Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * linux/drivers/video/kyro/STG4000OverlayDevice.c |
| 3 | * |
| 4 | * Copyright (C) 2000 Imagination Technologies Ltd |
| 5 | * Copyright (C) 2002 STMicroelectronics |
| 6 | * |
| 7 | * This file is subject to the terms and conditions of the GNU General Public |
| 8 | * License. See the file COPYING in the main directory of this archive |
| 9 | * for more details. |
| 10 | */ |
| 11 | |
| 12 | #include <linux/kernel.h> |
| 13 | #include <linux/errno.h> |
| 14 | #include <linux/types.h> |
| 15 | |
| 16 | #include "STG4000Reg.h" |
Adrian Bunk | a0aa7d0 | 2006-01-09 20:54:04 -0800 | [diff] [blame] | 17 | #include "STG4000Interface.h" |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 18 | |
| 19 | /* HW Defines */ |
| 20 | |
| 21 | #define STG4000_NO_SCALING 0x800 |
| 22 | #define STG4000_NO_DECIMATION 0xFFFFFFFF |
| 23 | |
| 24 | /* Primary surface */ |
| 25 | #define STG4000_PRIM_NUM_PIX 5 |
| 26 | #define STG4000_PRIM_ALIGN 4 |
| 27 | #define STG4000_PRIM_ADDR_BITS 20 |
| 28 | |
| 29 | #define STG4000_PRIM_MIN_WIDTH 640 |
| 30 | #define STG4000_PRIM_MAX_WIDTH 1600 |
| 31 | #define STG4000_PRIM_MIN_HEIGHT 480 |
| 32 | #define STG4000_PRIM_MAX_HEIGHT 1200 |
| 33 | |
| 34 | /* Overlay surface */ |
| 35 | #define STG4000_OVRL_NUM_PIX 4 |
| 36 | #define STG4000_OVRL_ALIGN 2 |
| 37 | #define STG4000_OVRL_ADDR_BITS 20 |
| 38 | #define STG4000_OVRL_NUM_MODES 5 |
| 39 | |
| 40 | #define STG4000_OVRL_MIN_WIDTH 0 |
| 41 | #define STG4000_OVRL_MAX_WIDTH 720 |
| 42 | #define STG4000_OVRL_MIN_HEIGHT 0 |
| 43 | #define STG4000_OVRL_MAX_HEIGHT 576 |
| 44 | |
| 45 | /* Decimation and Scaling */ |
| 46 | static u32 adwDecim8[33] = { |
| 47 | 0xffffffff, 0xfffeffff, 0xffdffbff, 0xfefefeff, 0xfdf7efbf, |
| 48 | 0xfbdf7bdf, 0xf7bbddef, 0xeeeeeeef, 0xeeddbb77, 0xedb76db7, |
| 49 | 0xdb6db6db, 0xdb5b5b5b, 0xdab5ad6b, 0xd5ab55ab, 0xd555aaab, |
| 50 | 0xaaaaaaab, 0xaaaa5555, 0xaa952a55, 0xa94a5295, 0xa5252525, |
| 51 | 0xa4924925, 0x92491249, 0x91224489, 0x91111111, 0x90884211, |
| 52 | 0x88410821, 0x88102041, 0x81010101, 0x80800801, 0x80010001, |
| 53 | 0x80000001, 0x00000001, 0x00000000 |
| 54 | }; |
| 55 | |
| 56 | typedef struct _OVRL_SRC_DEST { |
| 57 | /*clipped on-screen pixel position of overlay */ |
| 58 | u32 ulDstX1; |
| 59 | u32 ulDstY1; |
| 60 | u32 ulDstX2; |
| 61 | u32 ulDstY2; |
| 62 | |
| 63 | /*clipped pixel pos of source data within buffer thses need to be 128 bit word aligned */ |
| 64 | u32 ulSrcX1; |
| 65 | u32 ulSrcY1; |
| 66 | u32 ulSrcX2; |
| 67 | u32 ulSrcY2; |
| 68 | |
| 69 | /* on-screen pixel position of overlay */ |
| 70 | s32 lDstX1; |
| 71 | s32 lDstY1; |
| 72 | s32 lDstX2; |
| 73 | s32 lDstY2; |
| 74 | } OVRL_SRC_DEST; |
| 75 | |
| 76 | static u32 ovlWidth, ovlHeight, ovlStride; |
| 77 | static int ovlLinear; |
| 78 | |
| 79 | void ResetOverlayRegisters(volatile STG4000REG __iomem *pSTGReg) |
| 80 | { |
| 81 | u32 tmp; |
| 82 | |
| 83 | /* Set Overlay address to default */ |
| 84 | tmp = STG_READ_REG(DACOverlayAddr); |
| 85 | CLEAR_BITS_FRM_TO(0, 20); |
| 86 | CLEAR_BIT(31); |
| 87 | STG_WRITE_REG(DACOverlayAddr, tmp); |
| 88 | |
| 89 | /* Set Overlay U address */ |
| 90 | tmp = STG_READ_REG(DACOverlayUAddr); |
| 91 | CLEAR_BITS_FRM_TO(0, 20); |
| 92 | STG_WRITE_REG(DACOverlayUAddr, tmp); |
| 93 | |
| 94 | /* Set Overlay V address */ |
| 95 | tmp = STG_READ_REG(DACOverlayVAddr); |
| 96 | CLEAR_BITS_FRM_TO(0, 20); |
| 97 | STG_WRITE_REG(DACOverlayVAddr, tmp); |
| 98 | |
| 99 | /* Set Overlay Size */ |
| 100 | tmp = STG_READ_REG(DACOverlaySize); |
| 101 | CLEAR_BITS_FRM_TO(0, 10); |
| 102 | CLEAR_BITS_FRM_TO(12, 31); |
| 103 | STG_WRITE_REG(DACOverlaySize, tmp); |
| 104 | |
| 105 | /* Set Overlay Vt Decimation */ |
| 106 | tmp = STG4000_NO_DECIMATION; |
| 107 | STG_WRITE_REG(DACOverlayVtDec, tmp); |
| 108 | |
| 109 | /* Set Overlay format to default value */ |
| 110 | tmp = STG_READ_REG(DACPixelFormat); |
| 111 | CLEAR_BITS_FRM_TO(4, 7); |
| 112 | CLEAR_BITS_FRM_TO(16, 22); |
| 113 | STG_WRITE_REG(DACPixelFormat, tmp); |
| 114 | |
| 115 | /* Set Vertical scaling to default */ |
| 116 | tmp = STG_READ_REG(DACVerticalScal); |
| 117 | CLEAR_BITS_FRM_TO(0, 11); |
| 118 | CLEAR_BITS_FRM_TO(16, 22); |
| 119 | tmp |= STG4000_NO_SCALING; /* Set to no scaling */ |
| 120 | STG_WRITE_REG(DACVerticalScal, tmp); |
| 121 | |
| 122 | /* Set Horizontal Scaling to default */ |
| 123 | tmp = STG_READ_REG(DACHorizontalScal); |
| 124 | CLEAR_BITS_FRM_TO(0, 11); |
| 125 | CLEAR_BITS_FRM_TO(16, 17); |
| 126 | tmp |= STG4000_NO_SCALING; /* Set to no scaling */ |
| 127 | STG_WRITE_REG(DACHorizontalScal, tmp); |
| 128 | |
| 129 | /* Set Blend mode to Alpha Blend */ |
| 130 | /* ????? SG 08/11/2001 Surely this isn't the alpha blend mode, |
| 131 | hopefully its overwrite |
| 132 | */ |
| 133 | tmp = STG_READ_REG(DACBlendCtrl); |
| 134 | CLEAR_BITS_FRM_TO(0, 30); |
| 135 | tmp = (GRAPHICS_MODE << 28); |
| 136 | STG_WRITE_REG(DACBlendCtrl, tmp); |
| 137 | |
| 138 | } |
| 139 | |
| 140 | int CreateOverlaySurface(volatile STG4000REG __iomem *pSTGReg, |
| 141 | u32 inWidth, |
| 142 | u32 inHeight, |
| 143 | int bLinear, |
| 144 | u32 ulOverlayOffset, |
| 145 | u32 * retStride, u32 * retUVStride) |
| 146 | { |
| 147 | u32 tmp; |
| 148 | u32 ulStride; |
| 149 | |
| 150 | if (inWidth > STG4000_OVRL_MAX_WIDTH || |
| 151 | inHeight > STG4000_OVRL_MAX_HEIGHT) { |
| 152 | return -EINVAL; |
| 153 | } |
| 154 | |
| 155 | /* Stride in 16 byte words - 16Bpp */ |
| 156 | if (bLinear) { |
| 157 | /* Format is 16bits so num 16 byte words is width/8 */ |
| 158 | if ((inWidth & 0x7) == 0) { /* inWidth % 8 */ |
| 159 | ulStride = (inWidth / 8); |
| 160 | } else { |
| 161 | /* Round up to next 16byte boundary */ |
| 162 | ulStride = ((inWidth + 8) / 8); |
| 163 | } |
| 164 | } else { |
| 165 | /* Y component is 8bits so num 16 byte words is width/16 */ |
| 166 | if ((inWidth & 0xf) == 0) { /* inWidth % 16 */ |
| 167 | ulStride = (inWidth / 16); |
| 168 | } else { |
| 169 | /* Round up to next 16byte boundary */ |
| 170 | ulStride = ((inWidth + 16) / 16); |
| 171 | } |
| 172 | } |
| 173 | |
| 174 | |
| 175 | /* Set Overlay address and Format mode */ |
| 176 | tmp = STG_READ_REG(DACOverlayAddr); |
| 177 | CLEAR_BITS_FRM_TO(0, 20); |
| 178 | if (bLinear) { |
| 179 | CLEAR_BIT(31); /* Overlay format to Linear */ |
| 180 | } else { |
| 181 | tmp |= SET_BIT(31); /* Overlay format to Planer */ |
| 182 | } |
| 183 | |
| 184 | /* Only bits 24:4 of the Overlay address */ |
| 185 | tmp |= (ulOverlayOffset >> 4); |
| 186 | STG_WRITE_REG(DACOverlayAddr, tmp); |
| 187 | |
| 188 | if (!bLinear) { |
| 189 | u32 uvSize = |
| 190 | (inWidth & 0x1) ? (inWidth + 1 / 2) : (inWidth / 2); |
| 191 | u32 uvStride; |
| 192 | u32 ulOffset; |
| 193 | /* Y component is 8bits so num 32 byte words is width/32 */ |
| 194 | if ((uvSize & 0xf) == 0) { /* inWidth % 16 */ |
| 195 | uvStride = (uvSize / 16); |
| 196 | } else { |
| 197 | /* Round up to next 32byte boundary */ |
| 198 | uvStride = ((uvSize + 16) / 16); |
| 199 | } |
| 200 | |
| 201 | ulOffset = ulOverlayOffset + (inHeight * (ulStride * 16)); |
| 202 | /* Align U,V data to 32byte boundary */ |
| 203 | if ((ulOffset & 0x1f) != 0) |
| 204 | ulOffset = (ulOffset + 32L) & 0xffffffE0L; |
| 205 | |
| 206 | tmp = STG_READ_REG(DACOverlayUAddr); |
| 207 | CLEAR_BITS_FRM_TO(0, 20); |
| 208 | tmp |= (ulOffset >> 4); |
| 209 | STG_WRITE_REG(DACOverlayUAddr, tmp); |
| 210 | |
| 211 | ulOffset += (inHeight / 2) * (uvStride * 16); |
| 212 | /* Align U,V data to 32byte boundary */ |
| 213 | if ((ulOffset & 0x1f) != 0) |
| 214 | ulOffset = (ulOffset + 32L) & 0xffffffE0L; |
| 215 | |
| 216 | tmp = STG_READ_REG(DACOverlayVAddr); |
| 217 | CLEAR_BITS_FRM_TO(0, 20); |
| 218 | tmp |= (ulOffset >> 4); |
| 219 | STG_WRITE_REG(DACOverlayVAddr, tmp); |
| 220 | |
| 221 | *retUVStride = uvStride * 16; |
| 222 | } |
| 223 | |
| 224 | |
| 225 | /* Set Overlay YUV pixel format |
| 226 | * Make sure that LUT not used - ?????? |
| 227 | */ |
| 228 | tmp = STG_READ_REG(DACPixelFormat); |
| 229 | /* Only support Planer or UYVY linear formats */ |
| 230 | CLEAR_BITS_FRM_TO(4, 9); |
| 231 | STG_WRITE_REG(DACPixelFormat, tmp); |
| 232 | |
| 233 | ovlWidth = inWidth; |
| 234 | ovlHeight = inHeight; |
| 235 | ovlStride = ulStride; |
| 236 | ovlLinear = bLinear; |
| 237 | *retStride = ulStride << 4; /* In bytes */ |
| 238 | |
| 239 | return 0; |
| 240 | } |
| 241 | |
| 242 | int SetOverlayBlendMode(volatile STG4000REG __iomem *pSTGReg, |
| 243 | OVRL_BLEND_MODE mode, |
| 244 | u32 ulAlpha, u32 ulColorKey) |
| 245 | { |
| 246 | u32 tmp; |
| 247 | |
| 248 | tmp = STG_READ_REG(DACBlendCtrl); |
| 249 | CLEAR_BITS_FRM_TO(28, 30); |
| 250 | tmp |= (mode << 28); |
| 251 | |
| 252 | switch (mode) { |
| 253 | case COLOR_KEY: |
| 254 | CLEAR_BITS_FRM_TO(0, 23); |
| 255 | tmp |= (ulColorKey & 0x00FFFFFF); |
| 256 | break; |
| 257 | |
| 258 | case GLOBAL_ALPHA: |
| 259 | CLEAR_BITS_FRM_TO(24, 27); |
| 260 | tmp |= ((ulAlpha & 0xF) << 24); |
| 261 | break; |
| 262 | |
| 263 | case CK_PIXEL_ALPHA: |
| 264 | CLEAR_BITS_FRM_TO(0, 23); |
| 265 | tmp |= (ulColorKey & 0x00FFFFFF); |
| 266 | break; |
| 267 | |
| 268 | case CK_GLOBAL_ALPHA: |
| 269 | CLEAR_BITS_FRM_TO(0, 23); |
| 270 | tmp |= (ulColorKey & 0x00FFFFFF); |
| 271 | CLEAR_BITS_FRM_TO(24, 27); |
| 272 | tmp |= ((ulAlpha & 0xF) << 24); |
| 273 | break; |
| 274 | |
| 275 | case GRAPHICS_MODE: |
| 276 | case PER_PIXEL_ALPHA: |
| 277 | break; |
| 278 | |
| 279 | default: |
| 280 | return -EINVAL; |
| 281 | } |
| 282 | |
| 283 | STG_WRITE_REG(DACBlendCtrl, tmp); |
| 284 | |
| 285 | return 0; |
| 286 | } |
| 287 | |
| 288 | void EnableOverlayPlane(volatile STG4000REG __iomem *pSTGReg) |
| 289 | { |
| 290 | u32 tmp; |
| 291 | /* Enable Overlay */ |
| 292 | tmp = STG_READ_REG(DACPixelFormat); |
| 293 | tmp |= SET_BIT(7); |
| 294 | STG_WRITE_REG(DACPixelFormat, tmp); |
| 295 | |
| 296 | /* Set video stream control */ |
| 297 | tmp = STG_READ_REG(DACStreamCtrl); |
| 298 | tmp |= SET_BIT(1); /* video stream */ |
| 299 | STG_WRITE_REG(DACStreamCtrl, tmp); |
| 300 | } |
| 301 | |
| 302 | static u32 Overlap(u32 ulBits, u32 ulPattern) |
| 303 | { |
| 304 | u32 ulCount = 0; |
| 305 | |
| 306 | while (ulBits) { |
| 307 | if (!(ulPattern & 1)) |
| 308 | ulCount++; |
| 309 | ulBits--; |
| 310 | ulPattern = ulPattern >> 1; |
| 311 | } |
| 312 | |
| 313 | return ulCount; |
| 314 | |
| 315 | } |
| 316 | |
| 317 | int SetOverlayViewPort(volatile STG4000REG __iomem *pSTGReg, |
| 318 | u32 left, u32 top, |
| 319 | u32 right, u32 bottom) |
| 320 | { |
| 321 | OVRL_SRC_DEST srcDest; |
| 322 | |
| 323 | u32 ulSrcTop, ulSrcBottom; |
| 324 | u32 ulSrc, ulDest; |
| 325 | u32 ulFxScale, ulFxOffset; |
| 326 | u32 ulHeight, ulWidth; |
| 327 | u32 ulPattern; |
| 328 | u32 ulDecimate, ulDecimated; |
| 329 | u32 ulApplied; |
| 330 | u32 ulDacXScale, ulDacYScale; |
| 331 | u32 ulScale; |
| 332 | u32 ulLeft, ulRight; |
| 333 | u32 ulSrcLeft, ulSrcRight; |
| 334 | u32 ulScaleLeft, ulScaleRight; |
| 335 | u32 ulhDecim; |
| 336 | u32 ulsVal; |
| 337 | u32 ulVertDecFactor; |
| 338 | int bResult; |
| 339 | u32 ulClipOff = 0; |
| 340 | u32 ulBits = 0; |
| 341 | u32 ulsAdd = 0; |
| 342 | u32 tmp, ulStride; |
| 343 | u32 ulExcessPixels, ulClip, ulExtraLines; |
| 344 | |
| 345 | |
| 346 | srcDest.ulSrcX1 = 0; |
| 347 | srcDest.ulSrcY1 = 0; |
| 348 | srcDest.ulSrcX2 = ovlWidth - 1; |
| 349 | srcDest.ulSrcY2 = ovlHeight - 1; |
| 350 | |
| 351 | srcDest.ulDstX1 = left; |
| 352 | srcDest.ulDstY1 = top; |
| 353 | srcDest.ulDstX2 = right; |
| 354 | srcDest.ulDstY2 = bottom; |
| 355 | |
| 356 | srcDest.lDstX1 = srcDest.ulDstX1; |
| 357 | srcDest.lDstY1 = srcDest.ulDstY1; |
| 358 | srcDest.lDstX2 = srcDest.ulDstX2; |
| 359 | srcDest.lDstY2 = srcDest.ulDstY2; |
| 360 | |
| 361 | /************* Vertical decimation/scaling ******************/ |
| 362 | |
| 363 | /* Get Src Top and Bottom */ |
| 364 | ulSrcTop = srcDest.ulSrcY1; |
| 365 | ulSrcBottom = srcDest.ulSrcY2; |
| 366 | |
| 367 | ulSrc = ulSrcBottom - ulSrcTop; |
| 368 | ulDest = srcDest.lDstY2 - srcDest.lDstY1; /* on-screen overlay */ |
| 369 | |
| 370 | if (ulSrc <= 1) |
| 371 | return -EINVAL; |
| 372 | |
| 373 | /* First work out the position we are to display as offset from the |
| 374 | * source of the buffer |
| 375 | */ |
| 376 | ulFxScale = (ulDest << 11) / ulSrc; /* fixed point scale factor */ |
| 377 | ulFxOffset = (srcDest.lDstY2 - srcDest.ulDstY2) << 11; |
| 378 | |
| 379 | ulSrcBottom = ulSrcBottom - (ulFxOffset / ulFxScale); |
| 380 | ulSrc = ulSrcBottom - ulSrcTop; |
| 381 | ulHeight = ulSrc; |
| 382 | |
| 383 | ulDest = srcDest.ulDstY2 - (srcDest.ulDstY1 - 1); |
| 384 | ulPattern = adwDecim8[ulBits]; |
| 385 | |
| 386 | /* At this point ulSrc represents the input decimator */ |
| 387 | if (ulSrc > ulDest) { |
| 388 | ulDecimate = ulSrc - ulDest; |
| 389 | ulBits = 0; |
| 390 | ulApplied = ulSrc / 32; |
| 391 | |
| 392 | while (((ulBits * ulApplied) + |
| 393 | Overlap((ulSrc % 32), |
| 394 | adwDecim8[ulBits])) < ulDecimate) |
| 395 | ulBits++; |
| 396 | |
| 397 | ulPattern = adwDecim8[ulBits]; |
| 398 | ulDecimated = |
| 399 | (ulBits * ulApplied) + Overlap((ulSrc % 32), |
| 400 | ulPattern); |
| 401 | ulSrc = ulSrc - ulDecimated; /* the number number of lines that will go into the scaler */ |
| 402 | } |
| 403 | |
| 404 | if (ulBits && (ulBits != 32)) { |
| 405 | ulVertDecFactor = (63 - ulBits) / (32 - ulBits); /* vertical decimation factor scaled up to nearest integer */ |
| 406 | } else { |
| 407 | ulVertDecFactor = 1; |
| 408 | } |
| 409 | |
| 410 | ulDacYScale = ((ulSrc - 1) * 2048) / (ulDest + 1); |
| 411 | |
| 412 | tmp = STG_READ_REG(DACOverlayVtDec); /* Decimation */ |
| 413 | CLEAR_BITS_FRM_TO(0, 31); |
| 414 | tmp = ulPattern; |
| 415 | STG_WRITE_REG(DACOverlayVtDec, tmp); |
| 416 | |
| 417 | /***************** Horizontal decimation/scaling ***************************/ |
| 418 | |
| 419 | /* |
Lucas De Marchi | 25985ed | 2011-03-30 22:57:33 -0300 | [diff] [blame] | 420 | * Now we handle the horizontal case, this is a simplified version of |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 421 | * the vertical case in that we decimate by factors of 2. as we are |
| 422 | * working in words we should always be able to decimate by these |
| 423 | * factors. as we always have to have a buffer which is aligned to a |
| 424 | * whole number of 128 bit words, we must align the left side to the |
| 425 | * lowest to the next lowest 128 bit boundary, and the right hand edge |
| 426 | * to the next largets boundary, (in a similar way to how we didi it in |
| 427 | * PMX1) as the left and right hand edges are aligned to these |
| 428 | * boundaries normally this only becomes an issue when we are chopping |
| 429 | * of one of the sides We shall work out vertical stuff first |
| 430 | */ |
| 431 | ulSrc = srcDest.ulSrcX2 - srcDest.ulSrcX1; |
| 432 | ulDest = srcDest.lDstX2 - srcDest.lDstX1; |
| 433 | #ifdef _OLDCODE |
| 434 | ulLeft = srcDest.ulDstX1; |
| 435 | ulRight = srcDest.ulDstX2; |
| 436 | #else |
| 437 | if (srcDest.ulDstX1 > 2) { |
| 438 | ulLeft = srcDest.ulDstX1 + 2; |
| 439 | ulRight = srcDest.ulDstX2 + 1; |
| 440 | } else { |
| 441 | ulLeft = srcDest.ulDstX1; |
| 442 | ulRight = srcDest.ulDstX2 + 1; |
| 443 | } |
| 444 | #endif |
| 445 | /* first work out the position we are to display as offset from the source of the buffer */ |
| 446 | bResult = 1; |
| 447 | |
| 448 | do { |
| 449 | if (ulDest == 0) |
| 450 | return -EINVAL; |
| 451 | |
| 452 | /* source pixels per dest pixel <<11 */ |
| 453 | ulFxScale = ((ulSrc - 1) << 11) / (ulDest); |
| 454 | |
| 455 | /* then number of destination pixels out we are */ |
| 456 | ulFxOffset = ulFxScale * ((srcDest.ulDstX1 - srcDest.lDstX1) + ulClipOff); |
| 457 | ulFxOffset >>= 11; |
| 458 | |
| 459 | /* this replaces the code which was making a decision as to use either ulFxOffset or ulSrcX1 */ |
| 460 | ulSrcLeft = srcDest.ulSrcX1 + ulFxOffset; |
| 461 | |
| 462 | /* then number of destination pixels out we are */ |
| 463 | ulFxOffset = ulFxScale * (srcDest.lDstX2 - srcDest.ulDstX2); |
| 464 | ulFxOffset >>= 11; |
| 465 | |
| 466 | ulSrcRight = srcDest.ulSrcX2 - ulFxOffset; |
| 467 | |
| 468 | /* |
| 469 | * we must align these to our 128 bit boundaries. we shall |
| 470 | * round down the pixel pos to the nearest 8 pixels. |
| 471 | */ |
| 472 | ulScaleLeft = ulSrcLeft; |
| 473 | ulScaleRight = ulSrcRight; |
| 474 | |
| 475 | /* shift fxscale until it is in the range of the scaler */ |
| 476 | ulhDecim = 0; |
| 477 | ulScale = (((ulSrcRight - ulSrcLeft) - 1) << (11 - ulhDecim)) / (ulRight - ulLeft + 2); |
| 478 | |
| 479 | while (ulScale > 0x800) { |
| 480 | ulhDecim++; |
| 481 | ulScale = (((ulSrcRight - ulSrcLeft) - 1) << (11 - ulhDecim)) / (ulRight - ulLeft + 2); |
| 482 | } |
| 483 | |
| 484 | /* |
| 485 | * to try and get the best values We first try and use |
| 486 | * src/dwdest for the scale factor, then we move onto src-1 |
| 487 | * |
| 488 | * we want to check to see if we will need to clip data, if so |
| 489 | * then we should clip our source so that we don't need to |
| 490 | */ |
| 491 | if (!ovlLinear) { |
| 492 | ulSrcLeft &= ~0x1f; |
| 493 | |
| 494 | /* |
| 495 | * we must align the right hand edge to the next 32 |
| 496 | * pixel` boundary, must be on a 256 boundary so u, and |
| 497 | * v are 128 bit aligned |
| 498 | */ |
| 499 | ulSrcRight = (ulSrcRight + 0x1f) & ~0x1f; |
| 500 | } else { |
| 501 | ulSrcLeft &= ~0x7; |
| 502 | |
| 503 | /* |
| 504 | * we must align the right hand edge to the next |
| 505 | * 8pixel` boundary |
| 506 | */ |
| 507 | ulSrcRight = (ulSrcRight + 0x7) & ~0x7; |
| 508 | } |
| 509 | |
| 510 | /* this is the input size line store needs to cope with */ |
| 511 | ulWidth = ulSrcRight - ulSrcLeft; |
| 512 | |
| 513 | /* |
| 514 | * use unclipped value to work out scale factror this is the |
| 515 | * scale factor we want we shall now work out the horizonal |
| 516 | * decimation and scaling |
| 517 | */ |
| 518 | ulsVal = ((ulWidth / 8) >> ulhDecim); |
| 519 | |
| 520 | if ((ulWidth != (ulsVal << ulhDecim) * 8)) |
| 521 | ulsAdd = 1; |
| 522 | |
| 523 | /* input pixels to scaler; */ |
| 524 | ulSrc = ulWidth >> ulhDecim; |
| 525 | |
| 526 | if (ulSrc <= 2) |
| 527 | return -EINVAL; |
| 528 | |
| 529 | ulExcessPixels = ((((ulScaleLeft - ulSrcLeft)) << (11 - ulhDecim)) / ulScale); |
| 530 | |
| 531 | ulClip = (ulSrc << 11) / ulScale; |
| 532 | ulClip -= (ulRight - ulLeft); |
| 533 | ulClip += ulExcessPixels; |
| 534 | |
| 535 | if (ulClip) |
| 536 | ulClip--; |
| 537 | |
| 538 | /* We may need to do more here if we really have a HW rev < 5 */ |
| 539 | } while (!bResult); |
| 540 | |
| 541 | ulExtraLines = (1 << ulhDecim) * ulVertDecFactor; |
| 542 | ulExtraLines += 64; |
| 543 | ulHeight += ulExtraLines; |
| 544 | |
| 545 | ulDacXScale = ulScale; |
| 546 | |
| 547 | |
| 548 | tmp = STG_READ_REG(DACVerticalScal); |
| 549 | CLEAR_BITS_FRM_TO(0, 11); |
| 550 | CLEAR_BITS_FRM_TO(16, 22); /* Vertical Scaling */ |
| 551 | |
| 552 | /* Calculate new output line stride, this is always the number of 422 |
| 553 | words in the line buffer, so it doesn't matter if the |
| 554 | mode is 420. Then set the vertical scale register. |
| 555 | */ |
| 556 | ulStride = (ulWidth >> (ulhDecim + 3)) + ulsAdd; |
| 557 | tmp |= ((ulStride << 16) | (ulDacYScale)); /* DAC_LS_CTRL = stride */ |
| 558 | STG_WRITE_REG(DACVerticalScal, tmp); |
| 559 | |
| 560 | /* Now set up the overlay size using the modified width and height |
| 561 | from decimate and scaling calculations |
| 562 | */ |
| 563 | tmp = STG_READ_REG(DACOverlaySize); |
| 564 | CLEAR_BITS_FRM_TO(0, 10); |
| 565 | CLEAR_BITS_FRM_TO(12, 31); |
| 566 | |
| 567 | if (ovlLinear) { |
| 568 | tmp |= |
| 569 | (ovlStride | ((ulHeight + 1) << 12) | |
| 570 | (((ulWidth / 8) - 1) << 23)); |
| 571 | } else { |
| 572 | tmp |= |
| 573 | (ovlStride | ((ulHeight + 1) << 12) | |
| 574 | (((ulWidth / 32) - 1) << 23)); |
| 575 | } |
| 576 | |
| 577 | STG_WRITE_REG(DACOverlaySize, tmp); |
| 578 | |
| 579 | /* Set Video Window Start */ |
| 580 | tmp = ((ulLeft << 16)) | (srcDest.ulDstY1); |
| 581 | STG_WRITE_REG(DACVidWinStart, tmp); |
| 582 | |
| 583 | /* Set Video Window End */ |
| 584 | tmp = ((ulRight) << 16) | (srcDest.ulDstY2); |
| 585 | STG_WRITE_REG(DACVidWinEnd, tmp); |
| 586 | |
| 587 | /* Finally set up the rest of the overlay regs in the order |
| 588 | done in the IMG driver |
| 589 | */ |
| 590 | tmp = STG_READ_REG(DACPixelFormat); |
| 591 | tmp = ((ulExcessPixels << 16) | tmp) & 0x7fffffff; |
| 592 | STG_WRITE_REG(DACPixelFormat, tmp); |
| 593 | |
| 594 | tmp = STG_READ_REG(DACHorizontalScal); |
| 595 | CLEAR_BITS_FRM_TO(0, 11); |
| 596 | CLEAR_BITS_FRM_TO(16, 17); |
| 597 | tmp |= ((ulhDecim << 16) | (ulDacXScale)); |
| 598 | STG_WRITE_REG(DACHorizontalScal, tmp); |
| 599 | |
| 600 | return 0; |
| 601 | } |