/************************************************************************** * * Copyright 2007 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. * **************************************************************************/ /** * Render target tile caching. * * Author: * Brian Paul */ #include "util/u_inlines.h" #include "util/format/u_format.h" #include "util/u_memory.h" #include "util/u_tile.h" #include "sp_tile_cache.h" static struct softpipe_cached_tile * sp_alloc_tile(struct softpipe_tile_cache *tc); /** * Return the position in the cache for the tile that contains win pos (x,y). * We currently use a direct mapped cache so this is like a hack key. * At some point we should investige something more sophisticated, like * a LRU replacement policy. */ #define CACHE_POS(x, y, l) \ (((x) + (y) * 5 + (l) * 10) % NUM_ENTRIES) static inline int addr_to_clear_pos(union tile_address addr) { int pos; pos = addr.bits.layer * (MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE); pos += addr.bits.y * (MAX_WIDTH / TILE_SIZE); pos += addr.bits.x; return pos; } /** * Is the tile at (x,y) in cleared state? */ static inline uint is_clear_flag_set(const uint *bitvec, union tile_address addr, unsigned max) { int pos, bit; pos = addr_to_clear_pos(addr); assert(pos / 32 < max); bit = bitvec[pos / 32] & (1 << (pos & 31)); return bit; } /** * Mark the tile at (x,y) as not cleared. */ static inline void clear_clear_flag(uint *bitvec, union tile_address addr, unsigned max) { int pos; pos = addr_to_clear_pos(addr); assert(pos / 32 < max); bitvec[pos / 32] &= ~(1 << (pos & 31)); } struct softpipe_tile_cache * sp_create_tile_cache( struct pipe_context *pipe ) { struct softpipe_tile_cache *tc; uint pos; /* sanity checking: max sure MAX_WIDTH/HEIGHT >= largest texture image */ assert(MAX_WIDTH >= pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_SIZE)); STATIC_ASSERT(sizeof(union tile_address) == 4); STATIC_ASSERT((TILE_SIZE << TILE_ADDR_BITS) >= MAX_WIDTH); tc = CALLOC_STRUCT( softpipe_tile_cache ); if (tc) { tc->pipe = pipe; for (pos = 0; pos < ARRAY_SIZE(tc->tile_addrs); pos++) { tc->tile_addrs[pos].bits.invalid = 1; } tc->last_tile_addr.bits.invalid = 1; /* this allocation allows us to guarantee that allocation * failures are never fatal later */ tc->tile = MALLOC_STRUCT( softpipe_cached_tile ); if (!tc->tile) { FREE(tc); return NULL; } /* XXX this code prevents valgrind warnings about use of uninitialized * memory in programs that don't clear the surface before rendering. * However, it breaks clearing in other situations (such as in * progs/tests/drawbuffers, see bug 24402). */ #if 0 /* set flags to indicate all the tiles are cleared */ memset(tc->clear_flags, 255, sizeof(tc->clear_flags)); #endif } return tc; } void sp_destroy_tile_cache(struct softpipe_tile_cache *tc) { if (tc) { uint pos; for (pos = 0; pos < ARRAY_SIZE(tc->entries); pos++) { /*assert(tc->entries[pos].x < 0);*/ FREE( tc->entries[pos] ); } FREE( tc->tile ); if (tc->num_maps) { int i; for (i = 0; i < tc->num_maps; i++) if (tc->transfer[i]) { tc->pipe->transfer_unmap(tc->pipe, tc->transfer[i]); } FREE(tc->transfer); FREE(tc->transfer_map); FREE(tc->clear_flags); } FREE( tc ); } } /** * Specify the surface to cache. */ void sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, struct pipe_surface *ps) { struct pipe_context *pipe = tc->pipe; int i; if (tc->num_maps) { if (ps == tc->surface) return; for (i = 0; i < tc->num_maps; i++) { pipe->transfer_unmap(pipe, tc->transfer[i]); tc->transfer[i] = NULL; tc->transfer_map[i] = NULL; } FREE(tc->transfer); FREE(tc->transfer_map); tc->num_maps = 0; FREE(tc->clear_flags); tc->clear_flags_size = 0; } tc->surface = ps; if (ps) { tc->num_maps = ps->u.tex.last_layer - ps->u.tex.first_layer + 1; tc->transfer = CALLOC(tc->num_maps, sizeof(struct pipe_transfer *)); tc->transfer_map = CALLOC(tc->num_maps, sizeof(void *)); tc->clear_flags_size = (MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) * tc->num_maps / 32 * sizeof(uint); tc->clear_flags = CALLOC(1, tc->clear_flags_size); if (ps->texture->target != PIPE_BUFFER) { for (i = 0; i < tc->num_maps; i++) { tc->transfer_map[i] = pipe_transfer_map(pipe, ps->texture, ps->u.tex.level, ps->u.tex.first_layer + i, PIPE_TRANSFER_READ_WRITE | PIPE_TRANSFER_UNSYNCHRONIZED, 0, 0, ps->width, ps->height, &tc->transfer[i]); } } else { /* can't render to buffers */ assert(0); } tc->depth_stencil = util_format_is_depth_or_stencil(ps->format); } } /** * Return the transfer being cached. */ struct pipe_surface * sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) { return tc->surface; } /** * Set pixels in a tile to the given clear color/value, float. */ static void clear_tile_rgba(struct softpipe_cached_tile *tile, enum pipe_format format, const union pipe_color_union *clear_value) { if (clear_value->f[0] == 0.0 && clear_value->f[1] == 0.0 && clear_value->f[2] == 0.0 && clear_value->f[3] == 0.0) { memset(tile->data.color, 0, sizeof(tile->data.color)); } else { uint i, j; if (util_format_is_pure_uint(format)) { for (i = 0; i < TILE_SIZE; i++) { for (j = 0; j < TILE_SIZE; j++) { tile->data.colorui128[i][j][0] = clear_value->ui[0]; tile->data.colorui128[i][j][1] = clear_value->ui[1]; tile->data.colorui128[i][j][2] = clear_value->ui[2]; tile->data.colorui128[i][j][3] = clear_value->ui[3]; } } } else if (util_format_is_pure_sint(format)) { for (i = 0; i < TILE_SIZE; i++) { for (j = 0; j < TILE_SIZE; j++) { tile->data.colori128[i][j][0] = clear_value->i[0]; tile->data.colori128[i][j][1] = clear_value->i[1]; tile->data.colori128[i][j][2] = clear_value->i[2]; tile->data.colori128[i][j][3] = clear_value->i[3]; } } } else { for (i = 0; i < TILE_SIZE; i++) { for (j = 0; j < TILE_SIZE; j++) { tile->data.color[i][j][0] = clear_value->f[0]; tile->data.color[i][j][1] = clear_value->f[1]; tile->data.color[i][j][2] = clear_value->f[2]; tile->data.color[i][j][3] = clear_value->f[3]; } } } } } /** * Set a tile to a solid value/color. */ static void clear_tile(struct softpipe_cached_tile *tile, enum pipe_format format, uint64_t clear_value) { uint i, j; switch (util_format_get_blocksize(format)) { case 1: memset(tile->data.any, (int) clear_value, TILE_SIZE * TILE_SIZE); break; case 2: if (clear_value == 0) { memset(tile->data.any, 0, 2 * TILE_SIZE * TILE_SIZE); } else { for (i = 0; i < TILE_SIZE; i++) { for (j = 0; j < TILE_SIZE; j++) { tile->data.depth16[i][j] = (ushort) clear_value; } } } break; case 4: if (clear_value == 0) { memset(tile->data.any, 0, 4 * TILE_SIZE * TILE_SIZE); } else { for (i = 0; i < TILE_SIZE; i++) { for (j = 0; j < TILE_SIZE; j++) { tile->data.depth32[i][j] = (uint) clear_value; } } } break; case 8: if (clear_value == 0) { memset(tile->data.any, 0, 8 * TILE_SIZE * TILE_SIZE); } else { for (i = 0; i < TILE_SIZE; i++) { for (j = 0; j < TILE_SIZE; j++) { tile->data.depth64[i][j] = clear_value; } } } break; default: assert(0); } } /** * Actually clear the tiles which were flagged as being in a clear state. */ static void sp_tile_cache_flush_clear(struct softpipe_tile_cache *tc, int layer) { struct pipe_transfer *pt = tc->transfer[layer]; const uint w = tc->transfer[layer]->box.width; const uint h = tc->transfer[layer]->box.height; uint x, y; uint numCleared = 0; assert(pt->resource); /* clear the scratch tile to the clear value */ if (tc->depth_stencil) { clear_tile(tc->tile, pt->resource->format, tc->clear_val); } else { clear_tile_rgba(tc->tile, pt->resource->format, &tc->clear_color); } /* push the tile to all positions marked as clear */ for (y = 0; y < h; y += TILE_SIZE) { for (x = 0; x < w; x += TILE_SIZE) { union tile_address addr = tile_address(x, y, layer); if (is_clear_flag_set(tc->clear_flags, addr, tc->clear_flags_size)) { /* write the scratch tile to the surface */ if (tc->depth_stencil) { pipe_put_tile_raw(pt, tc->transfer_map[layer], x, y, TILE_SIZE, TILE_SIZE, tc->tile->data.any, 0/*STRIDE*/); } else { pipe_put_tile_rgba(pt, tc->transfer_map[layer], x, y, TILE_SIZE, TILE_SIZE, tc->surface->format, tc->tile->data.color); } numCleared++; } } } #if 0 debug_printf("num cleared: %u\n", numCleared); #endif } static void sp_flush_tile(struct softpipe_tile_cache* tc, unsigned pos) { int layer = tc->tile_addrs[pos].bits.layer; if (!tc->tile_addrs[pos].bits.invalid) { if (tc->depth_stencil) { pipe_put_tile_raw(tc->transfer[layer], tc->transfer_map[layer], tc->tile_addrs[pos].bits.x * TILE_SIZE, tc->tile_addrs[pos].bits.y * TILE_SIZE, TILE_SIZE, TILE_SIZE, tc->entries[pos]->data.depth32, 0/*STRIDE*/); } else { pipe_put_tile_rgba(tc->transfer[layer], tc->transfer_map[layer], tc->tile_addrs[pos].bits.x * TILE_SIZE, tc->tile_addrs[pos].bits.y * TILE_SIZE, TILE_SIZE, TILE_SIZE, tc->surface->format, tc->entries[pos]->data.color); } tc->tile_addrs[pos].bits.invalid = 1; /* mark as empty */ } } /** * Flush the tile cache: write all dirty tiles back to the transfer. * any tiles "flagged" as cleared will be "really" cleared. */ void sp_flush_tile_cache(struct softpipe_tile_cache *tc) { int inuse = 0, pos; int i; if (tc->num_maps) { /* caching a drawing transfer */ for (pos = 0; pos < ARRAY_SIZE(tc->entries); pos++) { struct softpipe_cached_tile *tile = tc->entries[pos]; if (!tile) { assert(tc->tile_addrs[pos].bits.invalid); continue; } sp_flush_tile(tc, pos); ++inuse; } if (!tc->tile) tc->tile = sp_alloc_tile(tc); for (i = 0; i < tc->num_maps; i++) sp_tile_cache_flush_clear(tc, i); /* reset all clear flags to zero */ memset(tc->clear_flags, 0, tc->clear_flags_size); tc->last_tile_addr.bits.invalid = 1; } #if 0 debug_printf("flushed tiles in use: %d\n", inuse); #endif } static struct softpipe_cached_tile * sp_alloc_tile(struct softpipe_tile_cache *tc) { struct softpipe_cached_tile * tile = MALLOC_STRUCT(softpipe_cached_tile); if (!tile) { /* in this case, steal an existing tile */ if (!tc->tile) { unsigned pos; for (pos = 0; pos < ARRAY_SIZE(tc->entries); ++pos) { if (!tc->entries[pos]) continue; sp_flush_tile(tc, pos); tc->tile = tc->entries[pos]; tc->entries[pos] = NULL; break; } /* this should never happen */ if (!tc->tile) abort(); } tile = tc->tile; tc->tile = NULL; tc->last_tile_addr.bits.invalid = 1; } return tile; } /** * Get a tile from the cache. * \param x, y position of tile, in pixels */ struct softpipe_cached_tile * sp_find_cached_tile(struct softpipe_tile_cache *tc, union tile_address addr ) { struct pipe_transfer *pt; /* cache pos/entry: */ const int pos = CACHE_POS(addr.bits.x, addr.bits.y, addr.bits.layer); struct softpipe_cached_tile *tile = tc->entries[pos]; int layer; if (!tile) { tile = sp_alloc_tile(tc); tc->entries[pos] = tile; } if (addr.value != tc->tile_addrs[pos].value) { layer = tc->tile_addrs[pos].bits.layer; if (tc->tile_addrs[pos].bits.invalid == 0) { /* put dirty tile back in framebuffer */ if (tc->depth_stencil) { pipe_put_tile_raw(tc->transfer[layer], tc->transfer_map[layer], tc->tile_addrs[pos].bits.x * TILE_SIZE, tc->tile_addrs[pos].bits.y * TILE_SIZE, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } else { pipe_put_tile_rgba(tc->transfer[layer], tc->transfer_map[layer], tc->tile_addrs[pos].bits.x * TILE_SIZE, tc->tile_addrs[pos].bits.y * TILE_SIZE, TILE_SIZE, TILE_SIZE, tc->surface->format, tile->data.color); } } tc->tile_addrs[pos] = addr; layer = tc->tile_addrs[pos].bits.layer; pt = tc->transfer[layer]; assert(pt->resource); if (is_clear_flag_set(tc->clear_flags, addr, tc->clear_flags_size)) { /* don't get tile from framebuffer, just clear it */ if (tc->depth_stencil) { clear_tile(tile, pt->resource->format, tc->clear_val); } else { clear_tile_rgba(tile, pt->resource->format, &tc->clear_color); } clear_clear_flag(tc->clear_flags, addr, tc->clear_flags_size); } else { /* get new tile data from transfer */ if (tc->depth_stencil) { pipe_get_tile_raw(tc->transfer[layer], tc->transfer_map[layer], tc->tile_addrs[pos].bits.x * TILE_SIZE, tc->tile_addrs[pos].bits.y * TILE_SIZE, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } else { pipe_get_tile_rgba(tc->transfer[layer], tc->transfer_map[layer], tc->tile_addrs[pos].bits.x * TILE_SIZE, tc->tile_addrs[pos].bits.y * TILE_SIZE, TILE_SIZE, TILE_SIZE, tc->surface->format, tile->data.color); } } } tc->last_tile = tile; tc->last_tile_addr = addr; return tile; } /** * When a whole surface is being cleared to a value we can avoid * fetching tiles above. * Save the color and set a 'clearflag' for each tile of the screen. */ void sp_tile_cache_clear(struct softpipe_tile_cache *tc, const union pipe_color_union *color, uint64_t clearValue) { uint pos; tc->clear_color = *color; tc->clear_val = clearValue; /* set flags to indicate all the tiles are cleared */ memset(tc->clear_flags, 255, tc->clear_flags_size); for (pos = 0; pos < ARRAY_SIZE(tc->tile_addrs); pos++) { tc->tile_addrs[pos].bits.invalid = 1; } tc->last_tile_addr.bits.invalid = 1; }