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 "pipe/p_defines.h"
27 #include "util/u_bitmask.h"
28 #include "util/format/u_format.h"
29 #include "util/u_helpers.h"
30 #include "util/u_inlines.h"
31 #include "util/u_math.h"
32 #include "util/u_memory.h"
33 #include "util/u_transfer.h"
34 #include "tgsi/tgsi_parse.h"
35
36 #include "svga_context.h"
37 #include "svga_cmd.h"
38 #include "svga_format.h"
39 #include "svga_resource_buffer.h"
40 #include "svga_screen.h"
41
42
43 static void
svga_set_vertex_buffers(struct pipe_context * pipe,unsigned start_slot,unsigned count,unsigned unbind_num_trailing_slots,bool take_ownership,const struct pipe_vertex_buffer * buffers)44 svga_set_vertex_buffers(struct pipe_context *pipe,
45 unsigned start_slot, unsigned count,
46 unsigned unbind_num_trailing_slots,
47 bool take_ownership,
48 const struct pipe_vertex_buffer *buffers)
49 {
50 struct svga_context *svga = svga_context(pipe);
51
52 util_set_vertex_buffers_count(svga->curr.vb,
53 &svga->curr.num_vertex_buffers,
54 buffers, start_slot, count,
55 unbind_num_trailing_slots,
56 take_ownership);
57
58 svga->dirty |= SVGA_NEW_VBUFFER;
59 }
60
61
62 /**
63 * Does the given vertex attrib format need range adjustment in the VS?
64 * Range adjustment scales and biases values from [0,1] to [-1,1].
65 * This lets us avoid the swtnl path.
66 */
67 static boolean
attrib_needs_range_adjustment(enum pipe_format format)68 attrib_needs_range_adjustment(enum pipe_format format)
69 {
70 switch (format) {
71 case PIPE_FORMAT_R8G8B8_SNORM:
72 return TRUE;
73 default:
74 return FALSE;
75 }
76 }
77
78
79 /**
80 * Given a gallium vertex element format, return the corresponding
81 * SVGA3dDeclType.
82 */
83 static SVGA3dDeclType
translate_vertex_format_to_decltype(enum pipe_format format)84 translate_vertex_format_to_decltype(enum pipe_format format)
85 {
86 switch (format) {
87 case PIPE_FORMAT_R32_FLOAT: return SVGA3D_DECLTYPE_FLOAT1;
88 case PIPE_FORMAT_R32G32_FLOAT: return SVGA3D_DECLTYPE_FLOAT2;
89 case PIPE_FORMAT_R32G32B32_FLOAT: return SVGA3D_DECLTYPE_FLOAT3;
90 case PIPE_FORMAT_R32G32B32A32_FLOAT: return SVGA3D_DECLTYPE_FLOAT4;
91 case PIPE_FORMAT_B8G8R8A8_UNORM: return SVGA3D_DECLTYPE_D3DCOLOR;
92 case PIPE_FORMAT_R8G8B8A8_USCALED: return SVGA3D_DECLTYPE_UBYTE4;
93 case PIPE_FORMAT_R16G16_SSCALED: return SVGA3D_DECLTYPE_SHORT2;
94 case PIPE_FORMAT_R16G16B16A16_SSCALED: return SVGA3D_DECLTYPE_SHORT4;
95 case PIPE_FORMAT_R8G8B8A8_UNORM: return SVGA3D_DECLTYPE_UBYTE4N;
96 case PIPE_FORMAT_R16G16_SNORM: return SVGA3D_DECLTYPE_SHORT2N;
97 case PIPE_FORMAT_R16G16B16A16_SNORM: return SVGA3D_DECLTYPE_SHORT4N;
98 case PIPE_FORMAT_R16G16_UNORM: return SVGA3D_DECLTYPE_USHORT2N;
99 case PIPE_FORMAT_R16G16B16A16_UNORM: return SVGA3D_DECLTYPE_USHORT4N;
100 case PIPE_FORMAT_R10G10B10X2_USCALED: return SVGA3D_DECLTYPE_UDEC3;
101 case PIPE_FORMAT_R10G10B10X2_SNORM: return SVGA3D_DECLTYPE_DEC3N;
102 case PIPE_FORMAT_R16G16_FLOAT: return SVGA3D_DECLTYPE_FLOAT16_2;
103 case PIPE_FORMAT_R16G16B16A16_FLOAT: return SVGA3D_DECLTYPE_FLOAT16_4;
104
105 /* See attrib_needs_adjustment() and attrib_needs_w_to_1() above */
106 case PIPE_FORMAT_R8G8B8_SNORM: return SVGA3D_DECLTYPE_UBYTE4N;
107
108 /* See attrib_needs_w_to_1() above */
109 case PIPE_FORMAT_R16G16B16_SNORM: return SVGA3D_DECLTYPE_SHORT4N;
110 case PIPE_FORMAT_R16G16B16_UNORM: return SVGA3D_DECLTYPE_USHORT4N;
111 case PIPE_FORMAT_R8G8B8_UNORM: return SVGA3D_DECLTYPE_UBYTE4N;
112
113 default:
114 /* There are many formats without hardware support. This case
115 * will be hit regularly, meaning we'll need swvfetch.
116 */
117 return SVGA3D_DECLTYPE_MAX;
118 }
119 }
120
121
122 static void
define_input_element_object(struct svga_context * svga,struct svga_velems_state * velems)123 define_input_element_object(struct svga_context *svga,
124 struct svga_velems_state *velems)
125 {
126 SVGA3dInputElementDesc elements[PIPE_MAX_ATTRIBS];
127 unsigned i;
128
129 assert(velems->count <= PIPE_MAX_ATTRIBS);
130 assert(svga_have_vgpu10(svga));
131
132 for (i = 0; i < velems->count; i++) {
133 const struct pipe_vertex_element *elem = velems->velem + i;
134 SVGA3dSurfaceFormat svga_format;
135 unsigned vf_flags;
136
137 svga_translate_vertex_format_vgpu10(elem->src_format,
138 &svga_format, &vf_flags);
139
140 velems->decl_type[i] =
141 translate_vertex_format_to_decltype(elem->src_format);
142 elements[i].inputSlot = elem->vertex_buffer_index;
143 elements[i].alignedByteOffset = elem->src_offset;
144 elements[i].format = svga_format;
145
146 if (elem->instance_divisor) {
147 elements[i].inputSlotClass = SVGA3D_INPUT_PER_INSTANCE_DATA;
148 elements[i].instanceDataStepRate = elem->instance_divisor;
149 }
150 else {
151 elements[i].inputSlotClass = SVGA3D_INPUT_PER_VERTEX_DATA;
152 elements[i].instanceDataStepRate = 0;
153 }
154 elements[i].inputRegister = i;
155
156 if (elements[i].format == SVGA3D_FORMAT_INVALID) {
157 velems->need_swvfetch = TRUE;
158 }
159
160 if (util_format_is_pure_integer(elem->src_format)) {
161 velems->attrib_is_pure_int |= (1 << i);
162 }
163
164 if (vf_flags & VF_W_TO_1) {
165 velems->adjust_attrib_w_1 |= (1 << i);
166 }
167
168 if (vf_flags & VF_U_TO_F_CAST) {
169 velems->adjust_attrib_utof |= (1 << i);
170 }
171 else if (vf_flags & VF_I_TO_F_CAST) {
172 velems->adjust_attrib_itof |= (1 << i);
173 }
174
175 if (vf_flags & VF_BGRA) {
176 velems->attrib_is_bgra |= (1 << i);
177 }
178
179 if (vf_flags & VF_PUINT_TO_SNORM) {
180 velems->attrib_puint_to_snorm |= (1 << i);
181 }
182 else if (vf_flags & VF_PUINT_TO_USCALED) {
183 velems->attrib_puint_to_uscaled |= (1 << i);
184 }
185 else if (vf_flags & VF_PUINT_TO_SSCALED) {
186 velems->attrib_puint_to_sscaled |= (1 << i);
187 }
188 }
189
190 velems->id = util_bitmask_add(svga->input_element_object_id_bm);
191
192 SVGA_RETRY(svga, SVGA3D_vgpu10_DefineElementLayout(svga->swc, velems->count,
193 velems->id, elements));
194 }
195
196
197 /**
198 * Translate the vertex element types to SVGA3dDeclType and check
199 * for VS-based vertex attribute adjustments.
200 */
201 static void
translate_vertex_decls(struct svga_context * svga,struct svga_velems_state * velems)202 translate_vertex_decls(struct svga_context *svga,
203 struct svga_velems_state *velems)
204 {
205 unsigned i;
206
207 assert(!svga_have_vgpu10(svga));
208
209 for (i = 0; i < velems->count; i++) {
210 const enum pipe_format f = velems->velem[i].src_format;
211 SVGA3dSurfaceFormat svga_format;
212 unsigned vf_flags;
213
214 svga_translate_vertex_format_vgpu10(f, &svga_format, &vf_flags);
215
216 velems->decl_type[i] = translate_vertex_format_to_decltype(f);
217 if (velems->decl_type[i] == SVGA3D_DECLTYPE_MAX) {
218 /* Unsupported format - use software fetch */
219 velems->need_swvfetch = TRUE;
220 }
221
222 /* Check for VS-based adjustments */
223 if (attrib_needs_range_adjustment(f)) {
224 velems->adjust_attrib_range |= (1 << i);
225 }
226
227 if (vf_flags & VF_W_TO_1) {
228 velems->adjust_attrib_w_1 |= (1 << i);
229 }
230 }
231 }
232
233
234 static void *
svga_create_vertex_elements_state(struct pipe_context * pipe,unsigned count,const struct pipe_vertex_element * attribs)235 svga_create_vertex_elements_state(struct pipe_context *pipe,
236 unsigned count,
237 const struct pipe_vertex_element *attribs)
238 {
239 struct svga_context *svga = svga_context(pipe);
240 struct svga_velems_state *velems;
241
242 assert(count <= PIPE_MAX_ATTRIBS);
243 velems = (struct svga_velems_state *) MALLOC(sizeof(struct svga_velems_state));
244 if (velems) {
245 velems->count = count;
246 memcpy(velems->velem, attribs, sizeof(*attribs) * count);
247
248 velems->need_swvfetch = FALSE;
249 velems->adjust_attrib_range = 0x0;
250 velems->attrib_is_pure_int = 0x0;
251 velems->adjust_attrib_w_1 = 0x0;
252 velems->adjust_attrib_itof = 0x0;
253 velems->adjust_attrib_utof = 0x0;
254 velems->attrib_is_bgra = 0x0;
255 velems->attrib_puint_to_snorm = 0x0;
256 velems->attrib_puint_to_uscaled = 0x0;
257 velems->attrib_puint_to_sscaled = 0x0;
258
259 if (svga_have_vgpu10(svga)) {
260 define_input_element_object(svga, velems);
261 }
262 else {
263 translate_vertex_decls(svga, velems);
264 }
265 }
266
267 svga->hud.num_vertexelement_objects++;
268 SVGA_STATS_COUNT_INC(svga_screen(svga->pipe.screen)->sws,
269 SVGA_STATS_COUNT_VERTEXELEMENT);
270
271 return velems;
272 }
273
274
275 static void
svga_bind_vertex_elements_state(struct pipe_context * pipe,void * state)276 svga_bind_vertex_elements_state(struct pipe_context *pipe, void *state)
277 {
278 struct svga_context *svga = svga_context(pipe);
279 struct svga_velems_state *velems = (struct svga_velems_state *) state;
280
281 svga->curr.velems = velems;
282 svga->dirty |= SVGA_NEW_VELEMENT;
283 }
284
285
286 static void
svga_delete_vertex_elements_state(struct pipe_context * pipe,void * state)287 svga_delete_vertex_elements_state(struct pipe_context *pipe, void *state)
288 {
289 struct svga_context *svga = svga_context(pipe);
290 struct svga_velems_state *velems = (struct svga_velems_state *) state;
291
292 if (svga_have_vgpu10(svga)) {
293 svga_hwtnl_flush_retry(svga);
294
295 SVGA_RETRY(svga, SVGA3D_vgpu10_DestroyElementLayout(svga->swc,
296 velems->id));
297
298 if (velems->id == svga->state.hw_draw.layout_id)
299 svga->state.hw_draw.layout_id = SVGA3D_INVALID_ID;
300
301 util_bitmask_clear(svga->input_element_object_id_bm, velems->id);
302 velems->id = SVGA3D_INVALID_ID;
303 }
304
305 FREE(velems);
306 svga->hud.num_vertexelement_objects--;
307 }
308
309
310 void
svga_cleanup_vertex_state(struct svga_context * svga)311 svga_cleanup_vertex_state(struct svga_context *svga)
312 {
313 unsigned i;
314
315 for (i = 0 ; i < svga->curr.num_vertex_buffers; i++)
316 pipe_vertex_buffer_unreference(&svga->curr.vb[i]);
317
318 pipe_resource_reference(&svga->state.hw_draw.ib, NULL);
319
320 for (i = 0; i < svga->state.hw_draw.num_vbuffers; i++)
321 pipe_resource_reference(&svga->state.hw_draw.vbuffers[i], NULL);
322 }
323
324
325 void
svga_init_vertex_functions(struct svga_context * svga)326 svga_init_vertex_functions(struct svga_context *svga)
327 {
328 svga->pipe.set_vertex_buffers = svga_set_vertex_buffers;
329 svga->pipe.create_vertex_elements_state = svga_create_vertex_elements_state;
330 svga->pipe.bind_vertex_elements_state = svga_bind_vertex_elements_state;
331 svga->pipe.delete_vertex_elements_state = svga_delete_vertex_elements_state;
332 }
333