freedreno_guardband.h revision 7ec681f3 
1/*
2 * Copyright © 2020 Valve Corporation
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 * 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:
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 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 FROM,
20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 * SOFTWARE.
22 */
23
24#ifndef __FREEDRENO_GUARDBAND_H__
25#define __FREEDRENO_GUARDBAND_H__
26
27#include <assert.h>
28#include <math.h>
29#include <stdbool.h>
30
31static inline unsigned
32fd_calc_guardband(float offset, float scale, bool is_a3xx)
33{
34   /* On a3xx, the viewport max is 4k and the docs say the max guardband
35    * width is 8k. That is, GRAS cannot handle triangle coordinates more than
36    * 8k, positive or negative. On a4xx+ the viewport width was bumped to
37    * 16k, and so the guardband width was necessarily also bumped. Note that
38    * the numbers here should correspond to
39    * VkPhysicalDeviceLimits::viewportBoundsRange in Vulkan.
40    */
41   const float gb_min = is_a3xx ? -8192. : -32768.;
42   const float gb_max = is_a3xx ? 8191. : 32767.;
43
44   /* Clipping happens in normalized device coordinates, so we have to
45    * transform gb_min and gb_max to ndc using the inverse of the viewport
46    * transform. Avoid flipping min and max by using the absolute value of
47    * the scale.
48    */
49   const float gb_min_ndc = (gb_min - offset) / fabsf(scale);
50   const float gb_max_ndc = (gb_max - offset) / fabsf(scale);
51
52   /* There's only one GB_ADJ field, so presumably the guardband is
53    * [-GB_ADJ, GB_ADJ] like on Radeon. It's always safe to make the
54    * guardband smaller, so we have to take the min to get the largest range
55    * contained in [gb_min_ndc, gb_max_ndc].
56    */
57   const float gb_adj = fminf(-gb_min_ndc, gb_max_ndc);
58
59   /* The viewport should always be contained in the guardband. */
60   assert(gb_adj >= 1.0);
61
62   /* frexp returns an unspecified value if given an infinite value, which
63    * can happen if scale == 0.
64    */
65   if (isinf(gb_adj))
66      return 0x1ff;
67
68   /* Convert gb_adj to 3.6 floating point, rounding down since it's always
69    * safe to make the guard band smaller (but not the other way around!).
70    *
71    * Note: After converting back to a float, the value the blob returns here
72    * is sometimes a little smaller than the value we return. This seems to
73    * happen around the boundary between two different rounded values. For
74    * example, using the a6xx blob:
75    *
76    * min  | width  | unrounded gb_adj | blob result | mesa result
77    * ------------------------------------------------------------
78    * 0    | 510    |          127.498 |        127. |        127.
79    * 0    | 511    |          127.247 |        126. |        127.
80    * 0    | 512    |          126.996 |        126. |        126.
81    *
82    * The guardband must be 32767 wide, since that's what the blob reports
83    * for viewportBoundsRange, so I'm guessing that they're rounding slightly
84    * more conservatively somehow.
85    */
86   int gb_adj_exp;
87   float gb_adj_mantissa = frexpf(gb_adj, &gb_adj_exp);
88   assert(gb_adj_exp > 0);
89
90   /* Round non-representable numbers down to the largest possible number. */
91   if (gb_adj_exp > 8)
92      return 0x1ff;
93
94   return ((gb_adj_exp - 1) << 6) |
95          ((unsigned)truncf(gb_adj_mantissa * (1 << 7)) - (1 << 6));
96}
97
98#endif /* __FREEDRENO_GUARDBAND_H__ */
99