1/**************************************************************************
2 *
3 * Copyright 2007 VMware, Inc.
4 * All Rights Reserved.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
13 *
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
16 * of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25 *
26 **************************************************************************/
27
28 /*
29  * Authors:
30  *   Keith Whitwell <keithw@vmware.com>
31  *   Brian Paul
32  */
33
34#include "util/u_math.h"
35#include "util/u_memory.h"
36#include "util/ralloc.h"
37#include "pipe/p_shader_tokens.h"
38#include "pipe/p_context.h"
39#include "nir/nir_to_tgsi.h"
40
41#include "draw_private.h"
42#include "draw_context.h"
43#include "draw_vs.h"
44
45#include "tgsi/tgsi_parse.h"
46#include "tgsi/tgsi_scan.h"
47#include "tgsi/tgsi_exec.h"
48
49
50struct exec_vertex_shader {
51   struct draw_vertex_shader base;
52   struct tgsi_exec_machine *machine;
53};
54
55
56static struct exec_vertex_shader *
57exec_vertex_shader(struct draw_vertex_shader *vs)
58{
59   return (struct exec_vertex_shader *)vs;
60}
61
62
63/* Not required for run_linear.
64 */
65static void
66vs_exec_prepare(struct draw_vertex_shader *shader,
67                struct draw_context *draw)
68{
69   struct exec_vertex_shader *evs = exec_vertex_shader(shader);
70
71   debug_assert(!draw->llvm);
72   /* Specify the vertex program to interpret/execute.
73    * Avoid rebinding when possible.
74    */
75   if (evs->machine->Tokens != shader->state.tokens) {
76      tgsi_exec_machine_bind_shader(evs->machine,
77                                    shader->state.tokens,
78                                    draw->vs.tgsi.sampler,
79                                    draw->vs.tgsi.image,
80                                    draw->vs.tgsi.buffer);
81   }
82}
83
84
85
86/**
87 * Simplified vertex shader interface for the pt paths.  Given the
88 * complexity of code-generating all the above operations together,
89 * it's time to try doing all the other stuff separately.
90 */
91static void
92vs_exec_run_linear(struct draw_vertex_shader *shader,
93                   const float (*input)[4],
94                   float (*output)[4],
95                   const void *constants[PIPE_MAX_CONSTANT_BUFFERS],
96                    const unsigned const_size[PIPE_MAX_CONSTANT_BUFFERS],
97                   unsigned count,
98                   unsigned input_stride,
99                   unsigned output_stride,
100                   const unsigned *fetch_elts)
101{
102   struct exec_vertex_shader *evs = exec_vertex_shader(shader);
103   struct tgsi_exec_machine *machine = evs->machine;
104   unsigned int i, j;
105   unsigned slot;
106   boolean clamp_vertex_color = shader->draw->rasterizer->clamp_vertex_color;
107
108   debug_assert(!shader->draw->llvm);
109   tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
110                                  constants, const_size);
111
112   if (shader->info.uses_instanceid) {
113      unsigned i = machine->SysSemanticToIndex[TGSI_SEMANTIC_INSTANCEID];
114      assert(i < ARRAY_SIZE(machine->SystemValue));
115      for (j = 0; j < TGSI_QUAD_SIZE; j++)
116         machine->SystemValue[i].xyzw[0].i[j] = shader->draw->instance_id;
117   }
118
119   for (i = 0; i < count; i += MAX_TGSI_VERTICES) {
120      unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i);
121
122      /* Swizzle inputs.
123       */
124      for (j = 0; j < max_vertices; j++) {
125#if 0
126         debug_printf("%d) Input vert:\n", i + j);
127         for (slot = 0; slot < shader->info.num_inputs; slot++) {
128            debug_printf("\t%d: %f %f %f %f\n", slot,
129                         input[slot][0],
130                         input[slot][1],
131                         input[slot][2],
132                         input[slot][3]);
133         }
134#endif
135	 int basevertex = shader->draw->pt.user.eltSize ? shader->draw->pt.user.eltBias : shader->draw->start_index;
136
137         if (shader->info.uses_vertexid) {
138            unsigned vid = machine->SysSemanticToIndex[TGSI_SEMANTIC_VERTEXID];
139            assert(vid < ARRAY_SIZE(machine->SystemValue));
140            machine->SystemValue[vid].xyzw[0].i[j] = fetch_elts ? fetch_elts[i + j] : (i + j + basevertex);
141         }
142         if (shader->info.uses_basevertex) {
143            unsigned vid = machine->SysSemanticToIndex[TGSI_SEMANTIC_BASEVERTEX];
144            assert(vid < ARRAY_SIZE(machine->SystemValue));
145            machine->SystemValue[vid].xyzw[0].i[j] = basevertex;
146         }
147         if (shader->info.uses_vertexid_nobase) {
148            unsigned vid = machine->SysSemanticToIndex[TGSI_SEMANTIC_VERTEXID_NOBASE];
149            assert(vid < ARRAY_SIZE(machine->SystemValue));
150            machine->SystemValue[vid].xyzw[0].i[j] = fetch_elts ? (fetch_elts[i + j] - basevertex) : (i + j);
151         }
152
153         for (slot = 0; slot < shader->info.num_inputs; slot++) {
154#if 0
155            assert(!util_is_inf_or_nan(input[slot][0]));
156            assert(!util_is_inf_or_nan(input[slot][1]));
157            assert(!util_is_inf_or_nan(input[slot][2]));
158            assert(!util_is_inf_or_nan(input[slot][3]));
159#endif
160            machine->Inputs[slot].xyzw[0].f[j] = input[slot][0];
161            machine->Inputs[slot].xyzw[1].f[j] = input[slot][1];
162            machine->Inputs[slot].xyzw[2].f[j] = input[slot][2];
163            machine->Inputs[slot].xyzw[3].f[j] = input[slot][3];
164         }
165
166         input = (const float (*)[4])((const char *)input + input_stride);
167      }
168
169      machine->NonHelperMask = (1 << max_vertices) - 1;
170      /* run interpreter */
171      tgsi_exec_machine_run(machine, 0);
172
173      /* Unswizzle all output results.
174       */
175      for (j = 0; j < max_vertices; j++) {
176         for (slot = 0; slot < shader->info.num_outputs; slot++) {
177            enum tgsi_semantic name = shader->info.output_semantic_name[slot];
178            if (clamp_vertex_color &&
179                (name == TGSI_SEMANTIC_COLOR || name == TGSI_SEMANTIC_BCOLOR)) {
180               output[slot][0] = SATURATE(machine->Outputs[slot].xyzw[0].f[j]);
181               output[slot][1] = SATURATE(machine->Outputs[slot].xyzw[1].f[j]);
182               output[slot][2] = SATURATE(machine->Outputs[slot].xyzw[2].f[j]);
183               output[slot][3] = SATURATE(machine->Outputs[slot].xyzw[3].f[j]);
184            } else {
185               output[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
186               output[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
187               output[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
188               output[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
189            }
190         }
191
192#if 0
193         debug_printf("%d) Post xform vert:\n", i + j);
194         for (slot = 0; slot < shader->info.num_outputs; slot++) {
195            debug_printf("\t%d: %f %f %f %f\n", slot,
196                         output[slot][0],
197                         output[slot][1],
198                         output[slot][2],
199                         output[slot][3]);
200            assert(!util_is_inf_or_nan(output[slot][0]));
201         }
202#endif
203
204         output = (float (*)[4])((char *)output + output_stride);
205      }
206   }
207}
208
209
210static void
211vs_exec_delete(struct draw_vertex_shader *dvs)
212{
213   FREE((void*) dvs->state.tokens);
214   FREE(dvs);
215}
216
217
218struct draw_vertex_shader *
219draw_create_vs_exec(struct draw_context *draw,
220                    const struct pipe_shader_state *state)
221{
222   struct exec_vertex_shader *vs = CALLOC_STRUCT(exec_vertex_shader);
223
224   if (!vs)
225      return NULL;
226
227   if (state->type == PIPE_SHADER_IR_NIR) {
228      vs->base.state.type = PIPE_SHADER_IR_TGSI;
229      vs->base.state.tokens = nir_to_tgsi(state->ir.nir, draw->pipe->screen);
230   } else {
231      assert(state->type == PIPE_SHADER_IR_TGSI);
232      vs->base.state.type = state->type;
233
234      /* we need to keep a local copy of the tokens */
235      vs->base.state.tokens = tgsi_dup_tokens(state->tokens);
236      if (!vs->base.state.tokens) {
237         FREE(vs);
238         return NULL;
239      }
240   }
241
242   tgsi_scan_shader(vs->base.state.tokens, &vs->base.info);
243
244   vs->base.state.stream_output = state->stream_output;
245   vs->base.draw = draw;
246   vs->base.prepare = vs_exec_prepare;
247   vs->base.run_linear = vs_exec_run_linear;
248   vs->base.delete = vs_exec_delete;
249   vs->base.create_variant = draw_vs_create_variant_generic;
250   vs->machine = draw->vs.tgsi.machine;
251
252   return &vs->base;
253}
254