xref: /aosp_15_r20/external/mesa3d/src/gallium/auxiliary/draw/draw_vs_exec.c (revision 6104692788411f58d303aa86923a9ff6ecaded22)
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 <[email protected]>
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 
50 struct exec_vertex_shader {
51    struct draw_vertex_shader base;
52    struct tgsi_exec_machine *machine;
53 };
54 
55 
56 static struct exec_vertex_shader *
exec_vertex_shader(struct draw_vertex_shader * vs)57 exec_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  */
65 static void
vs_exec_prepare(struct draw_vertex_shader * shader,struct draw_context * draw)66 vs_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    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  */
91 static void
vs_exec_run_linear(struct draw_vertex_shader * shader,const float (* input)[4],float (* output)[4],const struct draw_buffer_info * constants,unsigned count,unsigned input_stride,unsigned output_stride,const unsigned * fetch_elts)92 vs_exec_run_linear(struct draw_vertex_shader *shader,
93                    const float (*input)[4],
94                    float (*output)[4],
95                    const struct draw_buffer_info *constants,
96                    unsigned count,
97                    unsigned input_stride,
98                    unsigned output_stride,
99                    const unsigned *fetch_elts)
100 {
101    struct exec_vertex_shader *evs = exec_vertex_shader(shader);
102    struct tgsi_exec_machine *machine = evs->machine;
103    unsigned int i, j;
104    unsigned slot;
105    bool clamp_vertex_color = shader->draw->rasterizer->clamp_vertex_color;
106 
107    assert(!shader->draw->llvm);
108    tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
109                                   (const struct tgsi_exec_consts_info *)constants);
110 
111    if (shader->info.uses_instanceid) {
112       unsigned i = machine->SysSemanticToIndex[TGSI_SEMANTIC_INSTANCEID];
113       assert(i < ARRAY_SIZE(machine->SystemValue));
114       for (j = 0; j < TGSI_QUAD_SIZE; j++)
115          machine->SystemValue[i].xyzw[0].i[j] = shader->draw->instance_id;
116    }
117 
118    for (i = 0; i < count; i += MAX_TGSI_VERTICES) {
119       unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i);
120 
121       /* Swizzle inputs.
122        */
123       for (j = 0; j < max_vertices; j++) {
124 #if 0
125          debug_printf("%d) Input vert:\n", i + j);
126          for (slot = 0; slot < shader->info.num_inputs; slot++) {
127             debug_printf("\t%d: %f %f %f %f\n", slot,
128                          input[slot][0],
129                          input[slot][1],
130                          input[slot][2],
131                          input[slot][3]);
132          }
133 #endif
134 	 int basevertex = shader->draw->pt.user.eltSize ? shader->draw->pt.user.eltBias : shader->draw->start_index;
135 
136          if (shader->info.uses_vertexid) {
137             unsigned vid = machine->SysSemanticToIndex[TGSI_SEMANTIC_VERTEXID];
138             assert(vid < ARRAY_SIZE(machine->SystemValue));
139             machine->SystemValue[vid].xyzw[0].i[j] = fetch_elts ? fetch_elts[i + j] : (i + j + basevertex);
140          }
141          if (shader->info.uses_basevertex) {
142             unsigned vid = machine->SysSemanticToIndex[TGSI_SEMANTIC_BASEVERTEX];
143             assert(vid < ARRAY_SIZE(machine->SystemValue));
144             machine->SystemValue[vid].xyzw[0].i[j] = basevertex;
145          }
146          if (shader->info.uses_vertexid_nobase) {
147             unsigned vid = machine->SysSemanticToIndex[TGSI_SEMANTIC_VERTEXID_NOBASE];
148             assert(vid < ARRAY_SIZE(machine->SystemValue));
149             machine->SystemValue[vid].xyzw[0].i[j] = fetch_elts ? (fetch_elts[i + j] - basevertex) : (i + j);
150          }
151 
152          for (slot = 0; slot < shader->info.num_inputs; slot++) {
153 #if 0
154             assert(!util_is_inf_or_nan(input[slot][0]));
155             assert(!util_is_inf_or_nan(input[slot][1]));
156             assert(!util_is_inf_or_nan(input[slot][2]));
157             assert(!util_is_inf_or_nan(input[slot][3]));
158 #endif
159             machine->Inputs[slot].xyzw[0].f[j] = input[slot][0];
160             machine->Inputs[slot].xyzw[1].f[j] = input[slot][1];
161             machine->Inputs[slot].xyzw[2].f[j] = input[slot][2];
162             machine->Inputs[slot].xyzw[3].f[j] = input[slot][3];
163          }
164 
165          input = (const float (*)[4])((const char *)input + input_stride);
166       }
167 
168       machine->NonHelperMask = (1 << max_vertices) - 1;
169       /* run interpreter */
170       tgsi_exec_machine_run(machine, 0);
171 
172       /* Unswizzle all output results.
173        */
174       for (j = 0; j < max_vertices; j++) {
175          for (slot = 0; slot < shader->info.num_outputs; slot++) {
176             enum tgsi_semantic name = shader->info.output_semantic_name[slot];
177             if (clamp_vertex_color &&
178                 (name == TGSI_SEMANTIC_COLOR || name == TGSI_SEMANTIC_BCOLOR)) {
179                output[slot][0] = SATURATE(machine->Outputs[slot].xyzw[0].f[j]);
180                output[slot][1] = SATURATE(machine->Outputs[slot].xyzw[1].f[j]);
181                output[slot][2] = SATURATE(machine->Outputs[slot].xyzw[2].f[j]);
182                output[slot][3] = SATURATE(machine->Outputs[slot].xyzw[3].f[j]);
183             } else {
184                output[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
185                output[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
186                output[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
187                output[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
188             }
189          }
190 
191 #if 0
192          debug_printf("%d) Post xform vert:\n", i + j);
193          for (slot = 0; slot < shader->info.num_outputs; slot++) {
194             debug_printf("\t%d: %f %f %f %f\n", slot,
195                          output[slot][0],
196                          output[slot][1],
197                          output[slot][2],
198                          output[slot][3]);
199             assert(!util_is_inf_or_nan(output[slot][0]));
200          }
201 #endif
202 
203          output = (float (*)[4])((char *)output + output_stride);
204       }
205    }
206 }
207 
208 
209 static void
vs_exec_delete(struct draw_vertex_shader * dvs)210 vs_exec_delete(struct draw_vertex_shader *dvs)
211 {
212    FREE((void*) dvs->state.tokens);
213    FREE(dvs);
214 }
215 
216 
217 struct draw_vertex_shader *
draw_create_vs_exec(struct draw_context * draw,const struct pipe_shader_state * state)218 draw_create_vs_exec(struct draw_context *draw,
219                     const struct pipe_shader_state *state)
220 {
221    struct exec_vertex_shader *vs = CALLOC_STRUCT(exec_vertex_shader);
222 
223    if (!vs)
224       return NULL;
225 
226    if (state->type == PIPE_SHADER_IR_NIR) {
227       vs->base.state.type = PIPE_SHADER_IR_TGSI;
228       vs->base.state.tokens = nir_to_tgsi(state->ir.nir, draw->pipe->screen);
229    } else {
230       assert(state->type == PIPE_SHADER_IR_TGSI);
231       vs->base.state.type = state->type;
232 
233       /* we need to keep a local copy of the tokens */
234       vs->base.state.tokens = tgsi_dup_tokens(state->tokens);
235       if (!vs->base.state.tokens) {
236          FREE(vs);
237          return NULL;
238       }
239    }
240 
241    tgsi_scan_shader(vs->base.state.tokens, &vs->base.info);
242 
243    vs->base.state.stream_output = state->stream_output;
244    vs->base.draw = draw;
245    vs->base.prepare = vs_exec_prepare;
246    vs->base.run_linear = vs_exec_run_linear;
247    vs->base.delete = vs_exec_delete;
248    vs->base.create_variant = draw_vs_create_variant_generic;
249    vs->machine = draw->vs.tgsi.machine;
250 
251    return &vs->base;
252 }
253