1 /*
2 * Copyright © 2016-2018 Broadcom
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 (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
13 * Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
21 * IN THE SOFTWARE.
22 */
23
24 #include "v3d_compiler.h"
25
26 /* We don't do any address packing. */
27 #define __gen_user_data void
28 #define __gen_address_type uint32_t
29 #define __gen_address_offset(reloc) (*reloc)
30 #define __gen_emit_reloc(cl, reloc)
31 #include "cle/v3d_packet_v33_pack.h"
32
33 void
v3d33_vir_emit_tex(struct v3d_compile * c,nir_tex_instr * instr)34 v3d33_vir_emit_tex(struct v3d_compile *c, nir_tex_instr *instr)
35 {
36 /* FIXME: We don't bother implementing pipelining for texture reads
37 * for any pre 4.x hardware. It should be straight forward to do but
38 * we are not really testing or even targetting this hardware at
39 * present.
40 */
41 ntq_flush_tmu(c);
42
43 unsigned unit = instr->texture_index;
44
45 struct V3D33_TEXTURE_UNIFORM_PARAMETER_0_CFG_MODE1 p0_unpacked = {
46 V3D33_TEXTURE_UNIFORM_PARAMETER_0_CFG_MODE1_header,
47
48 .fetch_sample_mode = instr->op == nir_texop_txf,
49 };
50
51 struct V3D33_TEXTURE_UNIFORM_PARAMETER_1_CFG_MODE1 p1_unpacked = {
52 };
53
54 switch (instr->sampler_dim) {
55 case GLSL_SAMPLER_DIM_1D:
56 if (instr->is_array)
57 p0_unpacked.lookup_type = TEXTURE_1D_ARRAY;
58 else
59 p0_unpacked.lookup_type = TEXTURE_1D;
60 break;
61 case GLSL_SAMPLER_DIM_2D:
62 case GLSL_SAMPLER_DIM_RECT:
63 if (instr->is_array)
64 p0_unpacked.lookup_type = TEXTURE_2D_ARRAY;
65 else
66 p0_unpacked.lookup_type = TEXTURE_2D;
67 break;
68 case GLSL_SAMPLER_DIM_3D:
69 p0_unpacked.lookup_type = TEXTURE_3D;
70 break;
71 case GLSL_SAMPLER_DIM_CUBE:
72 p0_unpacked.lookup_type = TEXTURE_CUBE_MAP;
73 break;
74 default:
75 unreachable("Bad sampler type");
76 }
77
78 struct qreg coords[5];
79 int next_coord = 0;
80 for (unsigned i = 0; i < instr->num_srcs; i++) {
81 switch (instr->src[i].src_type) {
82 case nir_tex_src_coord:
83 for (int j = 0; j < instr->coord_components; j++) {
84 coords[next_coord++] =
85 ntq_get_src(c, instr->src[i].src, j);
86 }
87 if (instr->coord_components < 2)
88 coords[next_coord++] = vir_uniform_f(c, 0.5);
89 break;
90 case nir_tex_src_bias:
91 coords[next_coord++] =
92 ntq_get_src(c, instr->src[i].src, 0);
93
94 p0_unpacked.bias_supplied = true;
95 break;
96 case nir_tex_src_lod:
97 coords[next_coord++] =
98 vir_FADD(c,
99 ntq_get_src(c, instr->src[i].src, 0),
100 vir_uniform(c, QUNIFORM_TEXTURE_FIRST_LEVEL,
101 unit));
102
103 if (instr->op != nir_texop_txf &&
104 instr->op != nir_texop_tg4) {
105 p0_unpacked.disable_autolod_use_bias_only = true;
106 }
107 break;
108 case nir_tex_src_comparator:
109 coords[next_coord++] =
110 ntq_get_src(c, instr->src[i].src, 0);
111
112 p0_unpacked.shadow = true;
113 break;
114
115 case nir_tex_src_offset: {
116 p0_unpacked.texel_offset_for_s_coordinate =
117 nir_src_comp_as_int(instr->src[i].src, 0);
118
119 if (instr->coord_components >= 2)
120 p0_unpacked.texel_offset_for_t_coordinate =
121 nir_src_comp_as_int(instr->src[i].src, 1);
122
123 if (instr->coord_components >= 3)
124 p0_unpacked.texel_offset_for_r_coordinate =
125 nir_src_comp_as_int(instr->src[i].src, 2);
126 break;
127 }
128
129 default:
130 unreachable("unknown texture source");
131 }
132 }
133
134 /* Limit the number of channels returned to both how many the NIR
135 * instruction writes and how many the instruction could produce.
136 */
137 p1_unpacked.return_words_of_texture_data =
138 instr->dest.is_ssa ?
139 nir_ssa_def_components_read(&instr->dest.ssa) :
140 (1 << instr->dest.reg.reg->num_components) - 1;
141
142 uint32_t p0_packed;
143 V3D33_TEXTURE_UNIFORM_PARAMETER_0_CFG_MODE1_pack(NULL,
144 (uint8_t *)&p0_packed,
145 &p0_unpacked);
146
147 uint32_t p1_packed;
148 V3D33_TEXTURE_UNIFORM_PARAMETER_1_CFG_MODE1_pack(NULL,
149 (uint8_t *)&p1_packed,
150 &p1_unpacked);
151 /* Load unit number into the address field, which will be be used by
152 * the driver to decide which texture to put in the actual address
153 * field.
154 */
155 p1_packed |= unit << 5;
156
157 /* There is no native support for GL texture rectangle coordinates, so
158 * we have to rescale from ([0, width], [0, height]) to ([0, 1], [0,
159 * 1]).
160 */
161 if (instr->sampler_dim == GLSL_SAMPLER_DIM_RECT) {
162 coords[0] = vir_FMUL(c, coords[0],
163 vir_uniform(c, QUNIFORM_TEXRECT_SCALE_X,
164 unit));
165 coords[1] = vir_FMUL(c, coords[1],
166 vir_uniform(c, QUNIFORM_TEXRECT_SCALE_Y,
167 unit));
168 }
169
170 int texture_u[] = {
171 vir_get_uniform_index(c, QUNIFORM_TEXTURE_CONFIG_P0_0 + unit, p0_packed),
172 vir_get_uniform_index(c, QUNIFORM_TEXTURE_CONFIG_P1, p1_packed),
173 };
174
175 for (int i = 0; i < next_coord; i++) {
176 struct qreg dst;
177
178 if (i == next_coord - 1)
179 dst = vir_reg(QFILE_MAGIC, V3D_QPU_WADDR_TMUL);
180 else
181 dst = vir_reg(QFILE_MAGIC, V3D_QPU_WADDR_TMU);
182
183 struct qinst *tmu = vir_MOV_dest(c, dst, coords[i]);
184
185 if (i < 2)
186 tmu->uniform = texture_u[i];
187 }
188
189 vir_emit_thrsw(c);
190
191 for (int i = 0; i < 4; i++) {
192 if (p1_unpacked.return_words_of_texture_data & (1 << i))
193 ntq_store_dest(c, &instr->dest, i, vir_LDTMU(c));
194 }
195 }
196