3 EXT_texture_compression_s3tc
7 GL_EXT_texture_compression_s3tc
11 Pat Brown, NVIDIA Corporation (pbrown 'at' nvidia.com)
19 1.1, 16 November 2001 (containing only clarifications relative to
20 version 1.0, dated 7 July 2000)
28 OpenGL 1.1 is required.
30 GL_ARB_texture_compression is required.
32 This extension is written against the OpenGL 1.2.1 Specification.
36 This extension provides additional texture compression functionality
37 specific to S3's S3TC format (called DXTC in Microsoft's DirectX API),
38 subject to all the requirements and limitations described by the extension
39 GL_ARB_texture_compression.
41 This extension supports DXT1, DXT3, and DXT5 texture compression formats.
42 For the DXT1 image format, this specification supports an RGB-only mode
43 and a special RGBA mode with single-bit "transparent" alpha.
47 Contact S3 Incorporated (http://www.s3.com) regarding any intellectual
48 property issues associated with implementing this extension.
50 WARNING: Vendors able to support S3TC texture compression in Direct3D
51 drivers do not necessarily have the right to use the same functionality in
56 (1) Should DXT2 and DXT4 (premultiplied alpha) formats be supported?
58 RESOLVED: No -- insufficient interest. Supporting DXT2 and DXT4
59 would require some rework to the TexEnv definition (maybe add a new
60 base internal format RGBA_PREMULTIPLIED_ALPHA) for these formats.
61 Note that the EXT_texture_env_combine extension (which extends normal
62 TexEnv modes) can be used to support textures with premultipled alpha.
64 (2) Should generic "RGB_S3TC_EXT" and "RGBA_S3TC_EXT" enums be supported
65 or should we use only the DXT<n> enums?
67 RESOLVED: No. A generic RGBA_S3TC_EXT is problematic because DXT3
68 and DXT5 are both nominally RGBA (and DXT1 with the 1-bit alpha is
69 also) yet one format must be chosen up front.
71 (3) Should TexSubImage support all block-aligned edits or just the minimal
72 functionality required by the ARB_texture_compression extension?
74 RESOLVED: Allow all valid block-aligned edits.
76 (4) A pre-compressed image with a DXT1 format can be used as either an
77 RGB_S3TC_DXT1 or an RGBA_S3TC_DXT1 image. If the image has
78 transparent texels, how are they treated in each format?
80 RESOLVED: The renderer has to make sure that an RGB_S3TC_DXT1 format
81 is decoded as RGB (where alpha is effectively one for all texels),
82 while RGBA_S3TC_DXT1 is decoded as RGBA (where alpha is zero for all
83 texels with "transparent" encodings). Otherwise, the formats are
86 (5) Is the encoding of the RGB components for DXT1 formats correct in this
87 spec? MSDN documentation does not specify an RGB color for the
88 "transparent" encoding. Is it really black?
90 RESOLVED: Yes. The specification for the DXT1 format initially
91 required black, but later changed that requirement to a
92 recommendation. All vendors involved in the definition of this
93 specification support black. In addition, specifying black has a
96 When blending multiple texels (GL_LINEAR filtering), mixing opaque and
97 transparent samples is problematic. Defining a black color on
98 transparent texels achieves a sensible result that works like a
99 texture with premultiplied alpha. For example, if three opaque white
100 and one transparent sample is being averaged, the result would be a
101 75% intensity gray (with an alpha of 75%). This is the same result on
102 the color channels as would be obtained using a white color, 75%
103 alpha, and a SRC_ALPHA blend factor.
105 (6) Is the encoding of the RGB components for DXT3 and DXT5 formats
106 correct in this spec? MSDN documentation suggests that the RGB blocks
107 for DXT3 and DXT5 are decoded as described by the DXT1 format.
109 RESOLVED: Yes -- this appears to be a bug in the MSDN documentation.
110 The specification for the DXT2-DXT5 formats require decoding using the
111 opaque block encoding, regardless of the relative values of "color0"
114 New Procedures and Functions
120 Accepted by the <internalformat> parameter of TexImage2D, CopyTexImage2D,
121 and CompressedTexImage2DARB and the <format> parameter of
122 CompressedTexSubImage2DARB:
124 COMPRESSED_RGB_S3TC_DXT1_EXT 0x83F0
125 COMPRESSED_RGBA_S3TC_DXT1_EXT 0x83F1
126 COMPRESSED_RGBA_S3TC_DXT3_EXT 0x83F2
127 COMPRESSED_RGBA_S3TC_DXT5_EXT 0x83F3
129 Additions to Chapter 2 of the OpenGL 1.2.1 Specification (OpenGL Operation)
133 Additions to Chapter 3 of the OpenGL 1.2.1 Specification (Rasterization)
135 Add to Table 3.16.1: Specific Compressed Internal Formats
137 Compressed Internal Format Base Internal Format
138 ========================== ====================
139 COMPRESSED_RGB_S3TC_DXT1_EXT RGB
140 COMPRESSED_RGBA_S3TC_DXT1_EXT RGBA
141 COMPRESSED_RGBA_S3TC_DXT3_EXT RGBA
142 COMPRESSED_RGBA_S3TC_DXT5_EXT RGBA
145 Modify Section 3.8.2, Alternate Image Specification
147 (add to end of TexSubImage discussion, p.123 -- after edit from the
148 ARB_texture_compression spec)
150 If the internal format of the texture image being modified is
151 COMPRESSED_RGB_S3TC_DXT1_EXT, COMPRESSED_RGBA_S3TC_DXT1_EXT,
152 COMPRESSED_RGBA_S3TC_DXT3_EXT, or COMPRESSED_RGBA_S3TC_DXT5_EXT, the
153 texture is stored using one of the several S3TC compressed texture image
154 formats. Such images are easily edited along 4x4 texel boundaries, so the
155 limitations on TexSubImage2D or CopyTexSubImage2D parameters are relaxed.
156 TexSubImage2D and CopyTexSubImage2D will result in an INVALID_OPERATION
157 error only if one of the following conditions occurs:
159 * <width> is not a multiple of four or equal to TEXTURE_WIDTH,
160 unless <xoffset> and <yoffset> are both zero.
161 * <height> is not a multiple of four or equal to TEXTURE_HEIGHT,
162 unless <xoffset> and <yoffset> are both zero.
163 * <xoffset> or <yoffset> is not a multiple of four.
165 The contents of any 4x4 block of texels of an S3TC compressed texture
166 image that does not intersect the area being modified are preserved during
167 valid TexSubImage2D and CopyTexSubImage2D calls.
170 Add to Section 3.8.2, Alternate Image Specification (adding to the end of
171 the CompressedTexImage section introduced by the ARB_texture_compression
174 If <internalformat> is COMPRESSED_RGB_S3TC_DXT1_EXT,
175 COMPRESSED_RGBA_S3TC_DXT1_EXT, COMPRESSED_RGBA_S3TC_DXT3_EXT, or
176 COMPRESSED_RGBA_S3TC_DXT5_EXT, the compressed texture is stored using one
177 of several S3TC compressed texture image formats. The S3TC texture
178 compression algorithm supports only 2D images without borders.
179 CompressedTexImage1DARB and CompressedTexImage3DARB produce an
180 INVALID_ENUM error if <internalformat> is an S3TC format.
181 CompressedTexImage2DARB will produce an INVALID_OPERATION error if
182 <border> is non-zero.
185 Add to Section 3.8.2, Alternate Image Specification (adding to the end of
186 the CompressedTexSubImage section introduced by the
187 ARB_texture_compression spec)
189 If the internal format of the texture image being modified is
190 COMPRESSED_RGB_S3TC_DXT1_EXT, COMPRESSED_RGBA_S3TC_DXT1_EXT,
191 COMPRESSED_RGBA_S3TC_DXT3_EXT, or COMPRESSED_RGBA_S3TC_DXT5_EXT, the
192 texture is stored using one of the several S3TC compressed texture image
193 formats. Since the S3TC texture compression algorithm supports only 2D
194 images, CompressedTexSubImage1DARB and CompressedTexSubImage3DARB produce
195 an INVALID_ENUM error if <format> is an S3TC format. Since S3TC images
196 are easily edited along 4x4 texel boundaries, the limitations on
197 CompressedTexSubImage2D are relaxed. CompressedTexSubImage2D will result
198 in an INVALID_OPERATION error only if one of the following conditions
201 * <width> is not a multiple of four or equal to TEXTURE_WIDTH.
202 * <height> is not a multiple of four or equal to TEXTURE_HEIGHT.
203 * <xoffset> or <yoffset> is not a multiple of four.
205 The contents of any 4x4 block of texels of an S3TC compressed texture
206 image that does not intersect the area being modified are preserved during
207 valid TexSubImage2D and CopyTexSubImage2D calls.
209 Additions to Chapter 4 of the OpenGL 1.2.1 Specification (Per-Fragment
210 Operations and the Frame Buffer)
214 Additions to Chapter 5 of the OpenGL 1.2.1 Specification (Special Functions)
218 Additions to Chapter 6 of the OpenGL 1.2.1 Specification (State and
223 Additions to Appendix A of the OpenGL 1.2.1 Specification (Invariance)
227 Additions to the AGL/GLX/WGL Specifications
237 INVALID_ENUM is generated by CompressedTexImage1DARB or
238 CompressedTexImage3DARB if <internalformat> is
239 COMPRESSED_RGB_S3TC_DXT1_EXT, COMPRESSED_RGBA_S3TC_DXT1_EXT,
240 COMPRESSED_RGBA_S3TC_DXT3_EXT, or COMPRESSED_RGBA_S3TC_DXT5_EXT.
242 INVALID_OPERATION is generated by CompressedTexImage2DARB if
243 <internalformat> is COMPRESSED_RGB_S3TC_DXT1_EXT,
244 COMPRESSED_RGBA_S3TC_DXT1_EXT, COMPRESSED_RGBA_S3TC_DXT3_EXT, or
245 COMPRESSED_RGBA_S3TC_DXT5_EXT and <border> is not equal to zero.
247 INVALID_ENUM is generated by CompressedTexSubImage1DARB or
248 CompressedTexSubImage3DARB if <format> is COMPRESSED_RGB_S3TC_DXT1_EXT,
249 COMPRESSED_RGBA_S3TC_DXT1_EXT, COMPRESSED_RGBA_S3TC_DXT3_EXT, or
250 COMPRESSED_RGBA_S3TC_DXT5_EXT.
252 INVALID_OPERATION is generated by TexSubImage2D CopyTexSubImage2D, or
253 CompressedTexSubImage2D if TEXTURE_INTERNAL_FORMAT is
254 COMPRESSED_RGB_S3TC_DXT1_EXT, COMPRESSED_RGBA_S3TC_DXT1_EXT,
255 COMPRESSED_RGBA_S3TC_DXT3_EXT, or COMPRESSED_RGBA_S3TC_DXT5_EXT and any of
256 the following apply: <width> is not a multiple of four or equal to
257 TEXTURE_WIDTH; <height> is not a multiple of four or equal to
258 TEXTURE_HEIGHT; <xoffset> or <yoffset> is not a multiple of four.
261 The following restrictions from the ARB_texture_compression specification
262 do not apply to S3TC texture formats, since subimage modification is
263 straightforward as long as the subimage is properly aligned.
265 DELETE: INVALID_OPERATION is generated by TexSubImage1D, TexSubImage2D,
266 DELETE: TexSubImage3D, CopyTexSubImage1D, CopyTexSubImage2D, or
267 DELETE: CopyTexSubImage3D if the internal format of the texture image is
268 DELETE: compressed and <xoffset>, <yoffset>, or <zoffset> does not equal
269 DELETE: -b, where b is value of TEXTURE_BORDER.
271 DELETE: INVALID_VALUE is generated by CompressedTexSubImage1DARB,
272 DELETE: CompressedTexSubImage2DARB, or CompressedTexSubImage3DARB if the
273 DELETE: entire texture image is not being edited: if <xoffset>,
274 DELETE: <yoffset>, or <zoffset> is greater than -b, <xoffset> + <width> is
275 DELETE: less than w+b, <yoffset> + <height> is less than h+b, or <zoffset>
276 DELETE: + <depth> is less than d+b, where b is the value of
277 DELETE: TEXTURE_BORDER, w is the value of TEXTURE_WIDTH, h is the value of
278 DELETE: TEXTURE_HEIGHT, and d is the value of TEXTURE_DEPTH.
280 See also errors in the GL_ARB_texture_compression specification.
284 In the "Textures" state table, increment the TEXTURE_INTERNAL_FORMAT
285 subscript for Z by 4 in the "Type" row.
287 New Implementation Dependent State
293 S3TC Compressed Texture Image Formats
295 Compressed texture images stored using the S3TC compressed image formats
296 are represented as a collection of 4x4 texel blocks, where each block
297 contains 64 or 128 bits of texel data. The image is encoded as a normal
298 2D raster image in which each 4x4 block is treated as a single pixel. If
299 an S3TC image has a width or height less than four, the data corresponding
300 to texels outside the image are irrelevant and undefined.
302 When an S3TC image with a width of <w>, height of <h>, and block size of
303 <blocksize> (8 or 16 bytes) is decoded, the corresponding image size (in
306 ceil(<w>/4) * ceil(<h>/4) * blocksize.
308 When decoding an S3TC image, the block containing the texel at offset
309 (<x>, <y>) begins at an offset (in bytes) relative to the base of the
312 blocksize * (ceil(<w>/4) * floor(<y>/4) + floor(<x>/4)).
314 The data corresponding to a specific texel (<x>, <y>) are extracted from a
315 4x4 texel block using a relative (x,y) value of
317 (<x> modulo 4, <y> modulo 4).
319 There are four distinct S3TC image formats:
321 COMPRESSED_RGB_S3TC_DXT1_EXT: Each 4x4 block of texels consists of 64
322 bits of RGB image data.
324 Each RGB image data block is encoded as a sequence of 8 bytes, called (in
325 order of increasing address):
327 c0_lo, c0_hi, c1_lo, c1_hi, bits_0, bits_1, bits_2, bits_3
329 The 8 bytes of the block are decoded into three quantities:
331 color0 = c0_lo + c0_hi * 256
332 color1 = c1_lo + c1_hi * 256
333 bits = bits_0 + 256 * (bits_1 + 256 * (bits_2 + 256 * bits_3))
335 color0 and color1 are 16-bit unsigned integers that are unpacked to
336 RGB colors RGB0 and RGB1 as though they were 16-bit packed pixels with
337 a <format> of RGB and a type of UNSIGNED_SHORT_5_6_5.
339 bits is a 32-bit unsigned integer, from which a two-bit control code
340 is extracted for a texel at location (x,y) in the block using:
342 code(x,y) = bits[2*(4*y+x)+1..2*(4*y+x)+0]
344 where bit 31 is the most significant and bit 0 is the least
347 The RGB color for a texel at location (x,y) in the block is given by:
349 RGB0, if color0 > color1 and code(x,y) == 0
350 RGB1, if color0 > color1 and code(x,y) == 1
351 (2*RGB0+RGB1)/3, if color0 > color1 and code(x,y) == 2
352 (RGB0+2*RGB1)/3, if color0 > color1 and code(x,y) == 3
354 RGB0, if color0 <= color1 and code(x,y) == 0
355 RGB1, if color0 <= color1 and code(x,y) == 1
356 (RGB0+RGB1)/2, if color0 <= color1 and code(x,y) == 2
357 BLACK, if color0 <= color1 and code(x,y) == 3
359 Arithmetic operations are done per component, and BLACK refers to an
360 RGB color where red, green, and blue are all zero.
362 Since this image has an RGB format, there is no alpha component and the
363 image is considered fully opaque.
366 COMPRESSED_RGBA_S3TC_DXT1_EXT: Each 4x4 block of texels consists of 64
367 bits of RGB image data and minimal alpha information. The RGB components
368 of a texel are extracted in the same way as COMPRESSED_RGB_S3TC_DXT1_EXT.
370 The alpha component for a texel at location (x,y) in the block is
373 0.0, if color0 <= color1 and code(x,y) == 3
376 IMPORTANT: When encoding an RGBA image into a format using 1-bit
377 alpha, any texels with an alpha component less than 0.5 end up with an
378 alpha of 0.0 and any texels with an alpha component greater than or
379 equal to 0.5 end up with an alpha of 1.0. When encoding an RGBA image
380 into the COMPRESSED_RGBA_S3TC_DXT1_EXT format, the resulting red,
381 green, and blue components of any texels with a final alpha of 0.0
382 will automatically be zero (black). If this behavior is not desired
383 by an application, it should not use COMPRESSED_RGBA_S3TC_DXT1_EXT.
384 This format will never be used when a generic compressed internal
385 format (Table 3.16.2) is specified, although the nearly identical
386 format COMPRESSED_RGB_S3TC_DXT1_EXT (above) may be.
389 COMPRESSED_RGBA_S3TC_DXT3_EXT: Each 4x4 block of texels consists of 64
390 bits of uncompressed alpha image data followed by 64 bits of RGB image
393 Each RGB image data block is encoded according to the
394 COMPRESSED_RGB_S3TC_DXT1_EXT format, with the exception that the two code
395 bits always use the non-transparent encodings. In other words, they are
396 treated as though color0 > color1, regardless of the actual values of
399 Each alpha image data block is encoded as a sequence of 8 bytes, called
400 (in order of increasing address):
402 a0, a1, a2, a3, a4, a5, a6, a7
404 The 8 bytes of the block are decoded into one 64-bit integer:
406 alpha = a0 + 256 * (a1 + 256 * (a2 + 256 * (a3 + 256 * (a4 +
407 256 * (a5 + 256 * (a6 + 256 * a7))))))
409 alpha is a 64-bit unsigned integer, from which a four-bit alpha value
410 is extracted for a texel at location (x,y) in the block using:
412 alpha(x,y) = bits[4*(4*y+x)+3..4*(4*y+x)+0]
414 where bit 63 is the most significant and bit 0 is the least
417 The alpha component for a texel at location (x,y) in the block is
418 given by alpha(x,y) / 15.
421 COMPRESSED_RGBA_S3TC_DXT5_EXT: Each 4x4 block of texels consists of 64
422 bits of compressed alpha image data followed by 64 bits of RGB image data.
424 Each RGB image data block is encoded according to the
425 COMPRESSED_RGB_S3TC_DXT1_EXT format, with the exception that the two code
426 bits always use the non-transparent encodings. In other words, they are
427 treated as though color0 > color1, regardless of the actual values of
430 Each alpha image data block is encoded as a sequence of 8 bytes, called
431 (in order of increasing address):
433 alpha0, alpha1, bits_0, bits_1, bits_2, bits_3, bits_4, bits_5
435 The alpha0 and alpha1 are 8-bit unsigned bytes converted to alpha
436 components by multiplying by 1/255.
438 The 6 "bits" bytes of the block are decoded into one 48-bit integer:
440 bits = bits_0 + 256 * (bits_1 + 256 * (bits_2 + 256 * (bits_3 +
441 256 * (bits_4 + 256 * bits_5))))
443 bits is a 48-bit unsigned integer, from which a three-bit control code
444 is extracted for a texel at location (x,y) in the block using:
446 code(x,y) = bits[3*(4*y+x)+1..3*(4*y+x)+0]
448 where bit 47 is the most significant and bit 0 is the least
451 The alpha component for a texel at location (x,y) in the block is
454 alpha0, code(x,y) == 0
455 alpha1, code(x,y) == 1
457 (6*alpha0 + 1*alpha1)/7, alpha0 > alpha1 and code(x,y) == 2
458 (5*alpha0 + 2*alpha1)/7, alpha0 > alpha1 and code(x,y) == 3
459 (4*alpha0 + 3*alpha1)/7, alpha0 > alpha1 and code(x,y) == 4
460 (3*alpha0 + 4*alpha1)/7, alpha0 > alpha1 and code(x,y) == 5
461 (2*alpha0 + 5*alpha1)/7, alpha0 > alpha1 and code(x,y) == 6
462 (1*alpha0 + 6*alpha1)/7, alpha0 > alpha1 and code(x,y) == 7
464 (4*alpha0 + 1*alpha1)/5, alpha0 <= alpha1 and code(x,y) == 2
465 (3*alpha0 + 2*alpha1)/5, alpha0 <= alpha1 and code(x,y) == 3
466 (2*alpha0 + 3*alpha1)/5, alpha0 <= alpha1 and code(x,y) == 4
467 (1*alpha0 + 4*alpha1)/5, alpha0 <= alpha1 and code(x,y) == 5
468 0.0, alpha0 <= alpha1 and code(x,y) == 6
469 1.0, alpha0 <= alpha1 and code(x,y) == 7
474 1.1, 11/16/01 pbrown: Updated contact info, clarified where texels
475 fall within a single block.
477 1.0, 07/07/00 prbrown1: Published final version agreed to by working
480 0.9, 06/24/00 prbrown1: Documented that block-aligned TexSubImage calls
481 do not modify existing texels outside the
482 modified blocks. Added caveat to allow for a
483 (0,0)-anchored TexSubImage operation of
486 0.7, 04/11/00 prbrown1: Added issues on DXT1, DXT3, and DXT5 encodings
487 where the MSDN documentation doesn't match what
488 is really done. Added enum values from the
491 0.4, 03/28/00 prbrown1: Updated to reflect final version of the
492 ARB_texture_compression extension. Allowed
493 block-aligned TexSubImage calls.
495 0.3, 03/07/00 prbrown1: Resolved issues pertaining to the format of RGB
496 blocks in the DXT3 and DXT5 formats (they don't
497 ever use the "transparent" encoding). Fixed
498 decoding of DXT1 blocks. Pointed out issue of
499 "transparent" texels in DXT1 encodings having
500 different behaviors for RGB and RGBA internal
503 0.2, 02/23/00 prbrown1: Minor revisions; added several issues.
505 0.11, 02/17/00 prbrown1: Slight modification to error semantics
506 (INVALID_ENUM instead of INVALID_OPERATION).
508 0.1, 02/15/00 prbrown1: Initial revision.