diff options
author | Jason Ekstrand <[email protected]> | 2020-03-18 11:48:47 -0500 |
---|---|---|
committer | Jason Ekstrand <[email protected]> | 2020-03-20 15:31:10 -0500 |
commit | aee004a7c8900938d1c17f0ac299d40001b383b0 (patch) | |
tree | 36524dad187a66e25c26ff397315a52740e328c8 /src/util | |
parent | 6be65b077743fc80efe061b1e05cb13b2ff1a6b1 (diff) |
util/sparse_array: Stash the node level in the node pointer
This reworks the data structure a bit and, in my view, simplifies it.
Instead of each node having a header which has the node level in it, we
use the bottom 6 bits of the pointer for that. This requires us to
allocate with the os_malloc/free_aligned helpers (which call into
posix_memalign on Linux) but cache-line aligning our allocations is
actually probably a good thing given that we're doing atomics on them.
The primary advantages to doing this is that it changes the number of
memory accesses per tree level from 2 to 1 when walking the tree because
we no longer have to look at node->level.
Reviewed-by: Lionel Landwerlin <[email protected]>
Tested-by: Marge Bot <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4228>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4228>
Diffstat (limited to 'src/util')
-rw-r--r-- | src/util/sparse_array.c | 138 | ||||
-rw-r--r-- | src/util/sparse_array.h | 2 |
2 files changed, 80 insertions, 60 deletions
diff --git a/src/util/sparse_array.c b/src/util/sparse_array.c index 7f583cfe133..61dbd2acdff 100644 --- a/src/util/sparse_array.c +++ b/src/util/sparse_array.c @@ -22,12 +22,18 @@ */ #include "sparse_array.h" +#include "os_memory.h" -struct util_sparse_array_node { - uint32_t level; - uint32_t _pad; - uint64_t max_idx; -}; +/* Aligning our allocations to 64 has two advantages: + * + * 1. On x86 platforms, it means that they are cache-line aligned so we + * reduce the likelihood that one of our allocations shares a cache line + * with some other allocation. + * + * 2. It lets us use the bottom 6 bits of the pointer to store the tree level + * of the node so we can avoid some pointer indirections. + */ +#define NODE_ALLOC_ALIGN 64 void util_sparse_array_init(struct util_sparse_array *arr, @@ -39,27 +45,45 @@ util_sparse_array_init(struct util_sparse_array *arr, assert(node_size >= 2 && node_size == (1ull << arr->node_size_log2)); } +#define NODE_PTR_MASK (~((uintptr_t)NODE_ALLOC_ALIGN - 1)) +#define NODE_LEVEL_MASK ((uintptr_t)NODE_ALLOC_ALIGN - 1) +#define NULL_NODE 0 + +static inline uintptr_t +_util_sparse_array_node(void *data, unsigned level) +{ + assert(data != NULL); + assert(((uintptr_t)data & NODE_LEVEL_MASK) == 0); + assert((level & NODE_PTR_MASK) == 0); + return (uintptr_t)data | level; +} + static inline void * -_util_sparse_array_node_data(struct util_sparse_array_node *node) +_util_sparse_array_node_data(uintptr_t handle) +{ + return (void *)(handle & NODE_PTR_MASK); +} + +static inline unsigned +_util_sparse_array_node_level(uintptr_t handle) { - return node + 1; + return handle & NODE_LEVEL_MASK; } static inline void _util_sparse_array_node_finish(struct util_sparse_array *arr, - struct util_sparse_array_node *node) + uintptr_t node) { - if (node->level > 0) { - struct util_sparse_array_node **children = - _util_sparse_array_node_data(node); + if (_util_sparse_array_node_level(node) > 0) { + uintptr_t *children = _util_sparse_array_node_data(node); size_t node_size = 1ull << arr->node_size_log2; for (size_t i = 0; i < node_size; i++) { - if (children[i] != NULL) + if (children[i]) _util_sparse_array_node_finish(arr, children[i]); } } - free(node); + os_free_aligned(_util_sparse_array_node_data(node)); } void @@ -69,36 +93,35 @@ util_sparse_array_finish(struct util_sparse_array *arr) _util_sparse_array_node_finish(arr, arr->root); } -static inline struct util_sparse_array_node * -_util_sparse_array_alloc_node(struct util_sparse_array *arr, +static inline uintptr_t +_util_sparse_array_node_alloc(struct util_sparse_array *arr, unsigned level) { - size_t size = sizeof(struct util_sparse_array_node); + size_t size; if (level == 0) { - size += arr->elem_size << arr->node_size_log2; + size = arr->elem_size << arr->node_size_log2; } else { - size += sizeof(struct util_sparse_array_node *) << arr->node_size_log2; + size = sizeof(uintptr_t) << arr->node_size_log2; } - struct util_sparse_array_node *node = calloc(1, size); - node->level = level; + void *data = os_malloc_aligned(size, NODE_ALLOC_ALIGN); + memset(data, 0, size); - return node; + return _util_sparse_array_node(data, level); } -static inline struct util_sparse_array_node * -_util_sparse_array_set_or_free_node(struct util_sparse_array_node **node_ptr, - struct util_sparse_array_node *cmp_node, - struct util_sparse_array_node *node) +static inline uintptr_t +_util_sparse_array_set_or_free_node(uintptr_t *node_ptr, + uintptr_t cmp_node, + uintptr_t node) { - struct util_sparse_array_node *prev_node = - p_atomic_cmpxchg(node_ptr, cmp_node, node); + uintptr_t prev_node = p_atomic_cmpxchg(node_ptr, cmp_node, node); if (prev_node != cmp_node) { /* We lost the race. Free this one and return the one that was already * allocated. */ - free(node); + os_free_aligned(_util_sparse_array_node_data(node)); return prev_node; } else { return node; @@ -109,32 +132,31 @@ void * util_sparse_array_get(struct util_sparse_array *arr, uint64_t idx) { const unsigned node_size_log2 = arr->node_size_log2; - struct util_sparse_array_node *root = p_atomic_read(&arr->root); - if (unlikely(root == NULL)) { + uintptr_t root = p_atomic_read(&arr->root); + if (unlikely(!root)) { unsigned root_level = 0; uint64_t idx_iter = idx >> node_size_log2; while (idx_iter) { idx_iter >>= node_size_log2; root_level++; } - struct util_sparse_array_node *new_root = - _util_sparse_array_alloc_node(arr, root_level); - root = _util_sparse_array_set_or_free_node(&arr->root, NULL, new_root); + uintptr_t new_root = _util_sparse_array_node_alloc(arr, root_level); + root = _util_sparse_array_set_or_free_node(&arr->root, + NULL_NODE, new_root); } while (1) { - uint64_t root_idx = idx >> (root->level * node_size_log2); + unsigned root_level = _util_sparse_array_node_level(root); + uint64_t root_idx = idx >> (root_level * node_size_log2); if (likely(root_idx < (1ull << node_size_log2))) break; /* In this case, we have a root but its level is low enough that the * requested index is out-of-bounds. */ - struct util_sparse_array_node *new_root = - _util_sparse_array_alloc_node(arr, root->level + 1); + uintptr_t new_root = _util_sparse_array_node_alloc(arr, root_level + 1); - struct util_sparse_array_node **new_root_children = - _util_sparse_array_node_data(new_root); + uintptr_t *new_root_children = _util_sparse_array_node_data(new_root); new_root_children[0] = root; /* We only add one at a time instead of the whole tree because it's @@ -146,43 +168,40 @@ util_sparse_array_get(struct util_sparse_array *arr, uint64_t idx) root = _util_sparse_array_set_or_free_node(&arr->root, root, new_root); } - struct util_sparse_array_node *node = root; - while (node->level > 0) { - uint64_t child_idx = (idx >> (node->level * node_size_log2)) & + void *node_data = _util_sparse_array_node_data(root); + unsigned node_level = _util_sparse_array_node_level(root); + while (node_level > 0) { + uint64_t child_idx = (idx >> (node_level * node_size_log2)) & ((1ull << node_size_log2) - 1); - struct util_sparse_array_node **children = - _util_sparse_array_node_data(node); - struct util_sparse_array_node *child = - p_atomic_read(&children[child_idx]); + uintptr_t *children = node_data; + uintptr_t child = p_atomic_read(&children[child_idx]); - if (unlikely(child == NULL)) { - child = _util_sparse_array_alloc_node(arr, node->level - 1); + if (unlikely(!child)) { + child = _util_sparse_array_node_alloc(arr, node_level - 1); child = _util_sparse_array_set_or_free_node(&children[child_idx], - NULL, child); + NULL_NODE, child); } - node = child; + node_data = _util_sparse_array_node_data(child); + node_level = _util_sparse_array_node_level(child); } uint64_t elem_idx = idx & ((1ull << node_size_log2) - 1); - return (void *)((char *)_util_sparse_array_node_data(node) + - (elem_idx * arr->elem_size)); + return (void *)((char *)node_data + (elem_idx * arr->elem_size)); } static void validate_node_level(struct util_sparse_array *arr, - struct util_sparse_array_node *node, - unsigned level) + uintptr_t node, unsigned level) { - assert(node->level == level); + assert(_util_sparse_array_node_level(node) == level); - if (node->level > 0) { - struct util_sparse_array_node **children = - _util_sparse_array_node_data(node); + if (_util_sparse_array_node_level(node) > 0) { + uintptr_t *children = _util_sparse_array_node_data(node); size_t node_size = 1ull << arr->node_size_log2; for (size_t i = 0; i < node_size; i++) { - if (children[i] != NULL) + if (children[i]) validate_node_level(arr, children[i], level - 1); } } @@ -191,7 +210,8 @@ validate_node_level(struct util_sparse_array *arr, void util_sparse_array_validate(struct util_sparse_array *arr) { - validate_node_level(arr, arr->root, arr->root->level); + uintptr_t root = p_atomic_read(&arr->root); + validate_node_level(arr, root, _util_sparse_array_node_level(root)); } void diff --git a/src/util/sparse_array.h b/src/util/sparse_array.h index 3947a2fa81b..f91fe21dae2 100644 --- a/src/util/sparse_array.h +++ b/src/util/sparse_array.h @@ -69,7 +69,7 @@ struct util_sparse_array { size_t elem_size; unsigned node_size_log2; - struct util_sparse_array_node *root; + uintptr_t root; }; void util_sparse_array_init(struct util_sparse_array *arr, |