[igt-dev] [PATCH i-g-t 3/5] lib/intel_allocator_bst: Implement the allocator with a BST

Andrzej Turko andrzej.turko at linux.intel.com
Wed Jul 28 15:12:05 UTC 2021


This implements an alternative to the simple allocator.
The BST allocator follows best-fit strategy.
Usage of a binary search tree speeds up finding free
blocks of memory (compared to a list).

Signed-off-by: Andrzej Turko <andrzej.turko at linux.intel.com>
Cc: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>
---
 lib/intel_allocator.c     |   7 +
 lib/intel_allocator.h     |   1 +
 lib/intel_allocator_bst.c | 672 ++++++++++++++++++++++++++++++++++++++
 lib/meson.build           |   1 +
 4 files changed, 681 insertions(+)
 create mode 100644 lib/intel_allocator_bst.c

diff --git a/lib/intel_allocator.c b/lib/intel_allocator.c
index 133176ed4..2ef487a05 100644
--- a/lib/intel_allocator.c
+++ b/lib/intel_allocator.c
@@ -71,6 +71,9 @@ intel_allocator_random_create(int fd, uint64_t start, uint64_t end);
 struct intel_allocator *
 intel_allocator_simple_create(int fd, uint64_t start, uint64_t end,
 			      enum allocator_strategy strategy);
+struct intel_allocator *
+intel_allocator_bst_create(int fd, uint64_t start, uint64_t end,
+			   enum allocator_strategy strategy);
 
 /*
  * Instead of trying to find first empty handle just get new one. Assuming
@@ -311,6 +314,10 @@ static struct intel_allocator *intel_allocator_create(int fd,
 		ial = intel_allocator_simple_create(fd, start, end,
 						    allocator_strategy);
 		break;
+	case INTEL_ALLOCATOR_BST:
+		ial = intel_allocator_bst_create(fd, start, end,
+						 allocator_strategy);
+		break;
 	default:
 		igt_assert_f(ial, "Allocator type %d not implemented\n",
 			     allocator_type);
diff --git a/lib/intel_allocator.h b/lib/intel_allocator.h
index 7d9d01123..93ecbc3e4 100644
--- a/lib/intel_allocator.h
+++ b/lib/intel_allocator.h
@@ -211,6 +211,7 @@ void intel_allocator_print(uint64_t allocator_handle);
 #define INTEL_ALLOCATOR_RELOC  1
 #define INTEL_ALLOCATOR_RANDOM 2
 #define INTEL_ALLOCATOR_SIMPLE 3
+#define INTEL_ALLOCATOR_BST    4
 
 #define GEN8_GTT_ADDRESS_WIDTH 48
 
diff --git a/lib/intel_allocator_bst.c b/lib/intel_allocator_bst.c
new file mode 100644
index 000000000..4affa4927
--- /dev/null
+++ b/lib/intel_allocator_bst.c
@@ -0,0 +1,672 @@
+// SPDX-License-Identifier: MIT
+/*
+ * Copyright © 2021 Intel Corporation
+ */
+
+#include <stdlib.h>
+#include "igt.h"
+#include "igt_map.h"
+#include "igt_bst.h"
+#include "intel_allocator.h"
+
+/* The addresses of allocated objects will be
+ * strictly smaller than this value (the upper
+ * end of the address range cannot exceed it).
+ * This is to enable the use of canonical
+ * addresses.
+ */
+#define IALB_MAX_ADDR (1ull<<48)
+
+#define RESERVED 4096
+
+/* Avoid compilation warning */
+struct intel_allocator *
+intel_allocator_bst_create(int fd, uint64_t start, uint64_t end,
+			   enum allocator_strategy strategy);
+
+struct intel_allocator_bst {
+	struct igt_map *objects;
+	struct igt_map *reserved;
+	struct igt_bst *by_size;
+	struct igt_bst *by_offset;
+
+	uint64_t start;
+	uint64_t end;
+
+	/* statistics */
+	uint64_t total_size;
+	uint64_t allocated_size;
+	uint64_t allocated_objects;
+	uint64_t reserved_size;
+	uint64_t reserved_areas;
+};
+
+struct intel_allocator_bst_vma_hole {
+
+	/* pointer to the node in the by_size bst */
+	void *size_ptr;
+
+	/* pointer to the node in the by_offset bst */
+	void *offset_ptr;
+
+	uint64_t size;
+	uint64_t offset;
+};
+
+struct intel_allocator_record {
+	uint32_t handle;
+	uint64_t offset;
+	uint64_t size;
+};
+
+/* 2^31 + 2^29 - 2^25 + 2^22 - 2^19 - 2^16 + 1 */
+#define GOLDEN_RATIO_PRIME_32 0x9e370001UL
+
+/*  2^63 + 2^61 - 2^57 + 2^54 - 2^51 - 2^18 + 1 */
+#define GOLDEN_RATIO_PRIME_64 0x9e37fffffffc0001ULL
+
+static inline uint32_t hash_handles(const void *val)
+{
+	uint32_t hash = *(uint32_t *) val;
+
+	hash = hash * GOLDEN_RATIO_PRIME_32;
+	return hash;
+}
+
+static int equal_handles(const void *a, const void *b)
+{
+	uint32_t *key1 = (uint32_t *) a, *key2 = (uint32_t *) b;
+
+	return *key1 == *key2;
+}
+
+static inline uint32_t hash_offsets(const void *val)
+{
+	uint64_t hash = *(uint64_t *) val;
+
+	hash = hash * GOLDEN_RATIO_PRIME_64;
+	/* High bits are more random, so use them. */
+	return hash >> 32;
+}
+
+static int equal_offsets(const void *a, const void *b)
+{
+	uint64_t *key1 = (uint64_t *) a, *key2 = (uint64_t *) b;
+
+	return *key1 == *key2;
+}
+
+static void map_entry_free_func(struct igt_map_entry *entry)
+{
+	free(entry->data);
+}
+
+static void holes_bst_validate(struct intel_allocator_bst *ialb)
+{
+	struct igt_bst *offset_bst, *size_bst;
+	struct intel_allocator_bst_vma_hole *hole;
+	void *nodeptr;
+	int by_offset_size, by_size_size;
+	uint64_t prev_offset;
+
+	offset_bst = ialb->by_offset;
+	size_bst = ialb->by_size;
+	by_offset_size = offset_bst->validate(offset_bst);
+	by_size_size = size_bst->validate(size_bst);
+	igt_assert(by_offset_size == by_size_size);
+
+	prev_offset = IALB_MAX_ADDR;
+	igt_bst_for_each_node_rev(offset_bst, nodeptr) {
+		hole = igt_bst_get_value(offset_bst, nodeptr);
+		igt_assert(hole);
+		/* Make sure the hole is stored under correct keys in
+		 * the balanced search trees.
+		 */
+		igt_assert(hole->offset == igt_bst_get_key(offset_bst, hole->offset_ptr));
+		igt_assert(hole->size == igt_bst_get_key(size_bst, hole->size_ptr));
+
+		igt_assert(hole->offset + hole->size > hole->offset);
+		/* Make sure no pair of holes is adjacent. */
+		igt_assert(prev_offset > hole->offset + hole->size);
+
+		prev_offset = hole->offset;
+	}
+}
+
+static void __intel_allocator_bst_alloc(struct intel_allocator_bst *ialb,
+					struct intel_allocator_bst_vma_hole *hole,
+					uint64_t offset, uint64_t size)
+{
+	struct igt_bst *offset_bst, *size_bst;
+	struct intel_allocator_bst_vma_hole *newhole;
+	uint64_t lower, higher;
+
+	igt_assert(hole->offset <= offset);
+	igt_assert(hole->offset + hole->size >= offset + size);
+
+	offset_bst = ialb->by_offset;
+	size_bst = ialb->by_size;
+
+	lower = offset - hole->offset;
+	higher = hole->size - size - lower;
+
+	if (lower > 0 && higher > 0) {
+		/* There are leftovers on both lower and higher addresses.
+		 * Create a hole for higher addresses from scratch and
+		 * one for lower addresses by modifying the existing hole.
+		 */
+
+		newhole = malloc(sizeof(struct intel_allocator_bst_vma_hole));
+		igt_assert(newhole);
+
+		newhole->offset = offset + size;
+		newhole->size = higher;
+
+		newhole->offset_ptr = igt_bst_insert(offset_bst, newhole, newhole->offset);
+		newhole->size_ptr = igt_bst_insert(size_bst, newhole, newhole->size);
+
+		hole->size = lower;
+		igt_bst_erase(size_bst, hole->size_ptr);
+		hole->size_ptr = igt_bst_insert(size_bst, hole, hole->size);
+
+	} else if (higher > 0) {
+		/* There is no free address space in this hole left
+		 * below the allocated object. Only a portion of
+		 * higher addresses is still free, so just adjust the
+		 * existing hole accordingly (change offset and size).
+		 */
+
+		hole->offset = offset + size;
+		hole->size = higher;
+
+		igt_bst_erase(offset_bst, hole->offset_ptr);
+		igt_bst_erase(size_bst, hole->size_ptr);
+
+		hole->offset_ptr = igt_bst_insert(offset_bst, hole, hole->offset);
+		hole->size_ptr = igt_bst_insert(size_bst, hole, hole->size);
+
+	} else if (lower > 0) {
+		/* The allocated block is aligned to the end of the
+		 * hole and the lower addresses are still free,
+		 * so shrink the existing hole (change size,
+		 * but not the offset).
+		 */
+
+		hole->size = lower;
+		igt_bst_erase(size_bst, hole->size_ptr);
+		hole->size_ptr = igt_bst_insert(size_bst, hole, hole->size);
+
+	} else {
+		/* There are no leftovers, just erase the hole. */
+
+		igt_bst_erase(offset_bst, hole->offset_ptr);
+		igt_bst_erase(size_bst, hole->size_ptr);
+		free(hole);
+	}
+}
+
+static void __intel_allocator_bst_add_hole(struct intel_allocator_bst *ialb,
+					   uint64_t offset, uint64_t size)
+{
+	struct igt_bst *offset_bst, *size_bst;
+	struct intel_allocator_bst_vma_hole *hole, *nxt, *prv;
+
+	offset_bst = ialb->by_offset;
+	size_bst = ialb->by_size;
+
+	prv = igt_bst_query_predecessor(offset_bst, offset);
+	if (prv && prv->offset + prv->size < offset)
+		prv = NULL;
+
+	nxt = igt_bst_query_successor(offset_bst, offset);
+	if (nxt && nxt->offset > offset + size)
+		nxt = NULL;
+
+	if (nxt && prv) {
+		/* There are holes adjacent to the new one
+		 * both below and above it in the address space.
+		 * Merge them all into a single hole:
+		 * erase the upper and increase size of the
+		 * lower one.
+		 */
+		prv->size += size + nxt->size;
+
+		igt_bst_erase(offset_bst, nxt->offset_ptr);
+		igt_bst_erase(size_bst, nxt->size_ptr);
+		free(nxt);
+
+		igt_bst_erase(size_bst, prv->size_ptr);
+		prv->size_ptr = igt_bst_insert(size_bst, prv, prv->size);
+
+	} else if (nxt) {
+		/* The only adjacent hole is above the new one,
+		 * so increase its size and update its offset.
+		 */
+		nxt->size += size;
+		nxt->offset = offset;
+
+		igt_bst_erase(offset_bst, nxt->offset_ptr);
+		igt_bst_erase(size_bst, nxt->size_ptr);
+
+		nxt->offset_ptr = igt_bst_insert(offset_bst, nxt, nxt->offset);
+		nxt->size_ptr = igt_bst_insert(size_bst, nxt, nxt->size);
+
+	} else if (prv) {
+		/* The only adjacent hole is below the new one,
+		 * just increase its size.
+		 */
+		prv->size += size;
+		igt_bst_erase(size_bst, prv->size_ptr);
+		prv->size_ptr = igt_bst_insert(size_bst, prv, prv->size);
+
+	} else {
+		/* There are no holes neighbouring the new one,
+		 * so just create a new hole.
+		 */
+		hole = malloc(sizeof(struct intel_allocator_bst_vma_hole));
+		igt_assert(hole);
+
+		hole->offset = offset;
+		hole->size = size;
+		hole->offset_ptr = igt_bst_insert(offset_bst, hole, offset);
+		hole->size_ptr = igt_bst_insert(size_bst, hole, size);
+
+	}
+}
+
+
+static void intel_allocator_bst_get_address_range(struct intel_allocator *ial,
+				      uint64_t *startp, uint64_t *endp)
+{
+	struct intel_allocator_bst *ialb = ial->priv;
+
+	if (startp)
+		*startp = ialb->start;
+
+	if (endp)
+		*endp = ialb->end;
+}
+
+static uint64_t intel_allocator_bst_alloc(struct intel_allocator *ial, uint32_t handle,
+					  uint64_t size, uint64_t alignment,
+					  enum allocator_strategy strategy)
+{
+	struct intel_allocator_bst *ialb = ial->priv;
+	struct intel_allocator_record *record;
+	struct intel_allocator_bst_vma_hole *hole;
+	uint64_t misalignment, offset;
+
+	igt_assert(size > 0);
+	alignment = (alignment > 0 ? alignment : 1);
+
+	record = igt_map_search(ialb->objects, &handle);
+
+	if (record) {
+		/* This block has already been allocated,
+		 * check that it satifies the requirements.
+		 */
+
+		igt_assert(size == record->size);
+		igt_assert(record->offset % alignment == 0);
+
+		return record->offset;
+	}
+
+	/* Look for a slightly bigger hole to make sure
+	 * the block can be aligned properly.
+	 */
+	hole = igt_bst_query_successor(ialb->by_size,
+				       size + alignment - 1);
+
+	/* Look for a smaller hole, maybe proper alignment
+	 * will still be possible.
+	 */
+	if (!hole)
+		hole = igt_bst_query_successor(ialb->by_size, size);
+
+	if (!hole)
+		return ALLOC_INVALID_ADDRESS;
+
+	if (strategy == ALLOC_STRATEGY_NONE)
+		strategy = ial->strategy;
+
+	switch (strategy) {
+	case ALLOC_STRATEGY_LOW_TO_HIGH:
+		misalignment = alignment - (hole->offset % alignment);
+		misalignment = misalignment == alignment ? 0 : misalignment;
+		offset = hole->offset + misalignment;
+
+		if (misalignment + size > hole->size)
+			return ALLOC_INVALID_ADDRESS;
+
+		break;
+	case ALLOC_STRATEGY_HIGH_TO_LOW:
+		offset = hole->offset + hole->size - size;
+		offset = offset - offset % alignment;
+
+		if (offset < hole->offset)
+			return ALLOC_INVALID_ADDRESS;
+
+		break;
+	default:
+		igt_assert_f(0, "Unsupported strategy.");
+	}
+
+	__intel_allocator_bst_alloc(ialb, hole, offset, size);
+
+	record = malloc(sizeof(struct intel_allocator_record));
+	igt_assert(record);
+	record->handle = handle;
+	record->offset = offset;
+	record->size = size;
+
+	igt_map_insert(ialb->objects, &record->handle, record);
+	ialb->allocated_objects++;
+	ialb->allocated_size += size;
+
+	return offset;
+}
+
+static bool intel_allocator_bst_is_allocated(struct intel_allocator *ial, uint32_t handle,
+					     uint64_t size, uint64_t offset)
+{
+	struct intel_allocator_bst *ialb = ial->priv;
+	struct intel_allocator_record *record;
+	bool same;
+
+	offset = DECANONICAL(offset);
+	record = igt_map_search(ialb->objects, &handle);
+	if (record) {
+		same = (record->handle == handle && record->size == size &&
+			record->offset == offset);
+	} else {
+		same = false;
+	}
+
+	return same;
+}
+
+static bool intel_allocator_bst_free(struct intel_allocator *ial, uint32_t handle)
+{
+	struct intel_allocator_bst *ialb = ial->priv;
+	struct intel_allocator_record *record;
+
+	record = igt_map_search(ialb->objects, &handle);
+
+	if (!record)
+		return false;
+
+	__intel_allocator_bst_add_hole(ialb, record->offset, record->size);
+	ialb->allocated_objects--;
+	ialb->allocated_size -= record->size;
+	igt_map_remove(ialb->objects, &handle, map_entry_free_func);
+
+	return true;
+}
+
+static bool intel_allocator_bst_reserve(struct intel_allocator *ial, uint32_t handle,
+					uint64_t start, uint64_t end)
+{
+	struct intel_allocator_bst *ialb = ial->priv;
+	struct intel_allocator_bst_vma_hole *hole;
+	struct intel_allocator_record *record;
+
+	start = DECANONICAL(start);
+	if (end != (1ull << GEN8_GTT_ADDRESS_WIDTH))
+		end = DECANONICAL(end);
+
+	igt_assert(start < end);
+	igt_assert(start >= ialb->start);
+	igt_assert(end <= ialb->end);
+
+	hole = igt_bst_query_predecessor(ialb->by_offset, start);
+	if (!hole)
+		return false;
+
+	igt_assert(hole->offset <= start);
+
+	if (hole->offset + hole->size < end)
+		return false;
+
+	__intel_allocator_bst_alloc(ialb, hole, start, end - start);
+
+	record = malloc(sizeof(struct intel_allocator_bst_vma_hole));
+	igt_assert(record);
+	record->offset = start;
+	record->size = end - start;
+	record->handle = handle;
+	igt_map_insert(ialb->reserved, &record->offset, record);
+	ialb->reserved_areas++;
+	ialb->reserved_size += end - start;
+
+	return true;
+}
+
+static bool intel_allocator_bst_unreserve(struct intel_allocator *ial, uint32_t handle,
+					  uint64_t start, uint64_t end)
+{
+	struct intel_allocator_bst *ialb = ial->priv;
+	struct intel_allocator_record *record;
+
+	start = DECANONICAL(start);
+	if (end != (1ull << GEN8_GTT_ADDRESS_WIDTH))
+		end = DECANONICAL(end);
+
+	record = igt_map_search(ialb->reserved, &start);
+
+	if (!record) {
+		igt_warn("Only a reserved block can be unreserved.\n");
+		return false;
+	}
+
+	if (record->size != end - start) {
+		igt_warn("Size of the block does not match the passed value.\n");
+		return false;
+	}
+
+	if (record->offset != start) {
+		igt_warn("Offset of the block does not match the passed value.\n");
+		return false;
+	}
+
+	if (record->handle != handle) {
+		igt_warn("Handle %u does not match the passed handle %u.\n",
+			 record->handle, handle);
+		return false;
+	}
+
+	__intel_allocator_bst_add_hole(ialb, start, end - start);
+	ialb->reserved_areas--;
+	ialb->reserved_size -= record->size;
+	igt_map_remove(ialb->reserved, &start, map_entry_free_func);
+
+	return true;
+}
+
+static bool intel_allocator_bst_is_reserved(struct intel_allocator *ial,
+					    uint64_t start, uint64_t end)
+{
+	struct intel_allocator_bst *ialb = ial->priv;
+	struct intel_allocator_record *record;
+
+	start = DECANONICAL(start);
+	if (end != (1ull << GEN8_GTT_ADDRESS_WIDTH))
+		end = DECANONICAL(end);
+
+	record = igt_map_search(ialb->reserved, &start);
+
+	if (!record)
+		return false;
+
+	if (record->offset == start && record->size == end - start)
+		return true;
+
+	return false;
+}
+
+static void intel_allocator_bst_destroy(struct intel_allocator *ial)
+{
+	struct intel_allocator_bst *ialb = ial->priv;
+	struct igt_bst *size_bst, *offset_bst;
+	void *nodeptr;
+
+	size_bst = ialb->by_size;
+	offset_bst = ialb->by_offset;
+
+	igt_bst_for_each_node(offset_bst, nodeptr)
+		free(offset_bst->get_value(nodeptr));
+
+	igt_bst_free(size_bst);
+	igt_bst_free(offset_bst);
+	free(offset_bst);
+	free(size_bst);
+
+	igt_map_destroy(ialb->objects, map_entry_free_func);
+	igt_map_destroy(ialb->reserved, map_entry_free_func);
+
+	free(ialb);
+	free(ial);
+}
+
+static bool intel_allocator_bst_is_empty(struct intel_allocator *ial)
+{
+	struct intel_allocator_bst *ialb = ial->priv;
+
+	return ialb->allocated_objects == 0 && ialb->reserved_areas == 0;
+}
+
+static void intel_allocator_bst_print(struct intel_allocator *ial, bool full)
+{
+	struct intel_allocator_bst *ialb = ial->priv;
+	struct intel_allocator_record *record;
+	struct intel_allocator_bst_vma_hole *hole;
+	struct igt_bst *offset_bst, *size_bst;
+	struct igt_map_entry *entry;
+	void *nodeptr;
+	uint64_t total_free;
+
+	igt_info("intel_allocator_bst <ial: %p, fd: %d> on "
+		 "[0x%"PRIX64" : 0x%"PRIx64"]\n", ial,
+		 ial->fd, ialb->start, ialb->end);
+
+	total_free = 0;
+	if (full) {
+		offset_bst = ialb->by_offset;
+		size_bst = ialb->by_size;
+
+		igt_info("holes by offset:\n");
+		igt_bst_for_each_node(offset_bst, nodeptr) {
+			hole = igt_bst_get_value(offset_bst, nodeptr);
+			igt_info("offset = %"PRIu64" (0x%"PRIx64") "
+				 "size = %"PRIu64" (0x%"PRIx64")\n",
+				 hole->offset, hole->offset, hole->size,
+				 hole->size);
+			total_free += hole->size;
+		}
+
+		igt_info("holes by size:\n");
+		igt_bst_for_each_node(size_bst, nodeptr) {
+			hole = igt_bst_get_value(size_bst, nodeptr);
+			igt_info("offset = %"PRIu64" (0x%"PRIx64") "
+				 "size = %"PRIu64" (0x%"PRIx64")\n",
+				 hole->offset, hole->offset, hole->size,
+				 hole->size);
+
+		}
+
+		igt_info("allocated objects:\n");
+		igt_map_foreach(ialb->objects, entry) {
+			record = entry->data;
+			igt_info("handle = %d, offset = %"PRIu64" "
+				"(0x%"PRIx64", size = %"PRIu64" (0x%"PRIx64")\n",
+				 record->handle, record->offset, record->offset,
+				 record->size, record->size);
+
+		}
+
+		igt_info("reserved areas:\n");
+		igt_map_foreach(ialb->reserved, entry) {
+			record = entry->data;
+			igt_info("handle = %d, offset = %"PRIu64" "
+				"(0x%"PRIx64", size = %"PRIu64" (0x%"PRIx64")\n",
+				 record->handle, record->offset, record->offset,
+				 record->size, record->size);
+
+		}
+
+	} else {
+		offset_bst = ialb->by_offset;
+
+		igt_bst_for_each_node(offset_bst, nodeptr) {
+			igt_bst_get_value(offset_bst, nodeptr);
+			total_free += hole->size;
+		}
+	}
+
+	igt_info("free space: %"PRIu64"B (0x%"PRIx64") (%.2f%% full)\n"
+		 "allocated objects: %"PRIu64", size: %"PRIu64
+		 " (%"PRIx64")\n"
+		 "reserved areas: %"PRIu64", size: %"PRIu64
+		 " (%"PRIx64")\n", total_free, total_free,
+		 ((double) (ialb->total_size - total_free) /
+		  (double) ialb->total_size) * 100.0,
+		 ialb->allocated_objects, ialb->allocated_size,
+		 ialb->allocated_size, ialb->reserved_areas,
+		 ialb->reserved_areas, ialb->reserved_size);
+
+	holes_bst_validate(ialb);
+}
+
+struct intel_allocator *
+intel_allocator_bst_create(int fd, uint64_t start, uint64_t end,
+			   enum allocator_strategy strategy)
+{
+	struct intel_allocator *ial;
+	struct intel_allocator_bst *ialb;
+
+	igt_debug("Using BST allocator\n");
+
+	ialb = malloc(sizeof(struct intel_allocator_bst));
+	igt_assert(ialb);
+	ial = malloc(sizeof(struct intel_allocator));
+	igt_assert(ial);
+
+	/* Allocated object are stored by 4-byte-long handles. */
+	ialb->objects = igt_map_create(hash_handles, equal_handles);
+	/* Reserved object are stored by 8-byte-long offsets. */
+	ialb->reserved = igt_map_create(hash_offsets, equal_offsets);
+	igt_assert(ialb->objects && ialb->reserved);
+
+	ialb->by_offset = igt_bst_avl_create();
+	ialb->by_size = igt_bst_avl_create();
+
+	igt_assert(start < end);
+	igt_assert(end <= IALB_MAX_ADDR);
+	ialb->start = start;
+	ialb->end = end;
+
+	ialb->total_size = ialb->end - ialb->start;
+	ialb->allocated_size = 0;
+	ialb->allocated_objects = 0;
+	ialb->reserved_size = 0;
+	ialb->reserved_areas = 0;
+	__intel_allocator_bst_add_hole(ialb, ialb->start, ialb->total_size);
+
+	ial->fd = fd;
+	ial->type = INTEL_ALLOCATOR_BST;
+	ial->strategy = strategy;
+	ial->refcount = 0;
+	ial->priv = ialb;
+	ial->get_address_range = intel_allocator_bst_get_address_range;
+	ial->alloc = intel_allocator_bst_alloc;
+	ial->is_allocated = intel_allocator_bst_is_allocated;
+	ial->free = intel_allocator_bst_free;
+	ial->reserve = intel_allocator_bst_reserve;
+	ial->unreserve = intel_allocator_bst_unreserve;
+	ial->is_reserved = intel_allocator_bst_is_reserved;
+	ial->destroy = intel_allocator_bst_destroy;
+	ial->is_empty = intel_allocator_bst_is_empty;
+	ial->print = intel_allocator_bst_print;
+
+	return ial;
+}
diff --git a/lib/meson.build b/lib/meson.build
index a1aa0ee12..62c8dbf20 100644
--- a/lib/meson.build
+++ b/lib/meson.build
@@ -38,6 +38,7 @@ lib_sources = [
 	'igt_x86.c',
 	'instdone.c',
 	'intel_allocator.c',
+	'intel_allocator_bst.c',
 	'intel_allocator_msgchannel.c',
 	'intel_allocator_random.c',
 	'intel_allocator_reloc.c',
-- 
2.25.1



More information about the igt-dev mailing list