2 * Copyright © 2013 Intel Corporation
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:
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
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
21 * DEALINGS IN THE SOFTWARE.
31 vec4_vs_visitor::emit_prolog()
33 dst_reg sign_recovery_shift
;
34 dst_reg normalize_factor
;
35 dst_reg es3_normalize_factor
;
37 for (int i
= 0; i
< VERT_ATTRIB_MAX
; i
++) {
38 if (vs_prog_data
->inputs_read
& BITFIELD64_BIT(i
)) {
39 uint8_t wa_flags
= key
->gl_attrib_wa_flags
[i
];
42 reg_d
.type
= BRW_REGISTER_TYPE_D
;
44 reg_ud
.type
= BRW_REGISTER_TYPE_UD
;
46 /* Do GL_FIXED rescaling for GLES2.0. Our GL_FIXED attributes
47 * come in as floating point conversions of the integer values.
49 if (wa_flags
& BRW_ATTRIB_WA_COMPONENT_MASK
) {
51 dst
.type
= brw_type_for_base_type(glsl_type::vec4_type
);
52 dst
.writemask
= (1 << (wa_flags
& BRW_ATTRIB_WA_COMPONENT_MASK
)) - 1;
53 emit(MUL(dst
, src_reg(dst
), src_reg(1.0f
/ 65536.0f
)));
56 /* Do sign recovery for 2101010 formats if required. */
57 if (wa_flags
& BRW_ATTRIB_WA_SIGN
) {
58 if (sign_recovery_shift
.file
== BAD_FILE
) {
59 /* shift constant: <22,22,22,30> */
60 sign_recovery_shift
= dst_reg(this, glsl_type::uvec4_type
);
61 emit(MOV(writemask(sign_recovery_shift
, WRITEMASK_XYZ
), src_reg(22u)));
62 emit(MOV(writemask(sign_recovery_shift
, WRITEMASK_W
), src_reg(30u)));
65 emit(SHL(reg_ud
, src_reg(reg_ud
), src_reg(sign_recovery_shift
)));
66 emit(ASR(reg_d
, src_reg(reg_d
), src_reg(sign_recovery_shift
)));
69 /* Apply BGRA swizzle if required. */
70 if (wa_flags
& BRW_ATTRIB_WA_BGRA
) {
71 src_reg temp
= src_reg(reg
);
72 temp
.swizzle
= BRW_SWIZZLE4(2,1,0,3);
76 if (wa_flags
& BRW_ATTRIB_WA_NORMALIZE
) {
77 /* ES 3.0 has different rules for converting signed normalized
78 * fixed-point numbers than desktop GL.
80 if ((wa_flags
& BRW_ATTRIB_WA_SIGN
) && !use_legacy_snorm_formula
) {
81 /* According to equation 2.2 of the ES 3.0 specification,
82 * signed normalization conversion is done by:
86 if (es3_normalize_factor
.file
== BAD_FILE
) {
87 /* mul constant: 1 / (2^(b-1) - 1) */
88 es3_normalize_factor
= dst_reg(this, glsl_type::vec4_type
);
89 emit(MOV(writemask(es3_normalize_factor
, WRITEMASK_XYZ
),
90 src_reg(1.0f
/ ((1<<9) - 1))));
91 emit(MOV(writemask(es3_normalize_factor
, WRITEMASK_W
),
92 src_reg(1.0f
/ ((1<<1) - 1))));
96 dst
.type
= brw_type_for_base_type(glsl_type::vec4_type
);
97 emit(MOV(dst
, src_reg(reg_d
)));
98 emit(MUL(dst
, src_reg(dst
), src_reg(es3_normalize_factor
)));
99 emit_minmax(BRW_CONDITIONAL_GE
, dst
, src_reg(dst
), src_reg(-1.0f
));
101 /* The following equations are from the OpenGL 3.2 specification:
103 * 2.1 unsigned normalization
106 * 2.2 signed normalization
109 * Both of these share a common divisor, which is represented by
110 * "normalize_factor" in the code below.
112 if (normalize_factor
.file
== BAD_FILE
) {
113 /* 1 / (2^b - 1) for b=<10,10,10,2> */
114 normalize_factor
= dst_reg(this, glsl_type::vec4_type
);
115 emit(MOV(writemask(normalize_factor
, WRITEMASK_XYZ
),
116 src_reg(1.0f
/ ((1<<10) - 1))));
117 emit(MOV(writemask(normalize_factor
, WRITEMASK_W
),
118 src_reg(1.0f
/ ((1<<2) - 1))));
122 dst
.type
= brw_type_for_base_type(glsl_type::vec4_type
);
123 emit(MOV(dst
, src_reg((wa_flags
& BRW_ATTRIB_WA_SIGN
) ? reg_d
: reg_ud
)));
125 /* For signed normalization, we want the numerator to be 2c+1. */
126 if (wa_flags
& BRW_ATTRIB_WA_SIGN
) {
127 emit(MUL(dst
, src_reg(dst
), src_reg(2.0f
)));
128 emit(ADD(dst
, src_reg(dst
), src_reg(1.0f
)));
131 emit(MUL(dst
, src_reg(dst
), src_reg(normalize_factor
)));
135 if (wa_flags
& BRW_ATTRIB_WA_SCALE
) {
137 dst
.type
= brw_type_for_base_type(glsl_type::vec4_type
);
138 emit(MOV(dst
, src_reg((wa_flags
& BRW_ATTRIB_WA_SIGN
) ? reg_d
: reg_ud
)));
146 vec4_vs_visitor::make_reg_for_system_value(ir_variable
*ir
)
148 /* VertexID is stored by the VF as the last vertex element, but
149 * we don't represent it with a flag in inputs_read, so we call
150 * it VERT_ATTRIB_MAX, which setup_attributes() picks up on.
152 dst_reg
*reg
= new(mem_ctx
) dst_reg(ATTR
, VERT_ATTRIB_MAX
);
154 switch (ir
->data
.location
) {
155 case SYSTEM_VALUE_BASE_VERTEX
:
156 reg
->writemask
= WRITEMASK_X
;
157 vs_prog_data
->uses_vertexid
= true;
159 case SYSTEM_VALUE_VERTEX_ID
:
160 case SYSTEM_VALUE_VERTEX_ID_ZERO_BASE
:
161 reg
->writemask
= WRITEMASK_Z
;
162 vs_prog_data
->uses_vertexid
= true;
164 case SYSTEM_VALUE_INSTANCE_ID
:
165 reg
->writemask
= WRITEMASK_W
;
166 vs_prog_data
->uses_instanceid
= true;
169 unreachable("not reached");
177 vec4_vs_visitor::emit_urb_write_header(int mrf
)
179 /* No need to do anything for VS; an implied write to this MRF will be
180 * performed by VS_OPCODE_URB_WRITE.
187 vec4_vs_visitor::emit_urb_write_opcode(bool complete
)
189 /* For VS, the URB writes end the thread. */
191 if (INTEL_DEBUG
& DEBUG_SHADER_TIME
)
192 emit_shader_time_end();
195 vec4_instruction
*inst
= emit(VS_OPCODE_URB_WRITE
);
196 inst
->urb_write_flags
= complete
?
197 BRW_URB_WRITE_EOT_COMPLETE
: BRW_URB_WRITE_NO_FLAGS
;
204 vec4_vs_visitor::emit_thread_end()
206 /* For VS, we always end the thread by emitting a single vertex.
207 * emit_urb_write_opcode() will take care of setting the eot flag on the
214 vec4_vs_visitor::vec4_vs_visitor(const struct brw_compiler
*compiler
,
216 const struct brw_vs_prog_key
*key
,
217 struct brw_vs_prog_data
*vs_prog_data
,
218 struct gl_vertex_program
*vp
,
219 struct gl_shader_program
*prog
,
221 int shader_time_index
,
222 bool use_legacy_snorm_formula
)
223 : vec4_visitor(compiler
, log_data
,
224 &vp
->Base
, &key
->base
, &vs_prog_data
->base
, prog
,
226 mem_ctx
, false /* no_spills */,
229 vs_prog_data(vs_prog_data
),
231 use_legacy_snorm_formula(use_legacy_snorm_formula
)
236 } /* namespace brw */