1 /*
2 * Copyright 2012 Red Hat Inc.
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
17 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
18 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
19 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
20 * OTHER DEALINGS IN THE SOFTWARE.
21 *
22 * Authors: Ben Skeggs
23 *
24 */
25
26 #include "draw/draw_context.h"
27 #include "util/u_dynarray.h"
28 #include "tgsi/tgsi_parse.h"
29 #include "nir/nir_to_tgsi.h"
30
31 #include "nv_object.xml.h"
32 #include "nv30/nv30-40_3d.xml.h"
33 #include "nv30/nv30_context.h"
34 #include "nv30/nvfx_shader.h"
35 #include "nv30/nv30_state.h"
36 #include "nv30/nv30_winsys.h"
37
38 static void
nv30_vertprog_destroy(struct nv30_vertprog * vp)39 nv30_vertprog_destroy(struct nv30_vertprog *vp)
40 {
41 util_dynarray_fini(&vp->branch_relocs);
42 nouveau_heap_free(&vp->exec);
43 FREE(vp->insns);
44 vp->insns = NULL;
45 vp->nr_insns = 0;
46
47 util_dynarray_fini(&vp->const_relocs);
48 nouveau_heap_free(&vp->data);
49 FREE(vp->consts);
50 vp->consts = NULL;
51 vp->nr_consts = 0;
52
53 vp->translated = false;
54 }
55
56 void
nv30_vertprog_validate(struct nv30_context * nv30)57 nv30_vertprog_validate(struct nv30_context *nv30)
58 {
59 struct nouveau_pushbuf *push = nv30->base.pushbuf;
60 struct nouveau_object *eng3d = nv30->screen->eng3d;
61 struct nv30_vertprog *vp = nv30->vertprog.program;
62 struct nv30_fragprog *fp = nv30->fragprog.program;
63 bool upload_code = false;
64 bool upload_data = false;
65 unsigned i;
66
67 if (nv30->dirty & NV30_NEW_FRAGPROG) {
68 if (memcmp(vp->texcoord, fp->texcoord, sizeof(vp->texcoord))) {
69 if (vp->translated)
70 nv30_vertprog_destroy(vp);
71 memcpy(vp->texcoord, fp->texcoord, sizeof(vp->texcoord));
72 }
73 }
74
75 if (nv30->rast && nv30->rast->pipe.clip_plane_enable != vp->enabled_ucps) {
76 vp->enabled_ucps = nv30->rast->pipe.clip_plane_enable;
77 if (vp->translated)
78 nv30_vertprog_destroy(vp);
79 }
80
81 if (!vp->translated) {
82 vp->translated = _nvfx_vertprog_translate(eng3d->oclass, vp);
83 if (!vp->translated) {
84 nv30->draw_flags |= NV30_NEW_VERTPROG;
85 return;
86 }
87 nv30->dirty |= NV30_NEW_VERTPROG;
88 }
89
90 if (!vp->exec) {
91 struct nouveau_heap *heap = nv30->screen->vp_exec_heap;
92 struct nv30_shader_reloc *reloc = vp->branch_relocs.data;
93 unsigned nr_reloc = vp->branch_relocs.size / sizeof(*reloc);
94 uint32_t *inst, target;
95
96 if (nouveau_heap_alloc(heap, vp->nr_insns, &vp->exec, &vp->exec)) {
97 while (heap->next && heap->size < vp->nr_insns) {
98 struct nouveau_heap **evict = heap->next->priv;
99 nouveau_heap_free(evict);
100 }
101
102 if (nouveau_heap_alloc(heap, vp->nr_insns, &vp->exec, &vp->exec)) {
103 nv30->draw_flags |= NV30_NEW_VERTPROG;
104 return;
105 }
106 }
107
108 if (eng3d->oclass < NV40_3D_CLASS) {
109 while (nr_reloc--) {
110 inst = vp->insns[reloc->location].data;
111 target = vp->exec->start + reloc->target;
112
113 inst[2] &= ~0x000007fc;
114 inst[2] |= target << 2;
115 reloc++;
116 }
117 } else {
118 while (nr_reloc--) {
119 inst = vp->insns[reloc->location].data;
120 target = vp->exec->start + reloc->target;
121
122 inst[2] &= ~0x0000003f;
123 inst[2] |= target >> 3;
124 inst[3] &= ~0xe0000000;
125 inst[3] |= target << 29;
126 reloc++;
127 }
128 }
129
130 upload_code = true;
131 }
132
133 if (vp->nr_consts && !vp->data) {
134 struct nouveau_heap *heap = nv30->screen->vp_data_heap;
135 struct nv30_shader_reloc *reloc = vp->const_relocs.data;
136 unsigned nr_reloc = vp->const_relocs.size / sizeof(*reloc);
137 uint32_t *inst, target;
138
139 if (nouveau_heap_alloc(heap, vp->nr_consts, vp, &vp->data)) {
140 while (heap->next && heap->size < vp->nr_consts) {
141 struct nv30_vertprog *evp = heap->next->priv;
142 nouveau_heap_free(&evp->data);
143 }
144
145 if (nouveau_heap_alloc(heap, vp->nr_consts, vp, &vp->data)) {
146 nv30->draw_flags |= NV30_NEW_VERTPROG;
147 return;
148 }
149 }
150
151 if (eng3d->oclass < NV40_3D_CLASS) {
152 while (nr_reloc--) {
153 inst = vp->insns[reloc->location].data;
154 target = vp->data->start + reloc->target;
155
156 inst[1] &= ~0x0007fc000;
157 inst[1] |= (target & 0x1ff) << 14;
158 reloc++;
159 }
160 } else {
161 while (nr_reloc--) {
162 inst = vp->insns[reloc->location].data;
163 target = vp->data->start + reloc->target;
164
165 inst[1] &= ~0x0001ff000;
166 inst[1] |= (target & 0x1ff) << 12;
167 reloc++;
168 }
169 }
170
171 upload_code = true;
172 upload_data = true;
173 }
174
175 if (vp->nr_consts) {
176 struct nv04_resource *res = nv04_resource(nv30->vertprog.constbuf);
177
178 for (i = 0; i < vp->nr_consts; i++) {
179 struct nv30_vertprog_data *data = &vp->consts[i];
180
181 if (data->index < 0) {
182 if (!upload_data)
183 continue;
184 } else {
185 float *constbuf = (float *)res->data;
186 if (!upload_data &&
187 !memcmp(data->value, &constbuf[data->index * 4], 16))
188 continue;
189 memcpy(data->value, &constbuf[data->index * 4], 16);
190 }
191
192 BEGIN_NV04(push, NV30_3D(VP_UPLOAD_CONST_ID), 5);
193 PUSH_DATA (push, vp->data->start + i);
194 PUSH_DATAp(push, data->value, 4);
195 }
196 }
197
198 if (upload_code) {
199 BEGIN_NV04(push, NV30_3D(VP_UPLOAD_FROM_ID), 1);
200 PUSH_DATA (push, vp->exec->start);
201 for (i = 0; i < vp->nr_insns; i++) {
202 BEGIN_NV04(push, NV30_3D(VP_UPLOAD_INST(0)), 4);
203 PUSH_DATAp(push, vp->insns[i].data, 4);
204 }
205 }
206
207 if (nv30->dirty & (NV30_NEW_VERTPROG | NV30_NEW_FRAGPROG)) {
208 BEGIN_NV04(push, NV30_3D(VP_START_FROM_ID), 1);
209 PUSH_DATA (push, vp->exec->start);
210 if (eng3d->oclass < NV40_3D_CLASS) {
211 BEGIN_NV04(push, NV30_3D(ENGINE), 1);
212 PUSH_DATA (push, 0x00000013); /* vp instead of ff, somehow */
213 } else {
214 BEGIN_NV04(push, NV40_3D(VP_ATTRIB_EN), 2);
215 PUSH_DATA (push, vp->ir);
216 PUSH_DATA (push, vp->or | fp->vp_or);
217 BEGIN_NV04(push, NV30_3D(ENGINE), 1);
218 PUSH_DATA (push, 0x00000011);
219 }
220 }
221 }
222
223 static void *
nv30_vp_state_create(struct pipe_context * pipe,const struct pipe_shader_state * cso)224 nv30_vp_state_create(struct pipe_context *pipe,
225 const struct pipe_shader_state *cso)
226 {
227 struct nv30_vertprog *vp = CALLOC_STRUCT(nv30_vertprog);
228 if (!vp)
229 return NULL;
230
231 if (cso->type == PIPE_SHADER_IR_NIR) {
232 vp->pipe.tokens = nir_to_tgsi(cso->ir.nir, pipe->screen);
233 } else {
234 assert(cso->type == PIPE_SHADER_IR_TGSI);
235 /* we need to keep a local copy of the tokens */
236 vp->pipe.tokens = tgsi_dup_tokens(cso->tokens);
237 }
238
239 tgsi_scan_shader(vp->pipe.tokens, &vp->info);
240 return vp;
241 }
242
243 static void
nv30_vp_state_delete(struct pipe_context * pipe,void * hwcso)244 nv30_vp_state_delete(struct pipe_context *pipe, void *hwcso)
245 {
246 struct nv30_vertprog *vp = hwcso;
247
248 if (vp->translated)
249 nv30_vertprog_destroy(vp);
250
251 if (vp->draw)
252 draw_delete_vertex_shader(nv30_context(pipe)->draw, vp->draw);
253
254 FREE((void *)vp->pipe.tokens);
255 FREE(vp);
256 }
257
258 static void
nv30_vp_state_bind(struct pipe_context * pipe,void * hwcso)259 nv30_vp_state_bind(struct pipe_context *pipe, void *hwcso)
260 {
261 struct nv30_context *nv30 = nv30_context(pipe);
262
263 nv30->vertprog.program = hwcso;
264 nv30->dirty |= NV30_NEW_VERTPROG;
265 }
266
267 void
nv30_vertprog_init(struct pipe_context * pipe)268 nv30_vertprog_init(struct pipe_context *pipe)
269 {
270 pipe->create_vs_state = nv30_vp_state_create;
271 pipe->bind_vs_state = nv30_vp_state_bind;
272 pipe->delete_vs_state = nv30_vp_state_delete;
273 }
274