2 * linux/drivers/video/kyro/STG4000OverlayDevice.c
4 * Copyright (C) 2000 Imagination Technologies Ltd
5 * Copyright (C) 2002 STMicroelectronics
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
12 #include <linux/kernel.h>
13 #include <linux/errno.h>
14 #include <linux/types.h>
16 #include "STG4000Reg.h"
20 #define STG4000_NO_SCALING 0x800
21 #define STG4000_NO_DECIMATION 0xFFFFFFFF
24 #define STG4000_PRIM_NUM_PIX 5
25 #define STG4000_PRIM_ALIGN 4
26 #define STG4000_PRIM_ADDR_BITS 20
28 #define STG4000_PRIM_MIN_WIDTH 640
29 #define STG4000_PRIM_MAX_WIDTH 1600
30 #define STG4000_PRIM_MIN_HEIGHT 480
31 #define STG4000_PRIM_MAX_HEIGHT 1200
34 #define STG4000_OVRL_NUM_PIX 4
35 #define STG4000_OVRL_ALIGN 2
36 #define STG4000_OVRL_ADDR_BITS 20
37 #define STG4000_OVRL_NUM_MODES 5
39 #define STG4000_OVRL_MIN_WIDTH 0
40 #define STG4000_OVRL_MAX_WIDTH 720
41 #define STG4000_OVRL_MIN_HEIGHT 0
42 #define STG4000_OVRL_MAX_HEIGHT 576
44 /* Decimation and Scaling */
45 static u32 adwDecim8
[33] = {
46 0xffffffff, 0xfffeffff, 0xffdffbff, 0xfefefeff, 0xfdf7efbf,
47 0xfbdf7bdf, 0xf7bbddef, 0xeeeeeeef, 0xeeddbb77, 0xedb76db7,
48 0xdb6db6db, 0xdb5b5b5b, 0xdab5ad6b, 0xd5ab55ab, 0xd555aaab,
49 0xaaaaaaab, 0xaaaa5555, 0xaa952a55, 0xa94a5295, 0xa5252525,
50 0xa4924925, 0x92491249, 0x91224489, 0x91111111, 0x90884211,
51 0x88410821, 0x88102041, 0x81010101, 0x80800801, 0x80010001,
52 0x80000001, 0x00000001, 0x00000000
55 typedef struct _OVRL_SRC_DEST
{
56 /*clipped on-screen pixel position of overlay */
62 /*clipped pixel pos of source data within buffer thses need to be 128 bit word aligned */
68 /* on-screen pixel position of overlay */
75 static u32 ovlWidth
, ovlHeight
, ovlStride
;
78 void ResetOverlayRegisters(volatile STG4000REG __iomem
*pSTGReg
)
82 /* Set Overlay address to default */
83 tmp
= STG_READ_REG(DACOverlayAddr
);
84 CLEAR_BITS_FRM_TO(0, 20);
86 STG_WRITE_REG(DACOverlayAddr
, tmp
);
88 /* Set Overlay U address */
89 tmp
= STG_READ_REG(DACOverlayUAddr
);
90 CLEAR_BITS_FRM_TO(0, 20);
91 STG_WRITE_REG(DACOverlayUAddr
, tmp
);
93 /* Set Overlay V address */
94 tmp
= STG_READ_REG(DACOverlayVAddr
);
95 CLEAR_BITS_FRM_TO(0, 20);
96 STG_WRITE_REG(DACOverlayVAddr
, tmp
);
98 /* Set Overlay Size */
99 tmp
= STG_READ_REG(DACOverlaySize
);
100 CLEAR_BITS_FRM_TO(0, 10);
101 CLEAR_BITS_FRM_TO(12, 31);
102 STG_WRITE_REG(DACOverlaySize
, tmp
);
104 /* Set Overlay Vt Decimation */
105 tmp
= STG4000_NO_DECIMATION
;
106 STG_WRITE_REG(DACOverlayVtDec
, tmp
);
108 /* Set Overlay format to default value */
109 tmp
= STG_READ_REG(DACPixelFormat
);
110 CLEAR_BITS_FRM_TO(4, 7);
111 CLEAR_BITS_FRM_TO(16, 22);
112 STG_WRITE_REG(DACPixelFormat
, tmp
);
114 /* Set Vertical scaling to default */
115 tmp
= STG_READ_REG(DACVerticalScal
);
116 CLEAR_BITS_FRM_TO(0, 11);
117 CLEAR_BITS_FRM_TO(16, 22);
118 tmp
|= STG4000_NO_SCALING
; /* Set to no scaling */
119 STG_WRITE_REG(DACVerticalScal
, tmp
);
121 /* Set Horizontal Scaling to default */
122 tmp
= STG_READ_REG(DACHorizontalScal
);
123 CLEAR_BITS_FRM_TO(0, 11);
124 CLEAR_BITS_FRM_TO(16, 17);
125 tmp
|= STG4000_NO_SCALING
; /* Set to no scaling */
126 STG_WRITE_REG(DACHorizontalScal
, tmp
);
128 /* Set Blend mode to Alpha Blend */
129 /* ????? SG 08/11/2001 Surely this isn't the alpha blend mode,
130 hopefully its overwrite
132 tmp
= STG_READ_REG(DACBlendCtrl
);
133 CLEAR_BITS_FRM_TO(0, 30);
134 tmp
= (GRAPHICS_MODE
<< 28);
135 STG_WRITE_REG(DACBlendCtrl
, tmp
);
139 int CreateOverlaySurface(volatile STG4000REG __iomem
*pSTGReg
,
144 u32
* retStride
, u32
* retUVStride
)
149 if (inWidth
> STG4000_OVRL_MAX_WIDTH
||
150 inHeight
> STG4000_OVRL_MAX_HEIGHT
) {
154 /* Stride in 16 byte words - 16Bpp */
156 /* Format is 16bits so num 16 byte words is width/8 */
157 if ((inWidth
& 0x7) == 0) { /* inWidth % 8 */
158 ulStride
= (inWidth
/ 8);
160 /* Round up to next 16byte boundary */
161 ulStride
= ((inWidth
+ 8) / 8);
164 /* Y component is 8bits so num 16 byte words is width/16 */
165 if ((inWidth
& 0xf) == 0) { /* inWidth % 16 */
166 ulStride
= (inWidth
/ 16);
168 /* Round up to next 16byte boundary */
169 ulStride
= ((inWidth
+ 16) / 16);
174 /* Set Overlay address and Format mode */
175 tmp
= STG_READ_REG(DACOverlayAddr
);
176 CLEAR_BITS_FRM_TO(0, 20);
178 CLEAR_BIT(31); /* Overlay format to Linear */
180 tmp
|= SET_BIT(31); /* Overlay format to Planer */
183 /* Only bits 24:4 of the Overlay address */
184 tmp
|= (ulOverlayOffset
>> 4);
185 STG_WRITE_REG(DACOverlayAddr
, tmp
);
189 (inWidth
& 0x1) ? (inWidth
+ 1 / 2) : (inWidth
/ 2);
192 /* Y component is 8bits so num 32 byte words is width/32 */
193 if ((uvSize
& 0xf) == 0) { /* inWidth % 16 */
194 uvStride
= (uvSize
/ 16);
196 /* Round up to next 32byte boundary */
197 uvStride
= ((uvSize
+ 16) / 16);
200 ulOffset
= ulOverlayOffset
+ (inHeight
* (ulStride
* 16));
201 /* Align U,V data to 32byte boundary */
202 if ((ulOffset
& 0x1f) != 0)
203 ulOffset
= (ulOffset
+ 32L) & 0xffffffE0L
;
205 tmp
= STG_READ_REG(DACOverlayUAddr
);
206 CLEAR_BITS_FRM_TO(0, 20);
207 tmp
|= (ulOffset
>> 4);
208 STG_WRITE_REG(DACOverlayUAddr
, tmp
);
210 ulOffset
+= (inHeight
/ 2) * (uvStride
* 16);
211 /* Align U,V data to 32byte boundary */
212 if ((ulOffset
& 0x1f) != 0)
213 ulOffset
= (ulOffset
+ 32L) & 0xffffffE0L
;
215 tmp
= STG_READ_REG(DACOverlayVAddr
);
216 CLEAR_BITS_FRM_TO(0, 20);
217 tmp
|= (ulOffset
>> 4);
218 STG_WRITE_REG(DACOverlayVAddr
, tmp
);
220 *retUVStride
= uvStride
* 16;
224 /* Set Overlay YUV pixel format
225 * Make sure that LUT not used - ??????
227 tmp
= STG_READ_REG(DACPixelFormat
);
228 /* Only support Planer or UYVY linear formats */
229 CLEAR_BITS_FRM_TO(4, 9);
230 STG_WRITE_REG(DACPixelFormat
, tmp
);
233 ovlHeight
= inHeight
;
234 ovlStride
= ulStride
;
236 *retStride
= ulStride
<< 4; /* In bytes */
241 int SetOverlayBlendMode(volatile STG4000REG __iomem
*pSTGReg
,
242 OVRL_BLEND_MODE mode
,
243 u32 ulAlpha
, u32 ulColorKey
)
247 tmp
= STG_READ_REG(DACBlendCtrl
);
248 CLEAR_BITS_FRM_TO(28, 30);
253 CLEAR_BITS_FRM_TO(0, 23);
254 tmp
|= (ulColorKey
& 0x00FFFFFF);
258 CLEAR_BITS_FRM_TO(24, 27);
259 tmp
|= ((ulAlpha
& 0xF) << 24);
263 CLEAR_BITS_FRM_TO(0, 23);
264 tmp
|= (ulColorKey
& 0x00FFFFFF);
267 case CK_GLOBAL_ALPHA
:
268 CLEAR_BITS_FRM_TO(0, 23);
269 tmp
|= (ulColorKey
& 0x00FFFFFF);
270 CLEAR_BITS_FRM_TO(24, 27);
271 tmp
|= ((ulAlpha
& 0xF) << 24);
275 case PER_PIXEL_ALPHA
:
282 STG_WRITE_REG(DACBlendCtrl
, tmp
);
287 void EnableOverlayPlane(volatile STG4000REG __iomem
*pSTGReg
)
291 tmp
= STG_READ_REG(DACPixelFormat
);
293 STG_WRITE_REG(DACPixelFormat
, tmp
);
295 /* Set video stream control */
296 tmp
= STG_READ_REG(DACStreamCtrl
);
297 tmp
|= SET_BIT(1); /* video stream */
298 STG_WRITE_REG(DACStreamCtrl
, tmp
);
301 static u32
Overlap(u32 ulBits
, u32 ulPattern
)
306 if (!(ulPattern
& 1))
309 ulPattern
= ulPattern
>> 1;
316 int SetOverlayViewPort(volatile STG4000REG __iomem
*pSTGReg
,
318 u32 right
, u32 bottom
)
320 OVRL_SRC_DEST srcDest
;
322 u32 ulSrcTop
, ulSrcBottom
;
324 u32 ulFxScale
, ulFxOffset
;
325 u32 ulHeight
, ulWidth
;
327 u32 ulDecimate
, ulDecimated
;
329 u32 ulDacXScale
, ulDacYScale
;
332 u32 ulSrcLeft
, ulSrcRight
;
333 u32 ulScaleLeft
, ulScaleRight
;
342 u32 ulExcessPixels
, ulClip
, ulExtraLines
;
347 srcDest
.ulSrcX2
= ovlWidth
- 1;
348 srcDest
.ulSrcY2
= ovlHeight
- 1;
350 srcDest
.ulDstX1
= left
;
351 srcDest
.ulDstY1
= top
;
352 srcDest
.ulDstX2
= right
;
353 srcDest
.ulDstY2
= bottom
;
355 srcDest
.lDstX1
= srcDest
.ulDstX1
;
356 srcDest
.lDstY1
= srcDest
.ulDstY1
;
357 srcDest
.lDstX2
= srcDest
.ulDstX2
;
358 srcDest
.lDstY2
= srcDest
.ulDstY2
;
360 /************* Vertical decimation/scaling ******************/
362 /* Get Src Top and Bottom */
363 ulSrcTop
= srcDest
.ulSrcY1
;
364 ulSrcBottom
= srcDest
.ulSrcY2
;
366 ulSrc
= ulSrcBottom
- ulSrcTop
;
367 ulDest
= srcDest
.lDstY2
- srcDest
.lDstY1
; /* on-screen overlay */
372 /* First work out the position we are to display as offset from the
373 * source of the buffer
375 ulFxScale
= (ulDest
<< 11) / ulSrc
; /* fixed point scale factor */
376 ulFxOffset
= (srcDest
.lDstY2
- srcDest
.ulDstY2
) << 11;
378 ulSrcBottom
= ulSrcBottom
- (ulFxOffset
/ ulFxScale
);
379 ulSrc
= ulSrcBottom
- ulSrcTop
;
382 ulDest
= srcDest
.ulDstY2
- (srcDest
.ulDstY1
- 1);
383 ulPattern
= adwDecim8
[ulBits
];
385 /* At this point ulSrc represents the input decimator */
386 if (ulSrc
> ulDest
) {
387 ulDecimate
= ulSrc
- ulDest
;
389 ulApplied
= ulSrc
/ 32;
391 while (((ulBits
* ulApplied
) +
392 Overlap((ulSrc
% 32),
393 adwDecim8
[ulBits
])) < ulDecimate
)
396 ulPattern
= adwDecim8
[ulBits
];
398 (ulBits
* ulApplied
) + Overlap((ulSrc
% 32),
400 ulSrc
= ulSrc
- ulDecimated
; /* the number number of lines that will go into the scaler */
403 if (ulBits
&& (ulBits
!= 32)) {
404 ulVertDecFactor
= (63 - ulBits
) / (32 - ulBits
); /* vertical decimation factor scaled up to nearest integer */
409 ulDacYScale
= ((ulSrc
- 1) * 2048) / (ulDest
+ 1);
411 tmp
= STG_READ_REG(DACOverlayVtDec
); /* Decimation */
412 CLEAR_BITS_FRM_TO(0, 31);
414 STG_WRITE_REG(DACOverlayVtDec
, tmp
);
416 /***************** Horizontal decimation/scaling ***************************/
419 * Now we handle the horizontal case, this is a simplified verison of
420 * the vertical case in that we decimate by factors of 2. as we are
421 * working in words we should always be able to decimate by these
422 * factors. as we always have to have a buffer which is aligned to a
423 * whole number of 128 bit words, we must align the left side to the
424 * lowest to the next lowest 128 bit boundary, and the right hand edge
425 * to the next largets boundary, (in a similar way to how we didi it in
426 * PMX1) as the left and right hand edges are aligned to these
427 * boundaries normally this only becomes an issue when we are chopping
428 * of one of the sides We shall work out vertical stuff first
430 ulSrc
= srcDest
.ulSrcX2
- srcDest
.ulSrcX1
;
431 ulDest
= srcDest
.lDstX2
- srcDest
.lDstX1
;
433 ulLeft
= srcDest
.ulDstX1
;
434 ulRight
= srcDest
.ulDstX2
;
436 if (srcDest
.ulDstX1
> 2) {
437 ulLeft
= srcDest
.ulDstX1
+ 2;
438 ulRight
= srcDest
.ulDstX2
+ 1;
440 ulLeft
= srcDest
.ulDstX1
;
441 ulRight
= srcDest
.ulDstX2
+ 1;
444 /* first work out the position we are to display as offset from the source of the buffer */
451 /* source pixels per dest pixel <<11 */
452 ulFxScale
= ((ulSrc
- 1) << 11) / (ulDest
);
454 /* then number of destination pixels out we are */
455 ulFxOffset
= ulFxScale
* ((srcDest
.ulDstX1
- srcDest
.lDstX1
) + ulClipOff
);
458 /* this replaces the code which was making a decision as to use either ulFxOffset or ulSrcX1 */
459 ulSrcLeft
= srcDest
.ulSrcX1
+ ulFxOffset
;
461 /* then number of destination pixels out we are */
462 ulFxOffset
= ulFxScale
* (srcDest
.lDstX2
- srcDest
.ulDstX2
);
465 ulSrcRight
= srcDest
.ulSrcX2
- ulFxOffset
;
468 * we must align these to our 128 bit boundaries. we shall
469 * round down the pixel pos to the nearest 8 pixels.
471 ulScaleLeft
= ulSrcLeft
;
472 ulScaleRight
= ulSrcRight
;
474 /* shift fxscale until it is in the range of the scaler */
476 ulScale
= (((ulSrcRight
- ulSrcLeft
) - 1) << (11 - ulhDecim
)) / (ulRight
- ulLeft
+ 2);
478 while (ulScale
> 0x800) {
480 ulScale
= (((ulSrcRight
- ulSrcLeft
) - 1) << (11 - ulhDecim
)) / (ulRight
- ulLeft
+ 2);
484 * to try and get the best values We first try and use
485 * src/dwdest for the scale factor, then we move onto src-1
487 * we want to check to see if we will need to clip data, if so
488 * then we should clip our source so that we don't need to
494 * we must align the right hand edge to the next 32
495 * pixel` boundary, must be on a 256 boundary so u, and
496 * v are 128 bit aligned
498 ulSrcRight
= (ulSrcRight
+ 0x1f) & ~0x1f;
503 * we must align the right hand edge to the next
506 ulSrcRight
= (ulSrcRight
+ 0x7) & ~0x7;
509 /* this is the input size line store needs to cope with */
510 ulWidth
= ulSrcRight
- ulSrcLeft
;
513 * use unclipped value to work out scale factror this is the
514 * scale factor we want we shall now work out the horizonal
515 * decimation and scaling
517 ulsVal
= ((ulWidth
/ 8) >> ulhDecim
);
519 if ((ulWidth
!= (ulsVal
<< ulhDecim
) * 8))
522 /* input pixels to scaler; */
523 ulSrc
= ulWidth
>> ulhDecim
;
528 ulExcessPixels
= ((((ulScaleLeft
- ulSrcLeft
)) << (11 - ulhDecim
)) / ulScale
);
530 ulClip
= (ulSrc
<< 11) / ulScale
;
531 ulClip
-= (ulRight
- ulLeft
);
532 ulClip
+= ulExcessPixels
;
537 /* We may need to do more here if we really have a HW rev < 5 */
540 ulExtraLines
= (1 << ulhDecim
) * ulVertDecFactor
;
542 ulHeight
+= ulExtraLines
;
544 ulDacXScale
= ulScale
;
547 tmp
= STG_READ_REG(DACVerticalScal
);
548 CLEAR_BITS_FRM_TO(0, 11);
549 CLEAR_BITS_FRM_TO(16, 22); /* Vertical Scaling */
551 /* Calculate new output line stride, this is always the number of 422
552 words in the line buffer, so it doesn't matter if the
553 mode is 420. Then set the vertical scale register.
555 ulStride
= (ulWidth
>> (ulhDecim
+ 3)) + ulsAdd
;
556 tmp
|= ((ulStride
<< 16) | (ulDacYScale
)); /* DAC_LS_CTRL = stride */
557 STG_WRITE_REG(DACVerticalScal
, tmp
);
559 /* Now set up the overlay size using the modified width and height
560 from decimate and scaling calculations
562 tmp
= STG_READ_REG(DACOverlaySize
);
563 CLEAR_BITS_FRM_TO(0, 10);
564 CLEAR_BITS_FRM_TO(12, 31);
568 (ovlStride
| ((ulHeight
+ 1) << 12) |
569 (((ulWidth
/ 8) - 1) << 23));
572 (ovlStride
| ((ulHeight
+ 1) << 12) |
573 (((ulWidth
/ 32) - 1) << 23));
576 STG_WRITE_REG(DACOverlaySize
, tmp
);
578 /* Set Video Window Start */
579 tmp
= ((ulLeft
<< 16)) | (srcDest
.ulDstY1
);
580 STG_WRITE_REG(DACVidWinStart
, tmp
);
582 /* Set Video Window End */
583 tmp
= ((ulRight
) << 16) | (srcDest
.ulDstY2
);
584 STG_WRITE_REG(DACVidWinEnd
, tmp
);
586 /* Finally set up the rest of the overlay regs in the order
587 done in the IMG driver
589 tmp
= STG_READ_REG(DACPixelFormat
);
590 tmp
= ((ulExcessPixels
<< 16) | tmp
) & 0x7fffffff;
591 STG_WRITE_REG(DACPixelFormat
, tmp
);
593 tmp
= STG_READ_REG(DACHorizontalScal
);
594 CLEAR_BITS_FRM_TO(0, 11);
595 CLEAR_BITS_FRM_TO(16, 17);
596 tmp
|= ((ulhDecim
<< 16) | (ulDacXScale
));
597 STG_WRITE_REG(DACHorizontalScal
, tmp
);