fdo-mirrors/mesa
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/mesa/mesa.git synced 2026-05-09 02:28:10 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

vc4: Simplify the load/store utile functions.

Browse source 
They now have less of a dependency on the cpp, and don't have to do a
divide.

Hacking up mesa-demos teximage to do only one subtest and not draw
points, I saw 1024x1024 glTexSubImage2D() improve by 4.86939% +/-
1.40408% (n=30) and glGetTexImage() by 2.18978% +/- 0.140268% (n=5).
This commit is contained in:
Eric Anholt 2017-01-04 12:40:37 -08:00
parent 7b8c67b3cc
commit e64b1169d3
1 changed files with 22 additions and 10 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

32
src/gallium/drivers/vc4/vc4_tiling.c
View file

@ -87,6 +87,22 @@ vc4_utile_height(int cpp)
}
}
/** Returns the stride in bytes of a 64-byte microtile. */
static uint32_t
vc4_utile_stride(int cpp)
{
switch (cpp) {
case 1:
return 8;
case 2:
case 4:
case 8:
return 16;
default:
unreachable("bad cpp");
}
}
/**
* The texture unit decides what tiling format a particular miplevel is using
* this function, so we lay out our miptrees accordingly.
@ -101,25 +117,21 @@ vc4_size_is_lt(uint32_t width, uint32_t height, int cpp)
void
vc4_load_utile(void *dst, void *src, uint32_t dst_stride, uint32_t cpp)
{
uint32_t utile_h = vc4_utile_height(cpp);
uint32_t row_size = 64 / utile_h;
uint32_t src_stride = vc4_utile_stride(cpp);
for (int y = 0; y < utile_h; y++) {
memcpy(dst, src, row_size);
for (uint32_t src_offset = 0; src_offset < 64; src_offset += src_stride) {
memcpy(dst, src + src_offset, src_stride);
dst += dst_stride;
src += row_size;
}
}
void
vc4_store_utile(void *dst, void *src, uint32_t src_stride, uint32_t cpp)
{
uint32_t utile_h = vc4_utile_height(cpp);
uint32_t row_size = 64 / utile_h;
uint32_t dst_stride = vc4_utile_stride(cpp);
for (int y = 0; y < utile_h; y++) {
memcpy(dst, src, row_size);
dst += row_size;
for (uint32_t dst_offset = 0; dst_offset < 64; dst_offset += dst_stride) {
memcpy(dst + dst_offset, src, dst_stride);
src += src_stride;
}
}