2 * Copyright 2007-2008 Maarten Maathuis
3 * Copyright 2008 Stephane Marchesin
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 "nv30_shaders.h"
41 extern Atom xvSyncToVBlank, xvSetDefaults;
44 * The filtering function used for video scaling. We use a cubic filter as defined in
45 * "Reconstruction Filters in Computer Graphics"
46 * Mitchell & Netravali in SIGGRAPH '88
48 static float filter_func(float x)
51 const double C=(1.0-B)/2.0;
57 return ( (12.0-9.0*B-6.0*C)*x3+(-18.0+12.0*B+6.0*C)*x2+(6.0-2.0*B) )/6.0;
59 return ( (-B-6.0*C)*x3+(6.0*B+30.0*C)*x2+(-12.0*B-48.0*C)*x1+(8.0*B+24.0*C) )/6.0;
62 static int8_t f32tosb8(float v)
64 return (int8_t)(v*127.0);
68 * 512 means 2048 bytes of VRAM
70 #define TABLE_SIZE 512
71 static void compute_filter_table(int8_t *t) {
74 for(i=0;i<TABLE_SIZE;i++) {
77 float w0=filter_func(x+1.0);
78 float w1=filter_func(x);
79 float w2=filter_func(x-1.0);
80 float w3=filter_func(x-2.0);
82 t[4*i+2]=f32tosb8(1.0+x-w1/(w0+w1));
83 t[4*i+1]=f32tosb8(1.0-x+w3/(w2+w3));
84 t[4*i+0]=f32tosb8(w0+w1);
85 t[4*i+3]=f32tosb8(0.0);
90 NV30_LoadFilterTable(ScrnInfoPtr pScrn)
92 NVPtr pNv = NVPTR(pScrn);
94 if (!pNv->xv_filtertable_mem) {
95 if (nouveau_bo_new(pNv->dev, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART,
96 0, TABLE_SIZE*sizeof(float)*4, &pNv->xv_filtertable_mem)) {
97 xf86DrvMsg(pScrn->scrnIndex, X_ERROR,
98 "Couldn't alloc filter table!\n");
102 if (nouveau_bo_map(pNv->xv_filtertable_mem, NOUVEAU_BO_RDWR)) {
103 xf86DrvMsg(pScrn->scrnIndex, X_ERROR,
104 "Couldn't map filter table!\n");
108 int8_t *t=pNv->xv_filtertable_mem->map;
109 compute_filter_table(t);
113 #define SWIZZLE(ts0x,ts0y,ts0z,ts0w,ts1x,ts1y,ts1z,ts1w) \
115 NV34TCL_TX_SWIZZLE_S0_X_##ts0x | NV34TCL_TX_SWIZZLE_S0_Y_##ts0y | \
116 NV34TCL_TX_SWIZZLE_S0_Z_##ts0z | NV34TCL_TX_SWIZZLE_S0_W_##ts0w | \
117 NV34TCL_TX_SWIZZLE_S1_X_##ts1x | NV34TCL_TX_SWIZZLE_S1_Y_##ts1y | \
118 NV34TCL_TX_SWIZZLE_S1_Z_##ts1z | NV34TCL_TX_SWIZZLE_S1_W_##ts1w \
122 * Texture 0 : filter table
124 * Texture 2 : UV data
127 NV30VideoTexture(ScrnInfoPtr pScrn, struct nouveau_bo *src, int offset,
128 uint16_t width, uint16_t height, uint16_t src_pitch, int unit)
130 NVPtr pNv = NVPTR(pScrn);
131 struct nouveau_channel *chan = pNv->chan;
132 struct nouveau_grobj *rankine = pNv->Nv3D;
134 uint32_t card_fmt = 0;
135 uint32_t card_swz = 0;
139 card_fmt = NV34TCL_TX_FORMAT_FORMAT_A8R8G8B8;
140 card_swz = SWIZZLE(S1, S1, S1, S1, X, Y, Z, W);
143 card_fmt = NV34TCL_TX_FORMAT_FORMAT_A8_RECT2;
144 card_swz = SWIZZLE(S1, S1, S1, S1, X, X, X, X);
147 card_fmt = NV34TCL_TX_FORMAT_FORMAT_L8A8_RECT;
148 #if X_BYTE_ORDER == X_BIG_ENDIAN
149 card_swz = SWIZZLE(S1, S1, S1, S1, Z, W, X, Y); /* x = V, y = U */
151 card_swz = SWIZZLE(S1, S1, S1, S1, W, Z, Y, X); /* x = V, y = U */
156 BEGIN_RING(chan, rankine, NV34TCL_TX_OFFSET(unit), 8);
157 OUT_RELOCl(chan, src, offset, NOUVEAU_BO_VRAM | NOUVEAU_BO_RD);
159 OUT_RELOCd(chan, pNv->FB, NV34TCL_TX_FORMAT_DIMS_1D | card_fmt |
160 (1 << NV34TCL_TX_FORMAT_MIPMAP_LEVELS_SHIFT) |
162 NV34TCL_TX_FORMAT_BASE_SIZE_U_SHIFT) |
164 NV34TCL_TX_FORMAT_BASE_SIZE_V_SHIFT) |
166 NOUVEAU_BO_VRAM | NOUVEAU_BO_RD,
167 NV34TCL_TX_FORMAT_DMA0, 0);
168 OUT_RING (chan, NV34TCL_TX_WRAP_S_REPEAT |
169 NV34TCL_TX_WRAP_T_CLAMP_TO_EDGE |
170 NV34TCL_TX_WRAP_R_CLAMP_TO_EDGE);
172 OUT_RELOCd(chan, pNv->FB, NV34TCL_TX_FORMAT_DIMS_2D | card_fmt |
173 (1 << NV34TCL_TX_FORMAT_MIPMAP_LEVELS_SHIFT) |
175 NV34TCL_TX_FORMAT_BASE_SIZE_U_SHIFT) |
177 NV34TCL_TX_FORMAT_BASE_SIZE_V_SHIFT) |
179 NOUVEAU_BO_VRAM | NOUVEAU_BO_RD,
180 NV34TCL_TX_FORMAT_DMA0, 0);
181 OUT_RING (chan, NV34TCL_TX_WRAP_S_CLAMP_TO_EDGE |
182 NV34TCL_TX_WRAP_T_CLAMP_TO_EDGE |
183 NV34TCL_TX_WRAP_R_CLAMP_TO_EDGE);
186 OUT_RING (chan, NV34TCL_TX_ENABLE_ENABLE);
187 OUT_RING (chan, (src_pitch << NV34TCL_TX_SWIZZLE_RECT_PITCH_SHIFT) |
190 OUT_RING (chan, NV34TCL_TX_FILTER_SIGNED_ALPHA |
191 NV34TCL_TX_FILTER_SIGNED_RED |
192 NV34TCL_TX_FILTER_SIGNED_GREEN |
193 NV34TCL_TX_FILTER_SIGNED_BLUE |
194 NV34TCL_TX_FILTER_MINIFY_LINEAR |
195 NV34TCL_TX_FILTER_MAGNIFY_LINEAR | 0x2000);
197 OUT_RING (chan, NV34TCL_TX_FILTER_MINIFY_LINEAR |
198 NV34TCL_TX_FILTER_MAGNIFY_LINEAR | 0x2000);
199 OUT_RING (chan, (width << NV34TCL_TX_NPOT_SIZE_W_SHIFT) | height);
200 OUT_RING (chan, 0); /* border ARGB */
206 NV30GetSurfaceFormat(PixmapPtr ppix, int *fmt_ret)
208 switch (ppix->drawable.bitsPerPixel) {
210 *fmt_ret = NV34TCL_RT_FORMAT_COLOR_A8R8G8B8;
213 *fmt_ret = NV34TCL_RT_FORMAT_COLOR_X8R8G8B8;
216 *fmt_ret = NV34TCL_RT_FORMAT_COLOR_R5G6B5;
219 *fmt_ret = NV34TCL_RT_FORMAT_COLOR_B8;
229 NV30StopTexturedVideo(ScrnInfoPtr pScrn, pointer data, Bool Exit)
233 #define VERTEX_OUT(sx,sy,dx,dy) do { \
234 BEGIN_RING(chan, rankine, NV34TCL_VTX_ATTR_2F_X(8), 4); \
235 OUT_RINGf (chan, (sx)); OUT_RINGf (chan, (sy)); \
236 OUT_RINGf (chan, (sx)/2.0); OUT_RINGf (chan, (sy)/2.0); \
237 BEGIN_RING(chan, rankine, NV34TCL_VTX_ATTR_2I(0), 1); \
238 OUT_RING (chan, ((dy)<<16)|(dx)); \
242 NV30PutTextureImage(ScrnInfoPtr pScrn, struct nouveau_bo *src, int src_offset,
243 int src_offset2, int id, int src_pitch, BoxPtr dstBox,
244 int x1, int y1, int x2, int y2,
245 uint16_t width, uint16_t height,
246 uint16_t src_w, uint16_t src_h,
247 uint16_t drw_w, uint16_t drw_h,
252 NVPtr pNv = NVPTR(pScrn);
253 struct nouveau_channel *chan = pNv->chan;
254 struct nouveau_grobj *rankine = pNv->Nv3D;
255 Bool redirected = FALSE;
256 float X1, X2, Y1, Y2;
261 if (drw_w > 4096 || drw_h > 4096) {
262 xf86DrvMsg(pScrn->scrnIndex, X_ERROR,
263 "XV: Draw size too large.\n");
267 if (!NV30GetSurfaceFormat(ppix, &dst_format)) {
268 ErrorF("No surface format, bad.\n");
272 if (!NVExaPixmapIsOnscreen(ppix))
276 pbox = REGION_RECTS(clipBoxes);
277 nbox = REGION_NUM_RECTS(clipBoxes);
279 /* Disable blending */
280 BEGIN_RING(chan, rankine, NV34TCL_BLEND_FUNC_ENABLE, 1);
284 BEGIN_RING(chan, rankine, NV34TCL_RT_FORMAT, 3);
285 OUT_RING (chan, NV34TCL_RT_FORMAT_TYPE_LINEAR |
286 NV34TCL_RT_FORMAT_ZETA_Z24S8 |
288 OUT_RING (chan, (exaGetPixmapPitch(ppix) << 16) |
289 exaGetPixmapPitch(ppix));
290 OUT_PIXMAPl(chan, ppix, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
292 if (pNv->NVArch == 0x30) {
295 int w = ppix->drawable.x + ppix->drawable.width;
296 int h = ppix->drawable.y + ppix->drawable.height;
298 BEGIN_RING(chan, rankine, NV34TCL_VIEWPORT_HORIZ, 2);
299 OUT_RING (chan, (w<<16)|x);
300 OUT_RING (chan, (h<<16)|y);
301 BEGIN_RING(chan, rankine, NV34TCL_VIEWPORT_CLIP_HORIZ(0), 2);
302 OUT_RING (chan, (w-1+x)<<16);
303 OUT_RING (chan, (h-1+y)<<16);
304 BEGIN_RING(chan, rankine, NV34TCL_VIEWPORT_TX_ORIGIN, 1);
305 OUT_RING (chan, (y<<16)|x);
308 NV30_LoadFilterTable(pScrn);
310 BEGIN_RING(chan, rankine, NV34TCL_TX_UNITS_ENABLE, 1);
311 OUT_RING (chan, NV34TCL_TX_UNITS_ENABLE_TX0 |
312 NV34TCL_TX_UNITS_ENABLE_TX1);
314 NV30VideoTexture(pScrn, pNv->xv_filtertable_mem, 0, TABLE_SIZE, 1, 0 , 0);
315 NV30VideoTexture(pScrn, src, src_offset, src_w, src_h, src_pitch, 1);
316 /* We've got NV12 format, which means half width and half height texture of chroma channels. */
317 NV30VideoTexture(pScrn, src, src_offset2, src_w/2, src_h/2, src_pitch, 2);
319 BEGIN_RING(chan, rankine, NV34TCL_TX_ENABLE(3), 1);
320 OUT_RING (chan, 0x0);
323 NV30_LoadFragProg(pScrn, &nv30_fp_yv12_bicubic);
325 NV30_LoadFragProg(pScrn, &nv30_fp_yv12_bilinear);
327 /* Just before rendering we wait for vblank in the non-composited case. */
328 if (pPriv->SyncToVBlank && !redirected) {
329 uint8_t crtcs = nv_window_belongs_to_crtc(pScrn, dstBox->x1, dstBox->y1,
330 dstBox->x2 - dstBox->x1, dstBox->y2 - dstBox->y1);
334 NVWaitVSync(pScrn, 0);
335 else if (crtcs & 0x2)
336 NVWaitVSync(pScrn, 1);
339 /* These are fixed point values in the 16.16 format. */
340 X1 = (float)(x1>>16)+(float)(x1&0xFFFF)/(float)0x10000;
341 Y1 = (float)(y1>>16)+(float)(y1&0xFFFF)/(float)0x10000;
342 X2 = (float)(x2>>16)+(float)(x2&0xFFFF)/(float)0x10000;
343 Y2 = (float)(y2>>16)+(float)(y2&0xFFFF)/(float)0x10000;
345 BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
346 OUT_RING (chan, NV34TCL_VERTEX_BEGIN_END_TRIANGLES);
349 float tx1=X1+(float)(pbox->x1 - dstBox->x1)*(X2-X1)/(float)(drw_w);
350 float tx2=X1+(float)(pbox->x2 - dstBox->x1)*(src_w)/(float)(drw_w);
351 float ty1=Y1+(float)(pbox->y1 - dstBox->y1)*(Y2-Y1)/(float)(drw_h);
352 float ty2=Y1+(float)(pbox->y2 - dstBox->y1)*(src_h)/(float)(drw_h);
358 BEGIN_RING(chan, rankine, NV34TCL_SCISSOR_HORIZ, 2);
359 OUT_RING (chan, (sx2 << 16) | 0);
360 OUT_RING (chan, (sy2 << 16) | 0);
362 VERTEX_OUT(tx1, ty1, sx1, sy1);
363 VERTEX_OUT(tx2+(tx2-tx1), ty1, sx2+(sx2-sx1), sy1);
364 VERTEX_OUT(tx1, ty2+(ty2-ty1), sx1, sy2+(sy2-sy1));
369 BEGIN_RING(chan, rankine, NV34TCL_VERTEX_BEGIN_END, 1);
370 OUT_RING (chan, NV34TCL_VERTEX_BEGIN_END_STOP);
378 * NV30SetTexturePortAttribute
379 * sets the attribute "attribute" of port "data" to value "value"
380 * supported attributes:
384 * @param attribute attribute to set
385 * @param value value to which attribute is to be set
386 * @param data port from which the attribute is to be set
388 * @return Success, if setting is successful
389 * BadValue/BadMatch, if value/attribute are invalid
392 NV30SetTexturePortAttribute(ScrnInfoPtr pScrn, Atom attribute,
393 INT32 value, pointer data)
395 NVPortPrivPtr pPriv = (NVPortPrivPtr)data;
396 NVPtr pNv = NVPTR(pScrn);
398 if ((attribute == xvSyncToVBlank) && pNv->WaitVSyncPossible) {
399 if ((value < 0) || (value > 1))
401 pPriv->SyncToVBlank = value;
403 if (attribute == xvSetDefaults) {
404 pPriv->SyncToVBlank = pNv->WaitVSyncPossible;
412 * NV30GetTexturePortAttribute
413 * reads the value of attribute "attribute" from port "data" into INT32 "*value"
416 * @param pScrn unused
417 * @param attribute attribute to be read
418 * @param value value of attribute will be stored here
419 * @param data port from which attribute will be read
420 * @return Success, if queried attribute exists
423 NV30GetTexturePortAttribute(ScrnInfoPtr pScrn, Atom attribute,
424 INT32 *value, pointer data)
426 NVPortPrivPtr pPriv = (NVPortPrivPtr)data;
428 if(attribute == xvSyncToVBlank)
429 *value = (pPriv->SyncToVBlank) ? 1 : 0;