acb30409428322835fe29eeeaba7e337e2ab3713
[mesa.git] / src / gallium / drivers / r600 / eg_asm.c
1 /*
2 * Copyright 2010 Jerome Glisse <glisse@freedesktop.org>
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 * on the rights to use, copy, modify, merge, publish, distribute, sub
8 * license, and/or sell copies of the Software, and to permit persons to whom
9 * the 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 NON-INFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
19 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
20 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
21 * USE OR OTHER DEALINGS IN THE SOFTWARE.
22 */
23 #include "r600_pipe.h"
24 #include "r600_opcodes.h"
25 #include "r600_shader.h"
26
27 #include "util/u_memory.h"
28 #include "eg_sq.h"
29 #include <errno.h>
30
31 int eg_bytecode_cf_build(struct r600_bytecode *bc, struct r600_bytecode_cf *cf)
32 {
33 unsigned id = cf->id;
34
35 if (cf->op == CF_NATIVE) {
36 bc->bytecode[id++] = cf->isa[0];
37 bc->bytecode[id++] = cf->isa[1];
38 } else {
39 const struct cf_op_info *cfop = r600_isa_cf(cf->op);
40 unsigned opcode = r600_isa_cf_opcode(bc->isa->hw_class, cf->op);
41 if (cfop->flags & CF_ALU) { /* ALU clauses */
42
43 /* prepend ALU_EXTENDED if we need more than 2 kcache sets */
44 if (cf->eg_alu_extended) {
45 bc->bytecode[id++] =
46 S_SQ_CF_ALU_WORD0_EXT_KCACHE_BANK_INDEX_MODE0(V_SQ_CF_INDEX_NONE) |
47 S_SQ_CF_ALU_WORD0_EXT_KCACHE_BANK_INDEX_MODE1(V_SQ_CF_INDEX_NONE) |
48 S_SQ_CF_ALU_WORD0_EXT_KCACHE_BANK_INDEX_MODE2(V_SQ_CF_INDEX_NONE) |
49 S_SQ_CF_ALU_WORD0_EXT_KCACHE_BANK_INDEX_MODE3(V_SQ_CF_INDEX_NONE) |
50 S_SQ_CF_ALU_WORD0_EXT_KCACHE_BANK2(cf->kcache[2].bank) |
51 S_SQ_CF_ALU_WORD0_EXT_KCACHE_BANK3(cf->kcache[3].bank) |
52 S_SQ_CF_ALU_WORD0_EXT_KCACHE_MODE2(cf->kcache[2].mode);
53 bc->bytecode[id++] =
54 S_SQ_CF_ALU_WORD1_EXT_CF_INST(
55 r600_isa_cf_opcode(bc->isa->hw_class, CF_OP_ALU_EXT)) |
56 S_SQ_CF_ALU_WORD1_EXT_KCACHE_MODE3(cf->kcache[3].mode) |
57 S_SQ_CF_ALU_WORD1_EXT_KCACHE_ADDR2(cf->kcache[2].addr) |
58 S_SQ_CF_ALU_WORD1_EXT_KCACHE_ADDR3(cf->kcache[3].addr) |
59 S_SQ_CF_ALU_WORD1_EXT_BARRIER(1);
60 }
61 bc->bytecode[id++] = S_SQ_CF_ALU_WORD0_ADDR(cf->addr >> 1) |
62 S_SQ_CF_ALU_WORD0_KCACHE_MODE0(cf->kcache[0].mode) |
63 S_SQ_CF_ALU_WORD0_KCACHE_BANK0(cf->kcache[0].bank) |
64 S_SQ_CF_ALU_WORD0_KCACHE_BANK1(cf->kcache[1].bank);
65 bc->bytecode[id++] = S_SQ_CF_ALU_WORD1_CF_INST(opcode) |
66 S_SQ_CF_ALU_WORD1_KCACHE_MODE1(cf->kcache[1].mode) |
67 S_SQ_CF_ALU_WORD1_KCACHE_ADDR0(cf->kcache[0].addr) |
68 S_SQ_CF_ALU_WORD1_KCACHE_ADDR1(cf->kcache[1].addr) |
69 S_SQ_CF_ALU_WORD1_BARRIER(1) |
70 S_SQ_CF_ALU_WORD1_COUNT((cf->ndw / 2) - 1);
71 } else if (cfop->flags & CF_CLAUSE) {
72 /* CF_TEX/VTX (CF_ALU already handled above) */
73 bc->bytecode[id++] = S_SQ_CF_WORD0_ADDR(cf->addr >> 1);
74 bc->bytecode[id++] = S_SQ_CF_WORD1_CF_INST(opcode) |
75 S_SQ_CF_WORD1_BARRIER(1) |
76 S_SQ_CF_WORD1_COUNT((cf->ndw / 4) - 1);
77 } else if (cfop->flags & CF_EXP) {
78 /* EXPORT instructions */
79 bc->bytecode[id++] = S_SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR(cf->output.gpr) |
80 S_SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE(cf->output.elem_size) |
81 S_SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE(cf->output.array_base) |
82 S_SQ_CF_ALLOC_EXPORT_WORD0_TYPE(cf->output.type) |
83 S_SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR(cf->output.index_gpr);
84 bc->bytecode[id] =
85 S_SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT(cf->output.burst_count - 1) |
86 S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X(cf->output.swizzle_x) |
87 S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y(cf->output.swizzle_y) |
88 S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z(cf->output.swizzle_z) |
89 S_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W(cf->output.swizzle_w) |
90 S_SQ_CF_ALLOC_EXPORT_WORD1_BARRIER(cf->barrier) |
91 S_SQ_CF_ALLOC_EXPORT_WORD1_CF_INST(opcode);
92
93 if (bc->chip_class == EVERGREEN) /* no EOP on cayman */
94 bc->bytecode[id] |= S_SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM(cf->end_of_program);
95 id++;
96 } else if (cfop->flags & CF_MEM) {
97 /* MEM_STREAM, MEM_RING instructions */
98 bc->bytecode[id++] = S_SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR(cf->output.gpr) |
99 S_SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE(cf->output.elem_size) |
100 S_SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE(cf->output.array_base) |
101 S_SQ_CF_ALLOC_EXPORT_WORD0_TYPE(cf->output.type) |
102 S_SQ_CF_ALLOC_EXPORT_WORD0_INDEX_GPR(cf->output.index_gpr);
103 bc->bytecode[id] = S_SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT(cf->output.burst_count - 1) |
104 S_SQ_CF_ALLOC_EXPORT_WORD1_BARRIER(cf->barrier) |
105 S_SQ_CF_ALLOC_EXPORT_WORD1_CF_INST(opcode) |
106 S_SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK(cf->output.comp_mask) |
107 S_SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE(cf->output.array_size);
108 if (bc->chip_class == EVERGREEN) /* no EOP on cayman */
109 bc->bytecode[id] |= S_SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM(cf->end_of_program);
110 id++;
111 } else {
112 /* other instructions */
113 bc->bytecode[id++] = S_SQ_CF_WORD0_ADDR(cf->cf_addr >> 1);
114 bc->bytecode[id++] = S_SQ_CF_WORD1_CF_INST(opcode)|
115 S_SQ_CF_WORD1_BARRIER(1) |
116 S_SQ_CF_WORD1_COND(cf->cond) |
117 S_SQ_CF_WORD1_POP_COUNT(cf->pop_count) |
118 S_SQ_CF_WORD1_END_OF_PROGRAM(cf->end_of_program);
119 }
120 }
121 return 0;
122 }
123
124 #if 0
125 void eg_bytecode_export_read(struct r600_bytecode *bc,
126 struct r600_bytecode_output *output, uint32_t word0, uint32_t word1)
127 {
128 output->array_base = G_SQ_CF_ALLOC_EXPORT_WORD0_ARRAY_BASE(word0);
129 output->type = G_SQ_CF_ALLOC_EXPORT_WORD0_TYPE(word0);
130 output->gpr = G_SQ_CF_ALLOC_EXPORT_WORD0_RW_GPR(word0);
131 output->elem_size = G_SQ_CF_ALLOC_EXPORT_WORD0_ELEM_SIZE(word0);
132
133 output->swizzle_x = G_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_X(word1);
134 output->swizzle_y = G_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Y(word1);
135 output->swizzle_z = G_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_Z(word1);
136 output->swizzle_w = G_SQ_CF_ALLOC_EXPORT_WORD1_SWIZ_SEL_W(word1);
137 output->burst_count = G_SQ_CF_ALLOC_EXPORT_WORD1_BURST_COUNT(word1);
138 output->end_of_program = G_SQ_CF_ALLOC_EXPORT_WORD1_END_OF_PROGRAM(word1);
139 output->op = r600_isa_cf_by_opcode(bc->isa,
140 G_SQ_CF_ALLOC_EXPORT_WORD1_CF_INST(word1), /* is_cf_alu = */ 0 );
141 output->barrier = G_SQ_CF_ALLOC_EXPORT_WORD1_BARRIER(word1);
142 output->array_size = G_SQ_CF_ALLOC_EXPORT_WORD1_BUF_ARRAY_SIZE(word1);
143 output->comp_mask = G_SQ_CF_ALLOC_EXPORT_WORD1_BUF_COMP_MASK(word1);
144 }
145 #endif