2 * Mesa 3-D graphics library
4 * Copyright (C) 1999-2008 Brian Paul All Rights Reserved.
5 * Copyright (C) 2009-2010 VMware, Inc. All Rights Reserved.
7 * Permission is hereby granted, free of charge, to any person obtaining a
8 * copy of this software and associated documentation files (the "Software"),
9 * to deal in the Software without restriction, including without limitation
10 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
11 * and/or sell copies of the Software, and to permit persons to whom the
12 * Software is furnished to do so, subject to the following conditions:
14 * The above copyright notice and this permission notice shall be included
15 * in all copies or substantial portions of the Software.
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
18 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
19 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
20 * THEA AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
21 * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
22 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
28 * Image and pixel span packing and unpacking.
33 * XXX: MSVC takes forever to compile this module for x86_64 unless we disable
34 * this global optimization.
37 * - http://msdn.microsoft.com/en-us/library/1yk3ydd7.aspx
38 * - http://msdn.microsoft.com/en-us/library/chh3fb0k.aspx
40 #if defined(_MSC_VER) && defined(_M_X64)
41 # pragma optimize( "g", off )
52 #include "pixeltransfer.h"
54 #include "glformats.h"
55 #include "format_utils.h"
56 #include "format_pack.h"
60 * Flip the 8 bits in each byte of the given array.
63 * \param n number of bytes.
65 * \todo try this trick to flip bytes someday:
67 * v = ((v & 0x55555555) << 1) | ((v >> 1) & 0x55555555);
68 * v = ((v & 0x33333333) << 2) | ((v >> 2) & 0x33333333);
69 * v = ((v & 0x0f0f0f0f) << 4) | ((v >> 4) & 0x0f0f0f0f);
73 flip_bytes( GLubyte *p, GLuint n )
76 for (i = 0; i < n; i++) {
77 b = (GLuint) p[i]; /* words are often faster than bytes */
78 a = ((b & 0x01) << 7) |
93 * Unpack a 32x32 pixel polygon stipple from user memory using the
94 * current pixel unpack settings.
97 _mesa_unpack_polygon_stipple( const GLubyte *pattern, GLuint dest[32],
98 const struct gl_pixelstore_attrib *unpacking )
100 GLubyte *ptrn = (GLubyte *) _mesa_unpack_image(2, 32, 32, 1, GL_COLOR_INDEX,
101 GL_BITMAP, pattern, unpacking);
103 /* Convert pattern from GLubytes to GLuints and handle big/little
108 for (i = 0; i < 32; i++) {
109 dest[i] = (p[0] << 24)
121 * Pack polygon stipple into user memory given current pixel packing
125 _mesa_pack_polygon_stipple( const GLuint pattern[32], GLubyte *dest,
126 const struct gl_pixelstore_attrib *packing )
128 /* Convert pattern from GLuints to GLubytes to handle big/little
129 * endian differences.
133 for (i = 0; i < 32; i++) {
134 ptrn[i * 4 + 0] = (GLubyte) ((pattern[i] >> 24) & 0xff);
135 ptrn[i * 4 + 1] = (GLubyte) ((pattern[i] >> 16) & 0xff);
136 ptrn[i * 4 + 2] = (GLubyte) ((pattern[i] >> 8 ) & 0xff);
137 ptrn[i * 4 + 3] = (GLubyte) ((pattern[i] ) & 0xff);
140 _mesa_pack_bitmap(32, 32, ptrn, dest, packing);
148 _mesa_pack_bitmap( GLint width, GLint height, const GLubyte *source,
149 GLubyte *dest, const struct gl_pixelstore_attrib *packing )
151 GLint row, width_in_bytes;
157 width_in_bytes = DIV_ROUND_UP( width, 8 );
159 for (row = 0; row < height; row++) {
160 GLubyte *dst = (GLubyte *) _mesa_image_address2d(packing, dest,
161 width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0);
165 if ((packing->SkipPixels & 7) == 0) {
166 memcpy( dst, src, width_in_bytes );
167 if (packing->LsbFirst) {
168 flip_bytes( dst, width_in_bytes );
172 /* handling SkipPixels is a bit tricky (no pun intended!) */
174 if (packing->LsbFirst) {
175 GLubyte srcMask = 128;
176 GLubyte dstMask = 1 << (packing->SkipPixels & 0x7);
177 const GLubyte *s = src;
180 for (i = 0; i < width; i++) {
189 srcMask = srcMask >> 1;
191 if (dstMask == 128) {
197 dstMask = dstMask << 1;
202 GLubyte srcMask = 128;
203 GLubyte dstMask = 128 >> (packing->SkipPixels & 0x7);
204 const GLubyte *s = src;
207 for (i = 0; i < width; i++) {
216 srcMask = srcMask >> 1;
224 dstMask = dstMask >> 1;
229 src += width_in_bytes;
234 #define SWAP2BYTE(VALUE) \
236 GLubyte *bytes = (GLubyte *) &(VALUE); \
237 GLubyte tmp = bytes[0]; \
238 bytes[0] = bytes[1]; \
242 #define SWAP4BYTE(VALUE) \
244 GLubyte *bytes = (GLubyte *) &(VALUE); \
245 GLubyte tmp = bytes[0]; \
246 bytes[0] = bytes[3]; \
249 bytes[1] = bytes[2]; \
255 extract_uint_indexes(GLuint n, GLuint indexes[],
256 GLenum srcFormat, GLenum srcType, const GLvoid *src,
257 const struct gl_pixelstore_attrib *unpack )
259 assert(srcFormat == GL_COLOR_INDEX || srcFormat == GL_STENCIL_INDEX);
261 assert(srcType == GL_BITMAP ||
262 srcType == GL_UNSIGNED_BYTE ||
263 srcType == GL_BYTE ||
264 srcType == GL_UNSIGNED_SHORT ||
265 srcType == GL_SHORT ||
266 srcType == GL_UNSIGNED_INT ||
268 srcType == GL_UNSIGNED_INT_24_8_EXT ||
269 srcType == GL_HALF_FLOAT_ARB ||
270 srcType == GL_HALF_FLOAT_OES ||
271 srcType == GL_FLOAT ||
272 srcType == GL_FLOAT_32_UNSIGNED_INT_24_8_REV);
277 GLubyte *ubsrc = (GLubyte *) src;
278 if (unpack->LsbFirst) {
279 GLubyte mask = 1 << (unpack->SkipPixels & 0x7);
281 for (i = 0; i < n; i++) {
282 indexes[i] = (*ubsrc & mask) ? 1 : 0;
293 GLubyte mask = 128 >> (unpack->SkipPixels & 0x7);
295 for (i = 0; i < n; i++) {
296 indexes[i] = (*ubsrc & mask) ? 1 : 0;
308 case GL_UNSIGNED_BYTE:
311 const GLubyte *s = (const GLubyte *) src;
312 for (i = 0; i < n; i++)
319 const GLbyte *s = (const GLbyte *) src;
320 for (i = 0; i < n; i++)
324 case GL_UNSIGNED_SHORT:
327 const GLushort *s = (const GLushort *) src;
328 if (unpack->SwapBytes) {
329 for (i = 0; i < n; i++) {
330 GLushort value = s[i];
336 for (i = 0; i < n; i++)
344 const GLshort *s = (const GLshort *) src;
345 if (unpack->SwapBytes) {
346 for (i = 0; i < n; i++) {
347 GLshort value = s[i];
353 for (i = 0; i < n; i++)
358 case GL_UNSIGNED_INT:
361 const GLuint *s = (const GLuint *) src;
362 if (unpack->SwapBytes) {
363 for (i = 0; i < n; i++) {
370 for (i = 0; i < n; i++)
378 const GLint *s = (const GLint *) src;
379 if (unpack->SwapBytes) {
380 for (i = 0; i < n; i++) {
387 for (i = 0; i < n; i++)
395 const GLfloat *s = (const GLfloat *) src;
396 if (unpack->SwapBytes) {
397 for (i = 0; i < n; i++) {
398 GLfloat value = s[i];
400 indexes[i] = (GLuint) value;
404 for (i = 0; i < n; i++)
405 indexes[i] = (GLuint) s[i];
409 case GL_HALF_FLOAT_ARB:
410 case GL_HALF_FLOAT_OES:
413 const GLhalfARB *s = (const GLhalfARB *) src;
414 if (unpack->SwapBytes) {
415 for (i = 0; i < n; i++) {
416 GLhalfARB value = s[i];
418 indexes[i] = (GLuint) _mesa_half_to_float(value);
422 for (i = 0; i < n; i++)
423 indexes[i] = (GLuint) _mesa_half_to_float(s[i]);
427 case GL_UNSIGNED_INT_24_8_EXT:
430 const GLuint *s = (const GLuint *) src;
431 if (unpack->SwapBytes) {
432 for (i = 0; i < n; i++) {
435 indexes[i] = value & 0xff; /* lower 8 bits */
439 for (i = 0; i < n; i++)
440 indexes[i] = s[i] & 0xff; /* lower 8 bits */
444 case GL_FLOAT_32_UNSIGNED_INT_24_8_REV:
447 const GLuint *s = (const GLuint *) src;
448 if (unpack->SwapBytes) {
449 for (i = 0; i < n; i++) {
450 GLuint value = s[i*2+1];
452 indexes[i] = value & 0xff; /* lower 8 bits */
456 for (i = 0; i < n; i++)
457 indexes[i] = s[i*2+1] & 0xff; /* lower 8 bits */
463 _mesa_problem(NULL, "bad srcType in extract_uint_indexes");
470 clamp_float_to_uint(GLfloat f)
472 return f < 0.0F ? 0 : _mesa_lroundevenf(f);
477 clamp_half_to_uint(GLhalfARB h)
479 GLfloat f = _mesa_half_to_float(h);
480 return f < 0.0F ? 0 : _mesa_lroundevenf(f);
485 * Unpack a row of stencil data from a client buffer according to
486 * the pixel unpacking parameters.
487 * This is (or will be) used by glDrawPixels
489 * Args: ctx - the context
490 * n - number of pixels
491 * dstType - destination data type
492 * dest - destination array
493 * srcType - source pixel type
494 * source - source data pointer
495 * srcPacking - pixel unpacking parameters
496 * transferOps - apply offset/bias/lookup ops?
499 _mesa_unpack_stencil_span( struct gl_context *ctx, GLuint n,
500 GLenum dstType, GLvoid *dest,
501 GLenum srcType, const GLvoid *source,
502 const struct gl_pixelstore_attrib *srcPacking,
503 GLbitfield transferOps )
505 assert(srcType == GL_BITMAP ||
506 srcType == GL_UNSIGNED_BYTE ||
507 srcType == GL_BYTE ||
508 srcType == GL_UNSIGNED_SHORT ||
509 srcType == GL_SHORT ||
510 srcType == GL_UNSIGNED_INT ||
512 srcType == GL_UNSIGNED_INT_24_8_EXT ||
513 srcType == GL_HALF_FLOAT_ARB ||
514 srcType == GL_HALF_FLOAT_OES ||
515 srcType == GL_FLOAT ||
516 srcType == GL_FLOAT_32_UNSIGNED_INT_24_8_REV);
518 assert(dstType == GL_UNSIGNED_BYTE ||
519 dstType == GL_UNSIGNED_SHORT ||
520 dstType == GL_UNSIGNED_INT ||
521 dstType == GL_FLOAT_32_UNSIGNED_INT_24_8_REV);
523 /* only shift and offset apply to stencil */
524 transferOps &= IMAGE_SHIFT_OFFSET_BIT;
527 * Try simple cases first
529 if (transferOps == 0 &&
530 !ctx->Pixel.MapStencilFlag &&
531 srcType == GL_UNSIGNED_BYTE &&
532 dstType == GL_UNSIGNED_BYTE) {
533 memcpy(dest, source, n * sizeof(GLubyte));
535 else if (transferOps == 0 &&
536 !ctx->Pixel.MapStencilFlag &&
537 srcType == GL_UNSIGNED_INT &&
538 dstType == GL_UNSIGNED_INT &&
539 !srcPacking->SwapBytes) {
540 memcpy(dest, source, n * sizeof(GLuint));
546 GLuint *indexes = malloc(n * sizeof(GLuint));
549 _mesa_error(ctx, GL_OUT_OF_MEMORY, "stencil unpacking");
553 extract_uint_indexes(n, indexes, GL_STENCIL_INDEX, srcType, source,
556 if (transferOps & IMAGE_SHIFT_OFFSET_BIT) {
557 /* shift and offset indexes */
558 _mesa_shift_and_offset_ci(ctx, n, indexes);
561 if (ctx->Pixel.MapStencilFlag) {
562 /* Apply stencil lookup table */
563 const GLuint mask = ctx->PixelMaps.StoS.Size - 1;
565 for (i = 0; i < n; i++) {
566 indexes[i] = (GLuint)ctx->PixelMaps.StoS.Map[ indexes[i] & mask ];
570 /* convert to dest type */
572 case GL_UNSIGNED_BYTE:
574 GLubyte *dst = (GLubyte *) dest;
576 for (i = 0; i < n; i++) {
577 dst[i] = (GLubyte) (indexes[i] & 0xff);
581 case GL_UNSIGNED_SHORT:
583 GLuint *dst = (GLuint *) dest;
585 for (i = 0; i < n; i++) {
586 dst[i] = (GLushort) (indexes[i] & 0xffff);
590 case GL_UNSIGNED_INT:
591 memcpy(dest, indexes, n * sizeof(GLuint));
593 case GL_FLOAT_32_UNSIGNED_INT_24_8_REV:
595 GLuint *dst = (GLuint *) dest;
597 for (i = 0; i < n; i++) {
598 dst[i*2+1] = indexes[i] & 0xff; /* lower 8 bits */
603 _mesa_problem(ctx, "bad dstType in _mesa_unpack_stencil_span");
612 _mesa_pack_stencil_span( struct gl_context *ctx, GLuint n,
613 GLenum dstType, GLvoid *dest, const GLubyte *source,
614 const struct gl_pixelstore_attrib *dstPacking )
616 GLubyte *stencil = malloc(n * sizeof(GLubyte));
619 _mesa_error(ctx, GL_OUT_OF_MEMORY, "stencil packing");
623 if (ctx->Pixel.IndexShift || ctx->Pixel.IndexOffset ||
624 ctx->Pixel.MapStencilFlag) {
625 /* make a copy of input */
626 memcpy(stencil, source, n * sizeof(GLubyte));
627 _mesa_apply_stencil_transfer_ops(ctx, n, stencil);
632 case GL_UNSIGNED_BYTE:
633 memcpy(dest, source, n);
637 GLbyte *dst = (GLbyte *) dest;
640 dst[i] = (GLbyte) (source[i] & 0x7f);
644 case GL_UNSIGNED_SHORT:
646 GLushort *dst = (GLushort *) dest;
649 dst[i] = (GLushort) source[i];
651 if (dstPacking->SwapBytes) {
652 _mesa_swap2( (GLushort *) dst, n );
658 GLshort *dst = (GLshort *) dest;
661 dst[i] = (GLshort) source[i];
663 if (dstPacking->SwapBytes) {
664 _mesa_swap2( (GLushort *) dst, n );
668 case GL_UNSIGNED_INT:
670 GLuint *dst = (GLuint *) dest;
673 dst[i] = (GLuint) source[i];
675 if (dstPacking->SwapBytes) {
676 _mesa_swap4( (GLuint *) dst, n );
682 GLint *dst = (GLint *) dest;
685 dst[i] = (GLint) source[i];
687 if (dstPacking->SwapBytes) {
688 _mesa_swap4( (GLuint *) dst, n );
694 GLfloat *dst = (GLfloat *) dest;
697 dst[i] = (GLfloat) source[i];
699 if (dstPacking->SwapBytes) {
700 _mesa_swap4( (GLuint *) dst, n );
704 case GL_HALF_FLOAT_ARB:
705 case GL_HALF_FLOAT_OES:
707 GLhalfARB *dst = (GLhalfARB *) dest;
710 dst[i] = _mesa_float_to_half( (float) source[i] );
712 if (dstPacking->SwapBytes) {
713 _mesa_swap2( (GLushort *) dst, n );
718 if (dstPacking->LsbFirst) {
719 GLubyte *dst = (GLubyte *) dest;
722 for (i = 0; i < n; i++) {
725 *dst |= ((source[i] != 0) << shift);
734 GLubyte *dst = (GLubyte *) dest;
737 for (i = 0; i < n; i++) {
740 *dst |= ((source[i] != 0) << shift);
750 _mesa_problem(ctx, "bad type in _mesa_pack_index_span");
756 #define DEPTH_VALUES(GLTYPE, GLTYPE2FLOAT) \
759 const GLTYPE *src = (const GLTYPE *)source; \
760 for (i = 0; i < n; i++) { \
761 GLTYPE value = src[i]; \
762 if (srcPacking->SwapBytes) { \
763 if (sizeof(GLTYPE) == 2) { \
765 } else if (sizeof(GLTYPE) == 4) { \
769 depthValues[i] = GLTYPE2FLOAT(value); \
775 * Unpack a row of depth/z values from memory, returning GLushort, GLuint
777 * The glPixelTransfer (scale/bias) params will be applied.
779 * \param dstType one of GL_UNSIGNED_SHORT, GL_UNSIGNED_INT, GL_FLOAT
780 * \param depthMax max value for returned GLushort or GLuint values
781 * (ignored for GLfloat).
784 _mesa_unpack_depth_span( struct gl_context *ctx, GLuint n,
785 GLenum dstType, GLvoid *dest, GLuint depthMax,
786 GLenum srcType, const GLvoid *source,
787 const struct gl_pixelstore_attrib *srcPacking )
789 GLfloat *depthTemp = NULL, *depthValues;
790 GLboolean needClamp = GL_FALSE;
792 /* Look for special cases first.
793 * Not only are these faster, they're less prone to numeric conversion
794 * problems. Otherwise, converting from an int type to a float then
795 * back to an int type can introduce errors that will show up as
796 * artifacts in things like depth peeling which uses glCopyTexImage.
798 if (ctx->Pixel.DepthScale == 1.0F && ctx->Pixel.DepthBias == 0.0F) {
799 if (srcType == GL_UNSIGNED_INT && dstType == GL_UNSIGNED_SHORT) {
800 const GLuint *src = (const GLuint *) source;
801 GLushort *dst = (GLushort *) dest;
803 for (i = 0; i < n; i++) {
804 dst[i] = src[i] >> 16;
808 if (srcType == GL_UNSIGNED_SHORT
809 && dstType == GL_UNSIGNED_INT
810 && depthMax == 0xffffffff) {
811 const GLushort *src = (const GLushort *) source;
812 GLuint *dst = (GLuint *) dest;
814 for (i = 0; i < n; i++) {
815 dst[i] = src[i] | (src[i] << 16);
819 if (srcType == GL_UNSIGNED_INT_24_8
820 && dstType == GL_UNSIGNED_INT
821 && depthMax == 0xffffff) {
822 const GLuint *src = (const GLuint *) source;
823 GLuint *dst = (GLuint *) dest;
825 for (i = 0; i < n; i++) {
826 dst[i] = src[i] >> 8;
830 /* XXX may want to add additional cases here someday */
833 /* general case path follows */
835 if (dstType == GL_FLOAT) {
836 depthValues = (GLfloat *) dest;
839 depthTemp = malloc(n * sizeof(GLfloat));
841 _mesa_error(ctx, GL_OUT_OF_MEMORY, "pixel unpacking");
845 depthValues = depthTemp;
848 /* Convert incoming values to GLfloat. Some conversions will require
853 DEPTH_VALUES(GLbyte, BYTE_TO_FLOATZ);
856 case GL_UNSIGNED_BYTE:
857 DEPTH_VALUES(GLubyte, UBYTE_TO_FLOAT);
860 DEPTH_VALUES(GLshort, SHORT_TO_FLOATZ);
863 case GL_UNSIGNED_SHORT:
864 DEPTH_VALUES(GLushort, USHORT_TO_FLOAT);
867 DEPTH_VALUES(GLint, INT_TO_FLOAT);
870 case GL_UNSIGNED_INT:
871 DEPTH_VALUES(GLuint, UINT_TO_FLOAT);
873 case GL_UNSIGNED_INT_24_8_EXT: /* GL_EXT_packed_depth_stencil */
874 if (dstType == GL_UNSIGNED_INT_24_8_EXT &&
875 depthMax == 0xffffff &&
876 ctx->Pixel.DepthScale == 1.0F &&
877 ctx->Pixel.DepthBias == 0.0F) {
878 const GLuint *src = (const GLuint *) source;
879 GLuint *zValues = (GLuint *) dest;
881 for (i = 0; i < n; i++) {
882 GLuint value = src[i];
883 if (srcPacking->SwapBytes) {
886 zValues[i] = value & 0xffffff00;
892 const GLuint *src = (const GLuint *) source;
893 const GLfloat scale = 1.0f / 0xffffff;
895 for (i = 0; i < n; i++) {
896 GLuint value = src[i];
897 if (srcPacking->SwapBytes) {
900 depthValues[i] = (value >> 8) * scale;
904 case GL_FLOAT_32_UNSIGNED_INT_24_8_REV:
907 const GLfloat *src = (const GLfloat *)source;
908 for (i = 0; i < n; i++) {
909 GLfloat value = src[i * 2];
910 if (srcPacking->SwapBytes) {
913 depthValues[i] = value;
919 DEPTH_VALUES(GLfloat, 1*);
922 case GL_HALF_FLOAT_ARB:
923 case GL_HALF_FLOAT_OES:
926 const GLhalfARB *src = (const GLhalfARB *) source;
927 for (i = 0; i < n; i++) {
928 GLhalfARB value = src[i];
929 if (srcPacking->SwapBytes) {
932 depthValues[i] = _mesa_half_to_float(value);
938 _mesa_problem(NULL, "bad type in _mesa_unpack_depth_span()");
943 /* apply depth scale and bias */
945 const GLfloat scale = ctx->Pixel.DepthScale;
946 const GLfloat bias = ctx->Pixel.DepthBias;
947 if (scale != 1.0F || bias != 0.0F) {
949 for (i = 0; i < n; i++) {
950 depthValues[i] = depthValues[i] * scale + bias;
956 /* clamp to [0, 1] */
959 for (i = 0; i < n; i++) {
960 depthValues[i] = CLAMP(depthValues[i], 0.0F, 1.0F);
965 * Convert values to dstType
967 if (dstType == GL_UNSIGNED_INT) {
968 GLuint *zValues = (GLuint *) dest;
970 if (depthMax <= 0xffffff) {
971 /* no overflow worries */
972 for (i = 0; i < n; i++) {
973 zValues[i] = (GLuint) (depthValues[i] * (GLfloat) depthMax);
977 /* need to use double precision to prevent overflow problems */
978 for (i = 0; i < n; i++) {
979 GLdouble z = depthValues[i] * (GLdouble) depthMax;
980 if (z >= (GLdouble) 0xffffffff)
981 zValues[i] = 0xffffffff;
983 zValues[i] = (GLuint) z;
987 else if (dstType == GL_UNSIGNED_SHORT) {
988 GLushort *zValues = (GLushort *) dest;
990 assert(depthMax <= 0xffff);
991 for (i = 0; i < n; i++) {
992 zValues[i] = (GLushort) (depthValues[i] * (GLfloat) depthMax);
995 else if (dstType == GL_FLOAT) {
996 /* Nothing to do. depthValues is pointing to dest. */
998 else if (dstType == GL_FLOAT_32_UNSIGNED_INT_24_8_REV) {
999 GLfloat *zValues = (GLfloat*) dest;
1001 for (i = 0; i < n; i++) {
1002 zValues[i*2] = depthValues[i];
1014 * Pack an array of depth values. The values are floats in [0,1].
1017 _mesa_pack_depth_span( struct gl_context *ctx, GLuint n, GLvoid *dest,
1018 GLenum dstType, const GLfloat *depthSpan,
1019 const struct gl_pixelstore_attrib *dstPacking )
1021 GLfloat *depthCopy = malloc(n * sizeof(GLfloat));
1023 _mesa_error(ctx, GL_OUT_OF_MEMORY, "pixel packing");
1027 if (ctx->Pixel.DepthScale != 1.0F || ctx->Pixel.DepthBias != 0.0F) {
1028 memcpy(depthCopy, depthSpan, n * sizeof(GLfloat));
1029 _mesa_scale_and_bias_depth(ctx, n, depthCopy);
1030 depthSpan = depthCopy;
1034 case GL_UNSIGNED_BYTE:
1036 GLubyte *dst = (GLubyte *) dest;
1038 for (i = 0; i < n; i++) {
1039 dst[i] = FLOAT_TO_UBYTE( depthSpan[i] );
1045 GLbyte *dst = (GLbyte *) dest;
1047 for (i = 0; i < n; i++) {
1048 dst[i] = FLOAT_TO_BYTE( depthSpan[i] );
1052 case GL_UNSIGNED_SHORT:
1054 GLushort *dst = (GLushort *) dest;
1056 for (i = 0; i < n; i++) {
1057 CLAMPED_FLOAT_TO_USHORT(dst[i], depthSpan[i]);
1059 if (dstPacking->SwapBytes) {
1060 _mesa_swap2( (GLushort *) dst, n );
1066 GLshort *dst = (GLshort *) dest;
1068 for (i = 0; i < n; i++) {
1069 dst[i] = FLOAT_TO_SHORT( depthSpan[i] );
1071 if (dstPacking->SwapBytes) {
1072 _mesa_swap2( (GLushort *) dst, n );
1076 case GL_UNSIGNED_INT_24_8:
1078 const GLdouble scale = (GLdouble) 0xffffff;
1079 GLuint *dst = (GLuint *) dest;
1081 for (i = 0; i < n; i++) {
1082 GLuint z = (GLuint) (depthSpan[i] * scale);
1083 assert(z <= 0xffffff);
1086 if (dstPacking->SwapBytes) {
1087 _mesa_swap4( (GLuint *) dst, n );
1091 case GL_UNSIGNED_INT:
1093 GLuint *dst = (GLuint *) dest;
1095 for (i = 0; i < n; i++) {
1096 dst[i] = FLOAT_TO_UINT( depthSpan[i] );
1098 if (dstPacking->SwapBytes) {
1099 _mesa_swap4( (GLuint *) dst, n );
1105 GLint *dst = (GLint *) dest;
1107 for (i = 0; i < n; i++) {
1108 dst[i] = FLOAT_TO_INT( depthSpan[i] );
1110 if (dstPacking->SwapBytes) {
1111 _mesa_swap4( (GLuint *) dst, n );
1117 GLfloat *dst = (GLfloat *) dest;
1119 for (i = 0; i < n; i++) {
1120 dst[i] = depthSpan[i];
1122 if (dstPacking->SwapBytes) {
1123 _mesa_swap4( (GLuint *) dst, n );
1127 case GL_HALF_FLOAT_ARB:
1128 case GL_HALF_FLOAT_OES:
1130 GLhalfARB *dst = (GLhalfARB *) dest;
1132 for (i = 0; i < n; i++) {
1133 dst[i] = _mesa_float_to_half(depthSpan[i]);
1135 if (dstPacking->SwapBytes) {
1136 _mesa_swap2( (GLushort *) dst, n );
1141 _mesa_problem(ctx, "bad type in _mesa_pack_depth_span (%s)",
1142 _mesa_enum_to_string(dstType));
1151 * Pack depth and stencil values as GL_DEPTH_STENCIL (GL_UNSIGNED_INT_24_8 etc)
1154 _mesa_pack_depth_stencil_span(struct gl_context *ctx,GLuint n,
1155 GLenum dstType, GLuint *dest,
1156 const GLfloat *depthVals,
1157 const GLubyte *stencilVals,
1158 const struct gl_pixelstore_attrib *dstPacking)
1160 GLfloat *depthCopy = malloc(n * sizeof(GLfloat));
1161 GLubyte *stencilCopy = malloc(n * sizeof(GLubyte));
1164 if (!depthCopy || !stencilCopy) {
1165 _mesa_error(ctx, GL_OUT_OF_MEMORY, "pixel packing");
1171 if (ctx->Pixel.DepthScale != 1.0F || ctx->Pixel.DepthBias != 0.0F) {
1172 memcpy(depthCopy, depthVals, n * sizeof(GLfloat));
1173 _mesa_scale_and_bias_depth(ctx, n, depthCopy);
1174 depthVals = depthCopy;
1177 if (ctx->Pixel.IndexShift ||
1178 ctx->Pixel.IndexOffset ||
1179 ctx->Pixel.MapStencilFlag) {
1180 memcpy(stencilCopy, stencilVals, n * sizeof(GLubyte));
1181 _mesa_apply_stencil_transfer_ops(ctx, n, stencilCopy);
1182 stencilVals = stencilCopy;
1186 case GL_UNSIGNED_INT_24_8:
1187 for (i = 0; i < n; i++) {
1188 GLuint z = (GLuint) (depthVals[i] * 0xffffff);
1189 dest[i] = (z << 8) | (stencilVals[i] & 0xff);
1192 case GL_FLOAT_32_UNSIGNED_INT_24_8_REV:
1193 for (i = 0; i < n; i++) {
1194 ((GLfloat*)dest)[i*2] = depthVals[i];
1195 dest[i*2+1] = stencilVals[i] & 0xff;
1200 if (dstPacking->SwapBytes) {
1201 _mesa_swap4(dest, n);
1211 * Unpack image data. Apply byte swapping, byte flipping (bitmap).
1212 * Return all image data in a contiguous block. This is used when we
1213 * compile glDrawPixels, glTexImage, etc into a display list. We
1214 * need a copy of the data in a standard format.
1217 _mesa_unpack_image( GLuint dimensions,
1218 GLsizei width, GLsizei height, GLsizei depth,
1219 GLenum format, GLenum type, const GLvoid *pixels,
1220 const struct gl_pixelstore_attrib *unpack )
1222 GLint bytesPerRow, compsPerRow;
1223 GLboolean flipBytes, swap2, swap4;
1226 return NULL; /* not necessarily an error */
1228 if (width <= 0 || height <= 0 || depth <= 0)
1229 return NULL; /* generate error later */
1231 if (type == GL_BITMAP) {
1232 bytesPerRow = (width + 7) >> 3;
1233 flipBytes = unpack->LsbFirst;
1234 swap2 = swap4 = GL_FALSE;
1238 const GLint bytesPerPixel = _mesa_bytes_per_pixel(format, type);
1239 GLint components = _mesa_components_in_format(format);
1242 if (_mesa_type_is_packed(type))
1245 if (bytesPerPixel <= 0 || components <= 0)
1246 return NULL; /* bad format or type. generate error later */
1247 bytesPerRow = bytesPerPixel * width;
1248 bytesPerComp = bytesPerPixel / components;
1249 flipBytes = GL_FALSE;
1250 swap2 = (bytesPerComp == 2) && unpack->SwapBytes;
1251 swap4 = (bytesPerComp == 4) && unpack->SwapBytes;
1252 compsPerRow = components * width;
1253 assert(compsPerRow >= width);
1258 = malloc(bytesPerRow * height * depth);
1262 return NULL; /* generate GL_OUT_OF_MEMORY later */
1265 for (img = 0; img < depth; img++) {
1266 for (row = 0; row < height; row++) {
1267 const GLvoid *src = _mesa_image_address(dimensions, unpack, pixels,
1268 width, height, format, type, img, row, 0);
1270 if ((type == GL_BITMAP) && (unpack->SkipPixels & 0x7)) {
1272 flipBytes = GL_FALSE;
1273 if (unpack->LsbFirst) {
1274 GLubyte srcMask = 1 << (unpack->SkipPixels & 0x7);
1275 GLubyte dstMask = 128;
1276 const GLubyte *s = src;
1279 for (i = 0; i < width; i++) {
1283 if (srcMask == 128) {
1288 srcMask = srcMask << 1;
1296 dstMask = dstMask >> 1;
1301 GLubyte srcMask = 128 >> (unpack->SkipPixels & 0x7);
1302 GLubyte dstMask = 128;
1303 const GLubyte *s = src;
1306 for (i = 0; i < width; i++) {
1315 srcMask = srcMask >> 1;
1323 dstMask = dstMask >> 1;
1329 memcpy(dst, src, bytesPerRow);
1332 /* byte flipping/swapping */
1334 flip_bytes((GLubyte *) dst, bytesPerRow);
1337 _mesa_swap2((GLushort*) dst, compsPerRow);
1340 _mesa_swap4((GLuint*) dst, compsPerRow);
1350 _mesa_pack_luminance_from_rgba_float(GLuint n, GLfloat rgba[][4],
1351 GLvoid *dstAddr, GLenum dst_format,
1352 GLbitfield transferOps)
1355 GLfloat *dst = (GLfloat *) dstAddr;
1357 switch (dst_format) {
1359 if (transferOps & IMAGE_CLAMP_BIT) {
1360 for (i = 0; i < n; i++) {
1361 GLfloat sum = rgba[i][RCOMP] + rgba[i][GCOMP] + rgba[i][BCOMP];
1362 dst[i] = CLAMP(sum, 0.0F, 1.0F);
1365 for (i = 0; i < n; i++) {
1366 dst[i] = rgba[i][RCOMP] + rgba[i][GCOMP] + rgba[i][BCOMP];
1370 case GL_LUMINANCE_ALPHA:
1371 if (transferOps & IMAGE_CLAMP_BIT) {
1372 for (i = 0; i < n; i++) {
1373 GLfloat sum = rgba[i][RCOMP] + rgba[i][GCOMP] + rgba[i][BCOMP];
1374 dst[2*i] = CLAMP(sum, 0.0F, 1.0F);
1375 dst[2*i+1] = rgba[i][ACOMP];
1378 for (i = 0; i < n; i++) {
1379 dst[2*i] = rgba[i][RCOMP] + rgba[i][GCOMP] + rgba[i][BCOMP];
1380 dst[2*i+1] = rgba[i][ACOMP];
1385 assert(!"Unsupported format");
1390 clamp_sint64_to_sint32(int64_t src)
1392 return CLAMP(src, INT32_MIN, INT32_MAX);
1396 clamp_sint64_to_uint32(int64_t src)
1398 return CLAMP(src, 0, UINT32_MAX);
1402 clamp_uint64_to_uint32(uint64_t src)
1404 return MIN2(src, UINT32_MAX);
1408 clamp_uint64_to_sint32(uint64_t src)
1410 return MIN2(src, INT32_MAX);
1414 convert_integer_luminance64(int64_t src64, int bits,
1415 bool dst_is_signed, bool src_is_signed)
1419 /* Clamp Luminance value from 64-bit to 32-bit. Consider if we need
1420 * any signed<->unsigned conversion too.
1422 if (src_is_signed && dst_is_signed)
1423 src32 = clamp_sint64_to_sint32(src64);
1424 else if (src_is_signed && !dst_is_signed)
1425 src32 = clamp_sint64_to_uint32(src64);
1426 else if (!src_is_signed && dst_is_signed)
1427 src32 = clamp_uint64_to_sint32(src64);
1429 src32 = clamp_uint64_to_uint32(src64);
1431 /* If the dst type is < 32-bit, we need an extra clamp */
1436 return _mesa_signed_to_signed(src32, bits);
1438 return _mesa_unsigned_to_unsigned(src32, bits);
1443 convert_integer(int32_t src, int bits, bool dst_is_signed, bool src_is_signed)
1445 if (src_is_signed && dst_is_signed)
1446 return _mesa_signed_to_signed(src, bits);
1447 else if (src_is_signed && !dst_is_signed)
1448 return _mesa_signed_to_unsigned(src, bits);
1449 else if (!src_is_signed && dst_is_signed)
1450 return _mesa_unsigned_to_signed(src, bits);
1452 return _mesa_unsigned_to_unsigned(src, bits);
1456 _mesa_pack_luminance_from_rgba_integer(GLuint n,
1457 GLuint rgba[][4], bool rgba_is_signed,
1464 int32_t lum32, alpha;
1468 assert(dst_format == GL_LUMINANCE_INTEGER_EXT ||
1469 dst_format == GL_LUMINANCE_ALPHA_INTEGER_EXT);
1471 /* We first compute luminance values as a 64-bit addition of the
1472 * 32-bit R,G,B components, then we clamp the result to the dst type size.
1474 * Notice that this operation involves casting the 32-bit R,G,B components
1475 * to 64-bit before the addition. Since rgba is defined as a GLuint array
1476 * we need to be careful when rgba packs signed data and make sure
1477 * that we cast to a 32-bit signed integer values before casting them to
1478 * 64-bit signed integers.
1480 dst_is_signed = (dst_type == GL_BYTE || dst_type == GL_SHORT ||
1481 dst_type == GL_INT);
1483 dst_bits = _mesa_sizeof_type(dst_type) * 8;
1484 assert(dst_bits > 0);
1486 switch (dst_format) {
1487 case GL_LUMINANCE_INTEGER_EXT:
1488 for (i = 0; i < n; i++) {
1489 if (!rgba_is_signed) {
1490 lum64 = (uint64_t) rgba[i][RCOMP] +
1491 (uint64_t) rgba[i][GCOMP] +
1492 (uint64_t) rgba[i][BCOMP];
1494 lum64 = (int64_t) ((int32_t) rgba[i][RCOMP]) +
1495 (int64_t) ((int32_t) rgba[i][GCOMP]) +
1496 (int64_t) ((int32_t) rgba[i][BCOMP]);
1498 lum32 = convert_integer_luminance64(lum64, dst_bits,
1499 dst_is_signed, rgba_is_signed);
1502 case GL_UNSIGNED_BYTE: {
1503 GLbyte *dst = (GLbyte *) dstAddr;
1508 case GL_UNSIGNED_SHORT: {
1509 GLshort *dst = (GLshort *) dstAddr;
1514 case GL_UNSIGNED_INT: {
1515 GLint *dst = (GLint *) dstAddr;
1522 case GL_LUMINANCE_ALPHA_INTEGER_EXT:
1523 for (i = 0; i < n; i++) {
1524 if (!rgba_is_signed) {
1525 lum64 = (uint64_t) rgba[i][RCOMP] +
1526 (uint64_t) rgba[i][GCOMP] +
1527 (uint64_t) rgba[i][BCOMP];
1529 lum64 = (int64_t) ((int32_t) rgba[i][RCOMP]) +
1530 (int64_t) ((int32_t) rgba[i][GCOMP]) +
1531 (int64_t) ((int32_t) rgba[i][BCOMP]);
1533 lum32 = convert_integer_luminance64(lum64, dst_bits,
1534 dst_is_signed, rgba_is_signed);
1535 alpha = convert_integer(rgba[i][ACOMP], dst_bits,
1536 dst_is_signed, rgba_is_signed);
1539 case GL_UNSIGNED_BYTE: {
1540 GLbyte *dst = (GLbyte *) dstAddr;
1546 case GL_UNSIGNED_SHORT: {
1547 GLshort *dst = (GLshort *) dstAddr;
1553 case GL_UNSIGNED_INT: {
1554 GLint *dst = (GLint *) dstAddr;
1566 _mesa_unpack_color_index_to_rgba_float(struct gl_context *ctx, GLuint dims,
1567 const void *src, GLenum srcFormat, GLenum srcType,
1568 int srcWidth, int srcHeight, int srcDepth,
1569 const struct gl_pixelstore_attrib *srcPacking,
1570 GLbitfield transferOps)
1574 GLfloat *rgba, *dstPtr;
1576 count = srcWidth * srcHeight;
1577 indexes = malloc(count * sizeof(GLuint));
1579 _mesa_error(ctx, GL_OUT_OF_MEMORY, "pixel unpacking");
1583 rgba = malloc(4 * count * srcDepth * sizeof(GLfloat));
1586 _mesa_error(ctx, GL_OUT_OF_MEMORY, "pixel unpacking");
1590 /* Convert indexes to RGBA float */
1592 for (img = 0; img < srcDepth; img++) {
1593 const GLubyte *srcPtr =
1594 (const GLubyte *) _mesa_image_address(dims, srcPacking, src,
1595 srcWidth, srcHeight,
1599 extract_uint_indexes(count, indexes, srcFormat, srcType, srcPtr, srcPacking);
1601 if (transferOps & IMAGE_SHIFT_OFFSET_BIT)
1602 _mesa_shift_and_offset_ci(ctx, count, indexes);
1604 _mesa_map_ci_to_rgba(ctx, count, indexes, (float (*)[4])dstPtr);
1606 /* Don't do RGBA scale/bias or RGBA->RGBA mapping if starting
1607 * with color indexes.
1609 transferOps &= ~(IMAGE_SCALE_BIAS_BIT | IMAGE_MAP_COLOR_BIT);
1610 _mesa_apply_rgba_transfer_ops(ctx, transferOps, count, (float (*)[4])dstPtr);
1612 dstPtr += srcHeight * srcWidth * 4;
1621 _mesa_unpack_color_index_to_rgba_ubyte(struct gl_context *ctx, GLuint dims,
1622 const void *src, GLenum srcFormat, GLenum srcType,
1623 int srcWidth, int srcHeight, int srcDepth,
1624 const struct gl_pixelstore_attrib *srcPacking,
1625 GLbitfield transferOps)
1631 transferOps |= IMAGE_CLAMP_BIT;
1632 rgba = _mesa_unpack_color_index_to_rgba_float(ctx, dims,
1633 src, srcFormat, srcType,
1634 srcWidth, srcHeight, srcDepth,
1635 srcPacking, transferOps);
1637 count = srcWidth * srcHeight * srcDepth;
1638 dst = malloc(count * 4 * sizeof(GLubyte));
1639 for (i = 0; i < count; i++) {
1640 CLAMPED_FLOAT_TO_UBYTE(dst[i * 4 + 0], rgba[i * 4 + 0]);
1641 CLAMPED_FLOAT_TO_UBYTE(dst[i * 4 + 1], rgba[i * 4 + 1]);
1642 CLAMPED_FLOAT_TO_UBYTE(dst[i * 4 + 2], rgba[i * 4 + 2]);
1643 CLAMPED_FLOAT_TO_UBYTE(dst[i * 4 + 3], rgba[i * 4 + 3]);