Adaptive scalable texture compression

Adaptive scalable texture compression (ASTC) is a lossy block-based texture compression algorithm developed by Jørn Nystad et al. of ARM Ltd. and AMD.[1]

Full details of ASTC were first presented publicly at the High Performance Graphics 2012 conference, in a paper by Olson et al. entitled "Adaptive Scalable Texture Compression".[2]

ASTC was adopted as an official extension for both OpenGL and OpenGL ES by the Khronos Group on 6 August 2012.[3]

Hardware support

edit
Vendor/product Profile Generation
AMD Radeon ?
Apple GPUs LDR only A8 through A12[4]
Full Since A13[4]
Arm Mali Full Since Mali-T620/T720/T820[5]
Imagination PowerVR Full Since Series6XT[6]
Intel GPUs Full[7] From Skylake[8] ; Removed in Arc / Gen12.5[9]
Nvidia Tegra ? Since Kepler[10]
Qualcomm Adreno Full LDR since 4xx series,[11] at least 7xx series support GL_KHR_texture_compression_astc_hdr extension on Android 13

On Linux, all Gallium 3D drivers have a software fallback since 2018, so ASTC can be used on any AMD Radeon GPU.[12]

Overview

edit
 
Example image prior to compression
 
Detail from example image, after compression at 8, 3.56 and 2 bits/pixel

The method of compression is an evolution of Color Cell Compression with features including numerous closely spaced fractional bit rates, multiple color formats, support for high-dynamic-range (HDR) textures, and real 3D texture support.

The stated primary design goal for ASTC is to enable content developers to have better control over the space/quality tradeoff inherent in any lossy compression scheme. With ASTC, the ratio between adjacent bit rates is of the order of 25%, making it less expensive to increase quality for a given texture.

Encoding different assets often requires different color formats. ASTC allows a wide choice of input formats, including luminance-only, luminance-alpha, RGB, RGBA, and modes optimized for surface normals. The designer can thus choose the optimal format without having to support multiple different compression schemes.

The choices of bit rate and color format do not constrain each other, so that it's possible to choose from a large number of combinations.

Despite this flexibility, ASTC achieves better peak signal-to-noise ratios than PVRTC, S3TC, and ETC2 when measured at 2 and 3.56 bits per texel.[2] For HDR textures, it produces results comparable to BC6H at 8 bits per texel.[2]

Supported color formats

edit

ASTC supports anywhere from 1 to 4 channels. In modes with 2–4 channels, one of the channels can be treated as "uncorrelated" and be given a separate gradient for prediction. In any case, the data is decoded as RGBA.[13]

Channel count RGBA interpretation Description
1 L Luminance-only: RGB set to same value in decoded buffer, alpha set to 1
2 LA Luminance with transparency
2 L+A Luminance with uncorrelated transparency
3 RGB Full color, alpha set to 1
3 RG+B Full color with uncorrelated blue (not actually used for color purposes)
4 RGBA Full color with transparency
4 RGB+A Full color with uncorrelated transparency

Each of these may be encoded as low or high dynamic range. The encoder selects color formats independently for each block in the image.

In practice, ASTC may be used to represent data other than color. For example, the L+A format may be used to represent "X+Y", a normal map with uncorrelated components; the "RG+B" format can be used to represent XY+Z.[14] The astc-encoder software supplied by ARM supports "X+Y" generation with the -normal option. The shader is expected to treat the decoded output as a swizzled texture.[15]

2D block footprints and bit rates

edit

ASTC textures are compressed using a fixed block size of 128 bits, but with a variable block footprint ranging from 4×4 texels up to 12×12 texels. The available bit rates thus range from 8 bits per texel down to 0.89 bits per texel, with fine steps in between.

Block footprint Bit rate Increment
4×4 8.00 25%
5×4 6.40 25%
5×5 5.12 20%
6×5 4.27 20%
6×6 3.56 14%
8×5 3.20 20%
8×6 2.67 5%
10×5 2.56 20%
10×6 2.13 7%
8×8 2.00 25%
10×8 1.60 25%
10×10 1.28 20%
12×10 1.07 20%
12×12 0.89

In the above table, the "Increment" column shows the additional storage required to store a texture using this bit rate, as compared to the next smallest. Block footprints are presented as width × height.

3D block footprints and bit rates

edit

ASTC 3D textures are compressed using a fixed block size of 128 bits, as for 2D but with a variable block footprint ranging from 3×3×3 texels up to 6×6×6 texels. The available bit rates thus range from 4.74 bits per texel down to 0.59 bits per texel, with fine steps in between.

Block footprint Bit rate Increment
3×3×3 4.74 33%
4×3×3 3.56 33%
4×4×3 2.67 33%
4×4×4 2.00 25%
5×4×4 1.60 25%
5×5×4 1.28 50%
5×5×5 1.02 20%
6×5×5 0.85 20%
6×6×5 0.71 20%
6×6×6 0.59

Block footprints are presented as width × height × depth.

Universal ASTC

edit

UASTC (Universal ASTC) is a subset of ASTC specified by Binomial. The format is used in their Basis Universal "supercompressed" GPU texture format, which adds extra compression over compressed texture formats such as UASTC and ETC1S and allows for efficient conversion from UASTC/ETC1S to compressed texture formats directly usable by GPUs.[16] UASTC, as part of Basis Universal, is part of the KTX (Khronos Texture) file format.[17]

See also

edit

References

edit
  1. ^ "Adaptive Scalable Texture Compression (ASTC) technology developed by ARM and AMD".
  2. ^ a b c "Adaptive Scalable Texture Compression" (PDF). HPG 2012. Retrieved 2012-06-27.
  3. ^ "Khronos Releases ATSC Next-Generation Texture Compression Specification". The Khronos Group Inc. 2012-08-06. Retrieved 2012-08-06.
  4. ^ a b "Metal Feature Set Tables" (PDF). Apple Inc. 2020-10-21. Retrieved 2021-08-31.
  5. ^ "Arm Mali GPU Datasheet" (PDF). Arm Limited. 2021. Retrieved 2021-08-31.
  6. ^ "Imagination's new generation PowerVR Series6XT architecture delivers up to 50% higher performance and advanced power management". Imagination Technologies. 2014-01-06. Retrieved 2021-08-21.
  7. ^ "Intel Skylake Adds ASTC Texture Compression, Open-Source Support Coming". Phoronix. 2015-05-20. Retrieved 2021-08-31.
  8. ^ "Graphics API Developer's Guide For 6th Generation Intel Core Processors". Archived from the original on 2017-07-20.
  9. ^ Michael Larabel (2021-10-07). "Intel Removes ASTC Hardware From Gen12.5+ Graphics". Phoronix. Retrieved 2022-07-10.
  10. ^ "Vulkan API" (PDF).
  11. ^ "Qualcomm Adreno OpenGL ES Developer Guide" (PDF). Qualcomm. 2015-05-01. Retrieved 2021-08-31.
  12. ^ "[Mesa-dev] [PATCH 0/7] ASTC texture compression for all Gallium drivers". Lists.freedesktop.org. 23 July 2018. Retrieved 2022-09-01.
  13. ^ "Khronos Data Format Specification v1.1 rev 9". registry.khronos.org.
  14. ^ "Khronos Releases ASTC Next-Generation Texture Compression Specification". The Khronos Group. 6 August 2012.
  15. ^ "Effective ASTC Encoding [astc-encoder/Docs/Encoding.md at 2042cfc1a507c0414fb41dce1603ed53c503a0da · ARM-software/astc-encoder]". GitHub. The best way to store normal maps using ASTC is similar to the scheme used by BC5; store the X and Y components of a unit-length normal. The Z component of the normal can be reconstructed in shader code based on the knowledge that the vector is unit length. To encode this we need to store only two input components in the compressed data, and therefore use the rrrg coding swizzle to align the data with the ASTC luminance+alpha endpoint.
  16. ^ "UASTC Texture Specification". GitHub.
  17. ^ "KTX - GPU Texture Container Format". The Khronos Group. 16 April 2021.
edit