Around the same time, Allegorithmic was developing ProFX, the predecessor to Substance Designer, a middleware for real-time procedural texturing. ProFX also included a fast DXT encoder, allowing procedural textures to be converted into GPU-friendly formats at load time.<\/p>\n\n\n\n
Simon Brown<\/a> was working on PlayStation Home, Sony\u2019s 3D social virtual world where players could create and customize their avatars. To support this, he wrote a fast DXT encoder optimized for the PS3\u2019s SPUs, demonstrating the potential of offloading texture compression to parallel processors.<\/p>\n\n\n\n John Carmack had been talking about the megatexture technology for a while, but in 2006, when Jan Paul van Waveren<\/a> published the details of their Real-Time DXT Compression<\/a> implementation on the Intel Software Network, we at NVIDIA saw a potential problem: if Rage<\/em> ended up CPU-limited, it could push gamers toward CPU upgrades rather than GPUs. That made real-time texture compression in the GPU a strategic priority for us.<\/p>\n\n\n\n\n\n\n\n The first problem we encountered was that in Direct3D 10, updating a DXT texture without transferring data back to the CPU wasn\u2019t possible. Fortunately, OpenGL had recently introduced pixel buffer objects (PBOs). While not designed for this purpose, PBOs provided a workaround: we could render each block to an integer render target, copy the contents to a PBO, and then transfer the compressed data from the PBO to a DXT-compressed texture.<\/p>\n\n\n\n This is the technique that I employed in the examples that accompanied the papers that Jan Paul and I co-authored and that can still be found online:<\/p>\n\n\n\n These examples allowed us to demonstrate that GPU texture compression was possible and show its potential.<\/p>\n\n\n\n I then pushed for changes in graphics APIs to better support this use case. The first API to enable direct copies from uncompressed to compressed textures was Direct3D 10.1, which relaxed the restrictions in the For more details see: Format Conversion Using Direct3D 10.1<\/a><\/p>\n\n\n\n This functionality was also available on consoles at the time through low-level APIs provided by hardware vendors. In this GDC talk<\/a>, Jason Tranchida from Volition discusses their experience implementing real-time DXT compression using these APIs and Direct3D 10.1.<\/p>\n\n\n\n A year later, this feature arrived in OpenGL with NV_copy_image<\/a>. However, it wasn\u2019t until 2012, six years after its introduction in Direct3D, that it became more widely available via ARB_copy_image<\/a> and OpenGL 4.3.<\/p>\n\n\n\n In OpenGL, the equivalent of D3D\u2019s This feature could have facilitated the use of GPU texture compression in id Tech 5, enabling Rage<\/em> to offload more work from the CPU to the GPU. However, Rage<\/em> did not employ GPU texture compression at launch. Instead, it continued relying on the CPU-based approach described in Jan Paul’s paper.<\/p>\n\n\n\n At that point I was already working on The Witness, so it was Evan Hart from NVIDIA\u2019s Developer Technology team who revisited this problem and implemented GPU texture compression for id Tech 5, building on the foundations I had put in place, and addressing all the practical challenges that arise when integrating real-time GPU compression into a production engine.<\/p>\n\n\n\n I didn\u2019t pay much attention to these features again until I started developing Spark<\/a>. Since my focus was on mobile, I was particularly interested in OpenGL ES, which appeared to support the same functionality through the GL_EXT_copy_image<\/a> extension and had become a core feature in OpenGL ES 3.2.<\/p>\n\n\n\n While testing this feature, I encountered an unexpected limitation: OpenGL ES does not support the This requires packing the 64-bit That said, in a world where compute shaders are ubiquitous, there\u2019s now a much simpler way to accomplish the same task. By using a compute shader to write compressed data into a temporary buffer, we can bind that buffer to the This turned out to work well on all devices and run with similar performance.<\/p>\n\n\n\n Like OpenGL, Vulkan supports copying data from uncompressed to block-compressed textures. This functionality has been available since Vulkan 1.0 through the So imagine my surprise when I tested it and found that most devices didn’t implement it correctly. The following screenshots show the results on Adreno, PowerVR, and Mali devices. Only Mali produced the correct output!<\/p>\n\n\n\n\n
Direct3D 10<\/h2>\n\n\n\n
CopyResource<\/code><\/a> and CopySubresourceRegion<\/code><\/a> APIs to allow copies between prestructured-typed textures and block-compressed textures of the same bit width.<\/p>\n\n\n\nOpenGL<\/h2>\n\n\n\n
CopySubresourceRegion<\/code> is glCopyImageSubData<\/code><\/a>. Like its Direct3D 10 counterpart, it enables copying data from an uncompressed texture to a block-compressed texture, provided the formats are compatible, which generally means that the texel size matches the block size.<\/p>\n\n\n\nOpenGL ES<\/h2>\n\n\n\n
rg32ui<\/code> output image layout. Instead, to output 64-bit blocks, you have to use rgba16ui<\/code>:<\/p>\n\n\n\nlayout(rgba16ui) uniform restrict writeonly highp uimage2D dst;<\/code><\/pre>\n\n\n\nuvec2<\/code> block manually before passing it to imageStore<\/code>:<\/p>\n\n\n\nuvec4 pack_uvec2(uvec2 v) {\n return uvec4(v.x & 0xFFFFu, v.x >> 16, v.y & 0xFFFFu, v.y >> 16);\n}\n...\nimageStore(dst, uv, pack_uvec2(block));<\/code><\/pre>\n\n\n\nGL_PIXEL_UNPACK_BUFFER<\/code> target and source data directly from it using glCompressedTexSubImage2D<\/code><\/a>:<\/p>\n\n\n\nglBindBuffer(GL_PIXEL_UNPACK_BUFFER, tmp_buffer);\nglBindTexture(GL_TEXTURE_2D, dst_texture);\nglCompressedTexSubImage2D(GL_TEXTURE_2D, dst_level, dst_x, dst_y, width, height, \n gl_format, tmp_buffer_size, (const void*)tmp_buffer_offset);<\/code><\/pre>\n\n\n\nVulkan<\/h2>\n\n\n\n
vkCmdCopyImage<\/code><\/a> API.<\/p>\n\n\n\n![\"\"](\"https:\/\/www.ludicon.com\/castano\/blog\/wp-content\/uploads\/2025\/02\/Android-Vulkan-CmdCopyImage-700x581.png\")
<\/figure>\n\n\n\n
![\"\"](\"https:\/\/www.ludicon.com\/castano\/blog\/wp-content\/uploads\/2025\/02\/Android-Vulkan-RenderToBlockTexelView-700x516.png\")
<\/figure>\n\n\n\n