1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
|
/**************************************************************************
*
* Copyright 2010 VMware, Inc.
* All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the
* "Software"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sub license, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM,
* DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
* OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
* USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* The above copyright notice and this permission notice (including the
* next paragraph) shall be included in all copies or substantial portions
* of the Software.
*
**************************************************************************/
#include "lp_bld_type.h"
#include "lp_bld_arit.h"
#include "lp_bld_const.h"
#include "lp_bld_swizzle.h"
#include "lp_bld_quad.h"
#include "lp_bld_pack.h"
static const unsigned char
swizzle_left[4] = {
LP_BLD_QUAD_TOP_LEFT, LP_BLD_QUAD_TOP_LEFT,
LP_BLD_QUAD_BOTTOM_LEFT, LP_BLD_QUAD_BOTTOM_LEFT
};
static const unsigned char
swizzle_right[4] = {
LP_BLD_QUAD_TOP_RIGHT, LP_BLD_QUAD_TOP_RIGHT,
LP_BLD_QUAD_BOTTOM_RIGHT, LP_BLD_QUAD_BOTTOM_RIGHT
};
static const unsigned char
swizzle_top[4] = {
LP_BLD_QUAD_TOP_LEFT, LP_BLD_QUAD_TOP_RIGHT,
LP_BLD_QUAD_TOP_LEFT, LP_BLD_QUAD_TOP_RIGHT
};
static const unsigned char
swizzle_bottom[4] = {
LP_BLD_QUAD_BOTTOM_LEFT, LP_BLD_QUAD_BOTTOM_RIGHT,
LP_BLD_QUAD_BOTTOM_LEFT, LP_BLD_QUAD_BOTTOM_RIGHT
};
LLVMValueRef
lp_build_ddx(struct lp_build_context *bld,
LLVMValueRef a)
{
LLVMValueRef a_left = lp_build_swizzle_aos(bld, a, swizzle_left);
LLVMValueRef a_right = lp_build_swizzle_aos(bld, a, swizzle_right);
return lp_build_sub(bld, a_right, a_left);
}
LLVMValueRef
lp_build_ddy(struct lp_build_context *bld,
LLVMValueRef a)
{
LLVMValueRef a_top = lp_build_swizzle_aos(bld, a, swizzle_top);
LLVMValueRef a_bottom = lp_build_swizzle_aos(bld, a, swizzle_bottom);
return lp_build_sub(bld, a_bottom, a_top);
}
/*
* Helper for building packed ddx/ddy vector for one coord (scalar per quad
* values). The vector will look like this (8-wide):
* dr1dx _____ -dr1dy _____ dr2dx _____ -dr2dy _____
* This only requires one shuffle instead of two for more straightforward packing.
*/
LLVMValueRef
lp_build_packed_ddx_ddy_onecoord(struct lp_build_context *bld,
LLVMValueRef a)
{
struct gallivm_state *gallivm = bld->gallivm;
LLVMBuilderRef builder = gallivm->builder;
LLVMValueRef vec1, vec2;
/* use aos swizzle helper */
static const unsigned char swizzle1[] = { /* no-op swizzle */
LP_BLD_QUAD_TOP_LEFT, LP_BLD_SWIZZLE_DONTCARE,
LP_BLD_QUAD_BOTTOM_LEFT, LP_BLD_SWIZZLE_DONTCARE
};
static const unsigned char swizzle2[] = {
LP_BLD_QUAD_TOP_RIGHT, LP_BLD_SWIZZLE_DONTCARE,
LP_BLD_QUAD_TOP_LEFT, LP_BLD_SWIZZLE_DONTCARE
};
vec1 = lp_build_swizzle_aos(bld, a, swizzle1);
vec2 = lp_build_swizzle_aos(bld, a, swizzle2);
if (bld->type.floating)
return LLVMBuildFSub(builder, vec2, vec1, "ddxddy");
else
return LLVMBuildSub(builder, vec2, vec1, "ddxddy");
}
/*
* Helper for building packed ddx/ddy vector for one coord (scalar per quad
* values). The vector will look like this (8-wide):
* ds1dx ds1dy dt1dx dt1dy ds2dx ds2dy dt2dx dt2dy
* This only needs 2 (v)shufps.
*/
LLVMValueRef
lp_build_packed_ddx_ddy_twocoord(struct lp_build_context *bld,
LLVMValueRef a, LLVMValueRef b)
{
struct gallivm_state *gallivm = bld->gallivm;
LLVMBuilderRef builder = gallivm->builder;
LLVMValueRef shuffles1[LP_MAX_VECTOR_LENGTH/4];
LLVMValueRef shuffles2[LP_MAX_VECTOR_LENGTH/4];
LLVMValueRef vec1, vec2;
unsigned length, num_quads, i;
/* XXX: do hsub version */
length = bld->type.length;
num_quads = length / 4;
for (i = 0; i < num_quads; i++) {
unsigned s1 = 4 * i;
unsigned s2 = 4 * i + length;
shuffles1[4*i + 0] = lp_build_const_int32(gallivm, LP_BLD_QUAD_TOP_LEFT + s1);
shuffles1[4*i + 1] = lp_build_const_int32(gallivm, LP_BLD_QUAD_TOP_LEFT + s1);
shuffles1[4*i + 2] = lp_build_const_int32(gallivm, LP_BLD_QUAD_TOP_LEFT + s2);
shuffles1[4*i + 3] = lp_build_const_int32(gallivm, LP_BLD_QUAD_TOP_LEFT + s2);
shuffles2[4*i + 0] = lp_build_const_int32(gallivm, LP_BLD_QUAD_TOP_RIGHT + s1);
shuffles2[4*i + 1] = lp_build_const_int32(gallivm, LP_BLD_QUAD_BOTTOM_LEFT + s1);
shuffles2[4*i + 2] = lp_build_const_int32(gallivm, LP_BLD_QUAD_TOP_RIGHT + s2);
shuffles2[4*i + 3] = lp_build_const_int32(gallivm, LP_BLD_QUAD_BOTTOM_LEFT + s2);
}
vec1 = LLVMBuildShuffleVector(builder, a, b,
LLVMConstVector(shuffles1, length), "");
vec2 = LLVMBuildShuffleVector(builder, a, b,
LLVMConstVector(shuffles2, length), "");
if (bld->type.floating)
return LLVMBuildFSub(builder, vec2, vec1, "ddxddyddxddy");
else
return LLVMBuildSub(builder, vec2, vec1, "ddxddyddxddy");
}
/**
* Twiddle from quad format to row format
*
* src0 src1
* ######### ######### #################
* # 0 | 1 # # 4 | 5 # # 0 | 1 | 4 | 5 # src0
* #---+---# #---+---# -> #################
* # 2 | 3 # # 6 | 7 # # 2 | 3 | 6 | 7 # src1
* ######### ######### #################
*
*/
void
lp_bld_quad_twiddle(struct gallivm_state *gallivm,
struct lp_type lp_dst_type,
const LLVMValueRef* src,
unsigned src_count,
LLVMValueRef* dst)
{
LLVMBuilderRef builder = gallivm->builder;
LLVMTypeRef dst_type_ref;
LLVMTypeRef type2_ref;
struct lp_type type2;
unsigned i;
assert((src_count % 2) == 0);
/* Create a type with only 2 elements */
type2 = lp_dst_type;
type2.width = (lp_dst_type.width * lp_dst_type.length) / 2;
type2.length = 2;
type2.floating = 0;
type2_ref = lp_build_vec_type(gallivm, type2);
dst_type_ref = lp_build_vec_type(gallivm, lp_dst_type);
for (i = 0; i < src_count; i += 2) {
LLVMValueRef src0, src1;
src0 = LLVMBuildBitCast(builder, src[i + 0], type2_ref, "");
src1 = LLVMBuildBitCast(builder, src[i + 1], type2_ref, "");
dst[i + 0] = lp_build_interleave2(gallivm, type2, src0, src1, 0);
dst[i + 1] = lp_build_interleave2(gallivm, type2, src0, src1, 1);
dst[i + 0] = LLVMBuildBitCast(builder, dst[i + 0], dst_type_ref, "");
dst[i + 1] = LLVMBuildBitCast(builder, dst[i + 1], dst_type_ref, "");
}
}
|
