2 * Copyright 2014 Advanced Micro Devices, Inc.
5 * Permission is hereby granted, free of charge, to any person obtaining a
6 * copy of this software and associated documentation files (the "Software"),
7 * to deal in the Software without restriction, including without limitation
8 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
9 * and/or sell copies of the Software, and to permit persons to whom the
10 * Software is furnished to do so, subject to the following conditions:
12 * The above copyright notice and this permission notice (including the next
13 * paragraph) shall be included in all copies or substantial portions of the
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
25 #include "si_build_pm4.h"
27 /* For MSAA sample positions. */
28 #define FILL_SREG(s0x, s0y, s1x, s1y, s2x, s2y, s3x, s3y) \
29 ((((unsigned)(s0x) & 0xf) << 0) | (((unsigned)(s0y) & 0xf) << 4) | \
30 (((unsigned)(s1x) & 0xf) << 8) | (((unsigned)(s1y) & 0xf) << 12) | \
31 (((unsigned)(s2x) & 0xf) << 16) | (((unsigned)(s2y) & 0xf) << 20) | \
32 (((unsigned)(s3x) & 0xf) << 24) | (((unsigned)(s3y) & 0xf) << 28))
34 /* For obtaining location coordinates from registers */
35 #define SEXT4(x) ((int)((x) | ((x) & 0x8 ? 0xfffffff0 : 0)))
36 #define GET_SFIELD(reg, index) SEXT4(((reg) >> ((index) * 4)) & 0xf)
37 #define GET_SX(reg, index) GET_SFIELD((reg)[(index) / 4], ((index) % 4) * 2)
38 #define GET_SY(reg, index) GET_SFIELD((reg)[(index) / 4], ((index) % 4) * 2 + 1)
40 /* The following sample ordering is required by EQAA.
42 * Sample 0 is approx. in the top-left quadrant.
43 * Sample 1 is approx. in the bottom-right quadrant.
45 * Sample 2 is approx. in the bottom-left quadrant.
46 * Sample 3 is approx. in the top-right quadrant.
47 * (sample I={2,3} adds more detail to the vicinity of sample I-2)
49 * Sample 4 is approx. in the same quadrant as sample 0. (top-left)
50 * Sample 5 is approx. in the same quadrant as sample 1. (bottom-right)
51 * Sample 6 is approx. in the same quadrant as sample 2. (bottom-left)
52 * Sample 7 is approx. in the same quadrant as sample 3. (top-right)
53 * (sample I={4,5,6,7} adds more detail to the vicinity of sample I-4)
55 * The next 8 samples add more detail to the vicinity of the previous samples.
56 * (sample I (I >= 8) adds more detail to the vicinity of sample I-8)
58 * The ordering is specified such that:
59 * If we take the first 2 samples, we should get good 2x MSAA.
60 * If we add 2 more samples, we should get good 4x MSAA with the same sample locations.
61 * If we add 4 more samples, we should get good 8x MSAA with the same sample locations.
62 * If we add 8 more samples, we should get perfect 16x MSAA with the same sample locations.
64 * The ordering also allows finding samples in the same vicinity.
66 * Group N of 2 samples in the same vicinity in 16x MSAA: {N,N+8}
67 * Group N of 2 samples in the same vicinity in 8x MSAA: {N,N+4}
68 * Group N of 2 samples in the same vicinity in 4x MSAA: {N,N+2}
70 * Groups of 4 samples in the same vicinity in 16x MSAA:
71 * Top left: {0,4,8,12}
72 * Bottom right: {1,5,9,13}
73 * Bottom left: {2,6,10,14}
74 * Top right: {3,7,11,15}
76 * Groups of 4 samples in the same vicinity in 8x MSAA:
77 * Left half: {0,2,4,6}
78 * Right half: {1,3,5,7}
80 * Groups of 8 samples in the same vicinity in 16x MSAA:
81 * Left half: {0,2,4,6,8,10,12,14}
82 * Right half: {1,3,5,7,9,11,13,15}
86 static const uint32_t sample_locs_1x
=
87 FILL_SREG( 0, 0, 0, 0, 0, 0, 0, 0); /* S1, S2, S3 fields are not used by 1x */
88 static const uint64_t centroid_priority_1x
= 0x0000000000000000ull
;
91 static const uint32_t sample_locs_2x
=
92 FILL_SREG(-4,-4, 4, 4, 0, 0, 0, 0); /* S2 & S3 fields are not used by 2x MSAA */
93 static const uint64_t centroid_priority_2x
= 0x1010101010101010ull
;
95 /* 4x, 8x, and 16x MSAA
96 * - The first 4 locations happen to be optimal for 4x MSAA, better than
97 * the standard DX 4x locations.
98 * - The first 8 locations happen to be almost as good as 8x DX locations,
99 * but the DX locations are horrible for worst-case EQAA 8s4f and 8s2f.
101 static const uint32_t sample_locs_4x_8x_16x
[] = {
102 FILL_SREG(-5,-2, 5, 3, -2, 6, 3,-5),
103 FILL_SREG(-6,-7, 1, 1, -6, 4, 7,-3),
104 FILL_SREG(-1,-3, 6, 7, -3, 2, 0,-7),
105 FILL_SREG(-4,-6, 2, 5, -8, 0, 4,-1),
107 static const uint64_t centroid_priority_4x
= 0x2310231023102310ull
;
108 static const uint64_t centroid_priority_8x
= 0x4762310547623105ull
;
109 static const uint64_t centroid_priority_16x
= 0x49e7c6b231d0fa85ull
;
111 static void si_get_sample_position(struct pipe_context
*ctx
, unsigned sample_count
,
112 unsigned sample_index
, float *out_value
)
114 const uint32_t *sample_locs
;
116 switch (sample_count
) {
119 sample_locs
= &sample_locs_1x
;
122 sample_locs
= &sample_locs_2x
;
127 sample_locs
= sample_locs_4x_8x_16x
;
131 out_value
[0] = (GET_SX(sample_locs
, sample_index
) + 8) / 16.0f
;
132 out_value
[1] = (GET_SY(sample_locs
, sample_index
) + 8) / 16.0f
;
135 static void si_emit_max_4_sample_locs(struct radeon_cmdbuf
*cs
,
136 uint64_t centroid_priority
,
137 uint32_t sample_locs
)
139 radeon_set_context_reg_seq(cs
, R_028BD4_PA_SC_CENTROID_PRIORITY_0
, 2);
140 radeon_emit(cs
, centroid_priority
);
141 radeon_emit(cs
, centroid_priority
>> 32);
142 radeon_set_context_reg(cs
, R_028BF8_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0
, sample_locs
);
143 radeon_set_context_reg(cs
, R_028C08_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_0
, sample_locs
);
144 radeon_set_context_reg(cs
, R_028C18_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_0
, sample_locs
);
145 radeon_set_context_reg(cs
, R_028C28_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_0
, sample_locs
);
148 static void si_emit_max_16_sample_locs(struct radeon_cmdbuf
*cs
,
149 uint64_t centroid_priority
,
150 const uint32_t *sample_locs
,
151 unsigned num_samples
)
153 radeon_set_context_reg_seq(cs
, R_028BD4_PA_SC_CENTROID_PRIORITY_0
, 2);
154 radeon_emit(cs
, centroid_priority
);
155 radeon_emit(cs
, centroid_priority
>> 32);
156 radeon_set_context_reg_seq(cs
, R_028BF8_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0
,
157 num_samples
== 8 ? 14 : 16);
158 radeon_emit_array(cs
, sample_locs
, 4);
159 radeon_emit_array(cs
, sample_locs
, 4);
160 radeon_emit_array(cs
, sample_locs
, 4);
161 radeon_emit_array(cs
, sample_locs
, num_samples
== 8 ? 2 : 4);
164 void si_emit_sample_locations(struct radeon_cmdbuf
*cs
, int nr_samples
)
166 switch (nr_samples
) {
169 si_emit_max_4_sample_locs(cs
, centroid_priority_1x
, sample_locs_1x
);
172 si_emit_max_4_sample_locs(cs
, centroid_priority_2x
, sample_locs_2x
);
175 si_emit_max_4_sample_locs(cs
, centroid_priority_4x
, sample_locs_4x_8x_16x
[0]);
178 si_emit_max_16_sample_locs(cs
, centroid_priority_8x
, sample_locs_4x_8x_16x
, 8);
181 si_emit_max_16_sample_locs(cs
, centroid_priority_16x
, sample_locs_4x_8x_16x
, 16);
186 void si_init_msaa_functions(struct si_context
*sctx
)
190 sctx
->b
.get_sample_position
= si_get_sample_position
;
192 si_get_sample_position(&sctx
->b
, 1, 0, sctx
->sample_locations_1x
[0]);
194 for (i
= 0; i
< 2; i
++)
195 si_get_sample_position(&sctx
->b
, 2, i
, sctx
->sample_locations_2x
[i
]);
196 for (i
= 0; i
< 4; i
++)
197 si_get_sample_position(&sctx
->b
, 4, i
, sctx
->sample_locations_4x
[i
]);
198 for (i
= 0; i
< 8; i
++)
199 si_get_sample_position(&sctx
->b
, 8, i
, sctx
->sample_locations_8x
[i
]);
200 for (i
= 0; i
< 16; i
++)
201 si_get_sample_position(&sctx
->b
, 16, i
, sctx
->sample_locations_16x
[i
]);