node = _heap.first;
+	while (true)
+	{
+		if (node->used)
+			mon->used += node->size + sizeof(hnode_t);
+		else
+			mon->total += node->size + sizeof(hnode_t);
+
+		if (print_node_stats)
+			gfx_printf("%3d - %d, addr: 0x%08X, size: 0x%X\n",
+				count, node->used, (u32)node + sizeof(hnode_t), node->size);
+
+		count++;
+		
+		if (node->next)
+			node = node->next;
+		else
+			break;
+	}
+	mon->total += mon->used;
+}

nyx/nyx_gui/mem/heap.h

@@ -18,10 +18,12 @@
 #define _HEAP_H_
 
 #include "../utils/types.h"
+#include "../../../common/common_heap.h"
 
 void heap_init(u32 base);
 void *malloc(u32 size);
 void *calloc(u32 num, u32 size);
 void free(void *buf);
+void heap_monitor(heap_monitor_t *mon, bool print_node_stats);
 
 #endif