X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fmesa%2Fdrivers%2Fdri%2Fi965%2Fbrw_tex_layout.c;h=45aa85471afe12e0f588b852fef6fc986ec953c6;hb=877128505431adaf817dc8069172ebe4a1cdf5d8;hp=64a9535282bd43f8a0cf17fbf502bc559c859cfc;hpb=e32487b8a13a9efabb0359a9dde33e074e905e82;p=mesa.git diff --git a/src/mesa/drivers/dri/i965/brw_tex_layout.c b/src/mesa/drivers/dri/i965/brw_tex_layout.c index 64a9535282b..45aa85471af 100644 --- a/src/mesa/drivers/dri/i965/brw_tex_layout.c +++ b/src/mesa/drivers/dri/i965/brw_tex_layout.c @@ -1,223 +1,348 @@ /* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - 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, sublicense, 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 NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -/* Code to layout images in a mipmap tree for i965. + * Copyright 2006 VMware, Inc. + * Copyright © 2006 Intel Corporation + * + * 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, sublicense, 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 NONINFRINGEMENT. + * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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. + */ + +/** + * \file brw_tex_layout.cpp + * + * Code to lay out images in a mipmap tree. + * + * \author Keith Whitwell + * \author Michel Dänzer */ #include "intel_mipmap_tree.h" -#include "intel_tex_layout.h" -#include "intel_context.h" +#include "brw_context.h" #include "main/macros.h" -#include "intel_chipset.h" #define FILE_DEBUG_FLAG DEBUG_MIPTREE -GLboolean brw_miptree_layout(struct intel_context *intel, - struct intel_mipmap_tree *mt, - uint32_t tiling) +static unsigned int +intel_horizontal_texture_alignment_unit(struct brw_context *brw, + gl_format format) { - /* XXX: these vary depending on image format: */ - /* GLint align_w = 4; */ + /** + * From the "Alignment Unit Size" section of various specs, namely: + * - Gen3 Spec: "Memory Data Formats" Volume, Section 1.20.1.4 + * - i965 and G45 PRMs: Volume 1, Section 6.17.3.4. + * - Ironlake and Sandybridge PRMs: Volume 1, Part 1, Section 7.18.3.4 + * - BSpec (for Ivybridge and slight variations in separate stencil) + * + * +----------------------------------------------------------------------+ + * | | alignment unit width ("i") | + * | Surface Property |-----------------------------| + * | | 915 | 965 | ILK | SNB | IVB | + * +----------------------------------------------------------------------+ + * | YUV 4:2:2 format | 8 | 4 | 4 | 4 | 4 | + * | BC1-5 compressed format (DXTn/S3TC) | 4 | 4 | 4 | 4 | 4 | + * | FXT1 compressed format | 8 | 8 | 8 | 8 | 8 | + * | Depth Buffer (16-bit) | 4 | 4 | 4 | 4 | 8 | + * | Depth Buffer (other) | 4 | 4 | 4 | 4 | 4 | + * | Separate Stencil Buffer | N/A | N/A | 8 | 8 | 8 | + * | All Others | 4 | 4 | 4 | 4 | 4 | + * +----------------------------------------------------------------------+ + * + * On IVB+, non-special cases can be overridden by setting the SURFACE_STATE + * "Surface Horizontal Alignment" field to HALIGN_4 or HALIGN_8. + */ + if (_mesa_is_format_compressed(format)) { + /* The hardware alignment requirements for compressed textures + * happen to match the block boundaries. + */ + unsigned int i, j; + _mesa_get_format_block_size(format, &i, &j); + return i; + } + + if (format == MESA_FORMAT_S8) + return 8; + + if (brw->gen >= 7 && format == MESA_FORMAT_Z16) + return 8; + + return 4; +} - switch (mt->target) { - case GL_TEXTURE_CUBE_MAP: - if (intel->is_ironlake) { - GLuint align_h = 2, align_w = 4; - GLuint level; - GLuint x = 0; - GLuint y = 0; - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint qpitch = 0; - GLuint y_pitch = 0; - - mt->pitch = mt->width0; - intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h); - y_pitch = ALIGN(height, align_h); - - if (mt->compressed) { - mt->pitch = ALIGN(mt->width0, align_w); - } - - if (mt->first_level != mt->last_level) { - GLuint mip1_width; - - if (mt->compressed) { - mip1_width = ALIGN(minify(mt->width0), align_w) - + ALIGN(minify(minify(mt->width0)), align_w); - } else { - mip1_width = ALIGN(minify(mt->width0), align_w) - + minify(minify(mt->width0)); - } - - if (mip1_width > mt->pitch) { - mt->pitch = mip1_width; - } - } - - mt->pitch = intel_miptree_pitch_align(intel, mt, tiling, mt->pitch); - - if (mt->compressed) { - qpitch = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) / 4; - mt->total_height = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) / 4 * 6; - } else { - qpitch = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h); - mt->total_height = (y_pitch + ALIGN(minify(y_pitch), align_h) + 11 * align_h) * 6; - } - - for (level = mt->first_level; level <= mt->last_level; level++) { - GLuint img_height; - GLuint nr_images = 6; - GLuint q = 0; - - intel_miptree_set_level_info(mt, level, nr_images, x, y, width, - height, 1); - - for (q = 0; q < nr_images; q++) - intel_miptree_set_image_offset(mt, level, q, - x, y + q * qpitch); - - if (mt->compressed) - img_height = MAX2(1, height/4); - else - img_height = ALIGN(height, align_h); - - if (level == mt->first_level + 1) { - x += ALIGN(width, align_w); - } - else { - y += img_height; - } - - width = minify(width); - height = minify(height); - } - - break; - } +static unsigned int +intel_vertical_texture_alignment_unit(struct brw_context *brw, + gl_format format, bool multisampled) +{ + /** + * From the "Alignment Unit Size" section of various specs, namely: + * - Gen3 Spec: "Memory Data Formats" Volume, Section 1.20.1.4 + * - i965 and G45 PRMs: Volume 1, Section 6.17.3.4. + * - Ironlake and Sandybridge PRMs: Volume 1, Part 1, Section 7.18.3.4 + * - BSpec (for Ivybridge and slight variations in separate stencil) + * + * +----------------------------------------------------------------------+ + * | | alignment unit height ("j") | + * | Surface Property |-----------------------------| + * | | 915 | 965 | ILK | SNB | IVB | + * +----------------------------------------------------------------------+ + * | BC1-5 compressed format (DXTn/S3TC) | 4 | 4 | 4 | 4 | 4 | + * | FXT1 compressed format | 4 | 4 | 4 | 4 | 4 | + * | Depth Buffer | 2 | 2 | 2 | 4 | 4 | + * | Separate Stencil Buffer | N/A | N/A | N/A | 4 | 8 | + * | Multisampled (4x or 8x) render target | N/A | N/A | N/A | 4 | 4 | + * | All Others | 2 | 2 | 2 | * | * | + * +----------------------------------------------------------------------+ + * + * Where "*" means either VALIGN_2 or VALIGN_4 depending on the setting of + * the SURFACE_STATE "Surface Vertical Alignment" field. + */ + if (_mesa_is_format_compressed(format)) + return 4; + + if (format == MESA_FORMAT_S8) + return brw->gen >= 7 ? 8 : 4; + + /* Broadwell only supports VALIGN of 4, 8, and 16. The BSpec says 4 + * should always be used, except for stencil buffers, which should be 8. + */ + if (brw->gen >= 8) + return 4; + + if (multisampled) + return 4; + + GLenum base_format = _mesa_get_format_base_format(format); + + if (brw->gen >= 6 && + (base_format == GL_DEPTH_COMPONENT || + base_format == GL_DEPTH_STENCIL)) { + return 4; + } + + if (brw->gen == 7) { + /* On Gen7, we prefer a vertical alignment of 4 when possible, because + * that allows Y tiled render targets. + * + * From the Ivy Bridge PRM, Vol4 Part1 2.12.2.1 (SURFACE_STATE for most + * messages), on p64, under the heading "Surface Vertical Alignment": + * + * Value of 1 [VALIGN_4] is not supported for format YCRCB_NORMAL + * (0x182), YCRCB_SWAPUVY (0x183), YCRCB_SWAPUV (0x18f), YCRCB_SWAPY + * (0x190) + * + * VALIGN_4 is not supported for surface format R32G32B32_FLOAT. + */ + if (base_format == GL_YCBCR_MESA || format == MESA_FORMAT_RGB_FLOAT32) + return 2; + + return 4; + } + + return 2; +} + +static void +brw_miptree_layout_2d(struct intel_mipmap_tree *mt) +{ + unsigned x = 0; + unsigned y = 0; + unsigned width = mt->physical_width0; + unsigned height = mt->physical_height0; + unsigned depth = mt->physical_depth0; /* number of array layers. */ + + mt->total_width = mt->physical_width0; - case GL_TEXTURE_3D: { - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint depth = mt->depth0; - GLuint pack_x_pitch, pack_x_nr; - GLuint pack_y_pitch; - GLuint level; - GLuint align_h = 2; - GLuint align_w = 4; - - mt->total_height = 0; - intel_get_texture_alignment_unit(mt->internal_format, &align_w, &align_h); - - if (mt->compressed) { - mt->pitch = ALIGN(width, align_w); - pack_y_pitch = (height + 3) / 4; + if (mt->compressed) { + mt->total_width = ALIGN(mt->physical_width0, mt->align_w); + } + + /* May need to adjust width to accomodate the placement of + * the 2nd mipmap. This occurs when the alignment + * constraints of mipmap placement push the right edge of the + * 2nd mipmap out past the width of its parent. + */ + if (mt->first_level != mt->last_level) { + unsigned mip1_width; + + if (mt->compressed) { + mip1_width = ALIGN(minify(mt->physical_width0, 1), mt->align_w) + + ALIGN(minify(mt->physical_width0, 2), mt->align_w); + } else { + mip1_width = ALIGN(minify(mt->physical_width0, 1), mt->align_w) + + minify(mt->physical_width0, 2); + } + + if (mip1_width > mt->total_width) { + mt->total_width = mip1_width; + } + } + + mt->total_height = 0; + + for (unsigned level = mt->first_level; level <= mt->last_level; level++) { + unsigned img_height; + + intel_miptree_set_level_info(mt, level, x, y, width, + height, depth); + + img_height = ALIGN(height, mt->align_h); + if (mt->compressed) + img_height /= mt->align_h; + + /* Because the images are packed better, the final offset + * might not be the maximal one: + */ + mt->total_height = MAX2(mt->total_height, y + img_height); + + /* Layout_below: step right after second mipmap. + */ + if (level == mt->first_level + 1) { + x += ALIGN(width, mt->align_w); } else { - mt->pitch = intel_miptree_pitch_align (intel, mt, tiling, mt->width0); - pack_y_pitch = ALIGN(mt->height0, align_h); + y += img_height; } - pack_x_pitch = width; - pack_x_nr = 1; - - for (level = mt->first_level ; level <= mt->last_level ; level++) { - GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6; - GLint x = 0; - GLint y = 0; - GLint q, j; - - intel_miptree_set_level_info(mt, level, nr_images, - 0, mt->total_height, - width, height, depth); - - for (q = 0; q < nr_images;) { - for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { - intel_miptree_set_image_offset(mt, level, q, x, y); - x += pack_x_pitch; - } - - x = 0; - y += pack_y_pitch; - } - - - mt->total_height += y; - width = minify(width); - height = minify(height); - depth = minify(depth); - - if (mt->compressed) { - pack_y_pitch = (height + 3) / 4; - - if (pack_x_pitch > ALIGN(width, align_w)) { - pack_x_pitch = ALIGN(width, align_w); - pack_x_nr <<= 1; - } - } else { - if (pack_x_pitch > 4) { - pack_x_pitch >>= 1; - pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr <= mt->pitch); - } - - if (pack_y_pitch > 2) { - pack_y_pitch >>= 1; - pack_y_pitch = ALIGN(pack_y_pitch, align_h); - } - } + width = minify(width, 1); + height = minify(height, 1); + } +} + +static void +align_cube(struct intel_mipmap_tree *mt) +{ + /* The 965's sampler lays cachelines out according to how accesses + * in the texture surfaces run, so they may be "vertical" through + * memory. As a result, the docs say in Surface Padding Requirements: + * Sampling Engine Surfaces that two extra rows of padding are required. + */ + if (mt->target == GL_TEXTURE_CUBE_MAP) + mt->total_height += 2; +} +static void +brw_miptree_layout_texture_array(struct brw_context *brw, + struct intel_mipmap_tree *mt) +{ + int h0, h1; + + h0 = ALIGN(mt->physical_height0, mt->align_h); + h1 = ALIGN(minify(mt->physical_height0, 1), mt->align_h); + if (mt->array_spacing_lod0) + mt->qpitch = h0; + else + mt->qpitch = (h0 + h1 + (brw->gen >= 7 ? 12 : 11) * mt->align_h); + if (mt->compressed) + mt->qpitch /= 4; + + brw_miptree_layout_2d(mt); + + for (unsigned level = mt->first_level; level <= mt->last_level; level++) { + for (int q = 0; q < mt->physical_depth0; q++) { + intel_miptree_set_image_offset(mt, level, q, 0, q * mt->qpitch); } - /* The 965's sampler lays cachelines out according to how accesses - * in the texture surfaces run, so they may be "vertical" through - * memory. As a result, the docs say in Surface Padding Requirements: - * Sampling Engine Surfaces that two extra rows of padding are required. - * We don't know of similar requirements for pre-965, but given that - * those docs are silent on padding requirements in general, let's play - * it safe. - */ + } + mt->total_height = mt->qpitch * mt->physical_depth0; + + align_cube(mt); +} + +static void +brw_miptree_layout_texture_3d(struct brw_context *brw, + struct intel_mipmap_tree *mt) +{ + unsigned yscale = mt->compressed ? 4 : 1; + + mt->total_width = 0; + mt->total_height = 0; + + unsigned ysum = 0; + for (unsigned level = mt->first_level; level <= mt->last_level; level++) { + unsigned WL = MAX2(mt->physical_width0 >> level, 1); + unsigned HL = MAX2(mt->physical_height0 >> level, 1); + unsigned DL = MAX2(mt->physical_depth0 >> level, 1); + unsigned wL = ALIGN(WL, mt->align_w); + unsigned hL = ALIGN(HL, mt->align_h); + if (mt->target == GL_TEXTURE_CUBE_MAP) - mt->total_height += 2; - break; + DL = 6; + + intel_miptree_set_level_info(mt, level, 0, 0, WL, HL, DL); + + for (unsigned q = 0; q < DL; q++) { + unsigned x = (q % (1 << level)) * wL; + unsigned y = ysum + (q >> level) * hL; + + intel_miptree_set_image_offset(mt, level, q, x, y / yscale); + mt->total_width = MAX2(mt->total_width, x + wL); + mt->total_height = MAX2(mt->total_height, (y + hL) / yscale); + } + + ysum += ALIGN(DL, 1 << level) / (1 << level) * hL; } + align_cube(mt); +} + +void +brw_miptree_layout(struct brw_context *brw, struct intel_mipmap_tree *mt) +{ + bool multisampled = mt->num_samples > 1; + mt->align_w = intel_horizontal_texture_alignment_unit(brw, mt->format); + mt->align_h = + intel_vertical_texture_alignment_unit(brw, mt->format, multisampled); + + switch (mt->target) { + case GL_TEXTURE_CUBE_MAP: + if (brw->gen == 4) { + /* Gen4 stores cube maps as 3D textures. */ + assert(mt->physical_depth0 == 6); + brw_miptree_layout_texture_3d(brw, mt); + } else { + /* All other hardware stores cube maps as 2D arrays. */ + brw_miptree_layout_texture_array(brw, mt); + } + break; + + case GL_TEXTURE_3D: + brw_miptree_layout_texture_3d(brw, mt); + break; + + case GL_TEXTURE_1D_ARRAY: + case GL_TEXTURE_2D_ARRAY: + case GL_TEXTURE_2D_MULTISAMPLE_ARRAY: + case GL_TEXTURE_CUBE_MAP_ARRAY: + brw_miptree_layout_texture_array(brw, mt); + break; + default: - i945_miptree_layout_2d(intel, mt, tiling); + switch (mt->msaa_layout) { + case INTEL_MSAA_LAYOUT_UMS: + case INTEL_MSAA_LAYOUT_CMS: + brw_miptree_layout_texture_array(brw, mt); + break; + case INTEL_MSAA_LAYOUT_NONE: + case INTEL_MSAA_LAYOUT_IMS: + brw_miptree_layout_2d(mt); + break; + } break; } - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - mt->pitch, - mt->total_height, - mt->cpp, - mt->pitch * mt->total_height * mt->cpp ); - - return GL_TRUE; + DBG("%s: %dx%dx%d\n", __FUNCTION__, + mt->total_width, mt->total_height, mt->cpp); }