Merge branch 'gallium-nopointsizeminmax'
[mesa.git] / src / gallium / drivers / svga / svga_state_rss.c
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 "pipe/p_defines.h"
28 #include "util/u_math.h"
29
30 #include "svga_context.h"
31 #include "svga_state.h"
32 #include "svga_cmd.h"
33
34
35 struct rs_queue {
36 unsigned rs_count;
37 SVGA3dRenderState rs[SVGA3D_RS_MAX];
38 };
39
40
41 #define EMIT_RS(svga, value, token, fail) \
42 do { \
43 if (svga->state.hw_draw.rs[SVGA3D_RS_##token] != value) { \
44 svga_queue_rs( &queue, SVGA3D_RS_##token, value ); \
45 svga->state.hw_draw.rs[SVGA3D_RS_##token] = value; \
46 } \
47 } while (0)
48
49 #define EMIT_RS_FLOAT(svga, fvalue, token, fail) \
50 do { \
51 unsigned value = fui(fvalue); \
52 if (svga->state.hw_draw.rs[SVGA3D_RS_##token] != value) { \
53 svga_queue_rs( &queue, SVGA3D_RS_##token, value ); \
54 svga->state.hw_draw.rs[SVGA3D_RS_##token] = value; \
55 } \
56 } while (0)
57
58
59 static INLINE void
60 svga_queue_rs( struct rs_queue *q,
61 unsigned rss,
62 unsigned value )
63 {
64 q->rs[q->rs_count].state = rss;
65 q->rs[q->rs_count].uintValue = value;
66 q->rs_count++;
67 }
68
69
70 /* Compare old and new render states and emit differences between them
71 * to hardware. Simplest implementation would be to emit the whole of
72 * the "to" state.
73 */
74 static int emit_rss( struct svga_context *svga,
75 unsigned dirty )
76 {
77 struct rs_queue queue;
78
79 queue.rs_count = 0;
80
81 if (dirty & SVGA_NEW_BLEND) {
82 const struct svga_blend_state *curr = svga->curr.blend;
83
84 EMIT_RS( svga, curr->rt[0].writemask, COLORWRITEENABLE, fail );
85 EMIT_RS( svga, curr->rt[0].blend_enable, BLENDENABLE, fail );
86
87 if (curr->rt[0].blend_enable) {
88 EMIT_RS( svga, curr->rt[0].srcblend, SRCBLEND, fail );
89 EMIT_RS( svga, curr->rt[0].dstblend, DSTBLEND, fail );
90 EMIT_RS( svga, curr->rt[0].blendeq, BLENDEQUATION, fail );
91
92 EMIT_RS( svga, curr->rt[0].separate_alpha_blend_enable,
93 SEPARATEALPHABLENDENABLE, fail );
94
95 if (curr->rt[0].separate_alpha_blend_enable) {
96 EMIT_RS( svga, curr->rt[0].srcblend_alpha, SRCBLENDALPHA, fail );
97 EMIT_RS( svga, curr->rt[0].dstblend_alpha, DSTBLENDALPHA, fail );
98 EMIT_RS( svga, curr->rt[0].blendeq_alpha, BLENDEQUATIONALPHA, fail );
99 }
100 }
101 }
102
103
104 if (dirty & (SVGA_NEW_DEPTH_STENCIL | SVGA_NEW_RAST)) {
105 const struct svga_depth_stencil_state *curr = svga->curr.depth;
106 const struct svga_rasterizer_state *rast = svga->curr.rast;
107
108 if (!curr->stencil[0].enabled)
109 {
110 /* Stencil disabled
111 */
112 EMIT_RS( svga, FALSE, STENCILENABLE, fail );
113 EMIT_RS( svga, FALSE, STENCILENABLE2SIDED, fail );
114 }
115 else if (curr->stencil[0].enabled && !curr->stencil[1].enabled)
116 {
117 /* Regular stencil
118 */
119 EMIT_RS( svga, TRUE, STENCILENABLE, fail );
120 EMIT_RS( svga, FALSE, STENCILENABLE2SIDED, fail );
121
122 EMIT_RS( svga, curr->stencil[0].func, STENCILFUNC, fail );
123 EMIT_RS( svga, curr->stencil[0].fail, STENCILFAIL, fail );
124 EMIT_RS( svga, curr->stencil[0].zfail, STENCILZFAIL, fail );
125 EMIT_RS( svga, curr->stencil[0].pass, STENCILPASS, fail );
126
127 EMIT_RS( svga, curr->stencil_ref, STENCILREF, fail );
128 EMIT_RS( svga, curr->stencil_mask, STENCILMASK, fail );
129 EMIT_RS( svga, curr->stencil_writemask, STENCILWRITEMASK, fail );
130 }
131 else
132 {
133 int cw, ccw;
134
135 /* Hardware frontwinding is always CW, so if ours is also CW,
136 * then our definition of front face agrees with hardware.
137 * Otherwise need to flip.
138 */
139 if (rast->templ.front_winding == PIPE_WINDING_CW) {
140 cw = 0;
141 ccw = 1;
142 }
143 else {
144 cw = 1;
145 ccw = 0;
146 }
147
148 /* Twoside stencil
149 */
150 EMIT_RS( svga, TRUE, STENCILENABLE, fail );
151 EMIT_RS( svga, TRUE, STENCILENABLE2SIDED, fail );
152
153 EMIT_RS( svga, curr->stencil[cw].func, STENCILFUNC, fail );
154 EMIT_RS( svga, curr->stencil[cw].fail, STENCILFAIL, fail );
155 EMIT_RS( svga, curr->stencil[cw].zfail, STENCILZFAIL, fail );
156 EMIT_RS( svga, curr->stencil[cw].pass, STENCILPASS, fail );
157
158 EMIT_RS( svga, curr->stencil[ccw].func, CCWSTENCILFUNC, fail );
159 EMIT_RS( svga, curr->stencil[ccw].fail, CCWSTENCILFAIL, fail );
160 EMIT_RS( svga, curr->stencil[ccw].zfail, CCWSTENCILZFAIL, fail );
161 EMIT_RS( svga, curr->stencil[ccw].pass, CCWSTENCILPASS, fail );
162
163 EMIT_RS( svga, curr->stencil_ref, STENCILREF, fail );
164 EMIT_RS( svga, curr->stencil_mask, STENCILMASK, fail );
165 EMIT_RS( svga, curr->stencil_writemask, STENCILWRITEMASK, fail );
166 }
167
168 EMIT_RS( svga, curr->zenable, ZENABLE, fail );
169 if (curr->zenable) {
170 EMIT_RS( svga, curr->zfunc, ZFUNC, fail );
171 EMIT_RS( svga, curr->zwriteenable, ZWRITEENABLE, fail );
172 }
173
174 EMIT_RS( svga, curr->alphatestenable, ALPHATESTENABLE, fail );
175 if (curr->alphatestenable) {
176 EMIT_RS( svga, curr->alphafunc, ALPHAFUNC, fail );
177 EMIT_RS_FLOAT( svga, curr->alpharef, ALPHAREF, fail );
178 }
179 }
180
181
182 if (dirty & SVGA_NEW_RAST)
183 {
184 const struct svga_rasterizer_state *curr = svga->curr.rast;
185
186 /* Shademode: still need to rearrange index list to move
187 * flat-shading PV first vertex.
188 */
189 EMIT_RS( svga, curr->shademode, SHADEMODE, fail );
190 EMIT_RS( svga, curr->cullmode, CULLMODE, fail );
191 EMIT_RS( svga, curr->scissortestenable, SCISSORTESTENABLE, fail );
192 EMIT_RS( svga, curr->multisampleantialias, MULTISAMPLEANTIALIAS, fail );
193 EMIT_RS( svga, curr->lastpixel, LASTPIXEL, fail );
194 EMIT_RS( svga, curr->linepattern, LINEPATTERN, fail );
195 EMIT_RS_FLOAT( svga, curr->pointsize, POINTSIZE, fail );
196 /* XXX still need to set this? */
197 EMIT_RS_FLOAT( svga, 0.0, POINTSIZEMIN, fail );
198 EMIT_RS_FLOAT( svga, SVGA_MAX_POINTSIZE, POINTSIZEMAX, fail );
199 }
200
201 if (dirty & (SVGA_NEW_RAST | SVGA_NEW_FRAME_BUFFER | SVGA_NEW_NEED_PIPELINE))
202 {
203 const struct svga_rasterizer_state *curr = svga->curr.rast;
204 float slope = 0.0;
205 float bias = 0.0;
206
207 /* Need to modify depth bias according to bound depthbuffer
208 * format. Don't do hardware depthbias while the software
209 * pipeline is active.
210 */
211 if (!svga->state.sw.need_pipeline &&
212 svga->curr.framebuffer.zsbuf)
213 {
214 slope = curr->slopescaledepthbias;
215 bias = svga->curr.depthscale * curr->depthbias;
216 }
217
218 EMIT_RS_FLOAT( svga, slope, SLOPESCALEDEPTHBIAS, fail );
219 EMIT_RS_FLOAT( svga, bias, DEPTHBIAS, fail );
220 }
221
222
223 if (queue.rs_count) {
224 SVGA3dRenderState *rs;
225
226 if (SVGA3D_BeginSetRenderState( svga->swc,
227 &rs,
228 queue.rs_count ) != PIPE_OK)
229 goto fail;
230
231 memcpy( rs,
232 queue.rs,
233 queue.rs_count * sizeof queue.rs[0]);
234
235 SVGA_FIFOCommitAll( svga->swc );
236 }
237
238 /* Also blend color:
239 */
240
241 return 0;
242
243 fail:
244 /* XXX: need to poison cached hardware state on failure to ensure
245 * dirty state gets re-emitted. Fix this by re-instating partial
246 * FIFOCommit command and only updating cached hw state once the
247 * initial allocation has succeeded.
248 */
249 memset(svga->state.hw_draw.rs, 0xcd, sizeof(svga->state.hw_draw.rs));
250
251 return PIPE_ERROR_OUT_OF_MEMORY;
252 }
253
254
255 struct svga_tracked_state svga_hw_rss =
256 {
257 "hw rss state",
258
259 (SVGA_NEW_BLEND |
260 SVGA_NEW_DEPTH_STENCIL |
261 SVGA_NEW_RAST |
262 SVGA_NEW_FRAME_BUFFER |
263 SVGA_NEW_NEED_PIPELINE),
264
265 emit_rss
266 };