1 /********************************************************** 2 * Copyright 2008-2009 VMware, Inc. All rights reserved. 3 * 4 * Permission is hereby granted, free of charge, to any person 5 * obtaining a copy of this software and associated documentation 6 * files (the "Software"), to deal in the Software without 7 * restriction, including without limitation the rights to use, copy, 8 * modify, merge, publish, distribute, sublicense, and/or sell copies 9 * of the Software, and to permit persons to whom the Software is 10 * furnished to do so, subject to the following conditions: 11 * 12 * The above copyright notice and this permission notice shall be 13 * included in all copies or substantial portions of the Software. 14 * 15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 16 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 17 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 18 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS 19 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN 20 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN 21 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE 22 * SOFTWARE. 23 * 24 **********************************************************/ 25 26 #include "util/u_inlines.h" 27 #include "util/u_prim.h" 28 #include "util/u_upload_mgr.h" 29 #include "indices/u_indices.h" 30 31 #include "svga_cmd.h" 32 #include "svga_draw.h" 33 #include "svga_draw_private.h" 34 #include "svga_resource_buffer.h" 35 #include "svga_winsys.h" 36 #include "svga_context.h" 37 #include "svga_hw_reg.h" 38 39 40 /** 41 * Return a new index buffer which contains a translation of the original 42 * index buffer. An example of a translation is converting from QUAD 43 * primitives to TRIANGLE primitives. Each set of four indexes for a quad 44 * will be converted to six indices for two triangles. 45 * 46 * Before generating the new index buffer we'll check if the incoming 47 * buffer already has a translated buffer that can be re-used. 48 * This benefits demos like Cinebench R15 which has many 49 * glDrawElements(GL_QUADS) commands (we can't draw quads natively). 50 * 51 * \param offset offset in bytes to first index to translate in src buffer 52 * \param orig_prim original primitive type (like PIPE_PRIM_QUADS) 53 * \param gen_prim new/generated primitive type (like PIPE_PRIM_TRIANGLES) 54 * \param orig_nr number of indexes to translate in source buffer 55 * \param gen_nr number of indexes to write into new/dest buffer 56 * \param index_size bytes per index (2 or 4) 57 * \param translate the translation function from the u_translate module 58 * \param out_buf returns the new/translated index buffer 59 * \return error code to indicate success failure 60 */ 61 static enum pipe_error 62 translate_indices(struct svga_hwtnl *hwtnl, struct pipe_resource *src, 63 unsigned offset, 64 enum pipe_prim_type orig_prim, enum pipe_prim_type gen_prim, 65 unsigned orig_nr, unsigned gen_nr, 66 unsigned index_size, 67 u_translate_func translate, struct pipe_resource **out_buf) 68 { 69 struct pipe_context *pipe = &hwtnl->svga->pipe; 70 struct svga_screen *screen = svga_screen(pipe->screen); 71 struct svga_buffer *src_sbuf = svga_buffer(src); 72 struct pipe_transfer *src_transfer = NULL; 73 struct pipe_transfer *dst_transfer = NULL; 74 unsigned size = index_size * gen_nr; 75 const void *src_map = NULL; 76 struct pipe_resource *dst = NULL; 77 void *dst_map = NULL; 78 79 assert(index_size == 2 || index_size == 4); 80 81 if (!screen->debug.no_cache_index_buffers) { 82 /* Check if we already have a translated index buffer */ 83 if (src_sbuf->translated_indices.buffer && 84 src_sbuf->translated_indices.orig_prim == orig_prim && 85 src_sbuf->translated_indices.new_prim == gen_prim && 86 src_sbuf->translated_indices.offset == offset && 87 src_sbuf->translated_indices.count == orig_nr && 88 src_sbuf->translated_indices.index_size == index_size) { 89 pipe_resource_reference(out_buf, src_sbuf->translated_indices.buffer); 90 return PIPE_OK; 91 } 92 } 93 94 /* Need to trim vertex count to make sure we don't write too much data 95 * to the dst buffer in the translate() call. 96 */ 97 u_trim_pipe_prim(gen_prim, &gen_nr); 98 99 size = index_size * gen_nr; 100 101 dst = pipe_buffer_create(pipe->screen, 102 PIPE_BIND_INDEX_BUFFER, PIPE_USAGE_DEFAULT, size); 103 if (!dst) 104 goto fail; 105 106 src_map = pipe_buffer_map(pipe, src, PIPE_TRANSFER_READ, &src_transfer); 107 if (!src_map) 108 goto fail; 109 110 dst_map = pipe_buffer_map(pipe, dst, PIPE_TRANSFER_WRITE, &dst_transfer); 111 if (!dst_map) 112 goto fail; 113 114 translate((const char *) src_map + offset, 0, 0, gen_nr, 0, dst_map); 115 116 pipe_buffer_unmap(pipe, src_transfer); 117 pipe_buffer_unmap(pipe, dst_transfer); 118 119 *out_buf = dst; 120 121 if (!screen->debug.no_cache_index_buffers) { 122 /* Save the new, translated index buffer in the hope we can use it 123 * again in the future. 124 */ 125 pipe_resource_reference(&src_sbuf->translated_indices.buffer, dst); 126 src_sbuf->translated_indices.orig_prim = orig_prim; 127 src_sbuf->translated_indices.new_prim = gen_prim; 128 src_sbuf->translated_indices.offset = offset; 129 src_sbuf->translated_indices.count = orig_nr; 130 src_sbuf->translated_indices.index_size = index_size; 131 } 132 133 return PIPE_OK; 134 135 fail: 136 if (src_map) 137 pipe_buffer_unmap(pipe, src_transfer); 138 139 if (dst_map) 140 pipe_buffer_unmap(pipe, dst_transfer); 141 142 if (dst) 143 pipe->screen->resource_destroy(pipe->screen, dst); 144 145 return PIPE_ERROR_OUT_OF_MEMORY; 146 } 147 148 149 enum pipe_error 150 svga_hwtnl_simple_draw_range_elements(struct svga_hwtnl *hwtnl, 151 struct pipe_resource *index_buffer, 152 unsigned index_size, int index_bias, 153 unsigned min_index, unsigned max_index, 154 enum pipe_prim_type prim, unsigned start, 155 unsigned count, 156 unsigned start_instance, 157 unsigned instance_count) 158 { 159 SVGA3dPrimitiveRange range; 160 unsigned hw_prim; 161 unsigned hw_count; 162 unsigned index_offset = start * index_size; 163 164 hw_prim = svga_translate_prim(prim, count, &hw_count); 165 if (hw_count == 0) 166 return PIPE_OK; /* nothing to draw */ 167 168 range.primType = hw_prim; 169 range.primitiveCount = hw_count; 170 range.indexArray.offset = index_offset; 171 range.indexArray.stride = index_size; 172 range.indexWidth = index_size; 173 range.indexBias = index_bias; 174 175 return svga_hwtnl_prim(hwtnl, &range, count, 176 min_index, max_index, index_buffer, 177 start_instance, instance_count); 178 } 179 180 181 enum pipe_error 182 svga_hwtnl_draw_range_elements(struct svga_hwtnl *hwtnl, 183 struct pipe_resource *index_buffer, 184 unsigned index_size, int index_bias, 185 unsigned min_index, unsigned max_index, 186 enum pipe_prim_type prim, unsigned start, unsigned count, 187 unsigned start_instance, unsigned instance_count) 188 { 189 enum pipe_prim_type gen_prim; 190 unsigned gen_size, gen_nr; 191 enum indices_mode gen_type; 192 u_translate_func gen_func; 193 enum pipe_error ret = PIPE_OK; 194 195 SVGA_STATS_TIME_PUSH(svga_sws(hwtnl->svga), 196 SVGA_STATS_TIME_HWTNLDRAWELEMENTS); 197 198 if (svga_need_unfilled_fallback(hwtnl, prim)) { 199 gen_type = u_unfilled_translator(prim, 200 index_size, 201 count, 202 hwtnl->api_fillmode, 203 &gen_prim, 204 &gen_size, &gen_nr, &gen_func); 205 } 206 else { 207 gen_type = u_index_translator(svga_hw_prims, 208 prim, 209 index_size, 210 count, 211 hwtnl->api_pv, 212 hwtnl->hw_pv, 213 PR_DISABLE, 214 &gen_prim, &gen_size, &gen_nr, &gen_func); 215 } 216 217 if (gen_type == U_TRANSLATE_MEMCPY) { 218 /* No need for translation, just pass through to hardware: 219 */ 220 ret = svga_hwtnl_simple_draw_range_elements(hwtnl, index_buffer, 221 index_size, 222 index_bias, 223 min_index, 224 max_index, 225 gen_prim, start, count, 226 start_instance, 227 instance_count); 228 } 229 else { 230 struct pipe_resource *gen_buf = NULL; 231 232 /* Need to allocate a new index buffer and run the translate 233 * func to populate it. Could potentially cache this translated 234 * index buffer with the original to avoid future 235 * re-translations. Not much point if we're just accelerating 236 * GL though, as index buffers are typically used only once 237 * there. 238 */ 239 ret = translate_indices(hwtnl, 240 index_buffer, 241 start * index_size, 242 prim, gen_prim, 243 count, gen_nr, gen_size, 244 gen_func, &gen_buf); 245 if (ret != PIPE_OK) 246 goto done; 247 248 ret = svga_hwtnl_simple_draw_range_elements(hwtnl, 249 gen_buf, 250 gen_size, 251 index_bias, 252 min_index, 253 max_index, 254 gen_prim, 0, gen_nr, 255 start_instance, 256 instance_count); 257 done: 258 if (gen_buf) 259 pipe_resource_reference(&gen_buf, NULL); 260 } 261 262 SVGA_STATS_TIME_POP(svga_sws(hwtnl->svga)); 263 return ret; 264 } 265