svga: update dumping code with new GBS commands, etc
[mesa.git] / src / gallium / drivers / svga / svga_state_need_swtnl.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_state.h"
28 #include "svga_context.h"
29 #include "svga_state.h"
30 #include "svga_debug.h"
31 #include "svga_hw_reg.h"
32
33
34 /**
35 * Given a gallium vertex element format, return the corresponding SVGA3D
36 * format. Return SVGA3D_DECLTYPE_MAX for unsupported gallium formats.
37 */
38 static INLINE SVGA3dDeclType
39 svga_translate_vertex_format(enum pipe_format format)
40 {
41 switch (format) {
42 case PIPE_FORMAT_R32_FLOAT: return SVGA3D_DECLTYPE_FLOAT1;
43 case PIPE_FORMAT_R32G32_FLOAT: return SVGA3D_DECLTYPE_FLOAT2;
44 case PIPE_FORMAT_R32G32B32_FLOAT: return SVGA3D_DECLTYPE_FLOAT3;
45 case PIPE_FORMAT_R32G32B32A32_FLOAT: return SVGA3D_DECLTYPE_FLOAT4;
46 case PIPE_FORMAT_B8G8R8A8_UNORM: return SVGA3D_DECLTYPE_D3DCOLOR;
47 case PIPE_FORMAT_R8G8B8A8_USCALED: return SVGA3D_DECLTYPE_UBYTE4;
48 case PIPE_FORMAT_R16G16_SSCALED: return SVGA3D_DECLTYPE_SHORT2;
49 case PIPE_FORMAT_R16G16B16A16_SSCALED: return SVGA3D_DECLTYPE_SHORT4;
50 case PIPE_FORMAT_R8G8B8A8_UNORM: return SVGA3D_DECLTYPE_UBYTE4N;
51 case PIPE_FORMAT_R16G16_SNORM: return SVGA3D_DECLTYPE_SHORT2N;
52 case PIPE_FORMAT_R16G16B16A16_SNORM: return SVGA3D_DECLTYPE_SHORT4N;
53 case PIPE_FORMAT_R16G16_UNORM: return SVGA3D_DECLTYPE_USHORT2N;
54 case PIPE_FORMAT_R16G16B16A16_UNORM: return SVGA3D_DECLTYPE_USHORT4N;
55 case PIPE_FORMAT_R10G10B10X2_USCALED: return SVGA3D_DECLTYPE_UDEC3;
56 case PIPE_FORMAT_R10G10B10X2_SNORM: return SVGA3D_DECLTYPE_DEC3N;
57 case PIPE_FORMAT_R16G16_FLOAT: return SVGA3D_DECLTYPE_FLOAT16_2;
58 case PIPE_FORMAT_R16G16B16A16_FLOAT: return SVGA3D_DECLTYPE_FLOAT16_4;
59
60 default:
61 /* There are many formats without hardware support. This case
62 * will be hit regularly, meaning we'll need swvfetch.
63 */
64 return SVGA3D_DECLTYPE_MAX;
65 }
66 }
67
68
69 static enum pipe_error
70 update_need_swvfetch(struct svga_context *svga, unsigned dirty)
71 {
72 unsigned i;
73 boolean need_swvfetch = FALSE;
74
75 if (!svga->curr.velems) {
76 /* No vertex elements bound. */
77 return PIPE_OK;
78 }
79
80 for (i = 0; i < svga->curr.velems->count; i++) {
81 svga->state.sw.ve_format[i] = svga_translate_vertex_format(svga->curr.velems->velem[i].src_format);
82 if (svga->state.sw.ve_format[i] == SVGA3D_DECLTYPE_MAX) {
83 /* Unsupported format - use software fetch */
84 need_swvfetch = TRUE;
85 break;
86 }
87 }
88
89 if (need_swvfetch != svga->state.sw.need_swvfetch) {
90 svga->state.sw.need_swvfetch = need_swvfetch;
91 svga->dirty |= SVGA_NEW_NEED_SWVFETCH;
92 }
93
94 return PIPE_OK;
95 }
96
97 struct svga_tracked_state svga_update_need_swvfetch =
98 {
99 "update need_swvfetch",
100 ( SVGA_NEW_VELEMENT ),
101 update_need_swvfetch
102 };
103
104
105
106 static enum pipe_error
107 update_need_pipeline(struct svga_context *svga, unsigned dirty)
108 {
109 boolean need_pipeline = FALSE;
110 struct svga_vertex_shader *vs = svga->curr.vs;
111
112 /* SVGA_NEW_RAST, SVGA_NEW_REDUCED_PRIMITIVE
113 */
114 if (svga->curr.rast->need_pipeline & (1 << svga->curr.reduced_prim)) {
115 SVGA_DBG(DEBUG_SWTNL, "%s: rast need_pipeline (0x%x) & prim (0x%x)\n",
116 __FUNCTION__,
117 svga->curr.rast->need_pipeline,
118 (1 << svga->curr.reduced_prim) );
119 SVGA_DBG(DEBUG_SWTNL, "%s: rast need_pipeline tris (%s), lines (%s), points (%s)\n",
120 __FUNCTION__,
121 svga->curr.rast->need_pipeline_tris_str,
122 svga->curr.rast->need_pipeline_lines_str,
123 svga->curr.rast->need_pipeline_points_str);
124 need_pipeline = TRUE;
125 }
126
127 /* EDGEFLAGS
128 */
129 if (vs && vs->base.info.writes_edgeflag) {
130 SVGA_DBG(DEBUG_SWTNL, "%s: edgeflags\n", __FUNCTION__);
131 need_pipeline = TRUE;
132 }
133
134 /* SVGA_NEW_FS, SVGA_NEW_RAST, SVGA_NEW_REDUCED_PRIMITIVE
135 */
136 if (svga->curr.reduced_prim == PIPE_PRIM_POINTS) {
137 unsigned sprite_coord_gen = svga->curr.rast->templ.sprite_coord_enable;
138 unsigned generic_inputs =
139 svga->curr.fs ? svga->curr.fs->generic_inputs : 0;
140
141 if (sprite_coord_gen &&
142 (generic_inputs & ~sprite_coord_gen)) {
143 /* The fragment shader is using some generic inputs that are
144 * not being replaced by auto-generated point/sprite coords (and
145 * auto sprite coord generation is turned on).
146 * The SVGA3D interface does not support that: if we enable
147 * SVGA3D_RS_POINTSPRITEENABLE it gets enabled for _all_
148 * texture coordinate sets.
149 * To solve this, we have to use the draw-module's wide/sprite
150 * point stage.
151 */
152 need_pipeline = TRUE;
153 }
154 }
155
156 if (need_pipeline != svga->state.sw.need_pipeline) {
157 svga->state.sw.need_pipeline = need_pipeline;
158 svga->dirty |= SVGA_NEW_NEED_PIPELINE;
159 }
160
161 /* DEBUG */
162 if (0 && svga->state.sw.need_pipeline)
163 debug_printf("sw.need_pipeline = %d\n", svga->state.sw.need_pipeline);
164
165 return PIPE_OK;
166 }
167
168
169 struct svga_tracked_state svga_update_need_pipeline =
170 {
171 "need pipeline",
172 (SVGA_NEW_RAST |
173 SVGA_NEW_FS |
174 SVGA_NEW_VS |
175 SVGA_NEW_REDUCED_PRIMITIVE),
176 update_need_pipeline
177 };
178
179
180 static enum pipe_error
181 update_need_swtnl(struct svga_context *svga, unsigned dirty)
182 {
183 boolean need_swtnl;
184
185 if (svga->debug.no_swtnl) {
186 svga->state.sw.need_swvfetch = FALSE;
187 svga->state.sw.need_pipeline = FALSE;
188 }
189
190 need_swtnl = (svga->state.sw.need_swvfetch ||
191 svga->state.sw.need_pipeline);
192
193 if (svga->debug.force_swtnl) {
194 need_swtnl = TRUE;
195 }
196
197 /*
198 * Some state changes the draw module does makes us believe we
199 * we don't need swtnl. This causes the vdecl code to pickup
200 * the wrong buffers and vertex formats. Try trivial/line-wide.
201 */
202 if (svga->state.sw.in_swtnl_draw)
203 need_swtnl = TRUE;
204
205 if (need_swtnl != svga->state.sw.need_swtnl) {
206 SVGA_DBG(DEBUG_SWTNL|DEBUG_PERF,
207 "%s: need_swvfetch %s, need_pipeline %s\n",
208 __FUNCTION__,
209 svga->state.sw.need_swvfetch ? "true" : "false",
210 svga->state.sw.need_pipeline ? "true" : "false");
211
212 svga->state.sw.need_swtnl = need_swtnl;
213 svga->dirty |= SVGA_NEW_NEED_SWTNL;
214 svga->swtnl.new_vdecl = TRUE;
215 }
216
217 return PIPE_OK;
218 }
219
220
221 struct svga_tracked_state svga_update_need_swtnl =
222 {
223 "need swtnl",
224 (SVGA_NEW_NEED_PIPELINE |
225 SVGA_NEW_NEED_SWVFETCH),
226 update_need_swtnl
227 };