RaySollium99/picodrive
1
0
Fork 0
mirror of https://github.com/RaySollium99/picodrive.git synced 2025-09-06 15:48:05 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

sdl, complete overhaul of hardware/software scaling

Browse source
This commit is contained in:
kub 2021-09-10 18:46:05 +02:00
parent 6651998e9f
commit d5d1778252
21 changed files with 1233 additions and 547 deletions
Download patch file Download diff file Expand all files Collapse all files

2
platform/common/emu.c
View file

@ -1219,7 +1219,7 @@ void emu_cmn_forced_frame(int no_scale, int do_emu, void *buf)
memset32((short *)g_screen_ptr + g_screen_ppitch * y, 0,
g_screen_width * 2 / 4);
PicoIn.opt &= ~POPT_ALT_RENDERER;
PicoIn.opt &= ~(POPT_ALT_RENDERER|POPT_EN_SOFTSCALE);
PicoIn.opt |= POPT_ACC_SPRITES;
if (!no_scale && currentConfig.scaling)
PicoIn.opt |= POPT_EN_SOFTSCALE;

16
platform/common/emu.h
View file

@ -37,10 +37,19 @@ extern int g_screen_ppitch; // pitch in pixels
enum {
EOPT_SCALE_NONE = 0,
EOPT_SCALE_SW,
// linux, GP2X:
EOPT_SCALE_SW = 1,
EOPT_SCALE_HW,
};
enum {
EOPT_FILTER_NONE = 0,
// software scalers
EOPT_FILTER_SMOOTHER = 1,
EOPT_FILTER_BILINEAR1,
EOPT_FILTER_BILINEAR2,
};
enum {
EOPT_CONFIRM_NONE = 0,
EOPT_CONFIRM_SAVE = 1,
@ -63,7 +72,7 @@ typedef struct _currentConfig_t {
int CPUclock;
int volume;
int gamma;
int scaling; // gp2x: EOPT_SCALE_*; psp: bilinear filtering
int scaling; // EOPT_SCALE_*
int vscaling;
int rotation; // for UIQ
float scale; // psp: screen scale
@ -72,7 +81,7 @@ typedef struct _currentConfig_t {
int turbo_rate;
int renderer;
int renderer32x;
int filter; // pandora
int filter; // EOPT_FILTER_* video filter
int analog_deadzone;
int msh2_khz;
int ssh2_khz;
@ -180,6 +189,7 @@ void plat_update_volume(int has_changed, int is_up);
/* should be in libpicofe/plat.h */
void plat_video_clear_status(void);
void plat_video_clear_buffers(void);
void plat_video_set_size(int w, int h);
#ifdef __cplusplus
} // extern "C"

83
platform/common/plat_sdl.c
View file

@ -22,6 +22,7 @@
#include <pico/pico_int.h>
static void *shadow_fb;
static struct area { int w, h; } area;
static struct in_pdata in_sdl_platform_data = {
.defbinds = in_sdl_defbinds,
@ -81,54 +82,77 @@ void bgr_to_uyvy_init(void)
}
}
void rgb565_to_uyvy(void *d, const void *s, int pixels, int x2)
void rgb565_to_uyvy(void *d, const void *s, int w, int h, int pitch, int x2)
{
uint32_t *dst = d;
const uint16_t *src = s;
int i;
if (x2)
for (; pixels > 0; src += 4, dst += 4, pixels -= 4)
{
struct uyvy *uyvy0 = yuv_uyvy + src[0], *uyvy1 = yuv_uyvy + src[1];
struct uyvy *uyvy2 = yuv_uyvy + src[2], *uyvy3 = yuv_uyvy + src[3];
if (x2) while (h--) {
for (i = w; i > 0; src += 4, dst += 4, i -= 4)
{
struct uyvy *uyvy0 = yuv_uyvy + src[0], *uyvy1 = yuv_uyvy + src[1];
struct uyvy *uyvy2 = yuv_uyvy + src[2], *uyvy3 = yuv_uyvy + src[3];
#if CPU_IS_LE
dst[0] = (uyvy0->y << 24) | uyvy0->vyu;
dst[1] = (uyvy1->y << 24) | uyvy1->vyu;
dst[2] = (uyvy2->y << 24) | uyvy2->vyu;
dst[3] = (uyvy3->y << 24) | uyvy3->vyu;
dst[0] = (uyvy0->y << 24) | uyvy0->vyu;
dst[1] = (uyvy1->y << 24) | uyvy1->vyu;
dst[2] = (uyvy2->y << 24) | uyvy2->vyu;
dst[3] = (uyvy3->y << 24) | uyvy3->vyu;
#else
dst[0] = uyvy0->y | (uyvy0->vyu << 8);
dst[1] = uyvy1->y | (uyvy1->vyu << 8);
dst[2] = uyvy2->y | (uyvy2->vyu << 8);
dst[3] = uyvy3->y | (uyvy3->vyu << 8);
dst[0] = uyvy0->y | (uyvy0->vyu << 8);
dst[1] = uyvy1->y | (uyvy1->vyu << 8);
dst[2] = uyvy2->y | (uyvy2->vyu << 8);
dst[3] = uyvy3->y | (uyvy3->vyu << 8);
#endif
} else
for (; pixels > 0; src += 4, dst += 2, pixels -= 4)
{
struct uyvy *uyvy0 = yuv_uyvy + src[0], *uyvy1 = yuv_uyvy + src[1];
struct uyvy *uyvy2 = yuv_uyvy + src[2], *uyvy3 = yuv_uyvy + src[3];
}
src += pitch - w;
} else while (h--) {
for (i = w; i > 0; src += 4, dst += 2, i -= 4)
{
struct uyvy *uyvy0 = yuv_uyvy + src[0], *uyvy1 = yuv_uyvy + src[1];
struct uyvy *uyvy2 = yuv_uyvy + src[2], *uyvy3 = yuv_uyvy + src[3];
#if CPU_IS_LE
dst[0] = (uyvy1->y << 24) | uyvy0->vyu;
dst[1] = (uyvy3->y << 24) | uyvy2->vyu;
dst[0] = (uyvy1->y << 24) | uyvy0->vyu;
dst[1] = (uyvy3->y << 24) | uyvy2->vyu;
#else
dst[0] = uyvy1->y | (uyvy0->vyu << 8);
dst[1] = uyvy3->y | (uyvy2->vyu << 8);
dst[0] = uyvy1->y | (uyvy0->vyu << 8);
dst[1] = uyvy3->y | (uyvy2->vyu << 8);
#endif
}
src += pitch - w;
}
}
static int clear_buf_cnt, clear_stat_cnt;
void plat_video_set_size(int w, int h)
{
if (area.w != w || area.h != h) {
area = (struct area) { w, h };
if (plat_sdl_change_video_mode(w, h, 0) < 0) {
// failed, revert to original resolution
plat_sdl_change_video_mode(g_screen_width, g_screen_height, 0);
w = g_screen_width, h = g_screen_height;
}
if (!plat_sdl_overlay && !plat_sdl_gl_active) {
g_screen_width = w;
g_screen_height = h;
g_screen_ppitch = w;
g_screen_ptr = plat_sdl_screen->pixels;
}
}
}
void plat_video_flip(void)
{
if (plat_sdl_overlay != NULL) {
SDL_Rect dstrect =
{ 0, 0, plat_sdl_screen->w, plat_sdl_screen->h };
SDL_LockYUVOverlay(plat_sdl_overlay);
rgb565_to_uyvy(plat_sdl_overlay->pixels[0], shadow_fb,
g_screen_ppitch * g_screen_height,
plat_sdl_overlay->w > 2*plat_sdl_overlay->h);
area.w, area.h, g_screen_ppitch,
plat_sdl_overlay->w >= 2*area.w);
SDL_UnlockYUVOverlay(plat_sdl_overlay);
SDL_DisplayYUVOverlay(plat_sdl_overlay, &dstrect);
}
@ -205,7 +229,7 @@ void plat_video_menu_end(void)
SDL_LockYUVOverlay(plat_sdl_overlay);
rgb565_to_uyvy(plat_sdl_overlay->pixels[0], shadow_fb,
g_menuscreen_pp * g_menuscreen_h, 0);
g_menuscreen_w, g_menuscreen_h, g_menuscreen_pp, 0);
SDL_UnlockYUVOverlay(plat_sdl_overlay);
SDL_DisplayYUVOverlay(plat_sdl_overlay, &dstrect);
@ -227,10 +251,10 @@ void plat_video_menu_leave(void)
void plat_video_loop_prepare(void)
{
// take over any new vout settings XXX ask plat_sdl for scaling instead!
// take over any new vout settings
plat_sdl_change_video_mode(g_menuscreen_w, g_menuscreen_h, 0);
// switch over to scaled output if available
if (plat_sdl_overlay != NULL || plat_sdl_gl_active || currentConfig.scaling != EOPT_SCALE_NONE) {
if (plat_sdl_overlay != NULL || plat_sdl_gl_active) {
g_screen_width = 320;
g_screen_height = 240;
g_screen_ppitch = g_screen_width;
@ -246,6 +270,7 @@ void plat_video_loop_prepare(void)
g_screen_ptr = plat_sdl_screen->pixels;
}
plat_video_set_buffer(g_screen_ptr);
plat_video_set_size(g_screen_width, g_screen_height);
}
void plat_early_init(void)

638
platform/common/upscale.c
View file

@ -7,7 +7,7 @@
* nn: nearest neighbour
* snn: "smoothed" nearest neighbour (see below)
* bln: n-level-bilinear with n quantized weights
* quantization: 0: a<1/2*n, 1/n: 1/2*n<=a<3/2*n, etc
* quantization: 0: a<1/(2*n), 1/n: 1/(2*n)<=a<3/(2*n), etc
* currently n=2, n=4 are implemented (there's n=8 mixing, but no filters)
* [NB this has been brought to my attn, which is probably the same as bl2:
* https://www.drdobbs.com/image-scaling-with-bresenham/184405045?pgno=1]
@ -18,490 +18,586 @@
* a sharper look than a bilinear filter, at the price of some visible jags
* on diagonal edges.
*
* scaling modes:
* 256x___ -> 320x___ only horizontal scaling. Produces an aspect error of
* ~7% for NTSC 224 line modes, but is correct for PAL
* 256/320x224/240
* -> 320x240 always produces 320x240 at DAR 4:3
* 160x144 -> 320x240 game gear (currently unused)
* example scaling modes:
* 256x_Y_ -> 320x_Y_, H32/mode 4, PAR 5:4, for PAL DAR 4:3 (NTSC 7% aspect err)
* 256x224 -> 320x240, H32/mode 4, PAR 5:4, for NTSC DAR 4:3 (PAL 7% aspect err)
* 320x224 -> 320x240, PAR 1:1, for NTSC, DAR 4:3 (PAL 7% etc etc...)
* 160x144 -> 320x240: GG, PAR 6:5, scaling to 320x240 for DAR 4:3
*
* (C) 2021 kub <derkub@gmail.com>
*/
#include "upscale.h"
/* 256x___ -> 320x___, H32/mode 4, PAR 5:4, for PAL DAR 4:3 (wrong for NTSC) */
void upscale_clut_nn_256_320x___(u8 *__restrict di, int ds, u8 *__restrict si, int ss, int height)
/* X x Y -> X*5/4 x Y */
void upscale_clut_nn_x_4_5(u8 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height)
{
int y;
for (y = 0; y < height; y++) {
h_upscale_nn_4_5(di, ds, si, ss, 256, f_nop);
h_upscale_nn_4_5(di, ds, si, ss, width, f_nop);
}
}
void upscale_rgb_nn_256_320x___(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int height, u16 *pal)
void upscale_rgb_nn_x_4_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y;
for (y = 0; y < height; y++) {
h_upscale_nn_4_5(di, ds, si, ss, 256, f_pal);
h_upscale_nn_4_5(di, ds, si, ss, width, f_pal);
}
}
void upscale_rgb_snn_256_320x___(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int height, u16 *pal)
void upscale_rgb_snn_x_4_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y;
for (y = 0; y < height; y++) {
h_upscale_snn_4_5(di, ds, si, ss, 256, f_pal);
h_upscale_snn_4_5(di, ds, si, ss, width, f_pal);
}
}
void upscale_rgb_bl2_256_320x___(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int height, u16 *pal)
void upscale_rgb_bl2_x_4_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y;
for (y = 0; y < height; y++) {
h_upscale_bl2_4_5(di, ds, si, ss, 256, f_pal);
h_upscale_bl2_4_5(di, ds, si, ss, width, f_pal);
}
}
void upscale_rgb_bl4_256_320x___(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int height, u16 *pal)
void upscale_rgb_bl4_x_4_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y;
for (y = 0; y < height; y++) {
h_upscale_bl4_4_5(di, ds, si, ss, 256, f_pal);
h_upscale_bl4_4_5(di, ds, si, ss, width, f_pal);
}
}
/* 256x224 -> 320x240, H32/mode 4, PAR 5:4, for NTSC DAR 4:3 (wrong for PAL) */
void upscale_clut_nn_256_320x224_240(u8 *__restrict di, int ds, u8 *__restrict si, int ss)
/* X x Y -> X*5/4 x Y*17/16 */
void upscale_clut_nn_x_4_5_y_16_17(u8 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height)
{
int swidth = width * 5/4;
int y, j;
/* 14:15, 0 1 2 3 4 5 6 6 7 8 9 10 11 12 13 */
for (y = 0; y < 224; y += 14) {
/* lines 0-6 */
for (j = 0; j < 7; j++) {
h_upscale_nn_4_5(di, ds, si, ss, 256, f_nop);
}
/* lines 8-14 */
di += ds;
for (j = 0; j < 7; j++) {
h_upscale_nn_4_5(di, ds, si, ss, 256, f_nop);
}
/* line 7 */
di -= 8*ds;
v_copy(&di[0], &di[-ds], 320, f_nop);
di += 8*ds;
}
}
void upscale_rgb_nn_256_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
{
int y, j;
for (y = 0; y < 224; y += 14) {
for (j = 0; j < 7; j++) {
h_upscale_nn_4_5(di, ds, si, ss, 256, f_pal);
for (y = 0; y < height; y += 16) {
for (j = 0; j < 8; j++) {
h_upscale_nn_4_5(di, ds, si, ss, width, f_nop);
}
di += ds;
for (j = 0; j < 7; j++) {
h_upscale_nn_4_5(di, ds, si, ss, 256, f_pal);
for (j = 0; j < 8; j++) {
h_upscale_nn_4_5(di, ds, si, ss, width, f_nop);
}
di -= 8*ds;
v_copy(&di[0], &di[-ds], 320, f_nop);
di += 8*ds;
di -= 9*ds;
v_copy(&di[0], &di[-ds], swidth, f_nop);
di += 9*ds;
}
}
void upscale_rgb_snn_256_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
void upscale_rgb_nn_x_4_5_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int swidth = width * 5/4;
int y, j;
/* 14:15, 0 1 2 3 4 5 5+6 6+7 7+8 8 9 10 11 12 13 */
for (y = 0; y < 224; y += 14) {
for (j = 0; j < 7; j++) {
h_upscale_snn_4_5(di, ds, si, ss, 256, f_pal);
for (y = 0; y < height; y += 16) {
for (j = 0; j < 8; j++) {
h_upscale_nn_4_5(di, ds, si, ss, width, f_pal);
}
di += ds;
for (j = 0; j < 7; j++) {
h_upscale_snn_4_5(di, ds, si, ss, 256, f_pal);
for (j = 0; j < 8; j++) {
h_upscale_nn_4_5(di, ds, si, ss, width, f_pal);
}
di -= 9*ds;
v_copy(&di[0], &di[-ds], swidth, f_nop);
di += 9*ds;
}
}
void upscale_rgb_snn_x_4_5_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int swidth = width * 5/4;
int y, j;
for (y = 0; y < height; y += 16) {
for (j = 0; j < 8; j++) {
h_upscale_snn_4_5(di, ds, si, ss, width, f_pal);
}
di += ds;
for (j = 0; j < 8; j++) {
h_upscale_snn_4_5(di, ds, si, ss, width, f_pal);
}
/* mix lines 6-8 */
di -= 8*ds;
v_mix(&di[0], &di[-ds], &di[ds], 320, p_05, f_nop);
v_mix(&di[-ds], &di[-2*ds], &di[-ds], 320, p_05, f_nop);
v_mix(&di[ ds], &di[ ds], &di[ 2*ds], 320, p_05, f_nop);
di += 8*ds;
di -= 9*ds;
v_mix(&di[0], &di[-ds], &di[ds], swidth, p_05, f_nop);
v_mix(&di[-ds], &di[-2*ds], &di[-ds], swidth, p_05, f_nop);
v_mix(&di[ ds], &di[ ds], &di[ 2*ds], swidth, p_05, f_nop);
di += 9*ds;
}
}
void upscale_rgb_bln_256_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
void upscale_rgb_bl2_x_4_5_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int swidth = width * 5/4;
int y, j;
/* 14:15, 0 1 2 2+3 3+4 4+5 5+6 6+7 7+8 8+9 9+10 10+11 11 12 13 */
for (y = 0; y < 224; y += 14) {
/* lines 0-2 */
for (j = 0; j < 3; j++) {
h_upscale_bln_4_5(di, ds, si, ss, 256, f_pal);
for (y = 0; y < height; y += 16) {
for (j = 0; j < 4; j++) {
h_upscale_bl2_4_5(di, ds, si, ss, width, f_pal);
}
/* lines 3-11 mixing prep */
di += ds;
for (j = 0; j < 11; j++) {
h_upscale_bln_4_5(di, ds, si, ss, 256, f_pal);
di += ds;
for (j = 0; j < 12; j++) {
h_upscale_bl2_4_5(di, ds, si, ss, width, f_pal);
}
di -= 12*ds;
/* mixing line 3: line 2 = -ds, line 3 = +ds */
v_mix(&di[0], &di[-ds], &di[ds], 320, p_025, f_nop);
/* mix lines 3-10 */
di -= 13*ds;
v_mix(&di[0], &di[-ds], &di[ds], swidth, p_05, f_nop);
for (j = 0; j < 7; j++) {
di += ds;
/* mixing lines 4-5: line n-1 = 0, line n = +ds */
v_mix(&di[0], &di[0], &di[ds], swidth, p_05, f_nop);
}
di += 6*ds;
}
}
void upscale_rgb_bl4_x_4_5_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int swidth = width * 5/4;
int y, j;
for (y = 0; y < height; y += 16) {
for (j = 0; j < 2; j++) {
v_mix(&di[0], &di[0], &di[ds], 320, p_025, f_nop);
h_upscale_bl4_4_5(di, ds, si, ss, width, f_pal);
}
di += ds;
for (j = 0; j < 14; j++) {
h_upscale_bl4_4_5(di, ds, si, ss, width, f_pal);
}
di -= 15*ds;
/* mixing line 2: line 1 = -ds, line 2 = +ds */
v_mix(&di[0], &di[-ds], &di[ds], swidth, p_025, f_nop);
di += ds;
/* mixing lines 3-5: line n-1 = 0, line n = +ds */
for (j = 0; j < 3; j++) {
v_mix(&di[0], &di[0], &di[ds], swidth, p_025, f_nop);
di += ds;
}
/* mixing line 6-8 */
for (j = 0; j < 3; j++) {
v_mix(&di[0], &di[0], &di[ds], 320, p_05, f_nop);
/* mixing lines 6-9 */
for (j = 0; j < 4; j++) {
v_mix(&di[0], &di[0], &di[ds], swidth, p_05, f_nop);
di += ds;
}
/* mixing lines 9-11 */
for (j = 0; j < 3; j++) {
v_mix(&di[0], &di[0], &di[ds], 320, p_075, f_nop);
/* mixing lines 10-13 */
for (j = 0; j < 4; j++) {
v_mix(&di[0], &di[0], &di[ds], swidth, p_075, f_nop);
di += ds;
}
/* lines 12-14, already in place */
/* lines 14-16, already in place */
di += 3*ds;
}
}
void upscale_rgb_bl2_256_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
/* "classic" upscaler as found in several emulators. It's really more like a
* x*4/3, y*16/15 upscaler, with an additional 5th row/17th line just inserted
* from the source image. That gives nice n/4,n/16 alpha values plus better
* symmetry in each block and avoids "borrowing" a row/line between blocks.
*/
void upscale_rgb_bln_x_4_5_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int swidth = width * 5/4;
int y, j;
/* 14:15, 0 1 2 2+3 3+4 4+5 5+6 6+7 7+8 8+9 9+10 10 11 12 13 */
for (y = 0; y < 224; y += 14) {
for (j = 0; j < 3; j++) {
h_upscale_bl2_4_5(di, ds, si, ss, 256, f_pal);
for (y = 0; y < height; y += 16) {
for (j = 0; j < 4; j++) {
h_upscale_bln_4_5(di, ds, si, ss, width, f_pal);
}
di += ds;
for (j = 0; j < 11; j++) {
h_upscale_bl2_4_5(di, ds, si, ss, 256, f_pal);
}
/* mix lines 3-10 */
di -= 12*ds;
v_mix(&di[0], &di[-ds], &di[ds], 320, p_05, f_nop);
for (j = 0; j < 7; j++) {
di += ds;
v_mix(&di[0], &di[0], &di[ds], 320, p_05, f_nop);
}
di += 5*ds;
}
}
void upscale_rgb_bl4_256_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
{
int y, j;
/* 14:15, 0 0+1 1+2 2+3 3+4 4+5 5+6 6+7 7+8 8+9 9+10 10+11 11+12 12 13 */
for (y = 0; y < 224; y += 14) {
/* line 0 */
h_upscale_bl4_4_5(di, ds, si, ss, 256, f_pal);
/* lines 1-14 mixing prep */
di += ds;
for (j = 0; j < 13; j++) {
h_upscale_bl4_4_5(di, ds, si, ss, 256, f_pal);
for (j = 0; j < 12; j++) {
h_upscale_bln_4_5(di, ds, si, ss, width, f_pal);
}
di -= 14*ds;
/* mixing line 1: line 0 = -ds, line 1 = +ds */
v_mix(&di[0], &di[-ds], &di[ds], 320, p_025, f_nop);
di -= 13*ds;
/* mixing line 4: line 3 = -ds, line 4 = +ds */
v_mix(&di[0], &di[-ds], &di[ds], swidth, p_025, f_nop);
di += ds;
/* mixing lines 2-4: line n-1 = 0, line n = +ds */
for (j = 0; j < 3; j++) {
v_mix(&di[0], &di[0], &di[ds], 320, p_025, f_nop);
/* mixing lines 5-6: line n-1 = 0, line n = +ds */
for (j = 0; j < 2; j++) {
v_mix(&di[0], &di[0], &di[ds], swidth, p_025, f_nop);
di += ds;
}
/* mixing lines 5-8 */
for (j = 0; j < 4; j++) {
v_mix(&di[0], &di[0], &di[ds], 320, p_05, f_nop);
/* mixing line 7-9 */
for (j = 0; j < 3; j++) {
v_mix(&di[0], &di[0], &di[ds], swidth, p_05, f_nop);
di += ds;
}
/* mixing lines 9-12 */
for (j = 0; j < 4; j++) {
v_mix(&di[0], &di[0], &di[ds], 320, p_075, f_nop);
/* mixing lines 10-12 */
for (j = 0; j < 3; j++) {
v_mix(&di[0], &di[0], &di[ds], swidth, p_075, f_nop);
di += ds;
}
/* lines 13-14, already in place */
di += 2*ds;
/* lines 13-16, already in place */
di += 4*ds;
}
}
void upscale_rgb_bl8_256_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
/* experimental 8 level bilinear for quality assessment */
void upscale_rgb_bl8_x_4_5_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y, j, d;
int swidth = width * 5/4;
int y, j;
/* 14:15, -1+0 0+1 1+2 2+3 3+4 4+5 5+6 6+7 7+8 8+9 9+10 10+11 11+12 12+13 13 */
for (y = 0, d = ds; y < 224; y += 14, d = -ds) {
/* lines 0-14 mixing prep */
for (y = 0; y < 224; y += 16) {
for (j = 0; j < 2; j++) {
h_upscale_bl8_4_5(di, ds, si, ss, width, f_pal);
}
di += ds;
for (j = 0; j < 14; j++) {
h_upscale_bl8_4_5(di, ds, si, ss, 256, f_pal);
h_upscale_bl8_4_5(di, ds, si, ss, width, f_pal);
}
di -= 15*ds;
/* mixing line 0: line 0 = -ds, line 1 = +ds */
v_mix(&di[0], &di[d], &di[ds], 320, p_0125, f_nop);
/* mixing line 2: line 2 = -ds, line 3 = +ds */
v_mix(&di[0], &di[-ds], &di[ds], swidth, p_0125, f_nop);
di += ds;
/* mixing line 1: line 1 = 0, line 2 = +ds */
v_mix(&di[0], &di[0], &di[ds], 320, p_0125, f_nop);
/* mixing line 3: line 3 = 0, line 4 = +ds */
v_mix(&di[0], &di[0], &di[ds], swidth, p_0125, f_nop);
di += ds;
/* mixing lines 2-3: line n-1 = 0, line n = +ds */
/* mixing lines 4-5: line n-1 = 0, line n = +ds */
for (j = 0; j < 2; j++) {
v_mix(&di[0], &di[0], &di[ds], 320, p_025, f_nop);
v_mix(&di[0], &di[0], &di[ds], swidth, p_025, f_nop);
di += ds;
}
/* mixing lines 4-5 */
/* mixing lines 6-7 */
for (j = 0; j < 2; j++) {
v_mix(&di[0], &di[0], &di[ds], 320, p_0375, f_nop);
di += ds;
}
/* mixing lines 6-7 */
/* mixing lines 8-9 */
for (j = 0; j < 2; j++) {
v_mix(&di[0], &di[0], &di[ds], 320, p_05, f_nop);
di += ds;
}
/* mixing lines 8-9 */
/* mixing lines 10-11 */
for (j = 0; j < 2; j++) {
v_mix(&di[0], &di[0], &di[ds], 320, p_0625, f_nop);
di += ds;
}
/* mixing lines 10-11 */
/* mixing lines 12-13 */
for (j = 0; j < 2; j++) {
v_mix(&di[0], &di[0], &di[ds], 320, p_075, f_nop);
di += ds;
}
/* mixing lines 12-13 */
/* mixing lines 14-15 */
for (j = 0; j < 2; j++) {
v_mix(&di[0], &di[0], &di[ds], 320, p_0875, f_nop);
di += ds;
}
/* line 14, already in place */
/* line 16, already in place */
di += ds;
}
}
/* 320x224 -> 320x240, PAR 1:1, for NTSC, DAR 4:3 (wrong for PAL) */
void upscale_clut_nn_320x224_240(u8 *__restrict di, int ds, u8 *__restrict si, int ss)
/* X x Y -> X x Y*17/16 */
void upscale_clut_nn_y_16_17(u8 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height)
{
int y, j;
for (y = 0; y < 224; y += 14) {
for (j = 0; j < 7; j++) {
h_copy(di, ds, si, ss, 320, f_nop);
}
di += ds;
for (j = 0; j < 7; j++) {
h_copy(di, ds, si, ss, 320, f_nop);
}
di -= 8*ds;
v_copy(&di[0], &di[-ds], 320, f_nop);
di += 8*ds;
}
}
void upscale_rgb_nn_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
{
int y, j;
for (y = 0; y < 224; y += 14) {
for (j = 0; j < 7; j++) {
h_copy(di, ds, si, ss, 320, f_pal);
}
di += ds;
for (j = 0; j < 7; j++) {
h_copy(di, ds, si, ss, 320, f_pal);
}
di -= 8*ds;
v_copy(&di[0], &di[-ds], 320, f_nop);
di += 8*ds;
}
}
void upscale_rgb_snn_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
{
int y, j;
for (y = 0; y < 224; y += 14) {
for (j = 0; j < 7; j++) {
h_copy(di, ds, si, ss, 320, f_pal);
}
di += ds;
for (j = 0; j < 7; j++) {
h_copy(di, ds, si, ss, 320, f_pal);
}
di -= 8*ds;
v_mix(&di[ 0], &di[-ds], &di[ds], 320, p_05, f_nop);
v_mix(&di[-ds], &di[-2*ds], &di[-ds], 320, p_05, f_nop);
v_mix(&di[ ds], &di[ ds], &di[ 2*ds], 320, p_05, f_nop);
di += 8*ds;
}
}
void upscale_rgb_bl2_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
{
int y, j;
for (y = 0; y < 224; y += 14) {
for (j = 0; j < 3; j++) {
h_copy(di, ds, si, ss, 320, f_pal);
}
for (y = 0; y < height; y += 16) {
for (j = 0; j < 8; j++) {
v_mix(&di[0], &si[-ss], &si[0], 320, p_05, f_pal);
di += ds;
si += ss;
h_copy(di, ds, si, ss, width, f_nop);
}
si -= ss;
for (j = 0; j < 4; j++) {
h_copy(di, ds, si, ss, 320, f_pal);
di += ds;
for (j = 0; j < 8; j++) {
h_copy(di, ds, si, ss, width, f_nop);
}
di -= 9*ds;
v_copy(&di[0], &di[-ds], width, f_nop);
di += 9*ds;
}
}
void upscale_rgb_bl4_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
void upscale_rgb_nn_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y, j;
for (y = 0; y < 224; y += 14) {
h_copy(di, ds, si, ss, 320, f_pal);
for (j = 0; j < 4; j++) {
v_mix(&di[0], &si[-ss], &si[0], 320, p_025, f_pal);
di += ds;
si += ss;
for (y = 0; y < height; y += 16) {
for (j = 0; j < 8; j++) {
h_copy(di, ds, si, ss, width, f_pal);
}
for (j = 0; j < 4; j++) {
v_mix(&di[0], &si[-ss], &si[0], 320, p_05, f_pal);
di += ds;
si += ss;
di += ds;
for (j = 0; j < 8; j++) {
h_copy(di, ds, si, ss, width, f_pal);
}
for (j = 0; j < 4; j++) {
v_mix(&di[0], &si[-ss], &si[0], 320, p_075, f_pal);
di += ds;
si += ss;
di -= 9*ds;
v_copy(&di[0], &di[-ds], width, f_nop);
di += 9*ds;
}
}
void upscale_rgb_snn_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y, j;
for (y = 0; y < height; y += 16) {
for (j = 0; j < 8; j++) {
h_copy(di, ds, si, ss, width, f_pal);
}
si -= ss;
di += ds;
for (j = 0; j < 8; j++) {
h_copy(di, ds, si, ss, width, f_pal);
}
/* mix lines 6-8 */
di -= 9*ds;
v_mix(&di[0], &di[-ds], &di[ds], width, p_05, f_nop);
v_mix(&di[-ds], &di[-2*ds], &di[-ds], width, p_05, f_nop);
v_mix(&di[ ds], &di[ ds], &di[ 2*ds], width, p_05, f_nop);
di += 9*ds;
}
}
void upscale_rgb_bl2_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y, j;
for (y = 0; y < height; y += 16) {
for (j = 0; j < 4; j++) {
h_copy(di, ds, si, ss, width, f_pal);
}
di += ds;
for (j = 0; j < 12; j++) {
h_copy(di, ds, si, ss, width, f_pal);
}
/* mix lines 3-10 */
di -= 13*ds;
v_mix(&di[0], &di[-ds], &di[ds], width, p_05, f_nop);
for (j = 0; j < 7; j++) {
di += ds;
v_mix(&di[0], &di[0], &di[ds], width, p_05, f_nop);
}
di += 6*ds;
}
}
void upscale_rgb_bl4_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y, j;
for (y = 0; y < height; y += 16) {
for (j = 0; j < 2; j++) {
h_copy(di, ds, si, ss, 320, f_pal);
h_copy(di, ds, si, ss, width, f_pal);
}
di += ds;
for (j = 0; j < 14; j++) {
h_copy(di, ds, si, ss, width, f_pal);
}
di -= 15*ds;
/* mixing line 2: line 1 = -ds, line 2 = +ds */
v_mix(&di[0], &di[-ds], &di[ds], width, p_025, f_nop);
di += ds;
/* mixing lines 3-5: line n-1 = 0, line n = +ds */
for (j = 0; j < 3; j++) {
v_mix(&di[0], &di[0], &di[ds], width, p_025, f_nop);
di += ds;
}
/* mixing lines 6-9 */
for (j = 0; j < 4; j++) {
v_mix(&di[0], &di[0], &di[ds], width, p_05, f_nop);
di += ds;
}
/* mixing lines 10-13 */
for (j = 0; j < 4; j++) {
v_mix(&di[0], &di[0], &di[ds], width, p_075, f_nop);
di += ds;
}
/* lines 14-16, already in place */
di += 3*ds;
}
}
/* 160x144 -> 320x240: GG, PAR 6:5, scaling to 320x240 for DAR 4:3 */
/* NB for smoother image could scale to 288x216, x*9/5, y*3/2 ?
* h: 11111 11112 22222 22233 33333 33444 44444 45555 55555
* 1 1 2 2+3 3 3+4 4 5 5
* v: 11 12 22
* 1 1+2 2
*/
void upscale_clut_nn_160_320x144_240(u8 *__restrict di, int ds, u8 *__restrict si, int ss)
/* X x Y -> X*2/1 x Y, e.g. for X 160->320 (GG) */
void upscale_clut_nn_x_1_2(u8 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height)
{
int y;
for (y = 0; y < height; y++) {
h_upscale_nn_1_2(di, ds, si, ss, width, f_nop);
}
}
void upscale_rgb_nn_x_1_2(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y;
for (y = 0; y < height; y++) {
h_upscale_nn_1_2(di, ds, si, ss, width, f_pal);
}
}
void upscale_rgb_bl2_x_1_2(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y;
for (y = 0; y < height; y++) {
h_upscale_bl2_1_2(di, ds, si, ss, width, f_pal);
}
}
/* X x Y -> X*2/1 x Y*5/3 (GG) */
void upscale_clut_nn_x_1_2_y_3_5(u8 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height)
{
int swidth = width * 2;
int y, j;
/* 3:5, 0 0 1 1 2 */
for (y = 0; y < 144; y += 3) {
for (y = 0; y < height; y += 3) {
/* lines 0,2,4 */
for (j = 0; j < 3; j++) {
h_upscale_nn_1_2(di, ds, si, ss, 160, f_nop);
h_upscale_nn_1_2(di, ds, si, ss, width, f_nop);
di += ds;
}
/* lines 1,3 */
di -= 5*ds;
for (j = 0; j < 2; j++) {
v_copy(&di[0], &di[-ds], 320, f_nop);
v_copy(&di[0], &di[-ds], swidth, f_nop);
di += 2*ds;
}
}
}
void upscale_rgb_nn_160_320x144_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
void upscale_rgb_nn_x_1_2_y_3_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int swidth = width * 2;
int y, j;
for (y = 0; y < 144; y += 3) {
for (y = 0; y < height; y += 3) {
for (j = 0; j < 3; j++) {
h_upscale_nn_1_2(di, ds, si, ss, 160, f_pal);
h_upscale_nn_1_2(di, ds, si, ss, width, f_pal);
di += ds;
}
di -= 5*ds;
for (j = 0; j < 2; j++) {
v_copy(&di[0], &di[-ds], 320, f_nop);
v_copy(&di[0], &di[-ds], swidth, f_nop);
di += 2*ds;
}
}
}
void upscale_rgb_snn_160_320x144_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
void upscale_rgb_bl2_x_1_2_y_3_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int swidth = width * 2;
int y, j;
/* 3:5, 0 0+1 1 1+2 2 */
for (y = 0; y < 144; y += 3) {
for (y = 0; y < height; y += 3) {
for (j = 0; j < 3; j++) {
h_upscale_nn_1_2(di, ds, si, ss, 160, f_pal);
h_upscale_bl2_1_2(di, ds, si, ss, width, f_pal);
di += ds;
}
di -= 5*ds;
for (j = 0; j < 2; j++) {
v_mix(&di[0], &di[-ds], &di[ds], 320, p_05, f_nop);
v_mix(&di[0], &di[-ds], &di[ds], swidth, p_05, f_nop);
di += 2*ds;
}
}
}
void upscale_rgb_bl2_160_320x144_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
void upscale_rgb_bl4_x_1_2_y_3_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int swidth = width * 2;
int y, j, d;
/* for 1st block backwards reference virtually duplicate source line 0 */
for (y = 0, d = 2*ds; y < height; y += 3, d = -ds) {
di += 2*ds;
for (j = 0; j < 3; j++) {
h_upscale_bl2_1_2(di, ds, si, ss, width, f_pal);
}
di -= 5*ds;
v_mix(&di[0], &di[d ], &di[2*ds], swidth, p_05, f_nop); /*-1+0 */
di += ds;
v_mix(&di[0], &di[ds], &di[2*ds], swidth, p_075, f_nop);/* 0+1 */
di += ds;
v_mix(&di[0], &di[ 0], &di[ ds], swidth, p_025, f_nop);/* 0+1 */
di += ds;
v_mix(&di[0], &di[ 0], &di[ ds], swidth, p_05, f_nop); /* 1+2 */
di += 2*ds;
}
}
/* X x Y -> X x Y*5/3, e.g. for Y 144->240 (GG) */
void upscale_clut_nn_y_3_5(u8 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height)
{
int y, j;
/* 3:5, 0 0+1 1 1+2 2 */
for (y = 0; y < 144; y += 3) {
for (y = 0; y < height; y += 3) {
/* lines 0,2,4 */
for (j = 0; j < 3; j++) {
h_upscale_bl2_1_2(di, ds, si, ss, 160, f_pal);
h_copy(di, ds, si, ss, width, f_nop);
di += ds;
}
/* lines 1,3 */
di -= 5*ds;
for (j = 0; j < 2; j++) {
v_copy(&di[0], &di[-ds], width, f_nop);
di += 2*ds;
}
}
}
void upscale_rgb_nn_y_3_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y, j;
for (y = 0; y < height; y += 3) {
for (j = 0; j < 3; j++) {
h_copy(di, ds, si, ss, width, f_pal);
di += ds;
}
di -= 5*ds;
for (j = 0; j < 2; j++) {
v_mix(&di[0], &di[-ds], &di[ds], 320, p_05, f_nop);
v_copy(&di[0], &di[-ds], width, f_nop);
di += 2*ds;
}
}
}
void upscale_rgb_bl4_160_320x144_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal)
void upscale_rgb_bl2_y_3_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y, j;
for (y = 0; y < height; y += 3) {
for (j = 0; j < 3; j++) {
h_copy(di, ds, si, ss, width, f_pal);
di += ds;
}
di -= 5*ds;
for (j = 0; j < 2; j++) {
v_mix(&di[0], &di[-ds], &di[ds], width, p_05, f_nop);
di += 2*ds;
}
}
}
void upscale_rgb_bl4_y_3_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal)
{
int y, j, d;
/* 3:5, -1+0, 0+1 0+1 1+2 2
* for 1st block backwards reference virtually duplicate source line 0 */
for (y = 0, d = 2*ds; y < 144; y += 3, d = -ds) {
/* for 1st block backwards reference virtually duplicate source line 0 */
for (y = 0, d = 2*ds; y < height; y += 3, d = -ds) {
di += 2*ds;
for (j = 0; j < 3; j++) {
h_upscale_bl2_1_2(di, ds, si, ss, 160, f_pal);
h_copy(di, ds, si, ss, width, f_pal);
}
di -= 5*ds;
v_mix(&di[0], &di[d ], &di[2*ds], 320, p_05, f_nop); /*-1+0 */
v_mix(&di[0], &di[d ], &di[2*ds], width, p_05, f_nop); /*-1+0 */
di += ds;
v_mix(&di[0], &di[ds], &di[2*ds], 320, p_075, f_nop); /* 0+1 */
v_mix(&di[0], &di[ds], &di[2*ds], width, p_075, f_nop);/* 0+1 */
di += ds;
v_mix(&di[0], &di[ 0], &di[ ds], 320, p_025, f_nop); /* 0+1 */
v_mix(&di[0], &di[ 0], &di[ ds], width, p_025, f_nop);/* 0+1 */
di += ds;
v_mix(&di[0], &di[ 0], &di[ ds], 320, p_05, f_nop); /* 1+2 */
v_mix(&di[0], &di[ 0], &di[ ds], width, p_05, f_nop); /* 1+2 */
di += 2*ds;
}
}

375
platform/common/upscale.h
View file

@ -7,7 +7,7 @@
* nn: nearest neighbour
* snn: "smoothed" nearest neighbour (see below)
* bln: n-level-bilinear with n quantized weights
* quantization: 0: a<1/2*n, 1/n: 1/2*n<=a<3/2*n, etc
* quantization: 0: a<1/(2*n), 1/n: 1/(2*n)<=a<3/(2*n), etc
* currently n=2, n=4 are implemented (there's n=8 mixing, but no filters)
* [NB this has been brought to my attn, which is probably the same as bl2:
* https://www.drdobbs.com/image-scaling-with-bresenham/184405045?pgno=1]
@ -18,34 +18,37 @@
* a sharper look than a bilinear filter, at the price of some visible jags
* on diagonal edges.
*
* scaling modes:
* 256x___ -> 320x___ only horizontal scaling. Produces an aspect error of
* ~7% for NTSC 224 line modes, but is correct for PAL
* 256/320x224/240
* -> 320x240 always produces 320x240 at DAR 4:3
* 160x144 -> 320x240 game gear (currently unused)
*
* example scaling modes:
* 256x_Y_ -> 320x_Y_, H32/mode 4, PAR 5:4, for PAL DAR 4:3 (NTSC 7% aspect err)
* 256x224 -> 320x240, H32/mode 4, PAR 5:4, for NTSC DAR 4:3 (PAL 7% aspect err)
* 320x224 -> 320x240, PAR 1:1, for NTSC, DAR 4:3 (PAL 7% etc etc...)
* 160x144 -> 320x240: GG, PAR 6:5, scaling to 320x240 for DAR 4:3
*
*
* (C) 2021 kub <derkub@gmail.com>
*
* This work is licensed under the terms of MAME license.
* See COPYING file in the top-level directory.
*/
#include <pico/pico_types.h>
/* RGB565 pixel mixing, see https://www.compuphase.com/graphic/scale3.htm and
http://blargg.8bitalley.com/info/rgb_mixing.html */
/* 2-level mixing */
//#define p_05(p1,p2) (((p1)+(p2) + ( ((p1)^(p2))&0x0821))>>1) // round up
//#define p_05(p1,p2) (((p1)+(p2) - ( ((p1)^(p2))&0x0821))>>1) // round down
#define p_05(p1,p2) (((p1)&(p2)) + ((((p1)^(p2))&~0x0821)>>1))
//#define p_05(d,p1,p2) d=(((p1)+(p2) + ( ((p1)^(p2))&0x0821))>>1) // round up
//#define p_05(d,p1,p2) d=(((p1)+(p2) - ( ((p1)^(p2))&0x0821))>>1) // round down
#define p_05(d,p1,p2) d=(((p1)&(p2)) + ((((p1)^(p2))&~0x0821)>>1))
/* 4-level mixing, 2 times slower */
// 1/4*p1 + 3/4*p2 = 1/2*(1/2*(p1+p2) + p2)
#define p_025(p1,p2) (t=p_05(p1, p2), p_05( t, p2))
#define p_075(p1,p2) p_025(p2,p1)
#define p_025(d,p1,p2) p_05(t, p1, p2); p_05( d, t, p2)
#define p_075(d,p1,p2) p_025(d,p2,p1)
/* 8-level mixing, 3 times slower */
// 1/8*p1 + 7/8*p2 = 1/2*(1/2*(1/2*(p1+p2) + p2) + p2)
#define p_0125(p1,p2) (t=p_05(p1, p2), u=p_05( t, p2), p_05( u, p2))
#define p_0125(d,p1,p2) p_05(t, p1, p2); p_05( u, t, p2); p_05( d, u, p2)
// 3/8*p1 + 5/8*p2 = 1/2*(1/2*(1/2*(p1+p2) + p2) + 1/2*(p1+p2))
#define p_0375(p1,p2) (t=p_05(p1, p2), u=p_05( t, p2), p_05( u, t))
#define p_0625(p1,p2) p_0375(p2,p1)
#define p_0875(p1,p2) p_0125(p2,p1)
#define p_0375(d,p1,p2) p_05(t, p1, p2); p_05( u, t, p2); p_05( d, u, t)
#define p_0625(d,p1,p2) p_0375(d,p2,p1)
#define p_0875(d,p1,p2) p_0125(d,p2,p1)
/* pixel transforms */
#define f_pal(v) pal[v] // convert CLUT index -> RGB565
@ -55,7 +58,7 @@
/*
scalers h:
256->320: - (4:5) (256x224/240 -> 320x224/240)
256->299: - (6:7) (256x224 -> 299x224, DAR 4:3, 10.5 px border )
256->299: - (6:7) (256x224 -> 299x224, alt?)
160->320: - (1:2) 2x (160x144 -> 320x240, GG)
160->288: - (5:9) (160x144 -> 288x216, GG alt?)
*/
@ -95,7 +98,7 @@ scalers h:
for (i = w/4; i > 0; i--, si += 4, di += 5) { \
di[0] = f(si[0]); \
di[1] = f(si[1]); \
di[2] = p_05(f(si[1]),f(si[2])); \
p_05(di[2], f(si[1]),f(si[2])); \
di[3] = f(si[2]); \
di[4] = f(si[3]); \
} \
@ -104,12 +107,12 @@ scalers h:
} while (0)
#define h_upscale_bln_4_5(di,ds,si,ss,w,f) do { \
int i, t; \
int i; u16 t; \
for (i = w/4; i > 0; i--, si += 4, di += 5) { \
di[0] = f(si[0]); \
di[1] = p_025(f(si[0]),f(si[1])); \
di[2] = p_05 (f(si[1]),f(si[2])); \
di[3] = p_075(f(si[2]),f(si[3])); \
p_025(di[1], f(si[0]),f(si[1])); \
p_05 (di[2], f(si[1]),f(si[2])); \
p_075(di[3], f(si[2]),f(si[3])); \
di[4] = f(si[3]); \
} \
di += ds - w/4*5; \
@ -120,8 +123,8 @@ scalers h:
int i; \
for (i = w/4; i > 0; i--, si += 4, di += 5) { \
di[0] = f(si[0]); \
di[1] = p_05(f(si[0]),f(si[1])); \
di[2] = p_05(f(si[1]),f(si[2])); \
p_05(di[1], f(si[0]),f(si[1])); \
p_05(di[2], f(si[1]),f(si[2])); \
di[3] = f(si[2]); \
di[4] = f(si[3]); \
} \
@ -130,12 +133,12 @@ scalers h:
} while (0)
#define h_upscale_bl4_4_5(di,ds,si,ss,w,f) do { \
int i, t; uint p = f(si[0]); \
int i; u16 t, p = f(si[0]); \
for (i = w/4; i > 0; i--, si += 4, di += 5) { \
di[0] = p_025(p, f(si[0])); \
di[1] = p_05 (f(si[0]),f(si[1])); \
di[2] = p_05 (f(si[1]),f(si[2])); \
di[3] = p_075(f(si[2]),f(si[3])); \
p_025(di[0], p, f(si[0])); \
p_05 (di[1], f(si[0]),f(si[1])); \
p_05 (di[2], f(si[1]),f(si[2])); \
p_075(di[3], f(si[2]),f(si[3])); \
di[4] = p = f(si[3]); \
} \
di += ds - w/4*5; \
@ -143,12 +146,12 @@ scalers h:
} while (0)
#define h_upscale_bl8_4_5(di,ds,si,ss,w,f) do { \
int i, t, u; uint p = f(si[0]); \
int i; u16 t, u, p = f(si[0]); \
for (i = w/4; i > 0; i--, si += 4, di += 5) { \
di[0] = p_025(p, f(si[0])); \
di[1] = p_0375(f(si[0]),f(si[1])); \
di[2] = p_0625(f(si[1]),f(si[2])); \
di[3] = p_075(f(si[2]),f(si[3])); \
p_025 (di[0], p, f(si[0])); \
p_0375(di[1], f(si[0]),f(si[1])); \
p_0625(di[2], f(si[1]),f(si[2])); \
p_075 (di[3], f(si[2]),f(si[3])); \
di[4] = p = f(si[3]); \
} \
di += ds - w/4*5; \
@ -195,7 +198,7 @@ scalers h:
di[0] = f(si[0]); \
di[1] = f(si[1]); \
di[2] = f(si[2]); \
di[3] = p_05(f(si[2]),f(si[3])); \
p_05(di[3], f(si[2]),f(si[3])); \
di[4] = f(si[3]); \
di[5] = f(si[4]); \
di[6] = f(si[5]); \
@ -208,10 +211,10 @@ scalers h:
int i; \
for (i = w/6; i > 0; i--, si += 6, di += 7) { \
di[0] = f(si[0]); \
di[1] = p_05(f(si[0]),f(si[1])); \
di[2] = p_05(f(si[1]),f(si[2])); \
di[3] = p_05(f(si[2]),f(si[3])); \
di[4] = p_05(f(si[3]),f(si[4])); \
p_05(di[1], f(si[0]),f(si[1])); \
p_05(di[2], f(si[1]),f(si[2])); \
p_05(di[3], f(si[2]),f(si[3])); \
p_05(di[4], f(si[3]),f(si[4])); \
di[5] = f(si[4]); \
di[6] = f(si[5]); \
} \
@ -220,14 +223,14 @@ scalers h:
} while (0)
#define h_upscale_bl4_6_7(di,ds,si,ss,w,f) do { \
int i, t; uint p = f(si[0]); \
int i; u16 t p = f(si[0]); \
for (i = w/6; i > 0; i--, si += 6, di += 7) { \
di[0] = p_025(p,f(si[0])); \
di[1] = p_025(f(si[0]),f(si[1])); \
di[2] = p_05 (f(si[1]),f(si[2])); \
di[3] = p_05 (f(si[2]),f(si[3])); \
di[4] = p_075(f(si[3]),f(si[4])); \
di[5] = p_075(f(si[4]),f(si[5])); \
p_025(di[0], p, f(si[0])); \
p_025(di[1], f(si[0]),f(si[1])); \
p_05 (di[2], f(si[1]),f(si[2])); \
p_05 (di[3], f(si[2]),f(si[3])); \
p_075(di[4], f(si[3]),f(si[4])); \
p_075(di[5], f(si[4]),f(si[5])); \
di[6] = p = f(si[5]); \
} \
di += ds - w/6*7; \
@ -258,9 +261,9 @@ scalers h:
di[0] = f(si[0]); \
di[1] = f(si[0]); \
di[2] = f(si[1]); \
di[3] = p_05(f(si[1]),f(si[2])); \
p_05(di[3], f(si[1]),f(si[2])); \
di[4] = f(si[2]); \
di[5] = p_05(f(si[2]),f(si[3])); \
p_05(di[5], f(si[2]),f(si[3])); \
di[6] = f(si[3]); \
di[7] = f(si[4]); \
di[8] = f(si[4]); \
@ -273,13 +276,13 @@ scalers h:
int i; \
for (i = w/5; i > 0; i--, si += 5, di += 9) { \
di[0] = f(si[0]); \
di[1] = p_05(f(si[0]),f(si[1])); \
p_05(di[1], f(si[0]),f(si[1])); \
di[2] = f(si[1]); \
di[3] = p_05(f(si[1]),f(si[2])); \
p_05(di[3], f(si[1]),f(si[2])); \
di[4] = f(si[2]); \
di[5] = p_05(f(si[2]),f(si[3])); \
p_05(di[5], f(si[2]),f(si[3])); \
di[6] = f(si[3]); \
di[7] = p_05(f(si[3]),f(si[4])); \
p_05(di[7], f(si[3]),f(si[4])); \
di[8] = f(si[4]); \
} \
di += ds - w/5*9; \
@ -287,16 +290,16 @@ scalers h:
} while (0)
#define h_upscale_bl4_5_9(di,ds,si,ss,w,f) do { \
int i, t; uint p = f(si[0]); \
int i; u16 t, p = f(si[0]); \
for (i = w/5; i > 0; i--, si += 5, di += 9) { \
di[0] = p_05 (p,f(si[0])); \
p_05 (di[0], p, f(si[0])); \
di[1] = f(si[0]); \
di[2] = p_025(f(si[0]),f(si[1])); \
di[3] = p_075(f(si[1]),f(si[2])); \
di[4] = p_025(f(si[1]),f(si[2])); \
di[5] = p_075(f(si[2]),f(si[3])); \
p_025(di[2], f(si[0]),f(si[1])); \
p_075(di[3], f(si[1]),f(si[2])); \
p_025(di[4], f(si[1]),f(si[2])); \
p_075(di[5], f(si[2]),f(si[3])); \
di[6] = f(si[3]); \
di[7] = p_05 (f(si[3]),f(si[4])); \
p_05 (di[7], f(si[3]),f(si[4])); \
di[8] = p = f(si[4]); \
} \
di += ds - w/5*9; \
@ -319,9 +322,9 @@ scalers h:
#define h_upscale_bl2_1_2(di,ds,si,ss,w,f) do { \
int i; uint p = f(si[0]); \
for (i = w/2; i > 0; i--, si += 2, di += 4) { \
di[0] = p_05 (p, f(si[0])); \
p_05 (di[0], p, f(si[0])); \
di[1] = f(si[0]); \
di[2] = p_05 (f(si[0]), f(si[1])); \
p_05 (di[2], f(si[0]),f(si[1])); \
di[3] = p = f(si[1]); \
} \
di += ds - w*2; \
@ -350,12 +353,12 @@ scalers v:
*/
#define v_mix(di,li,ri,w,p_mix,f) do { \
int i, t, u; (void)t, (void)u; \
u16 i, t, u; (void)t, (void)u; \
for (i = 0; i < w; i += 4) { \
(di)[i ] = p_mix(f((li)[i ]), f((ri)[i ])); \
(di)[i+1] = p_mix(f((li)[i+1]), f((ri)[i+1])); \
(di)[i+2] = p_mix(f((li)[i+2]), f((ri)[i+2])); \
(di)[i+3] = p_mix(f((li)[i+3]), f((ri)[i+3])); \
p_mix((di)[i ], f((li)[i ]),f((ri)[i ])); \
p_mix((di)[i+1], f((li)[i+1]),f((ri)[i+1])); \
p_mix((di)[i+2], f((li)[i+2]),f((ri)[i+2])); \
p_mix((di)[i+3], f((li)[i+3]),f((ri)[i+3])); \
} \
} while (0)
@ -369,32 +372,222 @@ scalers v:
} \
} while (0)
/* scale 14:15 */
#define v_upscale_nn_14_15(di,ds,w,l) do { \
if (++l == 7) { \
di += ds; \
} else if (l >= 14) { \
l = 0; \
di -= 7*ds; \
v_copy(&di[0], &di[-ds], w, f_nop); \
di += 7*ds; \
} \
} while (0)
#define v_upscale_snn_14_15(di,ds,w,l) do { \
if (++l == 7) { \
di += ds; \
} else if (l >= 14) { \
l = 0; \
di -= 7*ds; \
v_mix(&di[0], &di[-ds], &di[ds], w, p_05, f_nop); \
v_mix(&di[-ds], &di[-2*ds], &di[-ds], w, p_05, f_nop); \
v_mix(&di[ ds], &di[ ds], &di[ 2*ds], w, p_05, f_nop); \
di += 7*ds; \
} \
} while (0)
#define v_upscale_bl2_14_15(di,ds,w,l) do { \
if (++l == 3) { \
di += ds; \
} else if (l >= 14) { \
int j; \
l = 0; \
di -= 11*ds; \
v_mix(&di[0], &di[-ds], &di[ds], w, p_05, f_nop); \
for (j = 0; j < 7; j++) { \
di += ds; \
v_mix(&di[0], &di[0], &di[ds], w, p_05, f_nop); \
} \
di += 4*ds; \
} \
} while (0)
#define v_upscale_bl4_14_15(di,ds,w,l) do { \
if (++l == 1) { \
di += ds; \
} else if (l >= 14) { \
int j; \
l = 0; \
di -= 13*ds; \
v_mix(&di[0], &di[-ds], &di[ds], w, p_025, f_nop); \
di += ds; \
for (j = 0; j < 3; j++) { \
v_mix(&di[0], &di[0], &di[ds], w, p_025, f_nop); \
di += ds; \
} \
for (j = 0; j < 4; j++) { \
v_mix(&di[0], &di[0], &di[ds], w, p_05, f_nop); \
di += ds; \
} \
for (j = 0; j < 4; j++) { \
v_mix(&di[0], &di[0], &di[ds], w, p_075, f_nop); \
di += ds; \
} \
di += 1*ds; \
} \
} while (0)
/* scale 16:17 */
#define v_upscale_nn_16_17(di,ds,w,l) do { \
if (++l == 8) { \
di += ds; \
} else if (l >= 16) { \
l = 0; \
di -= 8*ds; \
v_copy(&di[0], &di[-ds], w, f_nop); \
di += 8*ds; \
} \
} while (0)
#define v_upscale_snn_16_17(di,ds,w,l) do { \
if (++l == 8) { \
di += ds; \
} else if (l >= 16) { \
l = 0; \
di -= 8*ds; \
v_mix(&di[0], &di[-ds], &di[ds], w, p_05, f_nop); \
v_mix(&di[-ds], &di[-2*ds], &di[-ds], w, p_05, f_nop); \
v_mix(&di[ ds], &di[ ds], &di[ 2*ds], w, p_05, f_nop); \
di += 8*ds; \
} \
} while (0)
#define v_upscale_bl2_16_17(di,ds,w,l) do { \
if (++l == 4) { \
di += ds; \
} else if (l >= 16) { \
int j; \
l = 0; \
di -= 12*ds; \
v_mix(&di[0], &di[-ds], &di[ds], w, p_05, f_nop); \
for (j = 0; j < 7; j++) { \
di += ds; \
v_mix(&di[0], &di[0], &di[ds], w, p_05, f_nop); \
} \
di += 5*ds; \
} \
} while (0)
#define v_upscale_bl4_16_17(di,ds,w,l) do { \
if (++l == 2) { \
di += ds; \
} else if (l >= 16) { \
int j; \
l = 0; \
di -= 14*ds; \
v_mix(&di[0], &di[-ds], &di[ds], w, p_025, f_nop); \
di += ds; \
for (j = 0; j < 3; j++) { \
v_mix(&di[0], &di[0], &di[ds], w, p_025, f_nop); \
di += ds; \
} \
for (j = 0; j < 4; j++) { \
v_mix(&di[0], &di[0], &di[ds], w, p_05, f_nop); \
di += ds; \
} \
for (j = 0; j < 4; j++) { \
v_mix(&di[0], &di[0], &di[ds], w, p_075, f_nop); \
di += ds; \
} \
di += 2*ds; \
} \
} while (0)
/* scale 3:5 */
#define v_upscale_nn_3_5(di,ds,w,l) do { \
if (++l < 3) { \
di += ds; \
} else { \
int j; \
l = 0; \
di -= 4*ds; \
for (j = 0; j < 2; j++) { \
v_copy(&di[0], &di[-ds], w, f_nop); \
di += 2*ds; \
} \
} \
} while (0)
#define v_upscale_snn_3_5(di,ds,w,l) do { \
if (++l < 3) { \
di += ds; \
} else { \
int j; \
l = 0; \
di -= 4*ds; \
for (j = 0; j < 2; j++) { \
v_mix(&di[0], &di[-ds], &di[ds], w, p_05, f_nop); \
di += 2*ds; \
} \
} \
} while (0)
/* scale 2:3 */
#define v_upscale_nn_2_3(di,ds,w,l) do { \
if (++l < 2) { \
di += ds; \
} else { \
int j; \
l = 0; \
di -= 2*ds; \
v_copy(&di[0], &di[-ds], w, f_nop); \
di += 2*ds; \
} \
} while (0)
#define v_upscale_snn_2_3(di,ds,w,l) do { \
if (++l < 2) { \
di += ds; \
} else { \
int j; \
l = 0; \
di -= 2*ds; \
v_mix(&di[0], &di[-ds], &di[ds], w, p_05, f_nop); \
di += 2*ds; \
} \
} while (0)
/* 256x___ -> 320x___, H32/mode 4, PAR 5:4, for PAL DAR 4:3 (wrong for NTSC) */
void upscale_clut_nn_256_320x___(u8 *__restrict di, int ds, u8 *__restrict si, int ss, int height);
void upscale_rgb_nn_256_320x___(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int height, u16 *pal);
void upscale_rgb_snn_256_320x___(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int height, u16 *pal);
void upscale_rgb_bl2_256_320x___(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int height, u16 *pal);
void upscale_rgb_bl4_256_320x___(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int height, u16 *pal);
/* X x Y -> X*5/4 x Y, for X 256->320 */
void upscale_rgb_nn_x_4_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_snn_x_4_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_bl2_x_4_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_bl4_x_4_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
/* 256x224 -> 320x240, H32/mode 4, PAR 5:4, for NTSC DAR 4:3 (wrong for PAL) */
void upscale_clut_nn_256_320x224_240(u8 *__restrict di, int ds, u8 *__restrict si, int ss);
void upscale_rgb_nn_256_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal);
void upscale_rgb_snn_256_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal);
void upscale_rgb_bl2_256_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal);
void upscale_rgb_bl4_256_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal);
/* X x Y -> X x Y*17/16, for Y 224->238 or 192->204 (SMS) */
void upscale_rgb_nn_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_snn_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_bl2_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_bl4_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
/* 320x224 -> 320x240, PAR 1:1, for NTSC, DAR 4:3 (wrong for PAL) */
void upscale_clut_nn_320x224_240(u8 *__restrict di, int ds, u8 *__restrict si, int ss);
void upscale_rgb_nn_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal);
void upscale_rgb_snn_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal);
void upscale_rgb_bl2_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal);
void upscale_rgb_bl4_320x224_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal);
/* X x Y -> X*5/4 x Y*17/16 */
void upscale_rgb_nn_x_4_5_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_snn_x_4_5_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_bl2_x_4_5_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_bl4_x_4_5_y_16_17(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
/* X x Y -> X*2/1 x Y, e.g. for X 160->320 (GG) */
void upscale_rgb_nn_x_1_2(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_bl2_x_1_2(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
/* X x Y -> X x Y*5/3, e.g. for Y 144->240 (GG) */
void upscale_rgb_nn_y_3_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_bl2_y_3_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_bl4_y_3_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
/* X x Y -> X*2/1 x Y*5/3 (GG) */
void upscale_rgb_nn_x_1_2_y_3_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_bl2_x_1_2_y_3_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
void upscale_rgb_bl4_x_1_2_y_3_5(u16 *__restrict di, int ds, u8 *__restrict si, int ss, int width, int height, u16 *pal);
/* 160x144 -> 320x240: GG, PAR 6:5, scaling to 320x240 for DAR 4:3 */
void upscale_clut_nn_160_320x144_240(u8 *__restrict di, int ds, u8 *__restrict si, int ss);
void upscale_rgb_nn_160_320x144_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal);
void upscale_rgb_snn_160_320x144_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal);
void upscale_rgb_bl2_160_320x144_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal);
void upscale_rgb_bl4_160_320x144_240(u16 *__restrict di, int ds, u8 *__restrict si, int ss, u16 *pal);