asrael/pxl8
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

remove simd, scalar math + compiler optimizations are good enough

Browse source
This commit is contained in:
asrael 2025-11-11 12:26:22 -06:00
parent e2c7998663
commit 4d84122ef3
8 changed files with 41 additions and 509 deletions
Download patch file Download diff file Expand all files Collapse all files

43
src/pxl8_blit.c
View file

@ -1,28 +1,15 @@
#include "pxl8_blit.h"
#include "pxl8_simd.h"
void pxl8_blit_simd_hicolor(u32* fb, u32 fb_width, const u32* sprite, u32 atlas_width,
i32 x, i32 y, u32 w, u32 h) {
void pxl8_blit_hicolor(u32* fb, u32 fb_width, const u32* sprite, u32 atlas_width,
i32 x, i32 y, u32 w, u32 h) {
u32* dest_base = fb + y * fb_width + x;
const u32* src_base = sprite;
for (u32 row = 0; row < h; row++) {
u32* dest_row = dest_base + row * fb_width;
const u32* src_row = src_base + row * atlas_width;
u32 col = 0;
for (; col + PXL8_SIMD_WIDTH_U32 <= w; col += PXL8_SIMD_WIDTH_U32) {
pxl8_simd_vec src_vec = pxl8_simd_load_u32(src_row + col);
pxl8_simd_vec dest_vec = pxl8_simd_load_u32(dest_row + col);
pxl8_simd_vec alpha_mask = pxl8_simd_alpha_mask_u32();
pxl8_simd_vec has_alpha = pxl8_simd_and(src_vec, alpha_mask);
pxl8_simd_vec zero = pxl8_simd_zero_u8();
pxl8_simd_vec mask = pxl8_simd_cmpeq_u32(has_alpha, zero);
pxl8_simd_vec result = pxl8_simd_blendv_u32(src_vec, dest_vec, mask);
pxl8_simd_store_u32(dest_row + col, result);
}
for (; col < w; col++) {
for (u32 col = 0; col < w; col++) {
if (src_row[col] & 0xFF000000) {
dest_row[col] = src_row[col];
}
@ -30,26 +17,16 @@ void pxl8_blit_simd_hicolor(u32* fb, u32 fb_width, const u32* sprite, u32 atlas_
}
}
void pxl8_blit_simd_indexed(u8* fb, u32 fb_width, const u8* sprite, u32 atlas_width,
i32 x, i32 y, u32 w, u32 h) {
void pxl8_blit_indexed(u8* fb, u32 fb_width, const u8* sprite, u32 atlas_width,
i32 x, i32 y, u32 w, u32 h) {
u8* dest_base = fb + y * fb_width + x;
const u8* src_base = sprite;
for (u32 row = 0; row < h; row++) {
u8* dest_row = dest_base + row * fb_width;
const u8* src_row = src_base + row * atlas_width;
u32 col = 0;
for (; col + PXL8_SIMD_WIDTH_U8 <= w; col += PXL8_SIMD_WIDTH_U8) {
pxl8_simd_vec src_vec = pxl8_simd_load_u8(src_row + col);
pxl8_simd_vec dest_vec = pxl8_simd_load_u8(dest_row + col);
pxl8_simd_vec zero = pxl8_simd_zero_u8();
pxl8_simd_vec mask = pxl8_simd_cmpeq_u8(src_vec, zero);
pxl8_simd_vec result = pxl8_simd_blendv_u8(src_vec, dest_vec, mask);
pxl8_simd_store_u8(dest_row + col, result);
}
for (; col < w; col++) {
for (u32 col = 0; col < w; col++) {
if (src_row[col] != 0) {
dest_row[col] = src_row[col];
}