Name ARB_sparse_texture_clamp Name Strings GL_ARB_sparse_texture_clamp Contact Eric Werness, NVIDIA Corporation (ewerness 'at' nvidia.com) Contributors Pat Brown, NVIDIA Corporation Piers Daniell, NVIDIA Corporation Jeff Bolz, NVIDIA Corporation Mathias Heyer, NVIDIA Corporation Notice Copyright (c) 2015 The Khronos Group Inc. Copyright terms at http://www.khronos.org/registry/speccopyright.html Specification Update Policy Khronos-approved extension specifications are updated in response to issues and bugs prioritized by the Khronos OpenGL Working Group. For extensions which have been promoted to a core Specification, fixes will first appear in the latest version of that core Specification, and will eventually be backported to the extension document. This policy is described in more detail at https://www.khronos.org/registry/OpenGL/docs/update_policy.php Status Complete. Approved by the ARB on June 26, 2015. Ratified by the Khronos Board of Promoters on August 7, 2015. Version Last Modified Date: April 30, 2015 Revision: 2 Number ARB Extension #187 Dependencies This extension is written against the OpenGL 4.5 Specification (Compatibility Profile), dated February 2, 2015. This extension is written against the OpenGL Shading Language Specification, version 4.50, revision 5. This extension requires and extends ARB_sparse_texture2. This extension interacts with NV_gpu_program4 and NV_gpu_program5. Overview This extension builds on the ARB_sparse_texture2 extension, providing the following new functionality: * New built-in GLSL texture lookup functions are provided that specify a minimum level of detail to use for lookups where the level of detail is computed automatically. This allows shaders to avoid accessing unpopulated portions of high-resolution levels of detail when it knows that the memory accessed is unpopulated, either from a priori knowledge or from feedback provided by the return value of previously executed "sparse" texture lookup functions. New Procedures and Functions None. New Tokens None. Modifications to the OpenGL 4.5 Specification (Compatibility Profile) Modify Section 8.14.1, Scale Factor and Level of Detail, p. 280 (move the third paragraph, p. 281, describing lod_min and lod_max in equation 8.6, up one paragraph and modify it to read as follows) lod_min and lod_max indicate minimum and maximum clamps on the computed level of detail. lod_max is taken directly from the TEXTURE_MAX_LOD texture or sampler parameter. If a texture access is performed in a fragment shader with a minimum level of detail clamp specified in the built-in texture lookup function, lod_min is the larger of the TEXTURE_MIN_LOD texture or sampler parameter and the minimum level of detail provided by the shader. Otherwise, lod_min is taken directly from the TEXTURE_MIN_LOD texture or sampler parameter. The initial values of the TEXTURE_MIN_LOD and TEXTURE_MAX_LOD texture and sampler parameters are chosen so as to never clamp the range of lambda values. Modifications to the OpenGL Shading Language Specification, Version 4.50 Including the following line in a shader can be used to control the language features described in this extension: #extension GL_ARB_sparse_texture_clamp : where is as specified in section 3.3. New preprocessor #defines are added to the OpenGL Shading Language: #define GL_ARB_sparse_texture_clamp 1 Modify Section 8.9, Texture Functions, p. 158 (insert after first paragraph, p. 159) The texture lookup functions with an parameter specify a minimum clamp applied to the automatic level of detail computations. Since automatic level of detail calculations are only supported by fragment shaders, these lookup functions are also only available in fragment shaders. No functions with the parameter are provided for rectangle textures, multisample textures, and texture buffers because mipmaps are not allowed for these types of textures. Modify Section 8.9.2, Texel Lookup Functions, p. 162 (The combination of this extension and ARB_sparse_texture2 adds two new variants of texture lookup functions. The "sparse" functions are like normal texture lookup functions, except that they return a sparse texture residency status to the caller and return the actual filtered texel value in an "out" parameter. The "Clamp" variants are functions adding a new parameter specifying a minimum LOD to use for texture lookup functions where level of detail is computed automatically. For each set of texture functions, we provide one to three new variants based on whether sparse and LOD clamping functionality are desired. These new variants copy the existing functions, add suffixes to the function names, and add one or more new parameters. We create new variants only for the targets for which sparse storage is supported -- no new functions are added for the following sampler types: gsampler1D, sampler1DShadow, gsampler1DArray, sampler1DArrayShadow. Additionally, to reduce the number of new functions added, we are not including any new variants for textureProj*() built-ins. To use the new features with projective texture lookups, shaders can divide through by q and use non-projective variants. We also chose not to provide "Clamp" variants of functions like textureLod() expecting an explicit level-of-detail.) (insert new lookup function table cells, at the end of the section, p. 168) Syntax: int sparseTextureClampARB(gsampler2D sampler, vec2 P, float lodClamp, out gvec4 texel [, float bias]); int sparseTextureClampARB(gsampler3D sampler, vec3 P, float lodClamp, out gvec4 texel [, float bias]); int sparseTextureClampARB(gsamplerCube sampler, vec3 P, float lodClamp, out gvec4 texel [, float bias]); int sparseTextureClampARB(sampler2DShadow sampler, vec3 P, float lodClamp, out float texel [, float bias]); int sparseTextureClampARB(samplerCubeShadow sampler, vec4 P, float lodClamp, out float texel [, float bias]); int sparseTextureClampARB(gsampler2DArray sampler, vec3 P, float lodClamp, out gvec4 texel [, float bias]); int sparseTextureClampARB(gsamplerCubeArray sampler, vec4 P, float lodClamp, out gvec4 texel [, float bias]); int sparseTextureClampARB(sampler2DArrayShadow sampler, vec4 P, float lodClamp, out float texel); int sparseTextureClampARB(samplerCubeArrayShadow sampler, vec4 P, float compare, float lodClamp, out float texel); Description: Do a filtered texture lookup as in texture(), but return texture access residency information from the function and the filtered lookup result in the out parameter . Additionally, clamp the automatically computed level of detail to be greater than or equal to . -- Syntax: gvec4 textureClampARB(gsampler1D sampler, float P, float lodClamp [, float bias]); gvec4 textureClampARB(gsampler2D sampler, vec2 P, float lodClamp [, float bias]); gvec4 textureClampARB(gsampler3D sampler, vec3 P, float lodClamp [, float bias]); gvec4 textureClampARB(gsamplerCube sampler, vec3 P, float lodClamp [, float bias]); float textureClampARB(sampler1DShadow sampler, vec3 P, float lodClamp [, float bias]); float textureClampARB(sampler2DShadow sampler, vec3 P, float lodClamp [, float bias]); float textureClampARB(samplerCubeShadow sampler, vec4 P, float lodClamp [, float bias]); gvec4 textureClampARB(gsampler1DArray sampler, vec2 P, float lodClamp [, float bias]); gvec4 textureClampARB(gsampler2DArray sampler, vec3 P, float lodClamp [, float bias]); gvec4 textureClampARB(gsamplerCubeArray sampler, vec4 P, float lodClamp [, float bias]); float textureClampARB(sampler1DArrayShadow sampler, vec3 P, float lodClamp [, float bias]); float textureClampARB(sampler2DArrayShadow sampler, vec4 P, float lodClamp); float textureClampARB(samplerCubeArrayShadow sampler, vec4 P, float compare, float lodClamp); Description: Do a filtered texture lookup as in texture(), but clamp the automatically computed level of detail to be greater than or equal to . -- Syntax: int sparseTextureOffsetClampARB(gsampler2D sampler, vec2 P, ivec2 offset, float lodClamp, out gvec4 texel [, float bias]); int sparseTextureOffsetClampARB(gsampler3D sampler, vec3 P, ivec3 offset, float lodClamp, out gvec4 texel [, float bias]); int sparseTextureOffsetClampARB(sampler2DShadow sampler, vec3 P, ivec2 offset, float lodClamp, out float texel [, float bias]); int sparseTextureOffsetClampARB(gsampler2DArray sampler, vec3 P, ivec2 offset, float lodClamp, out gvec4 texel [, float bias]); int sparseTextureOffsetClampARB(sampler2DArrayShadow sampler, vec4 P, ivec2 offset, float lodClamp, out float texel); Description: Do a filtered texture lookup as in textureOffset(), but return texture access residency information from the function and the filtered lookup result in the out parameter . Additionally, clamp the automatically computed level of detail to be greater than or equal to . -- Syntax: gvec4 textureOffsetClampARB(gsampler1D sampler, float P, int offset, float lodClamp [, float bias]); gvec4 textureOffsetClampARB(gsampler2D sampler, vec2 P, ivec2 offset, float lodClamp [, float bias]); gvec4 textureOffsetClampARB(gsampler3D sampler, vec3 P, ivec3 offset, float lodClamp [, float bias]); float textureOffsetClampARB(sampler1DShadow sampler, vec3 P, int offset, float lodClamp [, float bias]); float textureOffsetClampARB(sampler2DShadow sampler, vec3 P, ivec2 offset, float lodClamp [, float bias]); gvec4 textureOffsetClampARB(gsampler1DArray sampler, vec2 P, int offset, float lodClamp [, float bias]); gvec4 textureOffsetClampARB(gsampler2DArray sampler, vec3 P, ivec2 offset, float lodClamp [, float bias]); float textureOffsetClampARB(sampler1DArrayShadow sampler, vec3 P, int offset, float lodClamp [, float bias]); float textureOffsetClampARB(sampler2DArrayShadow sampler, vec4 P, ivec2 offset, float lodClamp); Description: Do a filtered texture lookup as in textureOffset(), but clamp the automatically computed level of detail to be greater than or equal to . -- Syntax: int sparseTextureGradClampARB(gsampler2D sampler, vec2 P, vec2 dPdx, vec2 dPdy, float lodClamp, out gvec4 texel); int sparseTextureGradClampARB(gsampler3D sampler, vec3 P, vec3 dPdx, vec3 dPdy, float lodClamp, out gvec4 texel); int sparseTextureGradClampARB(gsamplerCube sampler, vec3 P, vec3 dPdx, vec3 dPdy, float lodClamp, out gvec4 texel); int sparseTextureGradClampARB(sampler2DShadow sampler, vec3 P, vec2 dPdx, vec2 dPdy, float lodClamp, out float texel); int sparseTextureGradClampARB(samplerCubeShadow sampler, vec4 P, vec3 dPdx, vec3 dPdy, float lodClamp, out float texel); int sparseTextureGradClampARB(gsampler2DArray sampler, vec3 P, vec2 dPdx, vec2 dPdy, float lodClamp, out gvec4 texel); int sparseTextureGradClampARB(sampler2DArrayShadow sampler, vec4 P, vec2 dPdx, vec2 dPdy, float lodClamp, out float texel); int sparseTextureGradClampARB(gsamplerCubeArray sampler, vec4 P, vec3 dPdx, vec3 dPdy, float lodClamp, out gvec4 texel); Description: Do a filtered texture lookup as in textureGrad(), but return texture access residency information from the function and the filtered lookup result in the out parameter . Additionally, clamp the automatically computed level of detail to be greater than or equal to . -- Syntax: gvec4 textureGradClampARB(gsampler1D sampler, float P, float dPdx, float dPdy, float lodClamp); gvec4 textureGradClampARB(gsampler2D sampler, vec2 P, vec2 dPdx, vec2 dPdy, float lodClamp); gvec4 textureGradClampARB(gsampler3D sampler, vec3 P, vec3 dPdx, vec3 dPdy, float lodClamp); gvec4 textureGradClampARB(gsamplerCube sampler, vec3 P, vec3 dPdx, vec3 dPdy, float lodClamp); float textureGradClampARB(sampler1DShadow sampler, vec3 P, float dPdx, float dPdy, float lodClamp); float textureGradClampARB(sampler2DShadow sampler, vec3 P, vec2 dPdx, vec2 dPdy, float lodClamp); float textureGradClampARB(samplerCubeShadow sampler, vec4 P, vec3 dPdx, vec3 dPdy, float lodClamp); gvec4 textureGradClampARB(gsampler1DArray sampler, vec2 P, float dPdx, float dPdy, float lodClamp); gvec4 textureGradClampARB(gsampler2DArray sampler, vec3 P, vec2 dPdx, vec2 dPdy, float lodClamp); float textureGradClampARB(sampler1DArrayShadow sampler, vec3 P, float dPdx, float dPdy, float lodClamp); float textureGradClampARB(sampler2DArrayShadow sampler, vec4 P, vec2 dPdx, vec2 dPdy, float lodClamp); gvec4 textureGradClampARB(gsamplerCubeArray sampler, vec4 P, vec3 dPdx, vec3 dPdy, float lodClamp); Description: Do a filtered texture lookup as in textureGrad(), but clamp the automatically computed level of detail to be greater than or equal to . -- Syntax: int sparseTextureGradOffsetClampARB(gsampler2D sampler, vec2 P, vec2 dPdx, vec2 dPdy, ivec2 offset, float lodClamp, out gvec4 texel); int sparseTextureGradOffsetClampARB(gsampler3D sampler, vec3 P, vec3 dPdx, vec3 dPdy, ivec3 offset, float lodClamp, out gvec4 texel); int sparseTextureGradOffsetClampARB(sampler2DShadow sampler, vec3 P, vec2 dPdx, vec2 dPdy, ivec2 offset, float lodClamp, out float texel); int sparseTextureGradOffsetClampARB(gsampler2DArray sampler, vec3 P, vec2 dPdx, vec2 dPdy, ivec2 offset, float lodClamp, out gvec4 texel); int sparseTextureGradOffsetClampARB(sampler2DArrayShadow sampler, vec4 P, vec2 dPdx, vec2 dPdy, ivec2 offset, float lodClamp, out float texel); Description: Do a filtered texture lookup as in textureGradOffset(), but return texture access residency information from the function and the filtered lookup result in the out parameter . Additionally, clamp the automatically computed level of detail to be greater than or equal to . -- Syntax: gvec4 textureGradOffsetClampARB(gsampler1D sampler, float P, float dPdx, float dPdy, int offset, float lodClamp); gvec4 textureGradOffsetClampARB(gsampler2D sampler, vec2 P, vec2 dPdx, vec2 dPdy, ivec2 offset, float lodClamp); gvec4 textureGradOffsetClampARB(gsampler3D sampler, vec3 P, vec3 dPdx, vec3 dPdy, ivec3 offset, float lodClamp); float textureGradOffsetClampARB(sampler1DShadow sampler, vec3 P, float dPdx, float dPdy, int offset, float lodClamp); float textureGradOffsetClampARB(sampler2DShadow sampler, vec3 P, vec2 dPdx, vec2 dPdy, ivec2 offset, float lodClamp); gvec4 textureGradOffsetClampARB(gsampler1DArray sampler, vec2 P, float dPdx, float dPdy, int offset, float lodClamp); gvec4 textureGradOffsetClampARB(gsampler2DArray sampler, vec3 P, vec2 dPdx, vec2 dPdy, ivec2 offset, float lodClamp); float textureGradOffsetClampARB(sampler1DArrayShadow sampler, vec3 P, float dPdx, float dPdy, int offset, float lodClamp); float textureGradOffsetClampARB(sampler2DArrayShadow sampler, vec4 P, vec2 dPdx, vec2 dPdy, ivec2 offset, float lodClamp); Description: Do a filtered texture lookup as in textureGrad(), but clamp the automatically computed level of detail to be greater than or equal to . Additions to the AGL/GLX/WGL Specifications None. Errors None. New State None. New Implementation Dependent State None. Dependencies on NV_gpu_program4 and NV_gpu_program5 Modify Section 2.X.2, Program Grammar ::= "LODCLAMP" Modify Section 2.X.4.1, Program Instruction Modifiers (Update the discussion of instruction precision modifiers. If GL_NV_gpu_program_fp64 is not found in the extension string, the "F64" instruction modifier described below is not supported.) (add to Table X.14 of the NV_gpu_program4 specification.) Modifier Description -------- ------------------------------------------------------ LODCLAMP Clamp the LOD used by texture lookups to a specified value For texture fetch instructions with implicit LOD calcuations (TEX, TXB, TXD), the "LODCLAMP" modifier specifies that the instruction includes an extra floating-point component indicating a minimum level of detail to be used for the texture lookup. If the implicitly computed level of detail is less than the level of detail provided in the instruction data, that level should be used instead. Modify Section 2.X.4.4, Program Texture Access (modify the prototype of the TextureSample utility function, adding and removing ) result_t_vec TextureSample(float_vec coord, float_vec coord2, float_vec ddx, float_vec ddy, int_vec offset); (modify the description of to add ) and are two four-component floating-point vectors from which the (s,t,r) texture coordinates used for the texture access, the layer used for array textures, and the reference value used for depth comparisons (section 3.8.14) are extracted according to Table X.17. ... (replace the paragraph discussing ) and specify partial derivatives (ds/dx, dt/dx, dr/dx, ds/dy, dt/dy, and dr/dy) for the texture coordinates, and may be used for level of detail calculations and to derive footprint shapes for anisotropic texture filtering. The level of detail used for the texture lookup is a function of the texture instruction type, texture target, LODCLAMP qualifier, and the inputs and . For TEX, TXB, TXD, and TXP instructions in a base level of detail is computed based on the partial derivatives and . For the TXB and TXL instruction, an additional level of detail value is taken from the component in or identified by the first entry in the "lod" column of Table X.17. For TXB, this value is added to the computed base level of detail; for TXL, it specifies the base level of detail. After that, per-texture and per-texture unit LOD biases are added to the level of detail. Finally, if the LODCLAMP opcode modifier is specified, an LOD clamp value is extracted from or according to the second entry in the "lod" column of Table X.17. The computed level of detail is clamped to be greater than or equal to this LOD clamp value. coordinates used texTarget Texture Type s t r lay shd lod ---------------- --------------------- ----- --- --- ----- 1D TEXTURE_1D x - - - - w,x2 2D TEXTURE_2D x y - - - w,x2 3D TEXTURE_3D x y z - - w,x2 CUBE TEXTURE_CUBE_MAP x y z - - w,x2 RECT TEXTURE_RECTANGLE_ARB x y - - - -,- ARRAY1D TEXTURE_1D_ARRAY_EXT x - - y - w,x2 ARRAY2D TEXTURE_2D_ARRAY_EXT x y - z - w,x2 ARRAYCUBE TEXTURE_CUBE_MAP_ARRAY x y z w - x2,y2 SHADOW1D TEXTURE_1D x - - - z w,x2 SHADOW2D TEXTURE_2D x y - - z w,x2 SHADOWRECT TEXTURE_RECTANGLE_ARB x y - - z -,- SHADOWCUBE TEXTURE_CUBE_MAP x y z - w x2,y2 SHADOWARRAY1D TEXTURE_1D_ARRAY_EXT x - - y z w,x2 SHADOWARRAY2D TEXTURE_2D_ARRAY_EXT x y - z w -,x2 SHADOWARRAYCUBE TEXTURE_CUBE_MAP_ARRAY x y z w x2 -,y2 BUFFER TEXTURE_BUFFER_EXT RENDERBUFFER TEXTURE_RENDERBUFFER 2DMS TEXTURE_2D_MULTISAMPLE ARRAY2DMS TEXTURE_2D_MULTISAMPLE_ ARRAY Table X.17: Texture types accessed for each of the , and coordinate mappings. Components "x", "y", "z", and "w" are taken from the first coordinate vector ; "x2" and "y2" are taken from the second vector . The "SHADOW" and "ARRAY" targets are special pseudo-targets described below. The "coordinates used" column indicate the input values used for each coordinate of the texture lookup, the layer selector for array textures, the reference value for texture comparisons, and up to two components of level-of-detail information. Buffer textures are not supported by normal texture lookup functions, but are supported by TXF and TXQ, described below. Renderbuffer and multisample textures are not supported by normal texture lookup functions, but are supported by TXFMS. The TXB and TXL instructions are not supported for the targets SHADOWARRAY2D and SHADOWARRAYCUBE, so the first column of "lod" is ignored. Modify Section 2.X.8.Z, TXD: Texture Sample with Partials ... The partial derivatives of the texture coordinates with respect to X and Y are specified by the second and third floating-point source vectors. If the LODCLAMP instruction modifier is specified, floating-point level-of-detail clamp value is specified in the component of the third floating-point source vector. The level of detail is computed automatically using the provided partial derivatives. Issues (1) How does this extension compare to the ARB_sparse_texture2 extension? RESOLVED: We extend the mechanisms provided by ARB_sparse_texture2 in adding built-in texture and image lookup functions clamping the final level of detail computed based on texture coordinates, derivatives, and LOD bias to a minimum LOD specified in the shader. (2) The combination of this extension and GL_ARB_sparse_texture2 add a *large* number of new built-in functions. What can we do to control this? RESOLVED: We chose not to add any "sparse" or "LOD clamp" variants of projective texture lookups (e.g., textureProj). If required, you can divide through by the "q" texture coordinate and use an equivalent non-projective lookup. We obviously don't support features that make no sense -- for example, LOD clamp on single-level rectangle textures. We considered the possibility of more significant GLSL syntax changes to reduce the cross-product of different features. For example, the AMD extension has a function: int sparseTextureProjGradOffset(...); that combines four separate "optional" features (sparse, projection, explicitly specified gradients, and texel offsets) and is supported for six separate texture targets. One might consider an approach like: #define TEX_IS_PROJECTIVE 0x1 #define TEX_HAS_GRADIENTS 0x2 #define TEX_HAS_TEXEL_OFFSET 0x4 #define TEX_WANTS_SPARSE_STATUS 0x8 struct TexLookup3D { uint flags; /* in */ float q; /* in */ vec3 ddx, ddy; /* in */ ivec3 texelOffset; /* in */ int sparseStatus; /* out */ }; ... TexLookup3D lookup; lookup.flags = (TEX_IS_PROJECTIVE | TEX_HAS_GRADIENTS | TEX_HAS_TEXEL_OFFSET | TEX_WANTS_SPARSE_STATUS); lookup.q = coords.w; lookup.ddx = ddx; lookup.ddy = ddy; lookup.texelOffset = ivec3(-1,+1,+2); texture(sampler, lookup); to handle all possible cases in one interface. Alternately, a "prettier" C++-style approach with methods on sampler classes could be used. Given that either such feature might involve a large change to the shading language, it seems more appropriate to address this issue in a future core version of a shading language rather than an extension. (3) For new "LOD clamp" functions, how does the LOD clamp interact with the LOD bias? RESOLVED: The LOD clamp is applied after the LOD bias. Clamping to the LOD provided in the shader is logically applied at the same point in the pipeline where the LOD clamps based on the texture/sampler parameters TEXTURE_{MIN,MAX}_LOD are applied. Revision History Revision 1 - Split clamp functions from ARB_sparse_texture2