diff options
author | Brian Paul <[email protected]> | 2010-10-23 09:15:01 -0600 |
---|---|---|
committer | Brian Paul <[email protected]> | 2010-10-23 10:19:29 -0600 |
commit | 1c131752c3e07ef91f49d4970dafca6d26585334 (patch) | |
tree | c6a153387c63e4489d22c3a939162b1cfdd3230b /src/mesa/main/pack.c | |
parent | e67f6ee96e1aaeaba70e24d057c21172fa36f165 (diff) |
mesa: split up the image.c file
New files:
pack.c - image/row packing/unpacking functions
pixeltransfer.c - pixel scale/bias/lookup functions
Diffstat (limited to 'src/mesa/main/pack.c')
-rw-r--r-- | src/mesa/main/pack.c | 3848 |
1 files changed, 3848 insertions, 0 deletions
diff --git a/src/mesa/main/pack.c b/src/mesa/main/pack.c new file mode 100644 index 00000000000..981d99c4752 --- /dev/null +++ b/src/mesa/main/pack.c @@ -0,0 +1,3848 @@ +/* + * Mesa 3-D graphics library + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * Copyright (C) 2009-2010 VMware, Inc. All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THEA AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * \file pack.c + * Image and pixel span packing and unpacking. + */ + + +#include "glheader.h" +#include "colormac.h" +#include "enums.h" +#include "image.h" +#include "pack.h" +#include "pixeltransfer.h" +#include "imports.h" + + +/** + * NOTE: + * Normally, BYTE_TO_FLOAT(0) returns 0.00392 That causes problems when + * we later convert the float to a packed integer value (such as for + * GL_RGB5_A1) because we'll wind up with a non-zero value. + * + * We redefine the macros here so zero is handled correctly. + */ +#undef BYTE_TO_FLOAT +#define BYTE_TO_FLOAT(B) ((B) == 0 ? 0.0F : ((2.0F * (B) + 1.0F) * (1.0F/255.0F))) + +#undef SHORT_TO_FLOAT +#define SHORT_TO_FLOAT(S) ((S) == 0 ? 0.0F : ((2.0F * (S) + 1.0F) * (1.0F/65535.0F))) + + + +/** Compute ceiling of integer quotient of A divided by B. */ +#define CEILING( A, B ) ( (A) % (B) == 0 ? (A)/(B) : (A)/(B)+1 ) + + +/** + * Flip the 8 bits in each byte of the given array. + * + * \param p array. + * \param n number of bytes. + * + * \todo try this trick to flip bytes someday: + * \code + * v = ((v & 0x55555555) << 1) | ((v >> 1) & 0x55555555); + * v = ((v & 0x33333333) << 2) | ((v >> 2) & 0x33333333); + * v = ((v & 0x0f0f0f0f) << 4) | ((v >> 4) & 0x0f0f0f0f); + * \endcode + */ +static void +flip_bytes( GLubyte *p, GLuint n ) +{ + GLuint i, a, b; + for (i = 0; i < n; i++) { + b = (GLuint) p[i]; /* words are often faster than bytes */ + a = ((b & 0x01) << 7) | + ((b & 0x02) << 5) | + ((b & 0x04) << 3) | + ((b & 0x08) << 1) | + ((b & 0x10) >> 1) | + ((b & 0x20) >> 3) | + ((b & 0x40) >> 5) | + ((b & 0x80) >> 7); + p[i] = (GLubyte) a; + } +} + + + +/* + * Unpack a 32x32 pixel polygon stipple from user memory using the + * current pixel unpack settings. + */ +void +_mesa_unpack_polygon_stipple( const GLubyte *pattern, GLuint dest[32], + const struct gl_pixelstore_attrib *unpacking ) +{ + GLubyte *ptrn = (GLubyte *) _mesa_unpack_bitmap(32, 32, pattern, unpacking); + if (ptrn) { + /* Convert pattern from GLubytes to GLuints and handle big/little + * endian differences + */ + GLubyte *p = ptrn; + GLint i; + for (i = 0; i < 32; i++) { + dest[i] = (p[0] << 24) + | (p[1] << 16) + | (p[2] << 8) + | (p[3] ); + p += 4; + } + free(ptrn); + } +} + + +/* + * Pack polygon stipple into user memory given current pixel packing + * settings. + */ +void +_mesa_pack_polygon_stipple( const GLuint pattern[32], GLubyte *dest, + const struct gl_pixelstore_attrib *packing ) +{ + /* Convert pattern from GLuints to GLubytes to handle big/little + * endian differences. + */ + GLubyte ptrn[32*4]; + GLint i; + for (i = 0; i < 32; i++) { + ptrn[i * 4 + 0] = (GLubyte) ((pattern[i] >> 24) & 0xff); + ptrn[i * 4 + 1] = (GLubyte) ((pattern[i] >> 16) & 0xff); + ptrn[i * 4 + 2] = (GLubyte) ((pattern[i] >> 8 ) & 0xff); + ptrn[i * 4 + 3] = (GLubyte) ((pattern[i] ) & 0xff); + } + + _mesa_pack_bitmap(32, 32, ptrn, dest, packing); +} + + +/* + * Unpack bitmap data. Resulting data will be in most-significant-bit-first + * order with row alignment = 1 byte. + */ +GLvoid * +_mesa_unpack_bitmap( GLint width, GLint height, const GLubyte *pixels, + const struct gl_pixelstore_attrib *packing ) +{ + GLint bytes, row, width_in_bytes; + GLubyte *buffer, *dst; + + if (!pixels) + return NULL; + + /* Alloc dest storage */ + bytes = ((width + 7) / 8 * height); + buffer = (GLubyte *) malloc( bytes ); + if (!buffer) + return NULL; + + width_in_bytes = CEILING( width, 8 ); + dst = buffer; + for (row = 0; row < height; row++) { + const GLubyte *src = (const GLubyte *) + _mesa_image_address2d(packing, pixels, width, height, + GL_COLOR_INDEX, GL_BITMAP, row, 0); + if (!src) { + free(buffer); + return NULL; + } + + if ((packing->SkipPixels & 7) == 0) { + memcpy( dst, src, width_in_bytes ); + if (packing->LsbFirst) { + flip_bytes( dst, width_in_bytes ); + } + } + else { + /* handling SkipPixels is a bit tricky (no pun intended!) */ + GLint i; + if (packing->LsbFirst) { + GLubyte srcMask = 1 << (packing->SkipPixels & 0x7); + GLubyte dstMask = 128; + const GLubyte *s = src; + GLubyte *d = dst; + *d = 0; + for (i = 0; i < width; i++) { + if (*s & srcMask) { + *d |= dstMask; + } + if (srcMask == 128) { + srcMask = 1; + s++; + } + else { + srcMask = srcMask << 1; + } + if (dstMask == 1) { + dstMask = 128; + d++; + *d = 0; + } + else { + dstMask = dstMask >> 1; + } + } + } + else { + GLubyte srcMask = 128 >> (packing->SkipPixels & 0x7); + GLubyte dstMask = 128; + const GLubyte *s = src; + GLubyte *d = dst; + *d = 0; + for (i = 0; i < width; i++) { + if (*s & srcMask) { + *d |= dstMask; + } + if (srcMask == 1) { + srcMask = 128; + s++; + } + else { + srcMask = srcMask >> 1; + } + if (dstMask == 1) { + dstMask = 128; + d++; + *d = 0; + } + else { + dstMask = dstMask >> 1; + } + } + } + } + dst += width_in_bytes; + } + + return buffer; +} + + +/* + * Pack bitmap data. + */ +void +_mesa_pack_bitmap( GLint width, GLint height, const GLubyte *source, + GLubyte *dest, const struct gl_pixelstore_attrib *packing ) +{ + GLint row, width_in_bytes; + const GLubyte *src; + + if (!source) + return; + + width_in_bytes = CEILING( width, 8 ); + src = source; + for (row = 0; row < height; row++) { + GLubyte *dst = (GLubyte *) _mesa_image_address2d(packing, dest, + width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0); + if (!dst) + return; + + if ((packing->SkipPixels & 7) == 0) { + memcpy( dst, src, width_in_bytes ); + if (packing->LsbFirst) { + flip_bytes( dst, width_in_bytes ); + } + } + else { + /* handling SkipPixels is a bit tricky (no pun intended!) */ + GLint i; + if (packing->LsbFirst) { + GLubyte srcMask = 128; + GLubyte dstMask = 1 << (packing->SkipPixels & 0x7); + const GLubyte *s = src; + GLubyte *d = dst; + *d = 0; + for (i = 0; i < width; i++) { + if (*s & srcMask) { + *d |= dstMask; + } + if (srcMask == 1) { + srcMask = 128; + s++; + } + else { + srcMask = srcMask >> 1; + } + if (dstMask == 128) { + dstMask = 1; + d++; + *d = 0; + } + else { + dstMask = dstMask << 1; + } + } + } + else { + GLubyte srcMask = 128; + GLubyte dstMask = 128 >> (packing->SkipPixels & 0x7); + const GLubyte *s = src; + GLubyte *d = dst; + *d = 0; + for (i = 0; i < width; i++) { + if (*s & srcMask) { + *d |= dstMask; + } + if (srcMask == 1) { + srcMask = 128; + s++; + } + else { + srcMask = srcMask >> 1; + } + if (dstMask == 1) { + dstMask = 128; + d++; + *d = 0; + } + else { + dstMask = dstMask >> 1; + } + } + } + } + src += width_in_bytes; + } +} + + + + + +/** + * Used to pack an array [][4] of RGBA float colors as specified + * by the dstFormat, dstType and dstPacking. Used by glReadPixels, + * glGetConvolutionFilter(), etc. + * Note: the rgba values will be modified by this function when any pixel + * transfer ops are enabled. + */ +void +_mesa_pack_rgba_span_float(struct gl_context *ctx, GLuint n, GLfloat rgba[][4], + GLenum dstFormat, GLenum dstType, + GLvoid *dstAddr, + const struct gl_pixelstore_attrib *dstPacking, + GLbitfield transferOps) +{ + GLfloat luminance[MAX_WIDTH]; + const GLint comps = _mesa_components_in_format(dstFormat); + GLuint i; + + /* XXX + * This test should probably go away. Have the caller set/clear the + * IMAGE_CLAMP_BIT as needed. + */ + if (dstType != GL_FLOAT || ctx->Color.ClampReadColor == GL_TRUE) { + /* need to clamp to [0, 1] */ + transferOps |= IMAGE_CLAMP_BIT; + } + + if (transferOps) { + _mesa_apply_rgba_transfer_ops(ctx, transferOps, n, rgba); + } + + if (dstFormat == GL_LUMINANCE || dstFormat == GL_LUMINANCE_ALPHA) { + /* compute luminance values */ + if (transferOps & IMAGE_CLAMP_BIT) { + for (i = 0; i < n; i++) { + GLfloat sum = rgba[i][RCOMP] + rgba[i][GCOMP] + rgba[i][BCOMP]; + luminance[i] = CLAMP(sum, 0.0F, 1.0F); + } + } + else { + for (i = 0; i < n; i++) { + luminance[i] = rgba[i][RCOMP] + rgba[i][GCOMP] + rgba[i][BCOMP]; + } + } + } + + /* + * Pack/store the pixels. Ugh! Lots of cases!!! + */ + switch (dstType) { + case GL_UNSIGNED_BYTE: + { + GLubyte *dst = (GLubyte *) dstAddr; + switch (dstFormat) { + case GL_RED: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_UBYTE(rgba[i][RCOMP]); + break; + case GL_GREEN: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_UBYTE(rgba[i][GCOMP]); + break; + case GL_BLUE: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_UBYTE(rgba[i][BCOMP]); + break; + case GL_ALPHA: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_UBYTE(rgba[i][ACOMP]); + break; + case GL_LUMINANCE: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_UBYTE(luminance[i]); + break; + case GL_LUMINANCE_ALPHA: + for (i=0;i<n;i++) { + dst[i*2+0] = FLOAT_TO_UBYTE(luminance[i]); + dst[i*2+1] = FLOAT_TO_UBYTE(rgba[i][ACOMP]); + } + break; + case GL_RGB: + for (i=0;i<n;i++) { + dst[i*3+0] = FLOAT_TO_UBYTE(rgba[i][RCOMP]); + dst[i*3+1] = FLOAT_TO_UBYTE(rgba[i][GCOMP]); + dst[i*3+2] = FLOAT_TO_UBYTE(rgba[i][BCOMP]); + } + break; + case GL_RGBA: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_UBYTE(rgba[i][RCOMP]); + dst[i*4+1] = FLOAT_TO_UBYTE(rgba[i][GCOMP]); + dst[i*4+2] = FLOAT_TO_UBYTE(rgba[i][BCOMP]); + dst[i*4+3] = FLOAT_TO_UBYTE(rgba[i][ACOMP]); + } + break; + case GL_BGR: + for (i=0;i<n;i++) { + dst[i*3+0] = FLOAT_TO_UBYTE(rgba[i][BCOMP]); + dst[i*3+1] = FLOAT_TO_UBYTE(rgba[i][GCOMP]); + dst[i*3+2] = FLOAT_TO_UBYTE(rgba[i][RCOMP]); + } + break; + case GL_BGRA: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_UBYTE(rgba[i][BCOMP]); + dst[i*4+1] = FLOAT_TO_UBYTE(rgba[i][GCOMP]); + dst[i*4+2] = FLOAT_TO_UBYTE(rgba[i][RCOMP]); + dst[i*4+3] = FLOAT_TO_UBYTE(rgba[i][ACOMP]); + } + break; + case GL_ABGR_EXT: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_UBYTE(rgba[i][ACOMP]); + dst[i*4+1] = FLOAT_TO_UBYTE(rgba[i][BCOMP]); + dst[i*4+2] = FLOAT_TO_UBYTE(rgba[i][GCOMP]); + dst[i*4+3] = FLOAT_TO_UBYTE(rgba[i][RCOMP]); + } + break; + case GL_DUDV_ATI: + case GL_DU8DV8_ATI: + for (i=0;i<n;i++) { + dst[i*2+0] = FLOAT_TO_UBYTE(rgba[i][RCOMP]); + dst[i*2+1] = FLOAT_TO_UBYTE(rgba[i][GCOMP]); + } + break; + default: + _mesa_problem(ctx, "bad format in _mesa_pack_rgba_span\n"); + } + } + break; + case GL_BYTE: + { + GLbyte *dst = (GLbyte *) dstAddr; + switch (dstFormat) { + case GL_RED: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_BYTE(rgba[i][RCOMP]); + break; + case GL_GREEN: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_BYTE(rgba[i][GCOMP]); + break; + case GL_BLUE: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_BYTE(rgba[i][BCOMP]); + break; + case GL_ALPHA: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_BYTE(rgba[i][ACOMP]); + break; + case GL_LUMINANCE: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_BYTE(luminance[i]); + break; + case GL_LUMINANCE_ALPHA: + for (i=0;i<n;i++) { + dst[i*2+0] = FLOAT_TO_BYTE(luminance[i]); + dst[i*2+1] = FLOAT_TO_BYTE(rgba[i][ACOMP]); + } + break; + case GL_RGB: + for (i=0;i<n;i++) { + dst[i*3+0] = FLOAT_TO_BYTE(rgba[i][RCOMP]); + dst[i*3+1] = FLOAT_TO_BYTE(rgba[i][GCOMP]); + dst[i*3+2] = FLOAT_TO_BYTE(rgba[i][BCOMP]); + } + break; + case GL_RGBA: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_BYTE(rgba[i][RCOMP]); + dst[i*4+1] = FLOAT_TO_BYTE(rgba[i][GCOMP]); + dst[i*4+2] = FLOAT_TO_BYTE(rgba[i][BCOMP]); + dst[i*4+3] = FLOAT_TO_BYTE(rgba[i][ACOMP]); + } + break; + case GL_BGR: + for (i=0;i<n;i++) { + dst[i*3+0] = FLOAT_TO_BYTE(rgba[i][BCOMP]); + dst[i*3+1] = FLOAT_TO_BYTE(rgba[i][GCOMP]); + dst[i*3+2] = FLOAT_TO_BYTE(rgba[i][RCOMP]); + } + break; + case GL_BGRA: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_BYTE(rgba[i][BCOMP]); + dst[i*4+1] = FLOAT_TO_BYTE(rgba[i][GCOMP]); + dst[i*4+2] = FLOAT_TO_BYTE(rgba[i][RCOMP]); + dst[i*4+3] = FLOAT_TO_BYTE(rgba[i][ACOMP]); + } + break; + case GL_ABGR_EXT: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_BYTE(rgba[i][ACOMP]); + dst[i*4+1] = FLOAT_TO_BYTE(rgba[i][BCOMP]); + dst[i*4+2] = FLOAT_TO_BYTE(rgba[i][GCOMP]); + dst[i*4+3] = FLOAT_TO_BYTE(rgba[i][RCOMP]); + } + break; + case GL_DUDV_ATI: + case GL_DU8DV8_ATI: + for (i=0;i<n;i++) { + dst[i*2+0] = FLOAT_TO_BYTE(rgba[i][RCOMP]); + dst[i*2+1] = FLOAT_TO_BYTE(rgba[i][GCOMP]); + } + break; + default: + _mesa_problem(ctx, "bad format in _mesa_pack_rgba_span\n"); + } + } + break; + case GL_UNSIGNED_SHORT: + { + GLushort *dst = (GLushort *) dstAddr; + switch (dstFormat) { + case GL_RED: + for (i=0;i<n;i++) + CLAMPED_FLOAT_TO_USHORT(dst[i], rgba[i][RCOMP]); + break; + case GL_GREEN: + for (i=0;i<n;i++) + CLAMPED_FLOAT_TO_USHORT(dst[i], rgba[i][GCOMP]); + break; + case GL_BLUE: + for (i=0;i<n;i++) + CLAMPED_FLOAT_TO_USHORT(dst[i], rgba[i][BCOMP]); + break; + case GL_ALPHA: + for (i=0;i<n;i++) + CLAMPED_FLOAT_TO_USHORT(dst[i], rgba[i][ACOMP]); + break; + case GL_LUMINANCE: + for (i=0;i<n;i++) + UNCLAMPED_FLOAT_TO_USHORT(dst[i], luminance[i]); + break; + case GL_LUMINANCE_ALPHA: + for (i=0;i<n;i++) { + UNCLAMPED_FLOAT_TO_USHORT(dst[i*2+0], luminance[i]); + CLAMPED_FLOAT_TO_USHORT(dst[i*2+1], rgba[i][ACOMP]); + } + break; + case GL_RGB: + for (i=0;i<n;i++) { + CLAMPED_FLOAT_TO_USHORT(dst[i*3+0], rgba[i][RCOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*3+1], rgba[i][GCOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*3+2], rgba[i][BCOMP]); + } + break; + case GL_RGBA: + for (i=0;i<n;i++) { + CLAMPED_FLOAT_TO_USHORT(dst[i*4+0], rgba[i][RCOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*4+1], rgba[i][GCOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*4+2], rgba[i][BCOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*4+3], rgba[i][ACOMP]); + } + break; + case GL_BGR: + for (i=0;i<n;i++) { + CLAMPED_FLOAT_TO_USHORT(dst[i*3+0], rgba[i][BCOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*3+1], rgba[i][GCOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*3+2], rgba[i][RCOMP]); + } + break; + case GL_BGRA: + for (i=0;i<n;i++) { + CLAMPED_FLOAT_TO_USHORT(dst[i*4+0], rgba[i][BCOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*4+1], rgba[i][GCOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*4+2], rgba[i][RCOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*4+3], rgba[i][ACOMP]); + } + break; + case GL_ABGR_EXT: + for (i=0;i<n;i++) { + CLAMPED_FLOAT_TO_USHORT(dst[i*4+0], rgba[i][ACOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*4+1], rgba[i][BCOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*4+2], rgba[i][GCOMP]); + CLAMPED_FLOAT_TO_USHORT(dst[i*4+3], rgba[i][RCOMP]); + } + break; + case GL_DUDV_ATI: + case GL_DU8DV8_ATI: + for (i=0;i<n;i++) { + dst[i*2+0] = FLOAT_TO_USHORT(rgba[i][RCOMP]); + dst[i*2+1] = FLOAT_TO_USHORT(rgba[i][GCOMP]); + } + break; + default: + _mesa_problem(ctx, "bad format in _mesa_pack_rgba_span\n"); + } + } + break; + case GL_SHORT: + { + GLshort *dst = (GLshort *) dstAddr; + switch (dstFormat) { + case GL_RED: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_SHORT(rgba[i][RCOMP]); + break; + case GL_GREEN: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_SHORT(rgba[i][GCOMP]); + break; + case GL_BLUE: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_SHORT(rgba[i][BCOMP]); + break; + case GL_ALPHA: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_SHORT(rgba[i][ACOMP]); + break; + case GL_LUMINANCE: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_SHORT(luminance[i]); + break; + case GL_LUMINANCE_ALPHA: + for (i=0;i<n;i++) { + dst[i*2+0] = FLOAT_TO_SHORT(luminance[i]); + dst[i*2+1] = FLOAT_TO_SHORT(rgba[i][ACOMP]); + } + break; + case GL_RGB: + for (i=0;i<n;i++) { + dst[i*3+0] = FLOAT_TO_SHORT(rgba[i][RCOMP]); + dst[i*3+1] = FLOAT_TO_SHORT(rgba[i][GCOMP]); + dst[i*3+2] = FLOAT_TO_SHORT(rgba[i][BCOMP]); + } + break; + case GL_RGBA: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_SHORT(rgba[i][RCOMP]); + dst[i*4+1] = FLOAT_TO_SHORT(rgba[i][GCOMP]); + dst[i*4+2] = FLOAT_TO_SHORT(rgba[i][BCOMP]); + dst[i*4+3] = FLOAT_TO_SHORT(rgba[i][ACOMP]); + } + break; + case GL_BGR: + for (i=0;i<n;i++) { + dst[i*3+0] = FLOAT_TO_SHORT(rgba[i][BCOMP]); + dst[i*3+1] = FLOAT_TO_SHORT(rgba[i][GCOMP]); + dst[i*3+2] = FLOAT_TO_SHORT(rgba[i][RCOMP]); + } + break; + case GL_BGRA: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_SHORT(rgba[i][BCOMP]); + dst[i*4+1] = FLOAT_TO_SHORT(rgba[i][GCOMP]); + dst[i*4+2] = FLOAT_TO_SHORT(rgba[i][RCOMP]); + dst[i*4+3] = FLOAT_TO_SHORT(rgba[i][ACOMP]); + } + break; + case GL_ABGR_EXT: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_SHORT(rgba[i][ACOMP]); + dst[i*4+1] = FLOAT_TO_SHORT(rgba[i][BCOMP]); + dst[i*4+2] = FLOAT_TO_SHORT(rgba[i][GCOMP]); + dst[i*4+3] = FLOAT_TO_SHORT(rgba[i][RCOMP]); + } + break; + case GL_DUDV_ATI: + case GL_DU8DV8_ATI: + for (i=0;i<n;i++) { + dst[i*2+0] = FLOAT_TO_SHORT(rgba[i][RCOMP]); + dst[i*2+1] = FLOAT_TO_SHORT(rgba[i][GCOMP]); + } + break; + default: + _mesa_problem(ctx, "bad format in _mesa_pack_rgba_span\n"); + } + } + break; + case GL_UNSIGNED_INT: + { + GLuint *dst = (GLuint *) dstAddr; + switch (dstFormat) { + case GL_RED: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_UINT(rgba[i][RCOMP]); + break; + case GL_GREEN: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_UINT(rgba[i][GCOMP]); + break; + case GL_BLUE: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_UINT(rgba[i][BCOMP]); + break; + case GL_ALPHA: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_UINT(rgba[i][ACOMP]); + break; + case GL_LUMINANCE: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_UINT(luminance[i]); + break; + case GL_LUMINANCE_ALPHA: + for (i=0;i<n;i++) { + dst[i*2+0] = FLOAT_TO_UINT(luminance[i]); + dst[i*2+1] = FLOAT_TO_UINT(rgba[i][ACOMP]); + } + break; + case GL_RGB: + for (i=0;i<n;i++) { + dst[i*3+0] = FLOAT_TO_UINT(rgba[i][RCOMP]); + dst[i*3+1] = FLOAT_TO_UINT(rgba[i][GCOMP]); + dst[i*3+2] = FLOAT_TO_UINT(rgba[i][BCOMP]); + } + break; + case GL_RGBA: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_UINT(rgba[i][RCOMP]); + dst[i*4+1] = FLOAT_TO_UINT(rgba[i][GCOMP]); + dst[i*4+2] = FLOAT_TO_UINT(rgba[i][BCOMP]); + dst[i*4+3] = FLOAT_TO_UINT(rgba[i][ACOMP]); + } + break; + case GL_BGR: + for (i=0;i<n;i++) { + dst[i*3+0] = FLOAT_TO_UINT(rgba[i][BCOMP]); + dst[i*3+1] = FLOAT_TO_UINT(rgba[i][GCOMP]); + dst[i*3+2] = FLOAT_TO_UINT(rgba[i][RCOMP]); + } + break; + case GL_BGRA: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_UINT(rgba[i][BCOMP]); + dst[i*4+1] = FLOAT_TO_UINT(rgba[i][GCOMP]); + dst[i*4+2] = FLOAT_TO_UINT(rgba[i][RCOMP]); + dst[i*4+3] = FLOAT_TO_UINT(rgba[i][ACOMP]); + } + break; + case GL_ABGR_EXT: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_UINT(rgba[i][ACOMP]); + dst[i*4+1] = FLOAT_TO_UINT(rgba[i][BCOMP]); + dst[i*4+2] = FLOAT_TO_UINT(rgba[i][GCOMP]); + dst[i*4+3] = FLOAT_TO_UINT(rgba[i][RCOMP]); + } + break; + case GL_DUDV_ATI: + case GL_DU8DV8_ATI: + for (i=0;i<n;i++) { + dst[i*2+0] = FLOAT_TO_UINT(rgba[i][RCOMP]); + dst[i*2+1] = FLOAT_TO_UINT(rgba[i][GCOMP]); + } + break; + default: + _mesa_problem(ctx, "bad format in _mesa_pack_rgba_span\n"); + } + } + break; + case GL_INT: + { + GLint *dst = (GLint *) dstAddr; + switch (dstFormat) { + case GL_RED: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_INT(rgba[i][RCOMP]); + break; + case GL_GREEN: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_INT(rgba[i][GCOMP]); + break; + case GL_BLUE: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_INT(rgba[i][BCOMP]); + break; + case GL_ALPHA: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_INT(rgba[i][ACOMP]); + break; + case GL_LUMINANCE: + for (i=0;i<n;i++) + dst[i] = FLOAT_TO_INT(luminance[i]); + break; + case GL_LUMINANCE_ALPHA: + for (i=0;i<n;i++) { + dst[i*2+0] = FLOAT_TO_INT(luminance[i]); + dst[i*2+1] = FLOAT_TO_INT(rgba[i][ACOMP]); + } + break; + case GL_RGB: + for (i=0;i<n;i++) { + dst[i*3+0] = FLOAT_TO_INT(rgba[i][RCOMP]); + dst[i*3+1] = FLOAT_TO_INT(rgba[i][GCOMP]); + dst[i*3+2] = FLOAT_TO_INT(rgba[i][BCOMP]); + } + break; + case GL_RGBA: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_INT(rgba[i][RCOMP]); + dst[i*4+1] = FLOAT_TO_INT(rgba[i][GCOMP]); + dst[i*4+2] = FLOAT_TO_INT(rgba[i][BCOMP]); + dst[i*4+3] = FLOAT_TO_INT(rgba[i][ACOMP]); + } + break; + case GL_BGR: + for (i=0;i<n;i++) { + dst[i*3+0] = FLOAT_TO_INT(rgba[i][BCOMP]); + dst[i*3+1] = FLOAT_TO_INT(rgba[i][GCOMP]); + dst[i*3+2] = FLOAT_TO_INT(rgba[i][RCOMP]); + } + break; + case GL_BGRA: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_INT(rgba[i][BCOMP]); + dst[i*4+1] = FLOAT_TO_INT(rgba[i][GCOMP]); + dst[i*4+2] = FLOAT_TO_INT(rgba[i][RCOMP]); + dst[i*4+3] = FLOAT_TO_INT(rgba[i][ACOMP]); + } + break; + case GL_ABGR_EXT: + for (i=0;i<n;i++) { + dst[i*4+0] = FLOAT_TO_INT(rgba[i][ACOMP]); + dst[i*4+1] = FLOAT_TO_INT(rgba[i][BCOMP]); + dst[i*4+2] = FLOAT_TO_INT(rgba[i][GCOMP]); + dst[i*4+3] = FLOAT_TO_INT(rgba[i][RCOMP]); + } + break; + case GL_DUDV_ATI: + case GL_DU8DV8_ATI: + for (i=0;i<n;i++) { + dst[i*2+0] = FLOAT_TO_INT(rgba[i][RCOMP]); + dst[i*2+1] = FLOAT_TO_INT(rgba[i][GCOMP]); + } + break; + default: + _mesa_problem(ctx, "bad format in _mesa_pack_rgba_span\n"); + } + } + break; + case GL_FLOAT: + { + GLfloat *dst = (GLfloat *) dstAddr; + switch (dstFormat) { + case GL_RED: + for (i=0;i<n;i++) + dst[i] = rgba[i][RCOMP]; + break; + case GL_GREEN: + for (i=0;i<n;i++) + dst[i] = rgba[i][GCOMP]; + break; + case GL_BLUE: + for (i=0;i<n;i++) + dst[i] = rgba[i][BCOMP]; + break; + case GL_ALPHA: + for (i=0;i<n;i++) + dst[i] = rgba[i][ACOMP]; + break; + case GL_LUMINANCE: + for (i=0;i<n;i++) + dst[i] = luminance[i]; + break; + case GL_LUMINANCE_ALPHA: + for (i=0;i<n;i++) { + dst[i*2+0] = luminance[i]; + dst[i*2+1] = rgba[i][ACOMP]; + } + break; + case GL_RGB: + for (i=0;i<n;i++) { + dst[i*3+0] = rgba[i][RCOMP]; + dst[i*3+1] = rgba[i][GCOMP]; + dst[i*3+2] = rgba[i][BCOMP]; + } + break; + case GL_RGBA: + for (i=0;i<n;i++) { + dst[i*4+0] = rgba[i][RCOMP]; + dst[i*4+1] = rgba[i][GCOMP]; + dst[i*4+2] = rgba[i][BCOMP]; + dst[i*4+3] = rgba[i][ACOMP]; + } + break; + case GL_BGR: + for (i=0;i<n;i++) { + dst[i*3+0] = rgba[i][BCOMP]; + dst[i*3+1] = rgba[i][GCOMP]; + dst[i*3+2] = rgba[i][RCOMP]; + } + break; + case GL_BGRA: + for (i=0;i<n;i++) { + dst[i*4+0] = rgba[i][BCOMP]; + dst[i*4+1] = rgba[i][GCOMP]; + dst[i*4+2] = rgba[i][RCOMP]; + dst[i*4+3] = rgba[i][ACOMP]; + } + break; + case GL_ABGR_EXT: + for (i=0;i<n;i++) { + dst[i*4+0] = rgba[i][ACOMP]; + dst[i*4+1] = rgba[i][BCOMP]; + dst[i*4+2] = rgba[i][GCOMP]; + dst[i*4+3] = rgba[i][RCOMP]; + } + break; + case GL_DUDV_ATI: + case GL_DU8DV8_ATI: + for (i=0;i<n;i++) { + dst[i*2+0] = rgba[i][RCOMP]; + dst[i*2+1] = rgba[i][GCOMP]; + } + break; + default: + _mesa_problem(ctx, "bad format in _mesa_pack_rgba_span\n"); + } + } + break; + case GL_HALF_FLOAT_ARB: + { + GLhalfARB *dst = (GLhalfARB *) dstAddr; + switch (dstFormat) { + case GL_RED: + for (i=0;i<n;i++) + dst[i] = _mesa_float_to_half(rgba[i][RCOMP]); + break; + case GL_GREEN: + for (i=0;i<n;i++) + dst[i] = _mesa_float_to_half(rgba[i][GCOMP]); + break; + case GL_BLUE: + for (i=0;i<n;i++) + dst[i] = _mesa_float_to_half(rgba[i][BCOMP]); + break; + case GL_ALPHA: + for (i=0;i<n;i++) + dst[i] = _mesa_float_to_half(rgba[i][ACOMP]); + break; + case GL_LUMINANCE: + for (i=0;i<n;i++) + dst[i] = _mesa_float_to_half(luminance[i]); + break; + case GL_LUMINANCE_ALPHA: + for (i=0;i<n;i++) { + dst[i*2+0] = _mesa_float_to_half(luminance[i]); + dst[i*2+1] = _mesa_float_to_half(rgba[i][ACOMP]); + } + break; + case GL_RGB: + for (i=0;i<n;i++) { + dst[i*3+0] = _mesa_float_to_half(rgba[i][RCOMP]); + dst[i*3+1] = _mesa_float_to_half(rgba[i][GCOMP]); + dst[i*3+2] = _mesa_float_to_half(rgba[i][BCOMP]); + } + break; + case GL_RGBA: + for (i=0;i<n;i++) { + dst[i*4+0] = _mesa_float_to_half(rgba[i][RCOMP]); + dst[i*4+1] = _mesa_float_to_half(rgba[i][GCOMP]); + dst[i*4+2] = _mesa_float_to_half(rgba[i][BCOMP]); + dst[i*4+3] = _mesa_float_to_half(rgba[i][ACOMP]); + } + break; + case GL_BGR: + for (i=0;i<n;i++) { + dst[i*3+0] = _mesa_float_to_half(rgba[i][BCOMP]); + dst[i*3+1] = _mesa_float_to_half(rgba[i][GCOMP]); + dst[i*3+2] = _mesa_float_to_half(rgba[i][RCOMP]); + } + break; + case GL_BGRA: + for (i=0;i<n;i++) { + dst[i*4+0] = _mesa_float_to_half(rgba[i][BCOMP]); + dst[i*4+1] = _mesa_float_to_half(rgba[i][GCOMP]); + dst[i*4+2] = _mesa_float_to_half(rgba[i][RCOMP]); + dst[i*4+3] = _mesa_float_to_half(rgba[i][ACOMP]); + } + break; + case GL_ABGR_EXT: + for (i=0;i<n;i++) { + dst[i*4+0] = _mesa_float_to_half(rgba[i][ACOMP]); + dst[i*4+1] = _mesa_float_to_half(rgba[i][BCOMP]); + dst[i*4+2] = _mesa_float_to_half(rgba[i][GCOMP]); + dst[i*4+3] = _mesa_float_to_half(rgba[i][RCOMP]); + } + break; + case GL_DUDV_ATI: + case GL_DU8DV8_ATI: + for (i=0;i<n;i++) { + dst[i*2+0] = _mesa_float_to_half(rgba[i][RCOMP]); + dst[i*2+1] = _mesa_float_to_half(rgba[i][GCOMP]); + } + break; + default: + _mesa_problem(ctx, "bad format in _mesa_pack_rgba_span\n"); + } + } + break; + case GL_UNSIGNED_BYTE_3_3_2: + if (dstFormat == GL_RGB) { + GLubyte *dst = (GLubyte *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][RCOMP] * 7.0F) << 5) + | (IROUND(rgba[i][GCOMP] * 7.0F) << 2) + | (IROUND(rgba[i][BCOMP] * 3.0F) ); + } + } + break; + case GL_UNSIGNED_BYTE_2_3_3_REV: + if (dstFormat == GL_RGB) { + GLubyte *dst = (GLubyte *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][RCOMP] * 7.0F) ) + | (IROUND(rgba[i][GCOMP] * 7.0F) << 3) + | (IROUND(rgba[i][BCOMP] * 3.0F) << 6); + } + } + break; + case GL_UNSIGNED_SHORT_5_6_5: + if (dstFormat == GL_RGB) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][RCOMP] * 31.0F) << 11) + | (IROUND(rgba[i][GCOMP] * 63.0F) << 5) + | (IROUND(rgba[i][BCOMP] * 31.0F) ); + } + } + break; + case GL_UNSIGNED_SHORT_5_6_5_REV: + if (dstFormat == GL_RGB) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][RCOMP] * 31.0F) ) + | (IROUND(rgba[i][GCOMP] * 63.0F) << 5) + | (IROUND(rgba[i][BCOMP] * 31.0F) << 11); + } + } + break; + case GL_UNSIGNED_SHORT_4_4_4_4: + if (dstFormat == GL_RGBA) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][RCOMP] * 15.0F) << 12) + | (IROUND(rgba[i][GCOMP] * 15.0F) << 8) + | (IROUND(rgba[i][BCOMP] * 15.0F) << 4) + | (IROUND(rgba[i][ACOMP] * 15.0F) ); + } + } + else if (dstFormat == GL_BGRA) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][BCOMP] * 15.0F) << 12) + | (IROUND(rgba[i][GCOMP] * 15.0F) << 8) + | (IROUND(rgba[i][RCOMP] * 15.0F) << 4) + | (IROUND(rgba[i][ACOMP] * 15.0F) ); + } + } + else if (dstFormat == GL_ABGR_EXT) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][ACOMP] * 15.0F) << 12) + | (IROUND(rgba[i][BCOMP] * 15.0F) << 8) + | (IROUND(rgba[i][GCOMP] * 15.0F) << 4) + | (IROUND(rgba[i][RCOMP] * 15.0F) ); + } + } + break; + case GL_UNSIGNED_SHORT_4_4_4_4_REV: + if (dstFormat == GL_RGBA) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][RCOMP] * 15.0F) ) + | (IROUND(rgba[i][GCOMP] * 15.0F) << 4) + | (IROUND(rgba[i][BCOMP] * 15.0F) << 8) + | (IROUND(rgba[i][ACOMP] * 15.0F) << 12); + } + } + else if (dstFormat == GL_BGRA) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][BCOMP] * 15.0F) ) + | (IROUND(rgba[i][GCOMP] * 15.0F) << 4) + | (IROUND(rgba[i][RCOMP] * 15.0F) << 8) + | (IROUND(rgba[i][ACOMP] * 15.0F) << 12); + } + } + else if (dstFormat == GL_ABGR_EXT) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][ACOMP] * 15.0F) ) + | (IROUND(rgba[i][BCOMP] * 15.0F) << 4) + | (IROUND(rgba[i][GCOMP] * 15.0F) << 8) + | (IROUND(rgba[i][RCOMP] * 15.0F) << 12); + } + } + break; + case GL_UNSIGNED_SHORT_5_5_5_1: + if (dstFormat == GL_RGBA) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][RCOMP] * 31.0F) << 11) + | (IROUND(rgba[i][GCOMP] * 31.0F) << 6) + | (IROUND(rgba[i][BCOMP] * 31.0F) << 1) + | (IROUND(rgba[i][ACOMP] * 1.0F) ); + } + } + else if (dstFormat == GL_BGRA) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][BCOMP] * 31.0F) << 11) + | (IROUND(rgba[i][GCOMP] * 31.0F) << 6) + | (IROUND(rgba[i][RCOMP] * 31.0F) << 1) + | (IROUND(rgba[i][ACOMP] * 1.0F) ); + } + } + else if (dstFormat == GL_ABGR_EXT) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][ACOMP] * 31.0F) << 11) + | (IROUND(rgba[i][BCOMP] * 31.0F) << 6) + | (IROUND(rgba[i][GCOMP] * 31.0F) << 1) + | (IROUND(rgba[i][RCOMP] * 1.0F) ); + } + } + break; + case GL_UNSIGNED_SHORT_1_5_5_5_REV: + if (dstFormat == GL_RGBA) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][RCOMP] * 31.0F) ) + | (IROUND(rgba[i][GCOMP] * 31.0F) << 5) + | (IROUND(rgba[i][BCOMP] * 31.0F) << 10) + | (IROUND(rgba[i][ACOMP] * 1.0F) << 15); + } + } + else if (dstFormat == GL_BGRA) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][BCOMP] * 31.0F) ) + | (IROUND(rgba[i][GCOMP] * 31.0F) << 5) + | (IROUND(rgba[i][RCOMP] * 31.0F) << 10) + | (IROUND(rgba[i][ACOMP] * 1.0F) << 15); + } + } + else if (dstFormat == GL_ABGR_EXT) { + GLushort *dst = (GLushort *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][ACOMP] * 31.0F) ) + | (IROUND(rgba[i][BCOMP] * 31.0F) << 5) + | (IROUND(rgba[i][GCOMP] * 31.0F) << 10) + | (IROUND(rgba[i][RCOMP] * 1.0F) << 15); + } + } + break; + case GL_UNSIGNED_INT_8_8_8_8: + if (dstFormat == GL_RGBA) { + GLuint *dst = (GLuint *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][RCOMP] * 255.F) << 24) + | (IROUND(rgba[i][GCOMP] * 255.F) << 16) + | (IROUND(rgba[i][BCOMP] * 255.F) << 8) + | (IROUND(rgba[i][ACOMP] * 255.F) ); + } + } + else if (dstFormat == GL_BGRA) { + GLuint *dst = (GLuint *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][BCOMP] * 255.F) << 24) + | (IROUND(rgba[i][GCOMP] * 255.F) << 16) + | (IROUND(rgba[i][RCOMP] * 255.F) << 8) + | (IROUND(rgba[i][ACOMP] * 255.F) ); + } + } + else if (dstFormat == GL_ABGR_EXT) { + GLuint *dst = (GLuint *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][ACOMP] * 255.F) << 24) + | (IROUND(rgba[i][BCOMP] * 255.F) << 16) + | (IROUND(rgba[i][GCOMP] * 255.F) << 8) + | (IROUND(rgba[i][RCOMP] * 255.F) ); + } + } + break; + case GL_UNSIGNED_INT_8_8_8_8_REV: + if (dstFormat == GL_RGBA) { + GLuint *dst = (GLuint *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][RCOMP] * 255.0F) ) + | (IROUND(rgba[i][GCOMP] * 255.0F) << 8) + | (IROUND(rgba[i][BCOMP] * 255.0F) << 16) + | (IROUND(rgba[i][ACOMP] * 255.0F) << 24); + } + } + else if (dstFormat == GL_BGRA) { + GLuint *dst = (GLuint *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][BCOMP] * 255.0F) ) + | (IROUND(rgba[i][GCOMP] * 255.0F) << 8) + | (IROUND(rgba[i][RCOMP] * 255.0F) << 16) + | (IROUND(rgba[i][ACOMP] * 255.0F) << 24); + } + } + else if (dstFormat == GL_ABGR_EXT) { + GLuint *dst = (GLuint *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][ACOMP] * 255.0F) ) + | (IROUND(rgba[i][BCOMP] * 255.0F) << 8) + | (IROUND(rgba[i][GCOMP] * 255.0F) << 16) + | (IROUND(rgba[i][RCOMP] * 255.0F) << 24); + } + } + break; + case GL_UNSIGNED_INT_10_10_10_2: + if (dstFormat == GL_RGBA) { + GLuint *dst = (GLuint *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][RCOMP] * 1023.0F) << 22) + | (IROUND(rgba[i][GCOMP] * 1023.0F) << 12) + | (IROUND(rgba[i][BCOMP] * 1023.0F) << 2) + | (IROUND(rgba[i][ACOMP] * 3.0F) ); + } + } + else if (dstFormat == GL_BGRA) { + GLuint *dst = (GLuint *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][BCOMP] * 1023.0F) << 22) + | (IROUND(rgba[i][GCOMP] * 1023.0F) << 12) + | (IROUND(rgba[i][RCOMP] * 1023.0F) << 2) + | (IROUND(rgba[i][ACOMP] * 3.0F) ); + } + } + else if (dstFormat == GL_ABGR_EXT) { + GLuint *dst = (GLuint *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][ACOMP] * 1023.0F) << 22) + | (IROUND(rgba[i][BCOMP] * 1023.0F) << 12) + | (IROUND(rgba[i][GCOMP] * 1023.0F) << 2) + | (IROUND(rgba[i][RCOMP] * 3.0F) ); + } + } + break; + case GL_UNSIGNED_INT_2_10_10_10_REV: + if (dstFormat == GL_RGBA) { + GLuint *dst = (GLuint *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][RCOMP] * 1023.0F) ) + | (IROUND(rgba[i][GCOMP] * 1023.0F) << 10) + | (IROUND(rgba[i][BCOMP] * 1023.0F) << 20) + | (IROUND(rgba[i][ACOMP] * 3.0F) << 30); + } + } + else if (dstFormat == GL_BGRA) { + GLuint *dst = (GLuint *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][BCOMP] * 1023.0F) ) + | (IROUND(rgba[i][GCOMP] * 1023.0F) << 10) + | (IROUND(rgba[i][RCOMP] * 1023.0F) << 20) + | (IROUND(rgba[i][ACOMP] * 3.0F) << 30); + } + } + else if (dstFormat == GL_ABGR_EXT) { + GLuint *dst = (GLuint *) dstAddr; + for (i=0;i<n;i++) { + dst[i] = (IROUND(rgba[i][ACOMP] * 1023.0F) ) + | (IROUND(rgba[i][BCOMP] * 1023.0F) << 10) + | (IROUND(rgba[i][GCOMP] * 1023.0F) << 20) + | (IROUND(rgba[i][RCOMP] * 3.0F) << 30); + } + } + break; + default: + _mesa_problem(ctx, "bad type in _mesa_pack_rgba_span_float"); + return; + } + + if (dstPacking->SwapBytes) { + GLint swapSize = _mesa_sizeof_packed_type(dstType); + if (swapSize == 2) { + if (dstPacking->SwapBytes) { + _mesa_swap2((GLushort *) dstAddr, n * comps); + } + } + else if (swapSize == 4) { + if (dstPacking->SwapBytes) { + _mesa_swap4((GLuint *) dstAddr, n * comps); + } + } + } +} + + + +#define SWAP2BYTE(VALUE) \ + { \ + GLubyte *bytes = (GLubyte *) &(VALUE); \ + GLubyte tmp = bytes[0]; \ + bytes[0] = bytes[1]; \ + bytes[1] = tmp; \ + } + +#define SWAP4BYTE(VALUE) \ + { \ + GLubyte *bytes = (GLubyte *) &(VALUE); \ + GLubyte tmp = bytes[0]; \ + bytes[0] = bytes[3]; \ + bytes[3] = tmp; \ + tmp = bytes[1]; \ + bytes[1] = bytes[2]; \ + bytes[2] = tmp; \ + } + + +static void +extract_uint_indexes(GLuint n, GLuint indexes[], + GLenum srcFormat, GLenum srcType, const GLvoid *src, + const struct gl_pixelstore_attrib *unpack ) +{ + ASSERT(srcFormat == GL_COLOR_INDEX || srcFormat == GL_STENCIL_INDEX); + + ASSERT(srcType == GL_BITMAP || + srcType == GL_UNSIGNED_BYTE || + srcType == GL_BYTE || + srcType == GL_UNSIGNED_SHORT || + srcType == GL_SHORT || + srcType == GL_UNSIGNED_INT || + srcType == GL_INT || + srcType == GL_UNSIGNED_INT_24_8_EXT || + srcType == GL_HALF_FLOAT_ARB || + srcType == GL_FLOAT); + + switch (srcType) { + case GL_BITMAP: + { + GLubyte *ubsrc = (GLubyte *) src; + if (unpack->LsbFirst) { + GLubyte mask = 1 << (unpack->SkipPixels & 0x7); + GLuint i; + for (i = 0; i < n; i++) { + indexes[i] = (*ubsrc & mask) ? 1 : 0; + if (mask == 128) { + mask = 1; + ubsrc++; + } + else { + mask = mask << 1; + } + } + } + else { + GLubyte mask = 128 >> (unpack->SkipPixels & 0x7); + GLuint i; + for (i = 0; i < n; i++) { + indexes[i] = (*ubsrc & mask) ? 1 : 0; + if (mask == 1) { + mask = 128; + ubsrc++; + } + else { + mask = mask >> 1; + } + } + } + } + break; + case GL_UNSIGNED_BYTE: + { + GLuint i; + const GLubyte *s = (const GLubyte *) src; + for (i = 0; i < n; i++) + indexes[i] = s[i]; + } + break; + case GL_BYTE: + { + GLuint i; + const GLbyte *s = (const GLbyte *) src; + for (i = 0; i < n; i++) + indexes[i] = s[i]; + } + break; + case GL_UNSIGNED_SHORT: + { + GLuint i; + const GLushort *s = (const GLushort *) src; + if (unpack->SwapBytes) { + for (i = 0; i < n; i++) { + GLushort value = s[i]; + SWAP2BYTE(value); + indexes[i] = value; + } + } + else { + for (i = 0; i < n; i++) + indexes[i] = s[i]; + } + } + break; + case GL_SHORT: + { + GLuint i; + const GLshort *s = (const GLshort *) src; + if (unpack->SwapBytes) { + for (i = 0; i < n; i++) { + GLshort value = s[i]; + SWAP2BYTE(value); + indexes[i] = value; + } + } + else { + for (i = 0; i < n; i++) + indexes[i] = s[i]; + } + } + break; + case GL_UNSIGNED_INT: + { + GLuint i; + const GLuint *s = (const GLuint *) src; + if (unpack->SwapBytes) { + for (i = 0; i < n; i++) { + GLuint value = s[i]; + SWAP4BYTE(value); + indexes[i] = value; + } + } + else { + for (i = 0; i < n; i++) + indexes[i] = s[i]; + } + } + break; + case GL_INT: + { + GLuint i; + const GLint *s = (const GLint *) src; + if (unpack->SwapBytes) { + for (i = 0; i < n; i++) { + GLint value = s[i]; + SWAP4BYTE(value); + indexes[i] = value; + } + } + else { + for (i = 0; i < n; i++) + indexes[i] = s[i]; + } + } + break; + case GL_FLOAT: + { + GLuint i; + const GLfloat *s = (const GLfloat *) src; + if (unpack->SwapBytes) { + for (i = 0; i < n; i++) { + GLfloat value = s[i]; + SWAP4BYTE(value); + indexes[i] = (GLuint) value; + } + } + else { + for (i = 0; i < n; i++) + indexes[i] = (GLuint) s[i]; + } + } + break; + case GL_HALF_FLOAT_ARB: + { + GLuint i; + const GLhalfARB *s = (const GLhalfARB *) src; + if (unpack->SwapBytes) { + for (i = 0; i < n; i++) { + GLhalfARB value = s[i]; + SWAP2BYTE(value); + indexes[i] = (GLuint) _mesa_half_to_float(value); + } + } + else { + for (i = 0; i < n; i++) + indexes[i] = (GLuint) _mesa_half_to_float(s[i]); + } + } + break; + case GL_UNSIGNED_INT_24_8_EXT: + { + GLuint i; + const GLuint *s = (const GLuint *) src; + if (unpack->SwapBytes) { + for (i = 0; i < n; i++) { + GLuint value = s[i]; + SWAP4BYTE(value); + indexes[i] = value & 0xff; /* lower 8 bits */ + } + } + else { + for (i = 0; i < n; i++) + indexes[i] = s[i] & 0xff; /* lower 8 bits */ + } + } + break; + + default: + _mesa_problem(NULL, "bad srcType in extract_uint_indexes"); + return; + } +} + + +/* + * This function extracts floating point RGBA values from arbitrary + * image data. srcFormat and srcType are the format and type parameters + * passed to glDrawPixels, glTexImage[123]D, glTexSubImage[123]D, etc. + * + * Refering to section 3.6.4 of the OpenGL 1.2 spec, this function + * implements the "Conversion to floating point", "Conversion to RGB", + * and "Final Expansion to RGBA" operations. + * + * Args: n - number of pixels + * rgba - output colors + * srcFormat - format of incoming data + * srcType - data type of incoming data + * src - source data pointer + * swapBytes - perform byteswapping of incoming data? + */ +static void +extract_float_rgba(GLuint n, GLfloat rgba[][4], + GLenum srcFormat, GLenum srcType, const GLvoid *src, + GLboolean swapBytes) +{ + GLint redIndex, greenIndex, blueIndex, alphaIndex; + GLint stride; + GLint rComp, bComp, gComp, aComp; + GLboolean intFormat; + GLfloat rs = 1.0f, gs = 1.0f, bs = 1.0f, as = 1.0f; /* scale factors */ + + ASSERT(srcFormat == GL_RED || + srcFormat == GL_GREEN || + srcFormat == GL_BLUE || + srcFormat == GL_ALPHA || + srcFormat == GL_LUMINANCE || + srcFormat == GL_LUMINANCE_ALPHA || + srcFormat == GL_INTENSITY || + srcFormat == GL_RG || + srcFormat == GL_RGB || + srcFormat == GL_BGR || + srcFormat == GL_RGBA || + srcFormat == GL_BGRA || + srcFormat == GL_ABGR_EXT || + srcFormat == GL_DU8DV8_ATI || + srcFormat == GL_DUDV_ATI || + srcFormat == GL_RED_INTEGER_EXT || + srcFormat == GL_GREEN_INTEGER_EXT || + srcFormat == GL_BLUE_INTEGER_EXT || + srcFormat == GL_ALPHA_INTEGER_EXT || + srcFormat == GL_RGB_INTEGER_EXT || + srcFormat == GL_RGBA_INTEGER_EXT || + srcFormat == GL_BGR_INTEGER_EXT || + srcFormat == GL_BGRA_INTEGER_EXT || + srcFormat == GL_LUMINANCE_INTEGER_EXT || + srcFormat == GL_LUMINANCE_ALPHA_INTEGER_EXT); + + ASSERT(srcType == GL_UNSIGNED_BYTE || + srcType == GL_BYTE || + srcType == GL_UNSIGNED_SHORT || + srcType == GL_SHORT || + srcType == GL_UNSIGNED_INT || + srcType == GL_INT || + srcType == GL_HALF_FLOAT_ARB || + srcType == GL_FLOAT || + srcType == GL_UNSIGNED_BYTE_3_3_2 || + srcType == GL_UNSIGNED_BYTE_2_3_3_REV || + srcType == GL_UNSIGNED_SHORT_5_6_5 || + srcType == GL_UNSIGNED_SHORT_5_6_5_REV || + srcType == GL_UNSIGNED_SHORT_4_4_4_4 || + srcType == GL_UNSIGNED_SHORT_4_4_4_4_REV || + srcType == GL_UNSIGNED_SHORT_5_5_5_1 || + srcType == GL_UNSIGNED_SHORT_1_5_5_5_REV || + srcType == GL_UNSIGNED_INT_8_8_8_8 || + srcType == GL_UNSIGNED_INT_8_8_8_8_REV || + srcType == GL_UNSIGNED_INT_10_10_10_2 || + srcType == GL_UNSIGNED_INT_2_10_10_10_REV); + + rComp = gComp = bComp = aComp = -1; + + switch (srcFormat) { + case GL_RED: + case GL_RED_INTEGER_EXT: + redIndex = 0; + greenIndex = blueIndex = alphaIndex = -1; + stride = 1; + break; + case GL_GREEN: + case GL_GREEN_INTEGER_EXT: + greenIndex = 0; + redIndex = blueIndex = alphaIndex = -1; + stride = 1; + break; + case GL_BLUE: + case GL_BLUE_INTEGER_EXT: + blueIndex = 0; + redIndex = greenIndex = alphaIndex = -1; + stride = 1; + break; + case GL_ALPHA: + case GL_ALPHA_INTEGER_EXT: + redIndex = greenIndex = blueIndex = -1; + alphaIndex = 0; + stride = 1; + break; + case GL_LUMINANCE: + case GL_LUMINANCE_INTEGER_EXT: + redIndex = greenIndex = blueIndex = 0; + alphaIndex = -1; + stride = 1; + break; + case GL_LUMINANCE_ALPHA: + case GL_LUMINANCE_ALPHA_INTEGER_EXT: + redIndex = greenIndex = blueIndex = 0; + alphaIndex = 1; + stride = 2; + break; + case GL_INTENSITY: + redIndex = greenIndex = blueIndex = alphaIndex = 0; + stride = 1; + break; + case GL_RG: + case GL_RG_INTEGER: + redIndex = 0; + greenIndex = 1; + blueIndex = -1; + alphaIndex = -1; + rComp = 0; + gComp = 1; + bComp = 2; + aComp = 3; + stride = 2; + break; + case GL_RGB: + case GL_RGB_INTEGER: + redIndex = 0; + greenIndex = 1; + blueIndex = 2; + alphaIndex = -1; + rComp = 0; + gComp = 1; + bComp = 2; + aComp = 3; + stride = 3; + break; + case GL_BGR: + redIndex = 2; + greenIndex = 1; + blueIndex = 0; + alphaIndex = -1; + rComp = 2; + gComp = 1; + bComp = 0; + aComp = 3; + stride = 3; + break; + case GL_RGBA: + case GL_RGBA_INTEGER: + redIndex = 0; + greenIndex = 1; + blueIndex = 2; + alphaIndex = 3; + rComp = 0; + gComp = 1; + bComp = 2; + aComp = 3; + stride = 4; + break; + case GL_BGRA: + redIndex = 2; + greenIndex = 1; + blueIndex = 0; + alphaIndex = 3; + rComp = 2; + gComp = 1; + bComp = 0; + aComp = 3; + stride = 4; + break; + case GL_ABGR_EXT: + redIndex = 3; + greenIndex = 2; + blueIndex = 1; + alphaIndex = 0; + rComp = 3; + gComp = 2; + bComp = 1; + aComp = 0; + stride = 4; + break; + case GL_DU8DV8_ATI: + case GL_DUDV_ATI: + redIndex = 0; + greenIndex = 1; + blueIndex = -1; + alphaIndex = -1; + stride = 2; + break; + default: + _mesa_problem(NULL, "bad srcFormat %s in extract float data", + _mesa_lookup_enum_by_nr(srcFormat)); + return; + } + + intFormat = _mesa_is_integer_format(srcFormat); + +#define PROCESS(INDEX, CHANNEL, DEFAULT, DEFAULT_INT, TYPE, CONVERSION) \ + if ((INDEX) < 0) { \ + GLuint i; \ + if (intFormat) { \ + for (i = 0; i < n; i++) { \ + rgba[i][CHANNEL] = DEFAULT_INT; \ + } \ + } \ + else { \ + for (i = 0; i < n; i++) { \ + rgba[i][CHANNEL] = DEFAULT; \ + } \ + } \ + } \ + else if (swapBytes) { \ + const TYPE *s = (const TYPE *) src; \ + GLuint i; \ + for (i = 0; i < n; i++) { \ + TYPE value = s[INDEX]; \ + if (sizeof(TYPE) == 2) { \ + SWAP2BYTE(value); \ + } \ + else if (sizeof(TYPE) == 4) { \ + SWAP4BYTE(value); \ + } \ + if (intFormat) \ + rgba[i][CHANNEL] = (GLfloat) value; \ + else \ + rgba[i][CHANNEL] = (GLfloat) CONVERSION(value); \ + s += stride; \ + } \ + } \ + else { \ + const TYPE *s = (const TYPE *) src; \ + GLuint i; \ + if (intFormat) { \ + for (i = 0; i < n; i++) { \ + rgba[i][CHANNEL] = (GLfloat) s[INDEX]; \ + s += stride; \ + } \ + } \ + else { \ + for (i = 0; i < n; i++) { \ + rgba[i][CHANNEL] = (GLfloat) CONVERSION(s[INDEX]); \ + s += stride; \ + } \ + } \ + } + + switch (srcType) { + case GL_UNSIGNED_BYTE: + PROCESS(redIndex, RCOMP, 0.0F, 0, GLubyte, UBYTE_TO_FLOAT); + PROCESS(greenIndex, GCOMP, 0.0F, 0, GLubyte, UBYTE_TO_FLOAT); + PROCESS(blueIndex, BCOMP, 0.0F, 0, GLubyte, UBYTE_TO_FLOAT); + PROCESS(alphaIndex, ACOMP, 1.0F, 255, GLubyte, UBYTE_TO_FLOAT); + break; + case GL_BYTE: + PROCESS(redIndex, RCOMP, 0.0F, 0, GLbyte, BYTE_TO_FLOAT); + PROCESS(greenIndex, GCOMP, 0.0F, 0, GLbyte, BYTE_TO_FLOAT); + PROCESS(blueIndex, BCOMP, 0.0F, 0, GLbyte, BYTE_TO_FLOAT); + PROCESS(alphaIndex, ACOMP, 1.0F, 127, GLbyte, BYTE_TO_FLOAT); + break; + case GL_UNSIGNED_SHORT: + PROCESS(redIndex, RCOMP, 0.0F, 0, GLushort, USHORT_TO_FLOAT); + PROCESS(greenIndex, GCOMP, 0.0F, 0, GLushort, USHORT_TO_FLOAT); + PROCESS(blueIndex, BCOMP, 0.0F, 0, GLushort, USHORT_TO_FLOAT); + PROCESS(alphaIndex, ACOMP, 1.0F, 0xffff, GLushort, USHORT_TO_FLOAT); + break; + case GL_SHORT: + PROCESS(redIndex, RCOMP, 0.0F, 0, GLshort, SHORT_TO_FLOAT); + PROCESS(greenIndex, GCOMP, 0.0F, 0, GLshort, SHORT_TO_FLOAT); + PROCESS(blueIndex, BCOMP, 0.0F, 0, GLshort, SHORT_TO_FLOAT); + PROCESS(alphaIndex, ACOMP, 1.0F, 32767, GLshort, SHORT_TO_FLOAT); + break; + case GL_UNSIGNED_INT: + PROCESS(redIndex, RCOMP, 0.0F, 0, GLuint, UINT_TO_FLOAT); + PROCESS(greenIndex, GCOMP, 0.0F, 0, GLuint, UINT_TO_FLOAT); + PROCESS(blueIndex, BCOMP, 0.0F, 0, GLuint, UINT_TO_FLOAT); + PROCESS(alphaIndex, ACOMP, 1.0F, 0xffffffff, GLuint, UINT_TO_FLOAT); + break; + case GL_INT: + PROCESS(redIndex, RCOMP, 0.0F, 0, GLint, INT_TO_FLOAT); + PROCESS(greenIndex, GCOMP, 0.0F, 0, GLint, INT_TO_FLOAT); + PROCESS(blueIndex, BCOMP, 0.0F, 0, GLint, INT_TO_FLOAT); + PROCESS(alphaIndex, ACOMP, 1.0F, 2147483647, GLint, INT_TO_FLOAT); + break; + case GL_FLOAT: + PROCESS(redIndex, RCOMP, 0.0F, 0.0F, GLfloat, (GLfloat)); + PROCESS(greenIndex, GCOMP, 0.0F, 0.0F, GLfloat, (GLfloat)); + PROCESS(blueIndex, BCOMP, 0.0F, 0.0F, GLfloat, (GLfloat)); + PROCESS(alphaIndex, ACOMP, 1.0F, 1.0F, GLfloat, (GLfloat)); + break; + case GL_HALF_FLOAT_ARB: + PROCESS(redIndex, RCOMP, 0.0F, 0.0F, GLhalfARB, _mesa_half_to_float); + PROCESS(greenIndex, GCOMP, 0.0F, 0.0F, GLhalfARB, _mesa_half_to_float); + PROCESS(blueIndex, BCOMP, 0.0F, 0.0F, GLhalfARB, _mesa_half_to_float); + PROCESS(alphaIndex, ACOMP, 1.0F, 1.0F, GLhalfARB, _mesa_half_to_float); + break; + case GL_UNSIGNED_BYTE_3_3_2: + { + const GLubyte *ubsrc = (const GLubyte *) src; + GLuint i; + if (!intFormat) { + rs = 1.0F / 7.0F; + gs = 1.0F / 7.0F; + bs = 1.0F / 3.0F; + } + for (i = 0; i < n; i ++) { + GLubyte p = ubsrc[i]; + rgba[i][rComp] = ((p >> 5) ) * rs; + rgba[i][gComp] = ((p >> 2) & 0x7) * gs; + rgba[i][bComp] = ((p ) & 0x3) * bs; + rgba[i][aComp] = 1.0F; + } + } + break; + case GL_UNSIGNED_BYTE_2_3_3_REV: + { + const GLubyte *ubsrc = (const GLubyte *) src; + GLuint i; + if (!intFormat) { + rs = 1.0F / 7.0F; + gs = 1.0F / 7.0F; + bs = 1.0F / 3.0F; + } + for (i = 0; i < n; i ++) { + GLubyte p = ubsrc[i]; + rgba[i][rComp] = ((p ) & 0x7) * rs; + rgba[i][gComp] = ((p >> 3) & 0x7) * gs; + rgba[i][bComp] = ((p >> 6) ) * bs; + rgba[i][aComp] = 1.0F; + } + } + break; + case GL_UNSIGNED_SHORT_5_6_5: + if (!intFormat) { + rs = 1.0F / 31.0F; + gs = 1.0F / 63.0F; + bs = 1.0F / 31.0F; + } + if (swapBytes) { + const GLushort *ussrc = (const GLushort *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLushort p = ussrc[i]; + SWAP2BYTE(p); + rgba[i][rComp] = ((p >> 11) ) * rs; + rgba[i][gComp] = ((p >> 5) & 0x3f) * gs; + rgba[i][bComp] = ((p ) & 0x1f) * bs; + rgba[i][aComp] = 1.0F; + } + } + else { + const GLushort *ussrc = (const GLushort *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLushort p = ussrc[i]; + rgba[i][rComp] = ((p >> 11) ) * rs; + rgba[i][gComp] = ((p >> 5) & 0x3f) * gs; + rgba[i][bComp] = ((p ) & 0x1f) * bs; + rgba[i][aComp] = 1.0F; + } + } + break; + case GL_UNSIGNED_SHORT_5_6_5_REV: + if (!intFormat) { + rs = 1.0F / 31.0F; + gs = 1.0F / 63.0F; + bs = 1.0F / 31.0F; + } + if (swapBytes) { + const GLushort *ussrc = (const GLushort *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLushort p = ussrc[i]; + SWAP2BYTE(p); + rgba[i][rComp] = ((p ) & 0x1f) * rs; + rgba[i][gComp] = ((p >> 5) & 0x3f) * gs; + rgba[i][bComp] = ((p >> 11) ) * bs; + rgba[i][aComp] = 1.0F; + } + } + else { + const GLushort *ussrc = (const GLushort *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLushort p = ussrc[i]; + rgba[i][rComp] = ((p ) & 0x1f) * rs; + rgba[i][gComp] = ((p >> 5) & 0x3f) * gs; + rgba[i][bComp] = ((p >> 11) ) * bs; + rgba[i][aComp] = 1.0F; + } + } + break; + case GL_UNSIGNED_SHORT_4_4_4_4: + if (!intFormat) { + rs = gs = bs = as = 1.0F / 15.0F; + } + if (swapBytes) { + const GLushort *ussrc = (const GLushort *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLushort p = ussrc[i]; + SWAP2BYTE(p); + rgba[i][rComp] = ((p >> 12) ) * rs; + rgba[i][gComp] = ((p >> 8) & 0xf) * gs; + rgba[i][bComp] = ((p >> 4) & 0xf) * bs; + rgba[i][aComp] = ((p ) & 0xf) * as; + } + } + else { + const GLushort *ussrc = (const GLushort *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLushort p = ussrc[i]; + rgba[i][rComp] = ((p >> 12) ) * rs; + rgba[i][gComp] = ((p >> 8) & 0xf) * gs; + rgba[i][bComp] = ((p >> 4) & 0xf) * bs; + rgba[i][aComp] = ((p ) & 0xf) * as; + } + } + break; + case GL_UNSIGNED_SHORT_4_4_4_4_REV: + if (!intFormat) { + rs = gs = bs = as = 1.0F / 15.0F; + } + if (swapBytes) { + const GLushort *ussrc = (const GLushort *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLushort p = ussrc[i]; + SWAP2BYTE(p); + rgba[i][rComp] = ((p ) & 0xf) * rs; + rgba[i][gComp] = ((p >> 4) & 0xf) * gs; + rgba[i][bComp] = ((p >> 8) & 0xf) * bs; + rgba[i][aComp] = ((p >> 12) ) * as; + } + } + else { + const GLushort *ussrc = (const GLushort *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLushort p = ussrc[i]; + rgba[i][rComp] = ((p ) & 0xf) * rs; + rgba[i][gComp] = ((p >> 4) & 0xf) * gs; + rgba[i][bComp] = ((p >> 8) & 0xf) * bs; + rgba[i][aComp] = ((p >> 12) ) * as; + } + } + break; + case GL_UNSIGNED_SHORT_5_5_5_1: + if (!intFormat) { + rs = gs = bs = 1.0F / 31.0F; + } + if (swapBytes) { + const GLushort *ussrc = (const GLushort *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLushort p = ussrc[i]; + SWAP2BYTE(p); + rgba[i][rComp] = ((p >> 11) ) * rs; + rgba[i][gComp] = ((p >> 6) & 0x1f) * gs; + rgba[i][bComp] = ((p >> 1) & 0x1f) * bs; + rgba[i][aComp] = ((p ) & 0x1) * as; + } + } + else { + const GLushort *ussrc = (const GLushort *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLushort p = ussrc[i]; + rgba[i][rComp] = ((p >> 11) ) * rs; + rgba[i][gComp] = ((p >> 6) & 0x1f) * gs; + rgba[i][bComp] = ((p >> 1) & 0x1f) * bs; + rgba[i][aComp] = ((p ) & 0x1) * as; + } + } + break; + case GL_UNSIGNED_SHORT_1_5_5_5_REV: + if (!intFormat) { + rs = gs = bs = 1.0F / 31.0F; + } + if (swapBytes) { + const GLushort *ussrc = (const GLushort *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLushort p = ussrc[i]; + SWAP2BYTE(p); + rgba[i][rComp] = ((p ) & 0x1f) * rs; + rgba[i][gComp] = ((p >> 5) & 0x1f) * gs; + rgba[i][bComp] = ((p >> 10) & 0x1f) * bs; + rgba[i][aComp] = ((p >> 15) ) * as; + } + } + else { + const GLushort *ussrc = (const GLushort *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLushort p = ussrc[i]; + rgba[i][rComp] = ((p ) & 0x1f) * rs; + rgba[i][gComp] = ((p >> 5) & 0x1f) * gs; + rgba[i][bComp] = ((p >> 10) & 0x1f) * bs; + rgba[i][aComp] = ((p >> 15) ) * as; + } + } + break; + case GL_UNSIGNED_INT_8_8_8_8: + if (swapBytes) { + const GLuint *uisrc = (const GLuint *) src; + GLuint i; + if (intFormat) { + for (i = 0; i < n; i ++) { + GLuint p = uisrc[i]; + rgba[i][rComp] = (GLfloat) ((p ) & 0xff); + rgba[i][gComp] = (GLfloat) ((p >> 8) & 0xff); + rgba[i][bComp] = (GLfloat) ((p >> 16) & 0xff); + rgba[i][aComp] = (GLfloat) ((p >> 24) ); + } + } + else { + for (i = 0; i < n; i ++) { + GLuint p = uisrc[i]; + rgba[i][rComp] = UBYTE_TO_FLOAT((p ) & 0xff); + rgba[i][gComp] = UBYTE_TO_FLOAT((p >> 8) & 0xff); + rgba[i][bComp] = UBYTE_TO_FLOAT((p >> 16) & 0xff); + rgba[i][aComp] = UBYTE_TO_FLOAT((p >> 24) ); + } + } + } + else { + const GLuint *uisrc = (const GLuint *) src; + GLuint i; + if (intFormat) { + for (i = 0; i < n; i ++) { + GLuint p = uisrc[i]; + rgba[i][rComp] = (GLfloat) ((p >> 24) ); + rgba[i][gComp] = (GLfloat) ((p >> 16) & 0xff); + rgba[i][bComp] = (GLfloat) ((p >> 8) & 0xff); + rgba[i][aComp] = (GLfloat) ((p ) & 0xff); + } + } + else { + for (i = 0; i < n; i ++) { + GLuint p = uisrc[i]; + rgba[i][rComp] = UBYTE_TO_FLOAT((p >> 24) ); + rgba[i][gComp] = UBYTE_TO_FLOAT((p >> 16) & 0xff); + rgba[i][bComp] = UBYTE_TO_FLOAT((p >> 8) & 0xff); + rgba[i][aComp] = UBYTE_TO_FLOAT((p ) & 0xff); + } + } + } + break; + case GL_UNSIGNED_INT_8_8_8_8_REV: + if (swapBytes) { + const GLuint *uisrc = (const GLuint *) src; + GLuint i; + if (intFormat) { + for (i = 0; i < n; i ++) { + GLuint p = uisrc[i]; + rgba[i][rComp] = (GLfloat) ((p >> 24) ); + rgba[i][gComp] = (GLfloat) ((p >> 16) & 0xff); + rgba[i][bComp] = (GLfloat) ((p >> 8) & 0xff); + rgba[i][aComp] = (GLfloat) ((p ) & 0xff); + } + } + else { + for (i = 0; i < n; i ++) { + GLuint p = uisrc[i]; + rgba[i][rComp] = UBYTE_TO_FLOAT((p >> 24) ); + rgba[i][gComp] = UBYTE_TO_FLOAT((p >> 16) & 0xff); + rgba[i][bComp] = UBYTE_TO_FLOAT((p >> 8) & 0xff); + rgba[i][aComp] = UBYTE_TO_FLOAT((p ) & 0xff); + } + } + } + else { + const GLuint *uisrc = (const GLuint *) src; + GLuint i; + if (intFormat) { + for (i = 0; i < n; i ++) { + GLuint p = uisrc[i]; + rgba[i][rComp] = (GLfloat) ((p ) & 0xff); + rgba[i][gComp] = (GLfloat) ((p >> 8) & 0xff); + rgba[i][bComp] = (GLfloat) ((p >> 16) & 0xff); + rgba[i][aComp] = (GLfloat) ((p >> 24) ); + } + } + else { + for (i = 0; i < n; i ++) { + GLuint p = uisrc[i]; + rgba[i][rComp] = UBYTE_TO_FLOAT((p ) & 0xff); + rgba[i][gComp] = UBYTE_TO_FLOAT((p >> 8) & 0xff); + rgba[i][bComp] = UBYTE_TO_FLOAT((p >> 16) & 0xff); + rgba[i][aComp] = UBYTE_TO_FLOAT((p >> 24) ); + } + } + } + break; + case GL_UNSIGNED_INT_10_10_10_2: + if (!intFormat) { + rs = 1.0F / 1023.0F; + gs = 1.0F / 1023.0F; + bs = 1.0F / 1023.0F; + as = 1.0F / 3.0F; + } + if (swapBytes) { + const GLuint *uisrc = (const GLuint *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLuint p = uisrc[i]; + SWAP4BYTE(p); + rgba[i][rComp] = ((p >> 22) ) * rs; + rgba[i][gComp] = ((p >> 12) & 0x3ff) * gs; + rgba[i][bComp] = ((p >> 2) & 0x3ff) * bs; + rgba[i][aComp] = ((p ) & 0x3 ) * as; + } + } + else { + const GLuint *uisrc = (const GLuint *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLuint p = uisrc[i]; + rgba[i][rComp] = ((p >> 22) ) * rs; + rgba[i][gComp] = ((p >> 12) & 0x3ff) * gs; + rgba[i][bComp] = ((p >> 2) & 0x3ff) * bs; + rgba[i][aComp] = ((p ) & 0x3 ) * as; + } + } + break; + case GL_UNSIGNED_INT_2_10_10_10_REV: + if (!intFormat) { + rs = 1.0F / 1023.0F; + gs = 1.0F / 1023.0F; + bs = 1.0F / 1023.0F; + as = 1.0F / 3.0F; + } + if (swapBytes) { + const GLuint *uisrc = (const GLuint *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLuint p = uisrc[i]; + SWAP4BYTE(p); + rgba[i][rComp] = ((p ) & 0x3ff) * rs; + rgba[i][gComp] = ((p >> 10) & 0x3ff) * gs; + rgba[i][bComp] = ((p >> 20) & 0x3ff) * bs; + rgba[i][aComp] = ((p >> 30) ) * as; + } + } + else { + const GLuint *uisrc = (const GLuint *) src; + GLuint i; + for (i = 0; i < n; i ++) { + GLuint p = uisrc[i]; + rgba[i][rComp] = ((p ) & 0x3ff) * rs; + rgba[i][gComp] = ((p >> 10) & 0x3ff) * gs; + rgba[i][bComp] = ((p >> 20) & 0x3ff) * bs; + rgba[i][aComp] = ((p >> 30) ) * as; + } + } + break; + default: + _mesa_problem(NULL, "bad srcType in extract float data"); + break; + } +} + + +/* + * Unpack a row of color image data from a client buffer according to + * the pixel unpacking parameters. + * Return GLchan values in the specified dest image format. + * This is used by glDrawPixels and glTexImage?D(). + * \param ctx - the context + * n - number of pixels in the span + * dstFormat - format of destination color array + * dest - the destination color array + * srcFormat - source image format + * srcType - source image data type + * source - source image pointer + * srcPacking - pixel unpacking parameters + * transferOps - bitmask of IMAGE_*_BIT values of operations to apply + * + * XXX perhaps expand this to process whole images someday. + */ +void +_mesa_unpack_color_span_chan( struct gl_context *ctx, + GLuint n, GLenum dstFormat, GLchan dest[], + GLenum srcFormat, GLenum srcType, + const GLvoid *source, + const struct gl_pixelstore_attrib *srcPacking, + GLbitfield transferOps ) +{ + ASSERT(dstFormat == GL_ALPHA || + dstFormat == GL_LUMINANCE || + dstFormat == GL_LUMINANCE_ALPHA || + dstFormat == GL_INTENSITY || + dstFormat == GL_RED || + dstFormat == GL_RG || + dstFormat == GL_RGB || + dstFormat == GL_RGBA || + dstFormat == GL_COLOR_INDEX); + + ASSERT(srcFormat == GL_RED || + srcFormat == GL_GREEN || + srcFormat == GL_BLUE || + srcFormat == GL_ALPHA || + srcFormat == GL_LUMINANCE || + srcFormat == GL_LUMINANCE_ALPHA || + srcFormat == GL_INTENSITY || + srcFormat == GL_RG || + srcFormat == GL_RGB || + srcFormat == GL_BGR || + srcFormat == GL_RGBA || + srcFormat == GL_BGRA || + srcFormat == GL_ABGR_EXT || + srcFormat == GL_COLOR_INDEX); + + ASSERT(srcType == GL_BITMAP || + srcType == GL_UNSIGNED_BYTE || + srcType == GL_BYTE || + srcType == GL_UNSIGNED_SHORT || + srcType == GL_SHORT || + srcType == GL_UNSIGNED_INT || + srcType == GL_INT || + srcType == GL_HALF_FLOAT_ARB || + srcType == GL_FLOAT || + srcType == GL_UNSIGNED_BYTE_3_3_2 || + srcType == GL_UNSIGNED_BYTE_2_3_3_REV || + srcType == GL_UNSIGNED_SHORT_5_6_5 || + srcType == GL_UNSIGNED_SHORT_5_6_5_REV || + srcType == GL_UNSIGNED_SHORT_4_4_4_4 || + srcType == GL_UNSIGNED_SHORT_4_4_4_4_REV || + srcType == GL_UNSIGNED_SHORT_5_5_5_1 || + srcType == GL_UNSIGNED_SHORT_1_5_5_5_REV || + srcType == GL_UNSIGNED_INT_8_8_8_8 || + srcType == GL_UNSIGNED_INT_8_8_8_8_REV || + srcType == GL_UNSIGNED_INT_10_10_10_2 || + srcType == GL_UNSIGNED_INT_2_10_10_10_REV); + + /* Try simple cases first */ + if (transferOps == 0) { + if (srcType == CHAN_TYPE) { + if (dstFormat == GL_RGBA) { + if (srcFormat == GL_RGBA) { + memcpy( dest, source, n * 4 * sizeof(GLchan) ); + return; + } + else if (srcFormat == GL_RGB) { + GLuint i; + const GLchan *src = (const GLchan *) source; + GLchan *dst = dest; + for (i = 0; i < n; i++) { + dst[0] = src[0]; + dst[1] = src[1]; + dst[2] = src[2]; + dst[3] = CHAN_MAX; + src += 3; + dst += 4; + } + return; + } + } + else if (dstFormat == GL_RGB) { + if (srcFormat == GL_RGB) { + memcpy( dest, source, n * 3 * sizeof(GLchan) ); + return; + } + else if (srcFormat == GL_RGBA) { + GLuint i; + const GLchan *src = (const GLchan *) source; + GLchan *dst = dest; + for (i = 0; i < n; i++) { + dst[0] = src[0]; + dst[1] = src[1]; + dst[2] = src[2]; + src += 4; + dst += 3; + } + return; + } + } + else if (dstFormat == srcFormat) { + GLint comps = _mesa_components_in_format(srcFormat); + assert(comps > 0); + memcpy( dest, source, n * comps * sizeof(GLchan) ); + return; + } + } + /* + * Common situation, loading 8bit RGBA/RGB source images + * into 16/32 bit destination. (OSMesa16/32) + */ + else if (srcType == GL_UNSIGNED_BYTE) { + if (dstFormat == GL_RGBA) { + if (srcFormat == GL_RGB) { + GLuint i; + const GLubyte *src = (const GLubyte *) source; + GLchan *dst = dest; + for (i = 0; i < n; i++) { + dst[0] = UBYTE_TO_CHAN(src[0]); + dst[1] = UBYTE_TO_CHAN(src[1]); + dst[2] = UBYTE_TO_CHAN(src[2]); + dst[3] = CHAN_MAX; + src += 3; + dst += 4; + } + return; + } + else if (srcFormat == GL_RGBA) { + GLuint i; + const GLubyte *src = (const GLubyte *) source; + GLchan *dst = dest; + for (i = 0; i < n; i++) { + dst[0] = UBYTE_TO_CHAN(src[0]); + dst[1] = UBYTE_TO_CHAN(src[1]); + dst[2] = UBYTE_TO_CHAN(src[2]); + dst[3] = UBYTE_TO_CHAN(src[3]); + src += 4; + dst += 4; + } + return; + } + } + else if (dstFormat == GL_RGB) { + if (srcFormat == GL_RGB) { + GLuint i; + const GLubyte *src = (const GLubyte *) source; + GLchan *dst = dest; + for (i = 0; i < n; i++) { + dst[0] = UBYTE_TO_CHAN(src[0]); + dst[1] = UBYTE_TO_CHAN(src[1]); + dst[2] = UBYTE_TO_CHAN(src[2]); + src += 3; + dst += 3; + } + return; + } + else if (srcFormat == GL_RGBA) { + GLuint i; + const GLubyte *src = (const GLubyte *) source; + GLchan *dst = dest; + for (i = 0; i < n; i++) { + dst[0] = UBYTE_TO_CHAN(src[0]); + dst[1] = UBYTE_TO_CHAN(src[1]); + dst[2] = UBYTE_TO_CHAN(src[2]); + src += 4; + dst += 3; + } + return; + } + } + } + } + + + /* general solution begins here */ + { + GLint dstComponents; + GLint dstRedIndex, dstGreenIndex, dstBlueIndex, dstAlphaIndex; + GLint dstLuminanceIndex, dstIntensityIndex; + GLfloat rgba[MAX_WIDTH][4]; + + dstComponents = _mesa_components_in_format( dstFormat ); + /* source & dest image formats should have been error checked by now */ + assert(dstComponents > 0); + + /* + * Extract image data and convert to RGBA floats + */ + assert(n <= MAX_WIDTH); + if (srcFormat == GL_COLOR_INDEX) { + GLuint indexes[MAX_WIDTH]; + extract_uint_indexes(n, indexes, srcFormat, srcType, source, + srcPacking); + + if (dstFormat == GL_COLOR_INDEX) { + GLuint i; + _mesa_apply_ci_transfer_ops(ctx, transferOps, n, indexes); + /* convert to GLchan and return */ + for (i = 0; i < n; i++) { + dest[i] = (GLchan) (indexes[i] & 0xff); + } + return; + } + else { + /* Convert indexes to RGBA */ + if (transferOps & IMAGE_SHIFT_OFFSET_BIT) { + _mesa_shift_and_offset_ci(ctx, n, indexes); + } + _mesa_map_ci_to_rgba(ctx, n, indexes, rgba); + } + + /* Don't do RGBA scale/bias or RGBA->RGBA mapping if starting + * with color indexes. + */ + transferOps &= ~(IMAGE_SCALE_BIAS_BIT | IMAGE_MAP_COLOR_BIT); + } + else { + /* non-color index data */ + extract_float_rgba(n, rgba, srcFormat, srcType, source, + srcPacking->SwapBytes); + } + + /* Need to clamp if returning GLubytes or GLushorts */ +#if CHAN_TYPE != GL_FLOAT + transferOps |= IMAGE_CLAMP_BIT; +#endif + + if (transferOps) { + _mesa_apply_rgba_transfer_ops(ctx, transferOps, n, rgba); + } + + /* Now determine which color channels we need to produce. + * And determine the dest index (offset) within each color tuple. + */ + switch (dstFormat) { + case GL_ALPHA: + dstAlphaIndex = 0; + dstRedIndex = dstGreenIndex = dstBlueIndex = -1; + dstLuminanceIndex = dstIntensityIndex = -1; + break; + case GL_LUMINANCE: + dstLuminanceIndex = 0; + dstRedIndex = dstGreenIndex = dstBlueIndex = dstAlphaIndex = -1; + dstIntensityIndex = -1; + break; + case GL_LUMINANCE_ALPHA: + dstLuminanceIndex = 0; + dstAlphaIndex = 1; + dstRedIndex = dstGreenIndex = dstBlueIndex = -1; + dstIntensityIndex = -1; + break; + case GL_INTENSITY: + dstIntensityIndex = 0; + dstRedIndex = dstGreenIndex = dstBlueIndex = dstAlphaIndex = -1; + dstLuminanceIndex = -1; + break; + case GL_RED: + dstRedIndex = 0; + dstGreenIndex = dstBlueIndex = -1; + dstAlphaIndex = dstLuminanceIndex = dstIntensityIndex = -1; + break; + case GL_RG: + dstRedIndex = 0; + dstGreenIndex = 1; + dstBlueIndex = -1; + dstAlphaIndex = dstLuminanceIndex = dstIntensityIndex = -1; + break; + case GL_RGB: + dstRedIndex = 0; + dstGreenIndex = 1; + dstBlueIndex = 2; + dstAlphaIndex = dstLuminanceIndex = dstIntensityIndex = -1; + break; + case GL_RGBA: + dstRedIndex = 0; + dstGreenIndex = 1; + dstBlueIndex = 2; + dstAlphaIndex = 3; + dstLuminanceIndex = dstIntensityIndex = -1; + break; + default: + _mesa_problem(ctx, "bad dstFormat in _mesa_unpack_chan_span()"); + return; + } + + + /* Now return the GLchan data in the requested dstFormat */ + + if (dstRedIndex >= 0) { + GLchan *dst = dest; + GLuint i; + for (i = 0; i < n; i++) { + CLAMPED_FLOAT_TO_CHAN(dst[dstRedIndex], rgba[i][RCOMP]); + dst += dstComponents; + } + } + + if (dstGreenIndex >= 0) { + GLchan *dst = dest; + GLuint i; + for (i = 0; i < n; i++) { + CLAMPED_FLOAT_TO_CHAN(dst[dstGreenIndex], rgba[i][GCOMP]); + dst += dstComponents; + } + } + + if (dstBlueIndex >= 0) { + GLchan *dst = dest; + GLuint i; + for (i = 0; i < n; i++) { + CLAMPED_FLOAT_TO_CHAN(dst[dstBlueIndex], rgba[i][BCOMP]); + dst += dstComponents; + } + } + + if (dstAlphaIndex >= 0) { + GLchan *dst = dest; + GLuint i; + for (i = 0; i < n; i++) { + CLAMPED_FLOAT_TO_CHAN(dst[dstAlphaIndex], rgba[i][ACOMP]); + dst += dstComponents; + } + } + + if (dstIntensityIndex >= 0) { + GLchan *dst = dest; + GLuint i; + assert(dstIntensityIndex == 0); + assert(dstComponents == 1); + for (i = 0; i < n; i++) { + /* Intensity comes from red channel */ + CLAMPED_FLOAT_TO_CHAN(dst[i], rgba[i][RCOMP]); + } + } + + if (dstLuminanceIndex >= 0) { + GLchan *dst = dest; + GLuint i; + assert(dstLuminanceIndex == 0); + for (i = 0; i < n; i++) { + /* Luminance comes from red channel */ + CLAMPED_FLOAT_TO_CHAN(dst[0], rgba[i][RCOMP]); + dst += dstComponents; + } + } + } +} + + +/** + * Same as _mesa_unpack_color_span_chan(), but return GLfloat data + * instead of GLchan. + */ +void +_mesa_unpack_color_span_float( struct gl_context *ctx, + GLuint n, GLenum dstFormat, GLfloat dest[], + GLenum srcFormat, GLenum srcType, + const GLvoid *source, + const struct gl_pixelstore_attrib *srcPacking, + GLbitfield transferOps ) +{ + ASSERT(dstFormat == GL_ALPHA || + dstFormat == GL_LUMINANCE || + dstFormat == GL_LUMINANCE_ALPHA || + dstFormat == GL_INTENSITY || + dstFormat == GL_RED || + dstFormat == GL_RG || + dstFormat == GL_RGB || + dstFormat == GL_RGBA || + dstFormat == GL_COLOR_INDEX); + + ASSERT(srcFormat == GL_RED || + srcFormat == GL_GREEN || + srcFormat == GL_BLUE || + srcFormat == GL_ALPHA || + srcFormat == GL_LUMINANCE || + srcFormat == GL_LUMINANCE_ALPHA || + srcFormat == GL_INTENSITY || + srcFormat == GL_RG || + srcFormat == GL_RGB || + srcFormat == GL_BGR || + srcFormat == GL_RGBA || + srcFormat == GL_BGRA || + srcFormat == GL_ABGR_EXT || + srcFormat == GL_RED_INTEGER_EXT || + srcFormat == GL_GREEN_INTEGER_EXT || + srcFormat == GL_BLUE_INTEGER_EXT || + srcFormat == GL_ALPHA_INTEGER_EXT || + srcFormat == GL_RGB_INTEGER_EXT || + srcFormat == GL_RGBA_INTEGER_EXT || + srcFormat == GL_BGR_INTEGER_EXT || + srcFormat == GL_BGRA_INTEGER_EXT || + srcFormat == GL_LUMINANCE_INTEGER_EXT || + srcFormat == GL_LUMINANCE_ALPHA_INTEGER_EXT || + srcFormat == GL_COLOR_INDEX); + + ASSERT(srcType == GL_BITMAP || + srcType == GL_UNSIGNED_BYTE || + srcType == GL_BYTE || + srcType == GL_UNSIGNED_SHORT || + srcType == GL_SHORT || + srcType == GL_UNSIGNED_INT || + srcType == GL_INT || + srcType == GL_HALF_FLOAT_ARB || + srcType == GL_FLOAT || + srcType == GL_UNSIGNED_BYTE_3_3_2 || + srcType == GL_UNSIGNED_BYTE_2_3_3_REV || + srcType == GL_UNSIGNED_SHORT_5_6_5 || + srcType == GL_UNSIGNED_SHORT_5_6_5_REV || + srcType == GL_UNSIGNED_SHORT_4_4_4_4 || + srcType == GL_UNSIGNED_SHORT_4_4_4_4_REV || + srcType == GL_UNSIGNED_SHORT_5_5_5_1 || + srcType == GL_UNSIGNED_SHORT_1_5_5_5_REV || + srcType == GL_UNSIGNED_INT_8_8_8_8 || + srcType == GL_UNSIGNED_INT_8_8_8_8_REV || + srcType == GL_UNSIGNED_INT_10_10_10_2 || + srcType == GL_UNSIGNED_INT_2_10_10_10_REV); + + /* general solution, no special cases, yet */ + { + GLint dstComponents; + GLint dstRedIndex, dstGreenIndex, dstBlueIndex, dstAlphaIndex; + GLint dstLuminanceIndex, dstIntensityIndex; + GLfloat rgba[MAX_WIDTH][4]; + + dstComponents = _mesa_components_in_format( dstFormat ); + /* source & dest image formats should have been error checked by now */ + assert(dstComponents > 0); + + /* + * Extract image data and convert to RGBA floats + */ + assert(n <= MAX_WIDTH); + if (srcFormat == GL_COLOR_INDEX) { + GLuint indexes[MAX_WIDTH]; + extract_uint_indexes(n, indexes, srcFormat, srcType, source, + srcPacking); + + if (dstFormat == GL_COLOR_INDEX) { + GLuint i; + _mesa_apply_ci_transfer_ops(ctx, transferOps, n, indexes); + /* convert to GLchan and return */ + for (i = 0; i < n; i++) { + dest[i] = (GLchan) (indexes[i] & 0xff); + } + return; + } + else { + /* Convert indexes to RGBA */ + if (transferOps & IMAGE_SHIFT_OFFSET_BIT) { + _mesa_shift_and_offset_ci(ctx, n, indexes); + } + _mesa_map_ci_to_rgba(ctx, n, indexes, rgba); + } + + /* Don't do RGBA scale/bias or RGBA->RGBA mapping if starting + * with color indexes. + */ + transferOps &= ~(IMAGE_SCALE_BIAS_BIT | IMAGE_MAP_COLOR_BIT); + } + else { + /* non-color index data */ + extract_float_rgba(n, rgba, srcFormat, srcType, source, + srcPacking->SwapBytes); + } + + if (transferOps) { + _mesa_apply_rgba_transfer_ops(ctx, transferOps, n, rgba); + } + + /* Now determine which color channels we need to produce. + * And determine the dest index (offset) within each color tuple. + */ + switch (dstFormat) { + case GL_ALPHA: + dstAlphaIndex = 0; + dstRedIndex = dstGreenIndex = dstBlueIndex = -1; + dstLuminanceIndex = dstIntensityIndex = -1; + break; + case GL_LUMINANCE: + dstLuminanceIndex = 0; + dstRedIndex = dstGreenIndex = dstBlueIndex = dstAlphaIndex = -1; + dstIntensityIndex = -1; + break; + case GL_LUMINANCE_ALPHA: + dstLuminanceIndex = 0; + dstAlphaIndex = 1; + dstRedIndex = dstGreenIndex = dstBlueIndex = -1; + dstIntensityIndex = -1; + break; + case GL_INTENSITY: + dstIntensityIndex = 0; + dstRedIndex = dstGreenIndex = dstBlueIndex = dstAlphaIndex = -1; + dstLuminanceIndex = -1; + break; + case GL_RED: + dstRedIndex = 0; + dstGreenIndex = dstBlueIndex = -1; + dstAlphaIndex = dstLuminanceIndex = dstIntensityIndex = -1; + break; + case GL_RG: + dstRedIndex = 0; + dstGreenIndex = 1; + dstBlueIndex = -1; + dstAlphaIndex = dstLuminanceIndex = dstIntensityIndex = -1; + break; + case GL_RGB: + dstRedIndex = 0; + dstGreenIndex = 1; + dstBlueIndex = 2; + dstAlphaIndex = dstLuminanceIndex = dstIntensityIndex = -1; + break; + case GL_RGBA: + dstRedIndex = 0; + dstGreenIndex = 1; + dstBlueIndex = 2; + dstAlphaIndex = 3; + dstLuminanceIndex = dstIntensityIndex = -1; + break; + default: + _mesa_problem(ctx, "bad dstFormat in _mesa_unpack_color_span_float()"); + return; + } + + /* Now pack results in the requested dstFormat */ + if (dstRedIndex >= 0) { + GLfloat *dst = dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[dstRedIndex] = rgba[i][RCOMP]; + dst += dstComponents; + } + } + + if (dstGreenIndex >= 0) { + GLfloat *dst = dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[dstGreenIndex] = rgba[i][GCOMP]; + dst += dstComponents; + } + } + + if (dstBlueIndex >= 0) { + GLfloat *dst = dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[dstBlueIndex] = rgba[i][BCOMP]; + dst += dstComponents; + } + } + + if (dstAlphaIndex >= 0) { + GLfloat *dst = dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[dstAlphaIndex] = rgba[i][ACOMP]; + dst += dstComponents; + } + } + + if (dstIntensityIndex >= 0) { + GLfloat *dst = dest; + GLuint i; + assert(dstIntensityIndex == 0); + assert(dstComponents == 1); + for (i = 0; i < n; i++) { + /* Intensity comes from red channel */ + dst[i] = rgba[i][RCOMP]; + } + } + + if (dstLuminanceIndex >= 0) { + GLfloat *dst = dest; + GLuint i; + assert(dstLuminanceIndex == 0); + for (i = 0; i < n; i++) { + /* Luminance comes from red channel */ + dst[0] = rgba[i][RCOMP]; + dst += dstComponents; + } + } + } +} + +/** + * Similar to _mesa_unpack_color_span_float(), but for dudv data instead of rgba, + * directly return GLbyte data, no transfer ops apply. + */ +void +_mesa_unpack_dudv_span_byte( struct gl_context *ctx, + GLuint n, GLenum dstFormat, GLbyte dest[], + GLenum srcFormat, GLenum srcType, + const GLvoid *source, + const struct gl_pixelstore_attrib *srcPacking, + GLbitfield transferOps ) +{ + ASSERT(dstFormat == GL_DUDV_ATI); + ASSERT(srcFormat == GL_DUDV_ATI); + + ASSERT(srcType == GL_UNSIGNED_BYTE || + srcType == GL_BYTE || + srcType == GL_UNSIGNED_SHORT || + srcType == GL_SHORT || + srcType == GL_UNSIGNED_INT || + srcType == GL_INT || + srcType == GL_HALF_FLOAT_ARB || + srcType == GL_FLOAT); + + /* general solution */ + { + GLint dstComponents; + GLfloat rgba[MAX_WIDTH][4]; + GLbyte *dst = dest; + GLuint i; + + dstComponents = _mesa_components_in_format( dstFormat ); + /* source & dest image formats should have been error checked by now */ + assert(dstComponents > 0); + + /* + * Extract image data and convert to RGBA floats + */ + assert(n <= MAX_WIDTH); + extract_float_rgba(n, rgba, srcFormat, srcType, source, + srcPacking->SwapBytes); + + + /* Now determine which color channels we need to produce. + * And determine the dest index (offset) within each color tuple. + */ + + /* Now pack results in the requested dstFormat */ + for (i = 0; i < n; i++) { + /* not sure - need clamp[-1,1] here? */ + dst[0] = FLOAT_TO_BYTE(rgba[i][RCOMP]); + dst[1] = FLOAT_TO_BYTE(rgba[i][GCOMP]); + dst += dstComponents; + } + } +} + +/* + * Unpack a row of color index data from a client buffer according to + * the pixel unpacking parameters. + * This is (or will be) used by glDrawPixels, glTexImage[123]D, etc. + * + * Args: ctx - the context + * n - number of pixels + * dstType - destination data type + * dest - destination array + * srcType - source pixel type + * source - source data pointer + * srcPacking - pixel unpacking parameters + * transferOps - the pixel transfer operations to apply + */ +void +_mesa_unpack_index_span( const struct gl_context *ctx, GLuint n, + GLenum dstType, GLvoid *dest, + GLenum srcType, const GLvoid *source, + const struct gl_pixelstore_attrib *srcPacking, + GLbitfield transferOps ) +{ + ASSERT(srcType == GL_BITMAP || + srcType == GL_UNSIGNED_BYTE || + srcType == GL_BYTE || + srcType == GL_UNSIGNED_SHORT || + srcType == GL_SHORT || + srcType == GL_UNSIGNED_INT || + srcType == GL_INT || + srcType == GL_HALF_FLOAT_ARB || + srcType == GL_FLOAT); + + ASSERT(dstType == GL_UNSIGNED_BYTE || + dstType == GL_UNSIGNED_SHORT || + dstType == GL_UNSIGNED_INT); + + + transferOps &= (IMAGE_MAP_COLOR_BIT | IMAGE_SHIFT_OFFSET_BIT); + + /* + * Try simple cases first + */ + if (transferOps == 0 && srcType == GL_UNSIGNED_BYTE + && dstType == GL_UNSIGNED_BYTE) { + memcpy(dest, source, n * sizeof(GLubyte)); + } + else if (transferOps == 0 && srcType == GL_UNSIGNED_INT + && dstType == GL_UNSIGNED_INT && !srcPacking->SwapBytes) { + memcpy(dest, source, n * sizeof(GLuint)); + } + else { + /* + * general solution + */ + GLuint indexes[MAX_WIDTH]; + assert(n <= MAX_WIDTH); + + extract_uint_indexes(n, indexes, GL_COLOR_INDEX, srcType, source, + srcPacking); + + if (transferOps) + _mesa_apply_ci_transfer_ops(ctx, transferOps, n, indexes); + + /* convert to dest type */ + switch (dstType) { + case GL_UNSIGNED_BYTE: + { + GLubyte *dst = (GLubyte *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = (GLubyte) (indexes[i] & 0xff); + } + } + break; + case GL_UNSIGNED_SHORT: + { + GLuint *dst = (GLuint *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = (GLushort) (indexes[i] & 0xffff); + } + } + break; + case GL_UNSIGNED_INT: + memcpy(dest, indexes, n * sizeof(GLuint)); + break; + default: + _mesa_problem(ctx, "bad dstType in _mesa_unpack_index_span"); + } + } +} + + +void +_mesa_pack_index_span( const struct gl_context *ctx, GLuint n, + GLenum dstType, GLvoid *dest, const GLuint *source, + const struct gl_pixelstore_attrib *dstPacking, + GLbitfield transferOps ) +{ + GLuint indexes[MAX_WIDTH]; + + ASSERT(n <= MAX_WIDTH); + + transferOps &= (IMAGE_MAP_COLOR_BIT | IMAGE_SHIFT_OFFSET_BIT); + + if (transferOps & (IMAGE_MAP_COLOR_BIT | IMAGE_SHIFT_OFFSET_BIT)) { + /* make a copy of input */ + memcpy(indexes, source, n * sizeof(GLuint)); + _mesa_apply_ci_transfer_ops(ctx, transferOps, n, indexes); + source = indexes; + } + + switch (dstType) { + case GL_UNSIGNED_BYTE: + { + GLubyte *dst = (GLubyte *) dest; + GLuint i; + for (i = 0; i < n; i++) { + *dst++ = (GLubyte) source[i]; + } + } + break; + case GL_BYTE: + { + GLbyte *dst = (GLbyte *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = (GLbyte) source[i]; + } + } + break; + case GL_UNSIGNED_SHORT: + { + GLushort *dst = (GLushort *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = (GLushort) source[i]; + } + if (dstPacking->SwapBytes) { + _mesa_swap2( (GLushort *) dst, n ); + } + } + break; + case GL_SHORT: + { + GLshort *dst = (GLshort *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = (GLshort) source[i]; + } + if (dstPacking->SwapBytes) { + _mesa_swap2( (GLushort *) dst, n ); + } + } + break; + case GL_UNSIGNED_INT: + { + GLuint *dst = (GLuint *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = (GLuint) source[i]; + } + if (dstPacking->SwapBytes) { + _mesa_swap4( (GLuint *) dst, n ); + } + } + break; + case GL_INT: + { + GLint *dst = (GLint *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = (GLint) source[i]; + } + if (dstPacking->SwapBytes) { + _mesa_swap4( (GLuint *) dst, n ); + } + } + break; + case GL_FLOAT: + { + GLfloat *dst = (GLfloat *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = (GLfloat) source[i]; + } + if (dstPacking->SwapBytes) { + _mesa_swap4( (GLuint *) dst, n ); + } + } + break; + case GL_HALF_FLOAT_ARB: + { + GLhalfARB *dst = (GLhalfARB *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = _mesa_float_to_half((GLfloat) source[i]); + } + if (dstPacking->SwapBytes) { + _mesa_swap2( (GLushort *) dst, n ); + } + } + break; + default: + _mesa_problem(ctx, "bad type in _mesa_pack_index_span"); + } +} + + +/* + * Unpack a row of stencil data from a client buffer according to + * the pixel unpacking parameters. + * This is (or will be) used by glDrawPixels + * + * Args: ctx - the context + * n - number of pixels + * dstType - destination data type + * dest - destination array + * srcType - source pixel type + * source - source data pointer + * srcPacking - pixel unpacking parameters + * transferOps - apply offset/bias/lookup ops? + */ +void +_mesa_unpack_stencil_span( const struct gl_context *ctx, GLuint n, + GLenum dstType, GLvoid *dest, + GLenum srcType, const GLvoid *source, + const struct gl_pixelstore_attrib *srcPacking, + GLbitfield transferOps ) +{ + ASSERT(srcType == GL_BITMAP || + srcType == GL_UNSIGNED_BYTE || + srcType == GL_BYTE || + srcType == GL_UNSIGNED_SHORT || + srcType == GL_SHORT || + srcType == GL_UNSIGNED_INT || + srcType == GL_INT || + srcType == GL_UNSIGNED_INT_24_8_EXT || + srcType == GL_HALF_FLOAT_ARB || + srcType == GL_FLOAT); + + ASSERT(dstType == GL_UNSIGNED_BYTE || + dstType == GL_UNSIGNED_SHORT || + dstType == GL_UNSIGNED_INT); + + /* only shift and offset apply to stencil */ + transferOps &= IMAGE_SHIFT_OFFSET_BIT; + + /* + * Try simple cases first + */ + if (transferOps == 0 && + !ctx->Pixel.MapStencilFlag && + srcType == GL_UNSIGNED_BYTE && + dstType == GL_UNSIGNED_BYTE) { + memcpy(dest, source, n * sizeof(GLubyte)); + } + else if (transferOps == 0 && + !ctx->Pixel.MapStencilFlag && + srcType == GL_UNSIGNED_INT && + dstType == GL_UNSIGNED_INT && + !srcPacking->SwapBytes) { + memcpy(dest, source, n * sizeof(GLuint)); + } + else { + /* + * general solution + */ + GLuint indexes[MAX_WIDTH]; + assert(n <= MAX_WIDTH); + + extract_uint_indexes(n, indexes, GL_STENCIL_INDEX, srcType, source, + srcPacking); + + if (transferOps & IMAGE_SHIFT_OFFSET_BIT) { + /* shift and offset indexes */ + _mesa_shift_and_offset_ci(ctx, n, indexes); + } + + if (ctx->Pixel.MapStencilFlag) { + /* Apply stencil lookup table */ + const GLuint mask = ctx->PixelMaps.StoS.Size - 1; + GLuint i; + for (i = 0; i < n; i++) { + indexes[i] = (GLuint)ctx->PixelMaps.StoS.Map[ indexes[i] & mask ]; + } + } + + /* convert to dest type */ + switch (dstType) { + case GL_UNSIGNED_BYTE: + { + GLubyte *dst = (GLubyte *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = (GLubyte) (indexes[i] & 0xff); + } + } + break; + case GL_UNSIGNED_SHORT: + { + GLuint *dst = (GLuint *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = (GLushort) (indexes[i] & 0xffff); + } + } + break; + case GL_UNSIGNED_INT: + memcpy(dest, indexes, n * sizeof(GLuint)); + break; + default: + _mesa_problem(ctx, "bad dstType in _mesa_unpack_stencil_span"); + } + } +} + + +void +_mesa_pack_stencil_span( const struct gl_context *ctx, GLuint n, + GLenum dstType, GLvoid *dest, const GLstencil *source, + const struct gl_pixelstore_attrib *dstPacking ) +{ + GLstencil stencil[MAX_WIDTH]; + + ASSERT(n <= MAX_WIDTH); + + if (ctx->Pixel.IndexShift || ctx->Pixel.IndexOffset || + ctx->Pixel.MapStencilFlag) { + /* make a copy of input */ + memcpy(stencil, source, n * sizeof(GLstencil)); + _mesa_apply_stencil_transfer_ops(ctx, n, stencil); + source = stencil; + } + + switch (dstType) { + case GL_UNSIGNED_BYTE: + if (sizeof(GLstencil) == 1) { + memcpy( dest, source, n ); + } + else { + GLubyte *dst = (GLubyte *) dest; + GLuint i; + for (i=0;i<n;i++) { + dst[i] = (GLubyte) source[i]; + } + } + break; + case GL_BYTE: + { + GLbyte *dst = (GLbyte *) dest; + GLuint i; + for (i=0;i<n;i++) { + dst[i] = (GLbyte) (source[i] & 0x7f); + } + } + break; + case GL_UNSIGNED_SHORT: + { + GLushort *dst = (GLushort *) dest; + GLuint i; + for (i=0;i<n;i++) { + dst[i] = (GLushort) source[i]; + } + if (dstPacking->SwapBytes) { + _mesa_swap2( (GLushort *) dst, n ); + } + } + break; + case GL_SHORT: + { + GLshort *dst = (GLshort *) dest; + GLuint i; + for (i=0;i<n;i++) { + dst[i] = (GLshort) source[i]; + } + if (dstPacking->SwapBytes) { + _mesa_swap2( (GLushort *) dst, n ); + } + } + break; + case GL_UNSIGNED_INT: + { + GLuint *dst = (GLuint *) dest; + GLuint i; + for (i=0;i<n;i++) { + dst[i] = (GLuint) source[i]; + } + if (dstPacking->SwapBytes) { + _mesa_swap4( (GLuint *) dst, n ); + } + } + break; + case GL_INT: + { + GLint *dst = (GLint *) dest; + GLuint i; + for (i=0;i<n;i++) { + dst[i] = (GLint) source[i]; + } + if (dstPacking->SwapBytes) { + _mesa_swap4( (GLuint *) dst, n ); + } + } + break; + case GL_FLOAT: + { + GLfloat *dst = (GLfloat *) dest; + GLuint i; + for (i=0;i<n;i++) { + dst[i] = (GLfloat) source[i]; + } + if (dstPacking->SwapBytes) { + _mesa_swap4( (GLuint *) dst, n ); + } + } + break; + case GL_HALF_FLOAT_ARB: + { + GLhalfARB *dst = (GLhalfARB *) dest; + GLuint i; + for (i=0;i<n;i++) { + dst[i] = _mesa_float_to_half( (float) source[i] ); + } + if (dstPacking->SwapBytes) { + _mesa_swap2( (GLushort *) dst, n ); + } + } + break; + case GL_BITMAP: + if (dstPacking->LsbFirst) { + GLubyte *dst = (GLubyte *) dest; + GLint shift = 0; + GLuint i; + for (i = 0; i < n; i++) { + if (shift == 0) + *dst = 0; + *dst |= ((source[i] != 0) << shift); + shift++; + if (shift == 8) { + shift = 0; + dst++; + } + } + } + else { + GLubyte *dst = (GLubyte *) dest; + GLint shift = 7; + GLuint i; + for (i = 0; i < n; i++) { + if (shift == 7) + *dst = 0; + *dst |= ((source[i] != 0) << shift); + shift--; + if (shift < 0) { + shift = 7; + dst++; + } + } + } + break; + default: + _mesa_problem(ctx, "bad type in _mesa_pack_index_span"); + } +} + +#define DEPTH_VALUES(GLTYPE, GLTYPE2FLOAT) \ + do { \ + GLuint i; \ + const GLTYPE *src = (const GLTYPE *)source; \ + for (i = 0; i < n; i++) { \ + GLTYPE value = src[i]; \ + if (srcPacking->SwapBytes) { \ + if (sizeof(GLTYPE) == 2) { \ + SWAP2BYTE(value); \ + } else if (sizeof(GLTYPE) == 4) { \ + SWAP4BYTE(value); \ + } \ + } \ + depthValues[i] = GLTYPE2FLOAT(value); \ + } \ + } while (0) + + +/** + * Unpack a row of depth/z values from memory, returning GLushort, GLuint + * or GLfloat values. + * The glPixelTransfer (scale/bias) params will be applied. + * + * \param dstType one of GL_UNSIGNED_SHORT, GL_UNSIGNED_INT, GL_FLOAT + * \param depthMax max value for returned GLushort or GLuint values + * (ignored for GLfloat). + */ +void +_mesa_unpack_depth_span( const struct gl_context *ctx, GLuint n, + GLenum dstType, GLvoid *dest, GLuint depthMax, + GLenum srcType, const GLvoid *source, + const struct gl_pixelstore_attrib *srcPacking ) +{ + GLfloat depthTemp[MAX_WIDTH], *depthValues; + GLboolean needClamp = GL_FALSE; + + /* Look for special cases first. + * Not only are these faster, they're less prone to numeric conversion + * problems. Otherwise, converting from an int type to a float then + * back to an int type can introduce errors that will show up as + * artifacts in things like depth peeling which uses glCopyTexImage. + */ + if (ctx->Pixel.DepthScale == 1.0 && ctx->Pixel.DepthBias == 0.0) { + if (srcType == GL_UNSIGNED_INT && dstType == GL_UNSIGNED_SHORT) { + const GLuint *src = (const GLuint *) source; + GLushort *dst = (GLushort *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = src[i] >> 16; + } + return; + } + if (srcType == GL_UNSIGNED_SHORT + && dstType == GL_UNSIGNED_INT + && depthMax == 0xffffffff) { + const GLushort *src = (const GLushort *) source; + GLuint *dst = (GLuint *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = src[i] | (src[i] << 16); + } + return; + } + if (srcType == GL_UNSIGNED_INT_24_8 + && dstType == GL_UNSIGNED_INT + && depthMax == 0xffffff) { + const GLuint *src = (const GLuint *) source; + GLuint *dst = (GLuint *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = src[i] >> 8; + } + return; + } + /* XXX may want to add additional cases here someday */ + } + + /* general case path follows */ + + if (dstType == GL_FLOAT) { + depthValues = (GLfloat *) dest; + } + else { + depthValues = depthTemp; + } + + /* Convert incoming values to GLfloat. Some conversions will require + * clamping, below. + */ + switch (srcType) { + case GL_BYTE: + DEPTH_VALUES(GLbyte, BYTE_TO_FLOAT); + needClamp = GL_TRUE; + break; + case GL_UNSIGNED_BYTE: + DEPTH_VALUES(GLubyte, UBYTE_TO_FLOAT); + break; + case GL_SHORT: + DEPTH_VALUES(GLshort, SHORT_TO_FLOAT); + needClamp = GL_TRUE; + break; + case GL_UNSIGNED_SHORT: + DEPTH_VALUES(GLushort, USHORT_TO_FLOAT); + break; + case GL_INT: + DEPTH_VALUES(GLint, INT_TO_FLOAT); + needClamp = GL_TRUE; + break; + case GL_UNSIGNED_INT: + DEPTH_VALUES(GLuint, UINT_TO_FLOAT); + break; + case GL_UNSIGNED_INT_24_8_EXT: /* GL_EXT_packed_depth_stencil */ + if (dstType == GL_UNSIGNED_INT_24_8_EXT && + depthMax == 0xffffff && + ctx->Pixel.DepthScale == 1.0 && + ctx->Pixel.DepthBias == 0.0) { + const GLuint *src = (const GLuint *) source; + GLuint *zValues = (GLuint *) dest; + GLuint i; + for (i = 0; i < n; i++) { + GLuint value = src[i]; + if (srcPacking->SwapBytes) { + SWAP4BYTE(value); + } + zValues[i] = value & 0xffffff00; + } + return; + } + else { + const GLuint *src = (const GLuint *) source; + const GLfloat scale = 1.0f / 0xffffff; + GLuint i; + for (i = 0; i < n; i++) { + GLuint value = src[i]; + if (srcPacking->SwapBytes) { + SWAP4BYTE(value); + } + depthValues[i] = (value >> 8) * scale; + } + } + break; + case GL_FLOAT: + DEPTH_VALUES(GLfloat, 1*); + needClamp = GL_TRUE; + break; + case GL_HALF_FLOAT_ARB: + { + GLuint i; + const GLhalfARB *src = (const GLhalfARB *) source; + for (i = 0; i < n; i++) { + GLhalfARB value = src[i]; + if (srcPacking->SwapBytes) { + SWAP2BYTE(value); + } + depthValues[i] = _mesa_half_to_float(value); + } + needClamp = GL_TRUE; + } + break; + default: + _mesa_problem(NULL, "bad type in _mesa_unpack_depth_span()"); + return; + } + + /* apply depth scale and bias */ + { + const GLfloat scale = ctx->Pixel.DepthScale; + const GLfloat bias = ctx->Pixel.DepthBias; + if (scale != 1.0 || bias != 0.0) { + GLuint i; + for (i = 0; i < n; i++) { + depthValues[i] = depthValues[i] * scale + bias; + } + needClamp = GL_TRUE; + } + } + + /* clamp to [0, 1] */ + if (needClamp) { + GLuint i; + for (i = 0; i < n; i++) { + depthValues[i] = (GLfloat)CLAMP(depthValues[i], 0.0, 1.0); + } + } + + /* + * Convert values to dstType + */ + if (dstType == GL_UNSIGNED_INT) { + GLuint *zValues = (GLuint *) dest; + GLuint i; + if (depthMax <= 0xffffff) { + /* no overflow worries */ + for (i = 0; i < n; i++) { + zValues[i] = (GLuint) (depthValues[i] * (GLfloat) depthMax); + } + } + else { + /* need to use double precision to prevent overflow problems */ + for (i = 0; i < n; i++) { + GLdouble z = depthValues[i] * (GLfloat) depthMax; + if (z >= (GLdouble) 0xffffffff) + zValues[i] = 0xffffffff; + else + zValues[i] = (GLuint) z; + } + } + } + else if (dstType == GL_UNSIGNED_SHORT) { + GLushort *zValues = (GLushort *) dest; + GLuint i; + ASSERT(depthMax <= 0xffff); + for (i = 0; i < n; i++) { + zValues[i] = (GLushort) (depthValues[i] * (GLfloat) depthMax); + } + } + else { + ASSERT(dstType == GL_FLOAT); + /*ASSERT(depthMax == 1.0F);*/ + } +} + + +/* + * Pack an array of depth values. The values are floats in [0,1]. + */ +void +_mesa_pack_depth_span( const struct gl_context *ctx, GLuint n, GLvoid *dest, + GLenum dstType, const GLfloat *depthSpan, + const struct gl_pixelstore_attrib *dstPacking ) +{ + GLfloat depthCopy[MAX_WIDTH]; + + ASSERT(n <= MAX_WIDTH); + + if (ctx->Pixel.DepthScale != 1.0 || ctx->Pixel.DepthBias != 0.0) { + memcpy(depthCopy, depthSpan, n * sizeof(GLfloat)); + _mesa_scale_and_bias_depth(ctx, n, depthCopy); + depthSpan = depthCopy; + } + + switch (dstType) { + case GL_UNSIGNED_BYTE: + { + GLubyte *dst = (GLubyte *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = FLOAT_TO_UBYTE( depthSpan[i] ); + } + } + break; + case GL_BYTE: + { + GLbyte *dst = (GLbyte *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = FLOAT_TO_BYTE( depthSpan[i] ); + } + } + break; + case GL_UNSIGNED_SHORT: + { + GLushort *dst = (GLushort *) dest; + GLuint i; + for (i = 0; i < n; i++) { + CLAMPED_FLOAT_TO_USHORT(dst[i], depthSpan[i]); + } + if (dstPacking->SwapBytes) { + _mesa_swap2( (GLushort *) dst, n ); + } + } + break; + case GL_SHORT: + { + GLshort *dst = (GLshort *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = FLOAT_TO_SHORT( depthSpan[i] ); + } + if (dstPacking->SwapBytes) { + _mesa_swap2( (GLushort *) dst, n ); + } + } + break; + case GL_UNSIGNED_INT: + { + GLuint *dst = (GLuint *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = FLOAT_TO_UINT( depthSpan[i] ); + } + if (dstPacking->SwapBytes) { + _mesa_swap4( (GLuint *) dst, n ); + } + } + break; + case GL_INT: + { + GLint *dst = (GLint *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = FLOAT_TO_INT( depthSpan[i] ); + } + if (dstPacking->SwapBytes) { + _mesa_swap4( (GLuint *) dst, n ); + } + } + break; + case GL_FLOAT: + { + GLfloat *dst = (GLfloat *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = depthSpan[i]; + } + if (dstPacking->SwapBytes) { + _mesa_swap4( (GLuint *) dst, n ); + } + } + break; + case GL_HALF_FLOAT_ARB: + { + GLhalfARB *dst = (GLhalfARB *) dest; + GLuint i; + for (i = 0; i < n; i++) { + dst[i] = _mesa_float_to_half(depthSpan[i]); + } + if (dstPacking->SwapBytes) { + _mesa_swap2( (GLushort *) dst, n ); + } + } + break; + default: + _mesa_problem(ctx, "bad type in _mesa_pack_depth_span"); + } +} + + + +/** + * Pack depth and stencil values as GL_DEPTH_STENCIL/GL_UNSIGNED_INT_24_8. + */ +void +_mesa_pack_depth_stencil_span(const struct gl_context *ctx, GLuint n, GLuint *dest, + const GLfloat *depthVals, + const GLstencil *stencilVals, + const struct gl_pixelstore_attrib *dstPacking) +{ + GLfloat depthCopy[MAX_WIDTH]; + GLstencil stencilCopy[MAX_WIDTH]; + GLuint i; + + ASSERT(n <= MAX_WIDTH); + + if (ctx->Pixel.DepthScale != 1.0 || ctx->Pixel.DepthBias != 0.0) { + memcpy(depthCopy, depthVals, n * sizeof(GLfloat)); + _mesa_scale_and_bias_depth(ctx, n, depthCopy); + depthVals = depthCopy; + } + + if (ctx->Pixel.IndexShift || + ctx->Pixel.IndexOffset || + ctx->Pixel.MapStencilFlag) { + memcpy(stencilCopy, stencilVals, n * sizeof(GLstencil)); + _mesa_apply_stencil_transfer_ops(ctx, n, stencilCopy); + stencilVals = stencilCopy; + } + + for (i = 0; i < n; i++) { + GLuint z = (GLuint) (depthVals[i] * 0xffffff); + dest[i] = (z << 8) | (stencilVals[i] & 0xff); + } + + if (dstPacking->SwapBytes) { + _mesa_swap4(dest, n); + } +} + + + + +/** + * Unpack image data. Apply byte swapping, byte flipping (bitmap). + * Return all image data in a contiguous block. This is used when we + * compile glDrawPixels, glTexImage, etc into a display list. We + * need a copy of the data in a standard format. + */ +void * +_mesa_unpack_image( GLuint dimensions, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLenum type, const GLvoid *pixels, + const struct gl_pixelstore_attrib *unpack ) +{ + GLint bytesPerRow, compsPerRow; + GLboolean flipBytes, swap2, swap4; + + if (!pixels) + return NULL; /* not necessarily an error */ + + if (width <= 0 || height <= 0 || depth <= 0) + return NULL; /* generate error later */ + + if (type == GL_BITMAP) { + bytesPerRow = (width + 7) >> 3; + flipBytes = unpack->LsbFirst; + swap2 = swap4 = GL_FALSE; + compsPerRow = 0; + } + else { + const GLint bytesPerPixel = _mesa_bytes_per_pixel(format, type); + GLint components = _mesa_components_in_format(format); + GLint bytesPerComp; + + if (_mesa_type_is_packed(type)) + components = 1; + + if (bytesPerPixel <= 0 || components <= 0) + return NULL; /* bad format or type. generate error later */ + bytesPerRow = bytesPerPixel * width; + bytesPerComp = bytesPerPixel / components; + flipBytes = GL_FALSE; + swap2 = (bytesPerComp == 2) && unpack->SwapBytes; + swap4 = (bytesPerComp == 4) && unpack->SwapBytes; + compsPerRow = components * width; + assert(compsPerRow >= width); + } + + { + GLubyte *destBuffer + = (GLubyte *) malloc(bytesPerRow * height * depth); + GLubyte *dst; + GLint img, row; + if (!destBuffer) + return NULL; /* generate GL_OUT_OF_MEMORY later */ + + dst = destBuffer; + for (img = 0; img < depth; img++) { + for (row = 0; row < height; row++) { + const GLvoid *src = _mesa_image_address(dimensions, unpack, pixels, + width, height, format, type, img, row, 0); + + if ((type == GL_BITMAP) && (unpack->SkipPixels & 0x7)) { + GLint i; + flipBytes = GL_FALSE; + if (unpack->LsbFirst) { + GLubyte srcMask = 1 << (unpack->SkipPixels & 0x7); + GLubyte dstMask = 128; + const GLubyte *s = src; + GLubyte *d = dst; + *d = 0; + for (i = 0; i < width; i++) { + if (*s & srcMask) { + *d |= dstMask; + } + if (srcMask == 128) { + srcMask = 1; + s++; + } + else { + srcMask = srcMask << 1; + } + if (dstMask == 1) { + dstMask = 128; + d++; + *d = 0; + } + else { + dstMask = dstMask >> 1; + } + } + } + else { + GLubyte srcMask = 128 >> (unpack->SkipPixels & 0x7); + GLubyte dstMask = 128; + const GLubyte *s = src; + GLubyte *d = dst; + *d = 0; + for (i = 0; i < width; i++) { + if (*s & srcMask) { + *d |= dstMask; + } + if (srcMask == 1) { + srcMask = 128; + s++; + } + else { + srcMask = srcMask >> 1; + } + if (dstMask == 1) { + dstMask = 128; + d++; + *d = 0; + } + else { + dstMask = dstMask >> 1; + } + } + } + } + else { + memcpy(dst, src, bytesPerRow); + } + + /* byte flipping/swapping */ + if (flipBytes) { + flip_bytes((GLubyte *) dst, bytesPerRow); + } + else if (swap2) { + _mesa_swap2((GLushort*) dst, compsPerRow); + } + else if (swap4) { + _mesa_swap4((GLuint*) dst, compsPerRow); + } + dst += bytesPerRow; + } + } + return destBuffer; + } +} + |