Home | History | Annotate | Download | only in a5xx
      1 /*
      2  * Copyright (C) 2018 Rob Clark <robclark (at) freedesktop.org>
      3  *
      4  * Permission is hereby granted, free of charge, to any person obtaining a
      5  * copy of this software and associated documentation files (the "Software"),
      6  * to deal in the Software without restriction, including without limitation
      7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
      8  * and/or sell copies of the Software, and to permit persons to whom the
      9  * Software is furnished to do so, subject to the following conditions:
     10  *
     11  * The above copyright notice and this permission notice (including the next
     12  * paragraph) shall be included in all copies or substantial portions of the
     13  * Software.
     14  *
     15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
     16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
     17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
     18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
     19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
     20  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
     21  * SOFTWARE.
     22  *
     23  * Authors:
     24  *    Rob Clark <robclark (at) freedesktop.org>
     25  */
     26 
     27 #include "fd5_resource.h"
     28 
     29 /* indexed by cpp: */
     30 static const struct {
     31 	unsigned pitchalign;
     32 	unsigned heightalign;
     33 } tile_alignment[] = {
     34 	[1]  = { 128, 32 },
     35 	[2]  = { 128, 16 },
     36 	[3]  = { 128, 16 },
     37 	[4]  = {  64, 16 },
     38 	[8]  = {  64, 16 },
     39 	[12] = {  64, 16 },
     40 	[16] = {  64, 16 },
     41 };
     42 
     43 /* NOTE: good way to test this is:  (for example)
     44  *  piglit/bin/texelFetch fs sampler2D 100x100x1-100x300x1
     45  */
     46 static uint32_t
     47 setup_slices(struct fd_resource *rsc, uint32_t alignment, enum pipe_format format)
     48 {
     49 	struct pipe_resource *prsc = &rsc->base;
     50 	struct fd_screen *screen = fd_screen(prsc->screen);
     51 	enum util_format_layout layout = util_format_description(format)->layout;
     52 	uint32_t pitchalign = screen->gmem_alignw;
     53 	uint32_t heightalign;
     54 	uint32_t level, size = 0;
     55 	uint32_t width = prsc->width0;
     56 	uint32_t height = prsc->height0;
     57 	uint32_t depth = prsc->depth0;
     58 	/* in layer_first layout, the level (slice) contains just one
     59 	 * layer (since in fact the layer contains the slices)
     60 	 */
     61 	uint32_t layers_in_level = rsc->layer_first ? 1 : prsc->array_size;
     62 
     63 	heightalign = tile_alignment[rsc->cpp].heightalign;
     64 
     65 	for (level = 0; level <= prsc->last_level; level++) {
     66 		struct fd_resource_slice *slice = fd_resource_slice(rsc, level);
     67 		bool linear_level = fd_resource_level_linear(prsc, level);
     68 		uint32_t aligned_height = height;
     69 		uint32_t blocks;
     70 
     71 		if (rsc->tile_mode && !linear_level) {
     72 			pitchalign = tile_alignment[rsc->cpp].pitchalign;
     73 			aligned_height = align(aligned_height, heightalign);
     74 		} else {
     75 			pitchalign = 64;
     76 
     77 			/* The blits used for mem<->gmem work at a granularity of
     78 			 * 32x32, which can cause faults due to over-fetch on the
     79 			 * last level.  The simple solution is to over-allocate a
     80 			 * bit the last level to ensure any over-fetch is harmless.
     81 			 * The pitch is already sufficiently aligned, but height
     82 			 * may not be:
     83 			 */
     84 			if ((level == prsc->last_level) && (prsc->target != PIPE_BUFFER))
     85 				aligned_height = align(aligned_height, 32);
     86 		}
     87 
     88 		if (layout == UTIL_FORMAT_LAYOUT_ASTC)
     89 			slice->pitch =
     90 				util_align_npot(width, pitchalign * util_format_get_blockwidth(format));
     91 		else
     92 			slice->pitch = align(width, pitchalign);
     93 
     94 		slice->offset = size;
     95 		blocks = util_format_get_nblocks(format, slice->pitch, aligned_height);
     96 
     97 		/* 1d array and 2d array textures must all have the same layer size
     98 		 * for each miplevel on a3xx. 3d textures can have different layer
     99 		 * sizes for high levels, but the hw auto-sizer is buggy (or at least
    100 		 * different than what this code does), so as soon as the layer size
    101 		 * range gets into range, we stop reducing it.
    102 		 */
    103 		if (prsc->target == PIPE_TEXTURE_3D && (
    104 					level == 1 ||
    105 					(level > 1 && rsc->slices[level - 1].size0 > 0xf000)))
    106 			slice->size0 = align(blocks * rsc->cpp, alignment);
    107 		else if (level == 0 || rsc->layer_first || alignment == 1)
    108 			slice->size0 = align(blocks * rsc->cpp, alignment);
    109 		else
    110 			slice->size0 = rsc->slices[level - 1].size0;
    111 
    112 #if 0
    113 		debug_printf("%s: %ux%ux%u@%u: %2u: stride=%4u, size=%7u, aligned_height=%3u\n",
    114 				util_format_name(prsc->format),
    115 				prsc->width0, prsc->height0, prsc->depth0, rsc->cpp,
    116 				level, slice->pitch * rsc->cpp,
    117 				slice->size0 * depth * layers_in_level,
    118 				aligned_height);
    119 #endif
    120 
    121 		size += slice->size0 * depth * layers_in_level;
    122 
    123 		width = u_minify(width, 1);
    124 		height = u_minify(height, 1);
    125 		depth = u_minify(depth, 1);
    126 	}
    127 
    128 	return size;
    129 }
    130 
    131 uint32_t
    132 fd5_setup_slices(struct fd_resource *rsc)
    133 {
    134 	uint32_t alignment;
    135 
    136 	switch (rsc->base.target) {
    137 	case PIPE_TEXTURE_3D:
    138 		rsc->layer_first = false;
    139 		alignment = 4096;
    140 		break;
    141 	default:
    142 		rsc->layer_first = true;
    143 		alignment = 1;
    144 		break;
    145 	}
    146 
    147 	return setup_slices(rsc, alignment, rsc->base.format);
    148 }
    149