mirrors
/
Mother2GbaTranslation
mirror of https://github.com/jeffman/Mother2GbaTranslation.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Improve buffer storage performance by 1/3 

Also improves buffer loading performance by 1/5.
This commit is contained in:
Lorenzooone 2019-09-18 16:51:53 +02:00
parent 2f2970aae0
commit 8402fa5ae9
2 changed files with 61 additions and 32 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

91
src/c/vwf.c
View File

@ -1696,43 +1696,70 @@ void print_blankstr_buffer(int x, int y, int width, byte *dest)
void load_pixels_overworld_buffer()
{
byte* buffer = (byte*)(OVERWORLD_BUFFER - ((*tile_offset) * TILESET_OFFSET_BUFFER_MULTIPLIER));
for(int y = 0; y < 0x10; y++)
for(int x = 0; x < 0x1C; x++)
{
int tile_buffer = get_tile_number_with_offset_buffer(x + 1, y + 1);
int tile = get_tile_number_with_offset(x + 1, y + 1);
int foregroundRow = 0xFFFFFFFF;
buffer[(tile_buffer * 8) + 0] = reduce_bit_depth(vram[(tile * 8) + 0], foregroundRow);
buffer[(tile_buffer * 8) + 1] = reduce_bit_depth(vram[(tile * 8) + 1], foregroundRow);
buffer[(tile_buffer * 8) + 2] = reduce_bit_depth(vram[(tile * 8) + 2], foregroundRow);
buffer[(tile_buffer * 8) + 3] = reduce_bit_depth(vram[(tile * 8) + 3], foregroundRow);
buffer[(tile_buffer * 8) + 4] = reduce_bit_depth(vram[(tile * 8) + 4], foregroundRow);
buffer[(tile_buffer * 8) + 5] = reduce_bit_depth(vram[(tile * 8) + 5], foregroundRow);
buffer[(tile_buffer * 8) + 6] = reduce_bit_depth(vram[(tile * 8) + 6], foregroundRow);
buffer[(tile_buffer * 8) + 7] = reduce_bit_depth(vram[(tile * 8) + 7], foregroundRow);
}
for(int i = 0; i < 8 * 0x1C; i++)
{
//Doing this saves about 100k cycles during load
int tile = m2_coord_table[i] + *tile_offset;
int addedValue = (i >> 5) << 6;
int tile_buffer = (i & 0x1F) + addedValue + *tile_offset;
int foregroundRow = 0xFFFFFFFF;
buffer[(tile_buffer * 8) + 0] = reduce_bit_depth(vram[(tile * 8) + 0], foregroundRow);
buffer[(tile_buffer * 8) + 1] = reduce_bit_depth(vram[(tile * 8) + 1], foregroundRow);
buffer[(tile_buffer * 8) + 2] = reduce_bit_depth(vram[(tile * 8) + 2], foregroundRow);
buffer[(tile_buffer * 8) + 3] = reduce_bit_depth(vram[(tile * 8) + 3], foregroundRow);
buffer[(tile_buffer * 8) + 4] = reduce_bit_depth(vram[(tile * 8) + 4], foregroundRow);
buffer[(tile_buffer * 8) + 5] = reduce_bit_depth(vram[(tile * 8) + 5], foregroundRow);
buffer[(tile_buffer * 8) + 6] = reduce_bit_depth(vram[(tile * 8) + 6], foregroundRow);
buffer[(tile_buffer * 8) + 7] = reduce_bit_depth(vram[(tile * 8) + 7], foregroundRow);
tile_buffer += 0x20;
tile += 0x20;
buffer[(tile_buffer * 8) + 0] = reduce_bit_depth(vram[(tile * 8) + 0], foregroundRow);
buffer[(tile_buffer * 8) + 1] = reduce_bit_depth(vram[(tile * 8) + 1], foregroundRow);
buffer[(tile_buffer * 8) + 2] = reduce_bit_depth(vram[(tile * 8) + 2], foregroundRow);
buffer[(tile_buffer * 8) + 3] = reduce_bit_depth(vram[(tile * 8) + 3], foregroundRow);
buffer[(tile_buffer * 8) + 4] = reduce_bit_depth(vram[(tile * 8) + 4], foregroundRow);
buffer[(tile_buffer * 8) + 5] = reduce_bit_depth(vram[(tile * 8) + 5], foregroundRow);
buffer[(tile_buffer * 8) + 6] = reduce_bit_depth(vram[(tile * 8) + 6], foregroundRow);
buffer[(tile_buffer * 8) + 7] = reduce_bit_depth(vram[(tile * 8) + 7], foregroundRow);
}
}
void store_pixels_overworld_buffer(int totalYs)
{
byte* buffer = (byte*)(OVERWORLD_BUFFER - ((*tile_offset) * TILESET_OFFSET_BUFFER_MULTIPLIER));
for(int y = 0; y < totalYs; y++)
for(int x = 0; x < 0x1C; x++)
{
int tile = get_tile_number_with_offset(x + 1, y + 1);
int tile_buffer = get_tile_number_with_offset_buffer(x + 1, y + 1);
int* bufferValues = (int*)(&buffer[(tile_buffer * 8)]);
unsigned int first_half = bufferValues[0];
unsigned int second_half = bufferValues[1];
vram[(tile * 8) + 0] = m2_bits_to_nybbles_fast[(first_half >> 0) & 0xFF];
vram[(tile * 8) + 1] = m2_bits_to_nybbles_fast[(first_half >> 8) & 0xFF];
vram[(tile * 8) + 2] = m2_bits_to_nybbles_fast[(first_half >> 0x10) & 0xFF];
vram[(tile * 8) + 3] = m2_bits_to_nybbles_fast[(first_half >> 0x18) & 0xFF];
vram[(tile * 8) + 4] = m2_bits_to_nybbles_fast[(second_half >> 0) & 0xFF];
vram[(tile * 8) + 5] = m2_bits_to_nybbles_fast[(second_half >> 8) & 0xFF];
vram[(tile * 8) + 6] = m2_bits_to_nybbles_fast[(second_half >> 0x10) & 0xFF];
vram[(tile * 8) + 7] = m2_bits_to_nybbles_fast[(second_half >> 0x18) & 0xFF];
}
totalYs >>= 1;
int total = totalYs * 0x1C;
for(int i = 0; i < total; i++)
{
//Not using functions for the tile values saves about 30k cycles on average
int tile = m2_coord_table[i] + *tile_offset;
int addedValue = (i >> 5) << 6;
int tile_buffer = (i & 0x1F) + addedValue + *tile_offset;
int* bufferValues = (int*)(&buffer[(tile_buffer * 8)]);
unsigned int first_half = bufferValues[0];
unsigned int second_half = bufferValues[1];
vram[(tile * 8) + 0] = m2_bits_to_nybbles_fast[(first_half >> 0) & 0xFF];
vram[(tile * 8) + 1] = m2_bits_to_nybbles_fast[(first_half >> 8) & 0xFF];
vram[(tile * 8) + 2] = m2_bits_to_nybbles_fast[(first_half >> 0x10) & 0xFF];
vram[(tile * 8) + 3] = m2_bits_to_nybbles_fast[(first_half >> 0x18) & 0xFF];
vram[(tile * 8) + 4] = m2_bits_to_nybbles_fast[(second_half >> 0) & 0xFF];
vram[(tile * 8) + 5] = m2_bits_to_nybbles_fast[(second_half >> 8) & 0xFF];
vram[(tile * 8) + 6] = m2_bits_to_nybbles_fast[(second_half >> 0x10) & 0xFF];
vram[(tile * 8) + 7] = m2_bits_to_nybbles_fast[(second_half >> 0x18) & 0xFF];
//Do the tile right below (Saves about 50k cycles on average)
tile += 0x20;
bufferValues += 0x40;
first_half = bufferValues[0];
second_half = bufferValues[1];
vram[(tile * 8) + 0] = m2_bits_to_nybbles_fast[(first_half >> 0) & 0xFF];
vram[(tile * 8) + 1] = m2_bits_to_nybbles_fast[(first_half >> 8) & 0xFF];
vram[(tile * 8) + 2] = m2_bits_to_nybbles_fast[(first_half >> 0x10) & 0xFF];
vram[(tile * 8) + 3] = m2_bits_to_nybbles_fast[(first_half >> 0x18) & 0xFF];
vram[(tile * 8) + 4] = m2_bits_to_nybbles_fast[(second_half >> 0) & 0xFF];
vram[(tile * 8) + 5] = m2_bits_to_nybbles_fast[(second_half >> 8) & 0xFF];
vram[(tile * 8) + 6] = m2_bits_to_nybbles_fast[(second_half >> 0x10) & 0xFF];
vram[(tile * 8) + 7] = m2_bits_to_nybbles_fast[(second_half >> 0x18) & 0xFF];
}
}
// x,y: tile coordinates

2
src/m2-vwf-entries.asm
View File

@ -2977,6 +2977,7 @@ pop {r0-r3,pc}
//Stores the buffer into the vram. This avoids screen tearing.
store_pixels_overworld:
push {r0-r3,lr}
swi #5 //The improved performances allow using a VBlank before the storage in order to prevent screen tearing effectively
mov r0,#0x10
bl store_pixels_overworld_buffer
pop {r0-r3,pc}
@ -2985,6 +2986,7 @@ pop {r0-r3,pc}
//Stores the buffer into the vram. This avoids screen tearing.
store_pixels_overworld_psi_window:
push {r0-r3,lr}
swi #5 //The improved performances allow using a VBlank before the storage in order to prevent screen tearing effectively
mov r0,#0xA
bl store_pixels_overworld_buffer
pop {r0-r3,pc}