/************************************************************************** * * Copyright 2006 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, sub license, 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 (including the * next paragraph) 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 NON-INFRINGEMENT. * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS 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. * **************************************************************************/ #ifndef INTEL_MIPMAP_TREE_H #define INTEL_MIPMAP_TREE_H #include #include "intel_regions.h" #include "intel_resolve_map.h" #include #ifdef __cplusplus extern "C" { #endif /* A layer on top of the intel_regions code which adds: * * - Code to size and layout a region to hold a set of mipmaps. * - Query to determine if a new image fits in an existing tree. * - More refcounting * - maybe able to remove refcounting from intel_region? * - ? * * The fixed mipmap layout of intel hardware where one offset * specifies the position of all images in a mipmap hierachy * complicates the implementation of GL texture image commands, * compared to hardware where each image is specified with an * independent offset. * * In an ideal world, each texture object would be associated with a * single bufmgr buffer or 2d intel_region, and all the images within * the texture object would slot into the tree as they arrive. The * reality can be a little messier, as images can arrive from the user * with sizes that don't fit in the existing tree, or in an order * where the tree layout cannot be guessed immediately. * * This structure encodes an idealized mipmap tree. The GL image * commands build these where possible, otherwise store the images in * temporary system buffers. */ struct intel_resolve_map; struct intel_texture_image; /** * When calling intel_miptree_map() on an ETC-transcoded-to-RGB miptree or a * depthstencil-split-to-separate-stencil miptree, we'll normally make a * tmeporary and recreate the kind of data requested by Mesa core, since we're * satisfying some glGetTexImage() request or something. * * However, occasionally you want to actually map the miptree's current data * without transcoding back. This flag to intel_miptree_map() gets you that. */ #define BRW_MAP_DIRECT_BIT 0x80000000 struct intel_miptree_map { /** Bitfield of GL_MAP_READ_BIT, GL_MAP_WRITE_BIT, GL_MAP_INVALIDATE_BIT */ GLbitfield mode; /** Region of interest for the map. */ int x, y, w, h; /** Possibly malloced temporary buffer for the mapping. */ void *buffer; /** Possible pointer to a temporary linear miptree for the mapping. */ struct intel_mipmap_tree *mt; /** Pointer to the start of (map_x, map_y) returned by the mapping. */ void *ptr; /** Stride of the mapping. */ int stride; /** * intel_mipmap_tree::singlesample_mt is temporary storage that persists * only for the duration of the map. */ bool singlesample_mt_is_tmp; }; /** * Describes the location of each texture image within a texture region. */ struct intel_mipmap_level { /** Offset to this miptree level, used in computing x_offset. */ GLuint level_x; /** Offset to this miptree level, used in computing y_offset. */ GLuint level_y; GLuint width; GLuint height; /** * \brief Number of 2D slices in this miplevel. * * The exact semantics of depth varies according to the texture target: * - For GL_TEXTURE_CUBE_MAP, depth is 6. * - For GL_TEXTURE_2D_ARRAY, depth is the number of array slices. It is * identical for all miplevels in the texture. * - For GL_TEXTURE_3D, it is the texture's depth at this miplevel. Its * value, like width and height, varies with miplevel. * - For other texture types, depth is 1. */ GLuint depth; /** * \brief List of 2D images in this mipmap level. * * This may be a list of cube faces, array slices in 2D array texture, or * layers in a 3D texture. The list's length is \c depth. */ struct intel_mipmap_slice { /** * \name Offset to slice * \{ * * Hardware formats are so diverse that that there is no unified way to * compute the slice offsets, so we store them in this table. * * The (x, y) offset to slice \c s at level \c l relative the miptrees * base address is * \code * x = mt->level[l].slice[s].x_offset * y = mt->level[l].slice[s].y_offset */ GLuint x_offset; GLuint y_offset; /** \} */ /** * Mapping information. Persistent for the duration of * intel_miptree_map/unmap on this slice. */ struct intel_miptree_map *map; /** * \brief Is HiZ enabled for this slice? * * If \c mt->level[l].slice[s].has_hiz is set, then (1) \c mt->hiz_mt * has been allocated and (2) the HiZ memory corresponding to this slice * resides at \c mt->hiz_mt->level[l].slice[s]. */ bool has_hiz; } *slice; }; /** * Enum for keeping track of the different MSAA layouts supported by Gen7. */ enum intel_msaa_layout { /** * Ordinary surface with no MSAA. */ INTEL_MSAA_LAYOUT_NONE, /** * Interleaved Multisample Surface. The additional samples are * accommodated by scaling up the width and the height of the surface so * that all the samples corresponding to a pixel are located at nearby * memory locations. */ INTEL_MSAA_LAYOUT_IMS, /** * Uncompressed Multisample Surface. The surface is stored as a 2D array, * with array slice n containing all pixel data for sample n. */ INTEL_MSAA_LAYOUT_UMS, /** * Compressed Multisample Surface. The surface is stored as in * INTEL_MSAA_LAYOUT_UMS, but there is an additional buffer called the MCS * (Multisample Control Surface) buffer. Each pixel in the MCS buffer * indicates the mapping from sample number to array slice. This allows * the common case (where all samples constituting a pixel have the same * color value) to be stored efficiently by just using a single array * slice. */ INTEL_MSAA_LAYOUT_CMS, }; /** * Enum for keeping track of the fast clear state of a buffer associated with * a miptree. * * Fast clear works by deferring the memory writes that would be used to clear * the buffer, so that instead of performing them at the time of the clear * operation, the hardware automatically performs them at the time that the * buffer is later accessed for rendering. The MCS buffer keeps track of * which regions of the buffer still have pending clear writes. * * This enum keeps track of the driver's knowledge of pending fast clears in * the MCS buffer. * * MCS buffers only exist on Gen7+. */ enum intel_fast_clear_state { /** * There is no MCS buffer for this miptree, and one should never be * allocated. */ INTEL_FAST_CLEAR_STATE_NO_MCS, /** * No deferred clears are pending for this miptree, and the contents of the * color buffer are entirely correct. An MCS buffer may or may not exist * for this miptree. If it does exist, it is entirely in the "no deferred * clears pending" state. If it does not exist, it will be created the * first time a fast color clear is executed. * * In this state, the color buffer can be used for purposes other than * rendering without needing a render target resolve. * * Since there is no such thing as a "fast color clear resolve" for MSAA * buffers, an MSAA buffer will never be in this state. */ INTEL_FAST_CLEAR_STATE_RESOLVED, /** * An MCS buffer exists for this miptree, and deferred clears are pending * for some regions of the color buffer, as indicated by the MCS buffer. * The contents of the color buffer are only correct for the regions where * the MCS buffer doesn't indicate a deferred clear. * * If a single-sample buffer is in this state, a render target resolve must * be performed before it can be used for purposes other than rendering. */ INTEL_FAST_CLEAR_STATE_UNRESOLVED, /** * An MCS buffer exists for this miptree, and deferred clears are pending * for the entire color buffer, and the contents of the MCS buffer reflect * this. The contents of the color buffer are undefined. * * If a single-sample buffer is in this state, a render target resolve must * be performed before it can be used for purposes other than rendering. * * If the client attempts to clear a buffer which is already in this state, * the clear can be safely skipped, since the buffer is already clear. */ INTEL_FAST_CLEAR_STATE_CLEAR, }; struct intel_mipmap_tree { /* Effectively the key: */ GLenum target; /** * Generally, this is just the same as the gl_texture_image->TexFormat or * gl_renderbuffer->Format. * * However, for textures and renderbuffers with packed depth/stencil formats * on hardware where we want or need to use separate stencil, there will be * two miptrees for storing the data. If the depthstencil texture or rb is * MESA_FORMAT_Z32_FLOAT_S8X24_UINT, then mt->format will be * MESA_FORMAT_Z_FLOAT32, otherwise for MESA_FORMAT_Z24_UNORM_X8_UINT objects it will be * MESA_FORMAT_Z24_UNORM_S8_UINT. * * For ETC1/ETC2 textures, this is one of the uncompressed mesa texture * formats if the hardware lacks support for ETC1/ETC2. See @ref wraps_etc. */ mesa_format format; /** This variable stores the value of ETC compressed texture format */ mesa_format etc_format; /** * The X offset of each image in the miptree must be aligned to this. * See the comments in brw_tex_layout.c. */ unsigned int align_w; unsigned int align_h; /**< \see align_w */ GLuint first_level; GLuint last_level; /** * Level zero image dimensions. These dimensions correspond to the * physical layout of data in memory. Accordingly, they account for the * extra width, height, and or depth that must be allocated in order to * accommodate multisample formats, and they account for the extra factor * of 6 in depth that must be allocated in order to accommodate cubemap * textures. */ GLuint physical_width0, physical_height0, physical_depth0; GLuint cpp; GLuint num_samples; bool compressed; /** * Level zero image dimensions. These dimensions correspond to the * logical width, height, and depth of the region as seen by client code. * Accordingly, they do not account for the extra width, height, and/or * depth that must be allocated in order to accommodate multisample * formats, nor do they account for the extra factor of 6 in depth that * must be allocated in order to accommodate cubemap textures. */ uint32_t logical_width0, logical_height0, logical_depth0; /** * For 1D array, 2D array, cube, and 2D multisampled surfaces on Gen7: true * if the surface only contains LOD 0, and hence no space is for LOD's * other than 0 in between array slices. * * Corresponds to the surface_array_spacing bit in gen7_surface_state. */ bool array_spacing_lod0; /** * The distance in rows between array slices in an uncompressed surface. * * For compressed surfaces, slices are stored closer together physically; * the real distance is (qpitch / block height). */ uint32_t qpitch; /** * MSAA layout used by this buffer. */ enum intel_msaa_layout msaa_layout; /* Derived from the above: */ GLuint total_width; GLuint total_height; /* The 3DSTATE_CLEAR_PARAMS value associated with the last depth clear to * this depth mipmap tree, if any. */ uint32_t depth_clear_value; /* Includes image offset tables: */ struct intel_mipmap_level level[MAX_TEXTURE_LEVELS]; /* The data is held here: */ struct intel_region *region; /* Offset into region bo where miptree starts: */ uint32_t offset; /** * \brief Singlesample miptree. * * This is used under two cases. * * --- Case 1: As persistent singlesample storage for multisample window * system front and back buffers --- * * Suppose that the window system FBO was created with a multisample * config. Let `back_irb` be the `intel_renderbuffer` for the FBO's back * buffer. Then `back_irb` contains two miptrees: a parent multisample * miptree (back_irb->mt) and a child singlesample miptree * (back_irb->mt->singlesample_mt). The DRM buffer shared with DRI2 * belongs to `back_irb->mt->singlesample_mt` and contains singlesample * data. The singlesample miptree is created at the same time as and * persists for the lifetime of its parent multisample miptree. * * When access to the singlesample data is needed, such as at * eglSwapBuffers and glReadPixels, an automatic downsample occurs from * `back_rb->mt` to `back_rb->mt->singlesample_mt` when necessary. * * This description of the back buffer applies analogously to the front * buffer. * * * --- Case 2: As temporary singlesample storage for mapping multisample * miptrees --- * * Suppose the intel_miptree_map is called on a multisample miptree, `mt`, * for which case 1 does not apply (that is, `mt` does not belong to * a front or back buffer). Then `mt->singlesample_mt` is null at the * start of the call. intel_miptree_map will create a temporary * singlesample miptree, store it at `mt->singlesample_mt`, downsample from * `mt` to `mt->singlesample_mt` if necessary, then map * `mt->singlesample_mt`. The temporary miptree is later deleted during * intel_miptree_unmap. */ struct intel_mipmap_tree *singlesample_mt; /** * \brief A downsample is needed from this miptree to singlesample_mt. */ bool need_downsample; /** * \brief HiZ miptree * * The hiz miptree contains the miptree's hiz buffer. To allocate the hiz * miptree, use intel_miptree_alloc_hiz(). * * To determine if hiz is enabled, do not check this pointer. Instead, use * intel_miptree_slice_has_hiz(). */ struct intel_mipmap_tree *hiz_mt; /** * \brief Map of miptree slices to needed resolves. * * This is used only when the miptree has a child HiZ miptree. * * Let \c mt be a depth miptree with HiZ enabled. Then the resolve map is * \c mt->hiz_map. The resolve map of the child HiZ miptree, \c * mt->hiz_mt->hiz_map, is unused. */ struct intel_resolve_map hiz_map; /** * \brief Stencil miptree for depthstencil textures. * * This miptree is used for depthstencil textures and renderbuffers that * require separate stencil. It always has the true copy of the stencil * bits, regardless of mt->format. * * \see intel_miptree_map_depthstencil() * \see intel_miptree_unmap_depthstencil() */ struct intel_mipmap_tree *stencil_mt; /** * \brief MCS miptree. * * This miptree contains the "multisample control surface", which stores * the necessary information to implement compressed MSAA * (INTEL_MSAA_FORMAT_CMS) and "fast color clear" behaviour on Gen7+. * * NULL if no MCS miptree is in use for this surface. */ struct intel_mipmap_tree *mcs_mt; /** * Fast clear state for this buffer. */ enum intel_fast_clear_state fast_clear_state; /** * The SURFACE_STATE bits associated with the last fast color clear to this * color mipmap tree, if any. * * This value will only ever contain ones in bits 28-31, so it is safe to * OR into dword 7 of SURFACE_STATE. */ uint32_t fast_clear_color_value; /* These are also refcounted: */ GLuint refcount; }; enum intel_miptree_tiling_mode { INTEL_MIPTREE_TILING_ANY, INTEL_MIPTREE_TILING_Y, INTEL_MIPTREE_TILING_NONE, }; bool intel_is_non_msrt_mcs_buffer_supported(struct brw_context *brw, struct intel_mipmap_tree *mt); void intel_get_non_msrt_mcs_alignment(struct brw_context *brw, struct intel_mipmap_tree *mt, unsigned *width_px, unsigned *height); bool intel_miptree_alloc_non_msrt_mcs(struct brw_context *brw, struct intel_mipmap_tree *mt); struct intel_mipmap_tree *intel_miptree_create(struct brw_context *brw, GLenum target, mesa_format format, GLuint first_level, GLuint last_level, GLuint width0, GLuint height0, GLuint depth0, bool expect_accelerated_upload, GLuint num_samples, enum intel_miptree_tiling_mode); struct intel_mipmap_tree * intel_miptree_create_layout(struct brw_context *brw, GLenum target, mesa_format format, GLuint first_level, GLuint last_level, GLuint width0, GLuint height0, GLuint depth0, bool for_bo, GLuint num_samples); struct intel_mipmap_tree * intel_miptree_create_for_bo(struct brw_context *brw, drm_intel_bo *bo, mesa_format format, uint32_t offset, uint32_t width, uint32_t height, int pitch, uint32_t tiling); struct intel_mipmap_tree* intel_miptree_create_for_dri2_buffer(struct brw_context *brw, unsigned dri_attachment, mesa_format format, uint32_t num_samples, struct intel_region *region); struct intel_mipmap_tree* intel_miptree_create_for_image_buffer(struct brw_context *intel, enum __DRIimageBufferMask buffer_type, mesa_format format, uint32_t num_samples, struct intel_region *region); /** * Create a miptree appropriate as the storage for a non-texture renderbuffer. * The miptree has the following properties: * - The target is GL_TEXTURE_2D. * - There are no levels other than the base level 0. * - Depth is 1. */ struct intel_mipmap_tree* intel_miptree_create_for_renderbuffer(struct brw_context *brw, mesa_format format, uint32_t width, uint32_t height, uint32_t num_samples); /** \brief Assert that the level and layer are valid for the miptree. */ static inline void intel_miptree_check_level_layer(struct intel_mipmap_tree *mt, uint32_t level, uint32_t layer) { assert(level >= mt->first_level); assert(level <= mt->last_level); assert(layer < mt->level[level].depth); } void intel_miptree_reference(struct intel_mipmap_tree **dst, struct intel_mipmap_tree *src); void intel_miptree_release(struct intel_mipmap_tree **mt); /* Check if an image fits an existing mipmap tree layout */ bool intel_miptree_match_image(struct intel_mipmap_tree *mt, struct gl_texture_image *image); void intel_miptree_get_image_offset(struct intel_mipmap_tree *mt, GLuint level, GLuint slice, GLuint *x, GLuint *y); void intel_miptree_get_dimensions_for_image(struct gl_texture_image *image, int *width, int *height, int *depth); uint32_t intel_miptree_get_tile_offsets(struct intel_mipmap_tree *mt, GLuint level, GLuint slice, uint32_t *tile_x, uint32_t *tile_y); void intel_miptree_set_level_info(struct intel_mipmap_tree *mt, GLuint level, GLuint x, GLuint y, GLuint w, GLuint h, GLuint d); void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, GLuint level, GLuint img, GLuint x, GLuint y); void intel_miptree_copy_teximage(struct brw_context *brw, struct intel_texture_image *intelImage, struct intel_mipmap_tree *dst_mt, bool invalidate); bool intel_miptree_alloc_mcs(struct brw_context *brw, struct intel_mipmap_tree *mt, GLuint num_samples); /** * \name Miptree HiZ functions * \{ * * It is safe to call the "slice_set_need_resolve" and "slice_resolve" * functions on a miptree without HiZ. In that case, each function is a no-op. */ /** * \brief Allocate the miptree's embedded HiZ miptree. * \see intel_mipmap_tree:hiz_mt * \return false if allocation failed */ bool intel_miptree_alloc_hiz(struct brw_context *brw, struct intel_mipmap_tree *mt); bool intel_miptree_slice_has_hiz(struct intel_mipmap_tree *mt, uint32_t level, uint32_t layer); void intel_miptree_slice_set_needs_hiz_resolve(struct intel_mipmap_tree *mt, uint32_t level, uint32_t depth); void intel_miptree_slice_set_needs_depth_resolve(struct intel_mipmap_tree *mt, uint32_t level, uint32_t depth); void intel_miptree_set_all_slices_need_depth_resolve(struct intel_mipmap_tree *mt, uint32_t level); /** * \return false if no resolve was needed */ bool intel_miptree_slice_resolve_hiz(struct brw_context *brw, struct intel_mipmap_tree *mt, unsigned int level, unsigned int depth); /** * \return false if no resolve was needed */ bool intel_miptree_slice_resolve_depth(struct brw_context *brw, struct intel_mipmap_tree *mt, unsigned int level, unsigned int depth); /** * \return false if no resolve was needed */ bool intel_miptree_all_slices_resolve_hiz(struct brw_context *brw, struct intel_mipmap_tree *mt); /** * \return false if no resolve was needed */ bool intel_miptree_all_slices_resolve_depth(struct brw_context *brw, struct intel_mipmap_tree *mt); /**\}*/ /** * Update the fast clear state for a miptree to indicate that it has been used * for rendering. */ static inline void intel_miptree_used_for_rendering(struct intel_mipmap_tree *mt) { /* If the buffer was previously in fast clear state, change it to * unresolved state, since it won't be guaranteed to be clear after * rendering occurs. */ if (mt->fast_clear_state == INTEL_FAST_CLEAR_STATE_CLEAR) mt->fast_clear_state = INTEL_FAST_CLEAR_STATE_UNRESOLVED; } void intel_miptree_resolve_color(struct brw_context *brw, struct intel_mipmap_tree *mt); void intel_miptree_make_shareable(struct brw_context *brw, struct intel_mipmap_tree *mt); void intel_miptree_downsample(struct brw_context *brw, struct intel_mipmap_tree *mt); void intel_miptree_upsample(struct brw_context *brw, struct intel_mipmap_tree *mt); void brw_miptree_layout(struct brw_context *brw, struct intel_mipmap_tree *mt); void *intel_miptree_map_raw(struct brw_context *brw, struct intel_mipmap_tree *mt); void intel_miptree_unmap_raw(struct brw_context *brw, struct intel_mipmap_tree *mt); void intel_miptree_map(struct brw_context *brw, struct intel_mipmap_tree *mt, unsigned int level, unsigned int slice, unsigned int x, unsigned int y, unsigned int w, unsigned int h, GLbitfield mode, void **out_ptr, int *out_stride); void intel_miptree_unmap(struct brw_context *brw, struct intel_mipmap_tree *mt, unsigned int level, unsigned int slice); void intel_hiz_exec(struct brw_context *brw, struct intel_mipmap_tree *mt, unsigned int level, unsigned int layer, enum gen6_hiz_op op); #ifdef __cplusplus } #endif #endif