3 * Copyright 2007-2008 Maarten Maathuis
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 * and/or sell copies of the Software, and to permit persons to whom the
10 * Software is furnished to do so, subject to the following conditions:
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
21 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
22 * DEALINGS IN THE SOFTWARE.
30 #include <X11/extensions/Xv.h>
33 #include "dixstruct.h"
36 #include "nv_include.h"
39 #include "nv_shaders.h"
41 extern Atom xvSyncToVBlank, xvSetDefaults;
43 static nv_shader_t nv40_video = {
44 .card_priv.NV30VP.vp_in_reg = 0x00000309,
45 .card_priv.NV30VP.vp_out_reg = 0x0000c001,
48 /* MOV result.position, vertex.position */
49 0x40041c6c, 0x0040000d, 0x8106c083, 0x6041ff80,
50 /* MOV result.texcoord[0], vertex.texcoord[0] */
51 0x401f9c6c, 0x0040080d, 0x8106c083, 0x6041ff9c,
52 /* MOV result.texcoord[1], vertex.texcoord[1] */
53 0x401f9c6c, 0x0040090d, 0x8106c083, 0x6041ffa1,
57 static nv_shader_t nv40_yv12 = {
58 .card_priv.NV30FP.num_regs = 2,
61 /* INST 0: TEXR R0.x (TR0.xyzw), attrib.texcoord[0], abs(texture[0]) */
62 0x17008200, 0x1c9dc801, 0x0001c800, 0x3fe1c800,
63 /* INST 1: MADR R1.xyz (TR0.xyzw), R0.xxxx, { 1.16, -0.87, 0.53, -1.08 }.xxxx, { 1.16, -0.87, 0.53, -1.08 }.yzww */
64 0x04000e02, 0x1c9c0000, 0x00000002, 0x0001f202,
65 0x3f9507c8, 0xbf5ee393, 0x3f078fef, 0xbf8a6762,
66 /* INST 2: TEXR R0.yz (TR0.xyzw), attrib.texcoord[1], abs(texture[1]) */
67 0x1702ac80, 0x1c9dc801, 0x0001c800, 0x3fe1c800,
68 /* INST 3: MADR R1.xyz (TR0.xyzw), R0.yyyy, { 0.00, -0.39, 2.02, 0.00 }, R1 */
69 0x04000e02, 0x1c9cab00, 0x0001c802, 0x0001c804,
70 0x00000000, 0xbec890d6, 0x40011687, 0x00000000,
71 /* INST 4: MADR R0.xyz (TR0.xyzw), R0.zzzz, { 1.60, -0.81, 0.00, 0.00 }, R1 + END */
72 0x04000e81, 0x1c9d5500, 0x0001c802, 0x0001c804,
73 0x3fcc432d, 0xbf501a37, 0x00000000, 0x00000000,
77 #define SWIZZLE(ts0x,ts0y,ts0z,ts0w,ts1x,ts1y,ts1z,ts1w) \
79 NV40TCL_TEX_SWIZZLE_S0_X_##ts0x | NV40TCL_TEX_SWIZZLE_S0_Y_##ts0y | \
80 NV40TCL_TEX_SWIZZLE_S0_Z_##ts0z | NV40TCL_TEX_SWIZZLE_S0_W_##ts0w | \
81 NV40TCL_TEX_SWIZZLE_S1_X_##ts1x | NV40TCL_TEX_SWIZZLE_S1_Y_##ts1y | \
82 NV40TCL_TEX_SWIZZLE_S1_Z_##ts1z | NV40TCL_TEX_SWIZZLE_S1_W_##ts1w \
86 NV40VideoTexture(ScrnInfoPtr pScrn, int offset, uint16_t width, uint16_t height, uint16_t src_pitch, int unit)
88 NVPtr pNv = NVPTR(pScrn);
90 uint32_t card_fmt = 0;
91 uint32_t card_swz = 0;
94 /* Pretend we've got a normal 8 bits format. */
95 card_fmt = NV40TCL_TEX_FORMAT_FORMAT_L8;
96 card_swz = SWIZZLE(S1, S1, S1, S1, X, X, X, X);
98 /* Pretend we've got a normal 2x8 bits format. */
99 card_fmt = NV40TCL_TEX_FORMAT_FORMAT_A8L8;
100 card_swz = SWIZZLE(S1, S1, S1, S1, W, Z, Y, X); /* x = V, y = U */
103 BEGIN_RING(Nv3D, NV40TCL_TEX_OFFSET(unit), 8);
104 /* We get an obsolute offset, which needs to be corrected. */
105 OUT_RELOCl(pNv->FB, (uint32_t)(offset - pNv->FB->offset), NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
106 OUT_RELOCd(pNv->FB, card_fmt | NV40TCL_TEX_FORMAT_LINEAR |
107 NV40TCL_TEX_FORMAT_DIMS_2D | NV40TCL_TEX_FORMAT_NO_BORDER |
108 (0x8000) | (1 << NV40TCL_TEX_FORMAT_MIPMAP_COUNT_SHIFT),
109 NOUVEAU_BO_VRAM | NOUVEAU_BO_RD,
110 NV40TCL_TEX_FORMAT_DMA0, 0);
112 OUT_RING(NV40TCL_TEX_WRAP_S_CLAMP_TO_EDGE |
113 NV40TCL_TEX_WRAP_T_CLAMP_TO_EDGE |
114 NV40TCL_TEX_WRAP_R_CLAMP_TO_EDGE);
115 OUT_RING(NV40TCL_TEX_ENABLE_ENABLE);
117 OUT_RING(NV40TCL_TEX_FILTER_MIN_LINEAR |
118 NV40TCL_TEX_FILTER_MAG_LINEAR |
120 OUT_RING((width << 16) | height);
121 OUT_RING(0); /* border ARGB */
122 BEGIN_RING(Nv3D, NV40TCL_TEX_SIZE1(unit), 1);
123 OUT_RING((1 << NV40TCL_TEX_SIZE1_DEPTH_SHIFT) |
124 (uint16_t) src_pitch);
130 NV40GetSurfaceFormat(PixmapPtr pPix, int *fmt_ret)
132 switch (pPix->drawable.bitsPerPixel) {
134 *fmt_ret = NV40TCL_RT_FORMAT_COLOR_A8R8G8B8;
137 *fmt_ret = NV40TCL_RT_FORMAT_COLOR_X8R8G8B8;
140 *fmt_ret = NV40TCL_RT_FORMAT_COLOR_R5G6B5;
143 *fmt_ret = NV40TCL_RT_FORMAT_COLOR_B8;
153 NV40StopTexturedVideo(ScrnInfoPtr pScrn, pointer data, Bool Exit)
157 /* To support EXA 2.0, 2.1 has this in the header */
158 #ifndef exaMoveInPixmap
159 extern void exaMoveInPixmap(PixmapPtr pPixmap);
162 #define SF(bf) (NV40TCL_BLEND_FUNC_SRC_RGB_##bf | \
163 NV40TCL_BLEND_FUNC_SRC_ALPHA_##bf)
164 #define DF(bf) (NV40TCL_BLEND_FUNC_DST_RGB_##bf | \
165 NV40TCL_BLEND_FUNC_DST_ALPHA_##bf)
167 #define VERTEX_OUT(sx,sy,dx,dy) do { \
168 BEGIN_RING(Nv3D, NV40TCL_VTX_ATTR_2F_X(8), 4); \
169 OUT_RINGf ((sx)); OUT_RINGf ((sy)); \
170 OUT_RINGf ((sx)); OUT_RINGf ((sy)); \
171 BEGIN_RING(Nv3D, NV40TCL_VTX_ATTR_2I(0), 1); \
172 OUT_RING (((dy)<<16)|(dx)); \
175 #define GET_TEXTURED_PRIVATE(pNv) \
176 (NVPortPrivPtr)((pNv)->blitAdaptor->pPortPrivates[0].ptr)
178 int NV40PutTextureImage(ScrnInfoPtr pScrn, int src_offset,
179 int src_offset2, int id,
180 int src_pitch, BoxPtr dstBox,
181 int x1, int y1, int x2, int y2,
182 uint16_t width, uint16_t height,
183 uint16_t src_w, uint16_t src_h,
184 uint16_t drw_w, uint16_t drw_h,
188 NVPtr pNv = NVPTR(pScrn);
189 NVPortPrivPtr pPriv = GET_TEXTURED_PRIVATE(pNv);
190 Bool redirected = FALSE;
192 /* Remove some warnings. */
193 /* This has to be done better at some point. */
194 (void)nv40_vp_exa_render;
195 (void)nv30_fp_pass_col0;
196 (void)nv30_fp_pass_tex0;
197 (void)nv30_fp_composite_mask;
198 (void)nv30_fp_composite_mask_sa_ca;
199 (void)nv30_fp_composite_mask_ca;
201 if (drw_w > 4096 || drw_h > 4096) {
202 xf86DrvMsg(pScrn->scrnIndex, X_ERROR,
203 "XV: Draw size too large.\n");
207 float X1, X2, Y1, Y2;
208 uint16_t p_width, p_height;
209 float pf_width, pf_height; /* fractions pbox/dstBox */
210 float pf_x1, pf_y1; /* fractions of x,y/pBox */
211 float df_x1, df_y1; /* fractions of x,y/dstBox */
212 float drf_x, drf_y; /* draw to src ratio. */
213 PixmapPtr pPix = NVGetDrawablePixmap(pDraw);
217 if (!NV40GetSurfaceFormat(pPix, &dst_format)) {
218 ErrorF("No surface format, bad.\n");
221 /* This has to be called always, since it does more than just migration. */
222 exaMoveInPixmap(pPix);
223 ExaOffscreenMarkUsed(pPix);
226 /* Adjust coordinates if drawing to an offscreen pixmap */
227 if (pPix->screen_x || pPix->screen_y) {
228 REGION_TRANSLATE(pScrn->pScreen, clipBoxes,
231 dstBox->x1 -= pPix->screen_x;
232 dstBox->x2 -= pPix->screen_x;
233 dstBox->y1 -= pPix->screen_y;
234 dstBox->y2 -= pPix->screen_y;
237 /* I suspect that pDraw itself is not offscreen, hence not suited for damage tracking. */
238 DamageDamageRegion(&pPix->drawable, clipBoxes);
240 /* This is test is unneeded for !COMPOSITE. */
241 if (!NVExaPixmapIsOnscreen(pPix))
245 pbox = REGION_RECTS(clipBoxes);
246 nbox = REGION_NUM_RECTS(clipBoxes);
248 /* Disable blending */
249 BEGIN_RING(Nv3D, NV40TCL_BLEND_ENABLE, 1);
253 BEGIN_RING(Nv3D, NV40TCL_RT_FORMAT, 3);
254 OUT_RING (NV40TCL_RT_FORMAT_TYPE_LINEAR |
255 NV40TCL_RT_FORMAT_ZETA_Z24S8 |
257 OUT_RING (exaGetPixmapPitch(pPix));
258 OUT_PIXMAPl(pPix, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
260 NV40VideoTexture(pScrn, src_offset, src_w, src_h, src_pitch, 0);
261 /* We've got NV12 format, which means half width and half height texture of chroma channels. */
262 NV40VideoTexture(pScrn, src_offset2, src_w/2, src_h/2, src_pitch, 1);
264 NV40_LoadVtxProg(pScrn, &nv40_video);
265 NV40_LoadFragProg(pScrn, &nv40_yv12);
267 /* Appears to be some kind of cache flush, needed here at least
268 * sometimes.. funky text rendering otherwise :)
270 BEGIN_RING(Nv3D, NV40TCL_TEX_CACHE_CTL, 1);
272 BEGIN_RING(Nv3D, NV40TCL_TEX_CACHE_CTL, 1);
275 /* These are fixed point values in the 16.16 format. */
281 X1 = (float)x1/(float)src_w;
282 Y1 = (float)y1/(float)src_h;
283 X2 = (float)x2/(float)src_w;
284 Y2 = (float)y2/(float)src_h;
286 /* The corrections here are emperical, i tried to explain them as best as possible. */
288 /* Fractions of the upperleft corner of the dstBox. */
289 /* Because all source coords are normalized. */
290 df_x1 = (float)(dstBox->x1)/(float)src_w;
291 df_y1 = (float)(dstBox->y1)/(float)src_h;
293 /* Draw ratio fractions. */
294 drf_x = (float)src_w/(float)drw_w;
295 drf_y = (float)src_h/(float)drw_h;
297 /* Just before rendering we wait for vblank in the non-composited case. */
298 if (pPriv->SyncToVBlank && !redirected) {
299 uint8_t crtcs = nv_window_belongs_to_crtc(pScrn, dstBox->x1, dstBox->y1,
300 dstBox->x2 - dstBox->x1, dstBox->y2 - dstBox->y1);
304 NVWaitVSync(pScrn, 0);
305 else if (crtcs & 0x2)
306 NVWaitVSync(pScrn, 1);
309 BEGIN_RING(Nv3D, NV40TCL_BEGIN_END, 1);
310 OUT_RING (NV40TCL_BEGIN_END_TRIANGLES);
313 p_width = pbox->x2 - pbox->x1;
314 p_height = pbox->y2 - pbox->y1;
315 pf_width = (float)p_width/(float)src_w;
316 pf_height = (float)p_height/(float)src_h;
317 pf_x1 = (float)(pbox->x1)/(float)src_w;
318 pf_y1 = (float)(pbox->y1)/(float)src_h;
320 /* Let's use scissors, since we are drawing larger than our required drawing area. */
321 /* Reminder: Even the pbox'es are drawn as triangles. */
322 BEGIN_RING(Nv3D, NV40TCL_SCISSOR_HORIZ, 2);
323 OUT_RING ((p_width << 16) | pbox->x1);
324 OUT_RING ((p_height << 16) | pbox->y1);
326 /* Submit the appropriate vertices. */
327 /* This submits the same vertices for the Y and the UV texture. */
328 /* All are screen coordinate references must be scaled to draw size. */
329 /* I've included a little (svg) drawing for future reference. */
330 VERTEX_OUT(X1 + (pf_x1 - df_x1) * drf_x, Y1 + (pf_y1 - df_y1 - pf_height) * drf_y, pbox->x1, pbox->y1 - p_height);
331 VERTEX_OUT(X1 + (pf_x1 - df_x1) * drf_x, Y1 + (pf_y1 - df_y1 + pf_height) * drf_y, pbox->x1, pbox->y2);
332 VERTEX_OUT(X1 + (pf_x1 - df_x1 + 2*pf_width) * drf_x, Y1 + (pf_y1 - df_y1 + pf_height) * drf_y, pbox->x2 + p_width, pbox->y2);
337 BEGIN_RING(Nv3D, NV40TCL_BEGIN_END, 1);
338 OUT_RING (NV40TCL_BEGIN_END_STOP);
340 /* We can't leak state to exa. */
341 BEGIN_RING(Nv3D, NV40TCL_SCISSOR_HORIZ, 2);
342 OUT_RING ((4096 << 16));
343 OUT_RING ((4096 << 16));
351 * NVSetTexturePortAttribute
352 * sets the attribute "attribute" of port "data" to value "value"
353 * supported attributes:
357 * @param attribute attribute to set
358 * @param value value to which attribute is to be set
359 * @param data port from which the attribute is to be set
361 * @return Success, if setting is successful
362 * BadValue/BadMatch, if value/attribute are invalid
365 NVSetTexturePortAttribute(ScrnInfoPtr pScrn, Atom attribute,
366 INT32 value, pointer data)
368 NVPortPrivPtr pPriv = (NVPortPrivPtr)data;
369 NVPtr pNv = NVPTR(pScrn);
371 if ((attribute == xvSyncToVBlank) && pNv->WaitVSyncPossible) {
372 if ((value < 0) || (value > 1))
374 pPriv->SyncToVBlank = value;
376 if (attribute == xvSetDefaults) {
377 pPriv->SyncToVBlank = pNv->WaitVSyncPossible;
385 * NVGetTexturePortAttribute
386 * reads the value of attribute "attribute" from port "data" into INT32 "*value"
389 * @param pScrn unused
390 * @param attribute attribute to be read
391 * @param value value of attribute will be stored here
392 * @param data port from which attribute will be read
393 * @return Success, if queried attribute exists
396 NVGetTexturePortAttribute(ScrnInfoPtr pScrn, Atom attribute,
397 INT32 *value, pointer data)
399 NVPortPrivPtr pPriv = (NVPortPrivPtr)data;
401 if(attribute == xvSyncToVBlank)
402 *value = (pPriv->SyncToVBlank) ? 1 : 0;