src/gallium/drivers/radeonsi/si_state_msaa.c

   1 /*
   2  * Copyright 2014 Advanced Micro Devices, Inc.
   3  * All Rights Reserved.
   4  *
   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:
  11  *
  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
  14  * Software.
  15  *
  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
  22  * SOFTWARE.
  23  */
  24
  25 #include "si_build_pm4.h"
  26
  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))
  33
  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)
  39
  40 /* The following sample ordering is required by EQAA.
  41  *
  42  * Sample 0 is approx. in the top-left quadrant.
  43  * Sample 1 is approx. in the bottom-right quadrant.
  44  *
  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)
  48  *
  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)
  54  *
  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)
  57  *
  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.
  63  *
  64  * The ordering also allows finding samples in the same vicinity.
  65  *
  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}
  69  *
  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}
  75  *
  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}
  79  *
  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}
  83  */
  84
  85 /* Important note: We have to use the standard DX positions, because
  86  * the primitive discard compute shader relies on them.
  87  */
  88
  89 /* 1x MSAA */
  90 static const uint32_t sample_locs_1x =
  91         FILL_SREG( 0, 0,   0, 0,   0, 0,   0, 0); /* S1, S2, S3 fields are not used by 1x */
  92 static const uint64_t centroid_priority_1x = 0x0000000000000000ull;
  93
  94 /* 2x MSAA (the positions are sorted for EQAA) */
  95 static const uint32_t sample_locs_2x =
  96         FILL_SREG(-4,-4,   4, 4,   0, 0,   0, 0); /* S2 & S3 fields are not used by 2x MSAA */
  97 static const uint64_t centroid_priority_2x = 0x1010101010101010ull;
  98
  99 /* 4x MSAA (the positions are sorted for EQAA) */
 100 static const uint32_t sample_locs_4x =
 101         FILL_SREG(-2,-6,   2, 6,   -6, 2,  6,-2);
 102 static const uint64_t centroid_priority_4x = 0x3210321032103210ull;
 103
 104 /* 8x MSAA (the positions are sorted for EQAA) */
 105 static const uint32_t sample_locs_8x[] = {
 106         FILL_SREG(-3,-5,   5, 1,  -1, 3,   7,-7),
 107         FILL_SREG(-7,-1,   3, 7,  -5, 5,   1,-3),
 108         /* The following are unused by hardware, but we emit them to IBs
 109          * instead of multiple SET_CONTEXT_REG packets. */
 110         0,
 111         0,
 112 };
 113 static const uint64_t centroid_priority_8x = 0x3546012735460127ull;
 114
 115 /* 16x MSAA (the positions are sorted for EQAA) */
 116 static const uint32_t sample_locs_16x[] = {
 117         FILL_SREG(-5,-2,   5, 3,  -2, 6,   3,-5),
 118         FILL_SREG(-4,-6,   1, 1,  -6, 4,   7,-4),
 119         FILL_SREG(-1,-3,   6, 7,  -3, 2,   0,-7),
 120         FILL_SREG(-7,-8,   2, 5,  -8, 0,   4,-1),
 121 };
 122 static const uint64_t centroid_priority_16x = 0xc97e64b231d0fa85ull;
 123
 124 static void si_get_sample_position(struct pipe_context *ctx, unsigned sample_count,
 125                                    unsigned sample_index, float *out_value)
 126 {
 127         const uint32_t *sample_locs;
 128
 129         switch (sample_count) {
 130         case 1:
 131         default:
 132                 sample_locs = &sample_locs_1x;
 133                 break;
 134         case 2:
 135                 sample_locs = &sample_locs_2x;
 136                 break;
 137         case 4:
 138                 sample_locs = &sample_locs_4x;
 139                 break;
 140         case 8:
 141                 sample_locs = sample_locs_8x;
 142                 break;
 143         case 16:
 144                 sample_locs = sample_locs_16x;
 145                 break;
 146         }
 147
 148         out_value[0] = (GET_SX(sample_locs, sample_index) + 8) / 16.0f;
 149         out_value[1] = (GET_SY(sample_locs, sample_index) + 8) / 16.0f;
 150 }
 151
 152 static void si_emit_max_4_sample_locs(struct radeon_cmdbuf *cs,
 153                                       uint64_t centroid_priority,
 154                                       uint32_t sample_locs)
 155 {
 156         radeon_set_context_reg_seq(cs, R_028BD4_PA_SC_CENTROID_PRIORITY_0, 2);
 157         radeon_emit(cs, centroid_priority);
 158         radeon_emit(cs, centroid_priority >> 32);
 159         radeon_set_context_reg(cs, R_028BF8_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0, sample_locs);
 160         radeon_set_context_reg(cs, R_028C08_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y0_0, sample_locs);
 161         radeon_set_context_reg(cs, R_028C18_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y1_0, sample_locs);
 162         radeon_set_context_reg(cs, R_028C28_PA_SC_AA_SAMPLE_LOCS_PIXEL_X1Y1_0, sample_locs);
 163 }
 164
 165 static void si_emit_max_16_sample_locs(struct radeon_cmdbuf *cs,
 166                                        uint64_t centroid_priority,
 167                                        const uint32_t *sample_locs,
 168                                        unsigned num_samples)
 169 {
 170         radeon_set_context_reg_seq(cs, R_028BD4_PA_SC_CENTROID_PRIORITY_0, 2);
 171         radeon_emit(cs, centroid_priority);
 172         radeon_emit(cs, centroid_priority >> 32);
 173         radeon_set_context_reg_seq(cs, R_028BF8_PA_SC_AA_SAMPLE_LOCS_PIXEL_X0Y0_0,
 174                                    num_samples == 8 ? 14 : 16);
 175         radeon_emit_array(cs, sample_locs, 4);
 176         radeon_emit_array(cs, sample_locs, 4);
 177         radeon_emit_array(cs, sample_locs, 4);
 178         radeon_emit_array(cs, sample_locs, num_samples == 8 ? 2 : 4);
 179 }
 180
 181 void si_emit_sample_locations(struct radeon_cmdbuf *cs, int nr_samples)
 182 {
 183         switch (nr_samples) {
 184         default:
 185         case 1:
 186                 si_emit_max_4_sample_locs(cs, centroid_priority_1x, sample_locs_1x);
 187                 break;
 188         case 2:
 189                 si_emit_max_4_sample_locs(cs, centroid_priority_2x, sample_locs_2x);
 190                 break;
 191         case 4:
 192                 si_emit_max_4_sample_locs(cs, centroid_priority_4x, sample_locs_4x);
 193                 break;
 194         case 8:
 195                 si_emit_max_16_sample_locs(cs, centroid_priority_8x, sample_locs_8x, 8);
 196                 break;
 197         case 16:
 198                 si_emit_max_16_sample_locs(cs, centroid_priority_16x, sample_locs_16x, 16);
 199                 break;
 200         }
 201 }
 202
 203 void si_init_msaa_functions(struct si_context *sctx)
 204 {
 205         int i;
 206
 207         sctx->b.get_sample_position = si_get_sample_position;
 208
 209         si_get_sample_position(&sctx->b, 1, 0, sctx->sample_positions.x1[0]);
 210
 211         for (i = 0; i < 2; i++)
 212                 si_get_sample_position(&sctx->b, 2, i, sctx->sample_positions.x2[i]);
 213         for (i = 0; i < 4; i++)
 214                 si_get_sample_position(&sctx->b, 4, i, sctx->sample_positions.x4[i]);
 215         for (i = 0; i < 8; i++)
 216                 si_get_sample_position(&sctx->b, 8, i, sctx->sample_positions.x8[i]);
 217         for (i = 0; i < 16; i++)
 218                 si_get_sample_position(&sctx->b, 16, i, sctx->sample_positions.x16[i]);
 219 }