|
| 1 | +/*[[[magic |
| 2 | +local gcc_opt = options.setdefault("GCC.options", []); |
| 3 | +gcc_opt.removeif(x -> x.startswith("-O")); |
| 4 | +gcc_opt.append("-O3"); // Force _all_ optimizations because stuff in here is performance-critical |
| 5 | +]]]*/ |
| 6 | +/* Copyright (c) 2019-2025 Griefer@Work * |
| 7 | + * * |
| 8 | + * This software is provided 'as-is', without any express or implied * |
| 9 | + * warranty. In no event will the authors be held liable for any damages * |
| 10 | + * arising from the use of this software. * |
| 11 | + * * |
| 12 | + * Permission is granted to anyone to use this software for any purpose, * |
| 13 | + * including commercial applications, and to alter it and redistribute it * |
| 14 | + * freely, subject to the following restrictions: * |
| 15 | + * * |
| 16 | + * 1. The origin of this software must not be misrepresented; you must not * |
| 17 | + * claim that you wrote the original software. If you use this software * |
| 18 | + * in a product, an acknowledgement (see the following) in the product * |
| 19 | + * documentation is required: * |
| 20 | + * Portions Copyright (c) 2019-2025 Griefer@Work * |
| 21 | + * 2. Altered source versions must be plainly marked as such, and must not be * |
| 22 | + * misrepresented as being the original software. * |
| 23 | + * 3. This notice may not be removed or altered from any source distribution. * |
| 24 | + */ |
| 25 | +#ifndef GUARD_LIBVIDEO_GFX_SWGFX_HL_BLIT_C |
| 26 | +#define GUARD_LIBVIDEO_GFX_SWGFX_HL_BLIT_C 1 |
| 27 | +#define _KOS_SOURCE 1 |
| 28 | + |
| 29 | +/************************************************************************/ |
| 30 | +/* HIGH-LEVEL, GENERIC SW-GFX OPERATOR IMPLS */ |
| 31 | +/************************************************************************/ |
| 32 | + |
| 33 | +#include "api.h" |
| 34 | +/**/ |
| 35 | + |
| 36 | +#include <hybrid/compiler.h> |
| 37 | + |
| 38 | +#include <assert.h> |
| 39 | + |
| 40 | +#include <libvideo/gfx/gfx.h> |
| 41 | + |
| 42 | +#include "gfx-empty.h" |
| 43 | +#include "gfx-utils.h" |
| 44 | +#include "swgfx.h" |
| 45 | + |
| 46 | +DECL_BEGIN |
| 47 | + |
| 48 | +static_assert(sizeof(struct blt_swdrv) <= (_VIDEO_BLITTER_N_DRIVER * sizeof(void (*)(void))), |
| 49 | + "sizeof(struct blt_swdrv) too large for '_VIDEO_BLITTER_N_DRIVER'"); |
| 50 | + |
| 51 | +/************************************************************************/ |
| 52 | +/* BLITTER INITIALIZATION */ |
| 53 | +/************************************************************************/ |
| 54 | + |
| 55 | +LOCAL ATTR_PURE WUNUSED ATTR_IN(1) ATTR_IN(2) bool FCC |
| 56 | +noblend_blit_compatible(struct video_codec const *dst, |
| 57 | + struct video_codec const *src) { |
| 58 | + if (dst == src) |
| 59 | + return true; |
| 60 | + |
| 61 | + /* Certain codecs can still be blit onto each other without color conversion. |
| 62 | + * This is possible whenever all color channels present in "dst" are also |
| 63 | + * present at the same positions in "src". |
| 64 | + * |
| 65 | + * This is usually the case when "dst" lacks an alpha-channel and instead |
| 66 | + * features some unused padding in its place. */ |
| 67 | +#ifndef __OPTIMIZE_SIZE__ |
| 68 | + switch (dst->vc_codec) { |
| 69 | + case VIDEO_CODEC_RGBX8888: |
| 70 | + return src->vc_codec == VIDEO_CODEC_RGBA8888; |
| 71 | + case VIDEO_CODEC_XRGB8888: |
| 72 | + return src->vc_codec == VIDEO_CODEC_ARGB8888; |
| 73 | + case VIDEO_CODEC_XBGR8888: |
| 74 | + return src->vc_codec == VIDEO_CODEC_ABGR8888; |
| 75 | + case VIDEO_CODEC_BGRX8888: |
| 76 | + return src->vc_codec == VIDEO_CODEC_BGRA8888; |
| 77 | + case VIDEO_CODEC_RGBX4444: |
| 78 | + return src->vc_codec == VIDEO_CODEC_RGBA4444; |
| 79 | + case VIDEO_CODEC_XRGB4444: |
| 80 | + return src->vc_codec == VIDEO_CODEC_ARGB4444; |
| 81 | + case VIDEO_CODEC_XBGR4444: |
| 82 | + return src->vc_codec == VIDEO_CODEC_ABGR4444; |
| 83 | + case VIDEO_CODEC_BGRX4444: |
| 84 | + return src->vc_codec == VIDEO_CODEC_BGRA4444; |
| 85 | + case VIDEO_CODEC_RGBX5551: |
| 86 | + return src->vc_codec == VIDEO_CODEC_RGBA5551; |
| 87 | + case VIDEO_CODEC_XRGB1555: |
| 88 | + return src->vc_codec == VIDEO_CODEC_ARGB1555; |
| 89 | + case VIDEO_CODEC_XBGR1555: |
| 90 | + return src->vc_codec == VIDEO_CODEC_ABGR1555; |
| 91 | + case VIDEO_CODEC_BGRX5551: |
| 92 | + return src->vc_codec == VIDEO_CODEC_BGRA5551; |
| 93 | + default: |
| 94 | + break; |
| 95 | + } |
| 96 | +#endif /* !__OPTIMIZE_SIZE__ */ |
| 97 | + if ((dst->vc_specs.vcs_flags & (VIDEO_CODEC_FLAG_PAL | VIDEO_CODEC_FLAG_LUM)) == |
| 98 | + (src->vc_specs.vcs_flags & (VIDEO_CODEC_FLAG_PAL | VIDEO_CODEC_FLAG_LUM))) { |
| 99 | + if ((dst->vc_specs.vcs_rmask == 0 || dst->vc_specs.vcs_rmask == src->vc_specs.vcs_rmask) && |
| 100 | + (dst->vc_specs.vcs_gmask == 0 || dst->vc_specs.vcs_gmask == src->vc_specs.vcs_gmask) && |
| 101 | + (dst->vc_specs.vcs_bmask == 0 || dst->vc_specs.vcs_bmask == src->vc_specs.vcs_bmask) && |
| 102 | + (dst->vc_specs.vcs_amask == 0 || dst->vc_specs.vcs_amask == src->vc_specs.vcs_amask)) |
| 103 | + return true; |
| 104 | + } |
| 105 | + return false; |
| 106 | +} |
| 107 | + |
| 108 | +INTERN ATTR_RETNONNULL ATTR_INOUT(1) struct video_blitter *FCC |
| 109 | +libvideo_swgfx_blitfrom(struct video_blitter *__restrict ctx) { |
| 110 | + struct video_gfx const *src_gfx = ctx->vbt_src; |
| 111 | + struct video_gfx const *dst_gfx = ctx->vbt_dst; |
| 112 | + struct video_buffer const *src_buffer = src_gfx->vx_buffer; |
| 113 | + struct video_buffer const *dst_buffer = dst_gfx->vx_buffer; |
| 114 | + struct blt_swdrv *drv = video_swblitter_getdrv(ctx); |
| 115 | + video_swblitter_setops(ctx); |
| 116 | + |
| 117 | + /* Check for special case: source and target buffers are the same */ |
| 118 | + if (video_gfx_getclipw(src_gfx) == 0 || video_gfx_getcliph(src_gfx) == 0) { |
| 119 | + ctx->vbt_ops = &libvideo_emptyblitter_ops; |
| 120 | + } else if (src_buffer == dst_buffer) { |
| 121 | + if (src_gfx->vx_flags & VIDEO_GFX_F_LINEAR) { |
| 122 | + drv->bsw_stretch = &libvideo_swblitter_samebuf__stretch_l; |
| 123 | + drv->bsw_stretch_imatrix = &libvideo_swblitter_samebuf__stretch_imatrix_l; |
| 124 | + } else { |
| 125 | + drv->bsw_stretch = &libvideo_swblitter_samebuf__stretch_n; |
| 126 | + drv->bsw_stretch_imatrix = &libvideo_swblitter_samebuf__stretch_imatrix_n; |
| 127 | + } |
| 128 | + if ((src_gfx->vx_flags & VIDEO_GFX_F_BLUR) == 0 && |
| 129 | + VIDEO_COLOR_ISTRANSPARENT(src_gfx->vx_colorkey) && |
| 130 | + GFX_BLENDMODE_GET_MODE(dst_gfx->vx_blend) == GFX_BLENDMODE_OVERRIDE) { |
| 131 | + drv->bsw_blit = &libvideo_swblitter_noblend_samebuf__blit; |
| 132 | + drv->bsw_blit_imatrix = &libvideo_swblitter_noblend_samebuf__blit_imatrix; |
| 133 | + } else { |
| 134 | + /* Need to use different impls here that essentially do a "memmove"-style blit, |
| 135 | + * rather than the usual "memcpy"-style one (since in this case, writing new |
| 136 | + * pixels in an incorrect order might clobber other pixels that have yet to |
| 137 | + * be read) */ |
| 138 | + drv->bsw_blit = &libvideo_swblitter_samebuf__blit; |
| 139 | + drv->bsw_blit_imatrix = &libvideo_swblitter_samebuf__blit_imatrix; |
| 140 | + } |
| 141 | + } else if (GFX_BLENDMODE_GET_MODE(dst_gfx->vx_blend) == GFX_BLENDMODE_OVERRIDE) { |
| 142 | + if ((src_gfx->vx_flags & VIDEO_GFX_F_BLUR) != 0) |
| 143 | + goto set_generic_operators; |
| 144 | + if (!VIDEO_COLOR_ISTRANSPARENT(src_gfx->vx_colorkey)) |
| 145 | + goto set_generic_operators; |
| 146 | + if (noblend_blit_compatible(dst_buffer->vb_format.vf_codec, |
| 147 | + src_buffer->vb_format.vf_codec) && |
| 148 | + src_buffer->vb_format.vf_pal == dst_buffer->vb_format.vf_pal) { |
| 149 | + /* Special optimization when not doing any blending, and both GFX contexts |
| 150 | + * share the same codec: in this case, we can try to directly copy pixel |
| 151 | + * data, either through video locks, or by directly reading/writing pixels */ |
| 152 | + drv->bsw_blit = &libvideo_swblitter_noblend_samefmt__blit; |
| 153 | + drv->bsw_blit_imatrix = &libvideo_swblitter_noblend_samefmt__blit_imatrix; |
| 154 | + if (src_gfx->vx_flags & VIDEO_GFX_F_LINEAR) { |
| 155 | + drv->bsw_stretch = &libvideo_swblitter_noblend_samefmt__stretch_l; |
| 156 | + drv->bsw_stretch_imatrix = &libvideo_swblitter_noblend_samefmt__stretch_imatrix_l; |
| 157 | + } else { |
| 158 | + drv->bsw_stretch = &libvideo_swblitter_noblend_samefmt__stretch_n; |
| 159 | + drv->bsw_stretch_imatrix = &libvideo_swblitter_noblend_samefmt__stretch_imatrix_n; |
| 160 | + } |
| 161 | + } else { |
| 162 | + /* Special optimization when not doing any blending, and both GFX contexts |
| 163 | + * share the same codec: in this case, we can try to directly copy pixel |
| 164 | + * data, either through video locks, or by directly reading/writing pixels */ |
| 165 | + video_converter_init(libvideo_swblitter_generic__conv(ctx), |
| 166 | + &src_buffer->vb_format, |
| 167 | + &dst_buffer->vb_format); |
| 168 | + drv->bsw_blit = &libvideo_swblitter_noblend_difffmt__blit; |
| 169 | + drv->bsw_blit_imatrix = &libvideo_swblitter_noblend_difffmt__blit_imatrix; |
| 170 | + if (src_gfx->vx_flags & VIDEO_GFX_F_LINEAR) { |
| 171 | + drv->bsw_stretch = &libvideo_swblitter_noblend_difffmt__stretch_l; |
| 172 | + drv->bsw_stretch_imatrix = &libvideo_swblitter_noblend_difffmt__stretch_imatrix_l; |
| 173 | + } else { |
| 174 | + drv->bsw_stretch = &libvideo_swblitter_noblend_difffmt__stretch_n; |
| 175 | + drv->bsw_stretch_imatrix = &libvideo_swblitter_noblend_difffmt__stretch_imatrix_n; |
| 176 | + } |
| 177 | + } |
| 178 | + } else { |
| 179 | +set_generic_operators: |
| 180 | + drv->bsw_blit = &libvideo_swblitter_generic__blit; |
| 181 | + drv->bsw_blit_imatrix = &libvideo_swblitter_generic__blit_imatrix; |
| 182 | + if (src_gfx->vx_flags & VIDEO_GFX_F_LINEAR) { |
| 183 | + drv->bsw_stretch = &libvideo_swblitter_generic__stretch_l; |
| 184 | + drv->bsw_stretch_imatrix = &libvideo_swblitter_generic__stretch_imatrix_l; |
| 185 | + } else { |
| 186 | + drv->bsw_stretch = &libvideo_swblitter_generic__stretch_n; |
| 187 | + drv->bsw_stretch_imatrix = &libvideo_swblitter_generic__stretch_imatrix_n; |
| 188 | + } |
| 189 | + } |
| 190 | + return ctx; |
| 191 | +} |
| 192 | + |
| 193 | + |
| 194 | +INTERN ATTR_RETNONNULL ATTR_INOUT(1) struct video_blitter3 *FCC |
| 195 | +libvideo_swgfx_blitfrom3(struct video_blitter3 *__restrict ctx) { |
| 196 | + struct video_gfx const *src_gfx = ctx->vbt3_src; |
| 197 | + struct video_gfx const *rddst_gfx = ctx->vbt3_rddst; |
| 198 | + struct video_gfx const *wrdst_gfx = ctx->vbt3_wrdst; |
| 199 | + struct blt3_swdrv *drv = video_swblitter3_getdrv(ctx); |
| 200 | + video_swblitter3_setops(ctx); |
| 201 | + (void)wrdst_gfx; |
| 202 | + assert(video_gfx_getclipw(wrdst_gfx) != 0); |
| 203 | + assert(video_gfx_getcliph(wrdst_gfx) != 0); |
| 204 | + if (video_gfx_getclipw(src_gfx) == 0 || video_gfx_getcliph(src_gfx) == 0 || |
| 205 | + video_gfx_getclipw(rddst_gfx) == 0 || video_gfx_getcliph(rddst_gfx) == 0) { |
| 206 | + ctx->vbt3_ops = &libvideo_emptyblitter3_ops; |
| 207 | + } else { |
| 208 | + drv->bsw3_blendmode = rddst_gfx->vx_blend; |
| 209 | + switch (__builtin_expect(GFX_BLENDMODE_GET_MODE(rddst_gfx->vx_blend), |
| 210 | + GFX_BLENDMODE_ALPHA)) { |
| 211 | +#define LINK_libvideo_swblt_generic__blend_FOO(name, mode) \ |
| 212 | + case mode: \ |
| 213 | + drv->bsw3_blend = &libvideo_swblitter3__blend_##name; \ |
| 214 | + break; |
| 215 | + GFX_FOREACH_DEDICATED_BLENDMODE(LINK_libvideo_swblt_generic__blend_FOO) |
| 216 | + GFX_FOREACH_DEDICATED_BLENDMODE_FACTOR(LINK_libvideo_swblt_generic__blend_FOO) |
| 217 | +#undef LINK_libvideo_swblt_generic__blend_FOO |
| 218 | + default: |
| 219 | + drv->bsw3_blend = &libvideo_swblitter3__blend; |
| 220 | + break; |
| 221 | + } |
| 222 | + |
| 223 | + /* TODO: Special handling when buffers overlap */ |
| 224 | + /* TODO: More dedicated optimizations */ |
| 225 | + |
| 226 | + if likely(GFX_BLENDMODE_GET_MODE(wrdst_gfx->vx_blend) == GFX_BLENDMODE_OVERRIDE) { |
| 227 | + /* Special optimization for likely case where "wrdst_gfx" doesn't do any blending */ |
| 228 | + drv->bsw3_blit = &libvideo_swblitter3__blit__blend1; |
| 229 | + drv->bsw3_blit_imatrix = &libvideo_swblitter3__blit_imatrix__blend1; |
| 230 | + if (src_gfx->vx_flags & VIDEO_GFX_F_LINEAR) { |
| 231 | + drv->bsw3_stretch = &libvideo_swblitter3__stretch__blend1_l; |
| 232 | + drv->bsw3_stretch_imatrix = &libvideo_swblitter3__stretch_imatrix__blend1_l; |
| 233 | + } else { |
| 234 | + drv->bsw3_stretch = &libvideo_swblitter3__stretch__blend1_n; |
| 235 | + drv->bsw3_stretch_imatrix = &libvideo_swblitter3__stretch_imatrix__blend1_n; |
| 236 | + } |
| 237 | + } else { |
| 238 | + drv->bsw3_blit = &libvideo_swblitter3__blit__generic; |
| 239 | + drv->bsw3_blit_imatrix = &libvideo_swblitter3__blit_imatrix__generic; |
| 240 | + if (src_gfx->vx_flags & VIDEO_GFX_F_LINEAR) { |
| 241 | + drv->bsw3_stretch = &libvideo_swblitter3__stretch__generic_l; |
| 242 | + drv->bsw3_stretch_imatrix = &libvideo_swblitter3__stretch_imatrix__generic_l; |
| 243 | + } else { |
| 244 | + drv->bsw3_stretch = &libvideo_swblitter3__stretch__generic_n; |
| 245 | + drv->bsw3_stretch_imatrix = &libvideo_swblitter3__stretch_imatrix__generic_n; |
| 246 | + } |
| 247 | + } |
| 248 | + } |
| 249 | + return ctx; |
| 250 | +} |
| 251 | + |
| 252 | +DECL_END |
| 253 | + |
| 254 | +#ifndef __INTELLISENSE__ |
| 255 | +#define DEFINE_libvideo_swblitter_blit |
| 256 | +#include "swgfx/hl_blit-nowrap.c.inl" |
| 257 | +#define DEFINE_libvideo_swblitter_stretch |
| 258 | +#include "swgfx/hl_blit-nowrap.c.inl" |
| 259 | +#define DEFINE_libvideo_swblitter_blit_wrap |
| 260 | +#include "swgfx/hl_blit.c.inl" |
| 261 | + |
| 262 | +#define DEFINE_libvideo_swblitter3_blit |
| 263 | +#include "swgfx/hl_blit-nowrap.c.inl" |
| 264 | +#define DEFINE_libvideo_swblitter3_stretch |
| 265 | +#include "swgfx/hl_blit-nowrap.c.inl" |
| 266 | +#define DEFINE_libvideo_swblitter3_blit_wrap |
| 267 | +#include "swgfx/hl_blit.c.inl" |
| 268 | +#endif /* !__INTELLISENSE__ */ |
| 269 | + |
| 270 | +#endif /* !GUARD_LIBVIDEO_GFX_SWGFX_HL_BLIT_C */ |
0 commit comments