blender
/
blender
129
397
Fork 571
Code Issues 6.2k Pull Requests 752 Projects 19 Wiki Activity

GPU: Vertex Format: Increase number of byte per attribute name 

This reduces the risks of hash collision while maintaining a small number
of character per attrib.
This commit is contained in:
Clément Foucault 2019-08-14 23:43:03 +02:00
parent 80e9eb66d5
commit 6fcd071c7b
2 changed files with 24 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
source/blender/gpu/GPU_vertex_format.h
View File

@ -35,7 +35,7 @@
#define GPU_VERT_ATTR_NAMES_BUF_LEN 256
#define GPU_VERT_FORMAT_MAX_NAMES 63 /* More than enough, actual max is ~30. */
/* Computed as GPU_VERT_ATTR_NAMES_BUF_LEN / 30 (actual max format name). */
#define GPU_MAX_SAFE_ATTRIB_NAME 8
#define GPU_MAX_SAFE_ATTRIB_NAME 12
typedef enum {
GPU_COMP_I8,

49
source/blender/gpu/intern/gpu_vertex_format.c
View File

@ -220,20 +220,6 @@ int GPU_vertformat_attr_id_get(const GPUVertFormat *format, const char *name)
return -1;
}
/* Encode 4 original bytes into 6 safe bytes. */
static void safe_bytes(char out[6], const char data[4])
{
char safe_chars[63] = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789_";
uint32_t in = *(uint32_t *)data;
for (int i = 0; i < 6; i++) {
/* Encoding in base63 */
out[i] = safe_chars[in % 63u];
in /= 63u;
}
}
#if 0 /* For when we can use 11chars names. */
/* Encode 8 original bytes into 11 safe bytes. */
static void safe_bytes(char out[11], const char data[8])
{
@ -246,7 +232,6 @@ static void safe_bytes(char out[11], const char data[8])
in /= 63lu;
}
}
#endif
/* Warning: Always add a prefix to the result of this function as
* the generated string can start with a number and not be a valid attribute name. */
@ -254,22 +239,34 @@ void GPU_vertformat_safe_attrib_name(const char *attrib_name,
char *r_safe_name,
uint UNUSED(max_len))
{
char data[4] = {0};
/* We use a hash to identify each data layer based on its name.
* NOTE: This is still prone to hash collision but the risks are very low.*/
*(uint *)data = BLI_ghashutil_strhash_p_murmur(attrib_name);
char data[8] = {0};
uint len = strlen(attrib_name);
if (len > 8) {
/* Start with the first 4 chars of the name; */
for (int i = 0; i < 4; i++) {
data[i] = attrib_name[i];
}
/* We use a hash to identify each data layer based on its name.
* NOTE: This is still prone to hash collision but the risks are very low.*/
/* Start hashing after the first 2 chars. */
*(uint *)&data[4] = BLI_ghashutil_strhash_p_murmur(attrib_name + 4);
}
else {
/* Copy the whole name. Collision is barelly possible
* (hash would have to be equal to the last 4 bytes). */
for (int i = 0; i < 8 && attrib_name[i] != '\0'; i++) {
data[i] = attrib_name[i];
}
}
/* Convert to safe bytes characters. */
safe_bytes(r_safe_name, data);
/* End the string */
r_safe_name[6] = '\0';
r_safe_name[11] = '\0';
/* TOOD(fclem) When UV and Tangent buffers will be separated, name buffer will have plenty more
* space. In this case, we can think of having ~13 chars for each name which would be enough to
* encode some of the input string along with the hash to reduce colision possibility. */
BLI_assert(GPU_MAX_SAFE_ATTRIB_NAME >= 7);
BLI_assert(GPU_MAX_SAFE_ATTRIB_NAME >= 12);
#if 0 /* For debugging */
printf("%s > %x > %s\n", attrib_name, *(uint32_t *)data, r_safe_name);
printf("%s > %lx > %s\n", attrib_name, *(uint64_t *)data, r_safe_name);
#endif
}