• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /*
2  * Copyright (C) 2019 Collabora, Ltd.
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  * Authors:
24  *   Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
25  */
26 
27 #include "kmod/pan_kmod.h"
28 #include "panfrost/util/pan_ir.h"
29 #include "pan_props.h"
30 
31 #include <genxml/gen_macros.h>
32 
33 /* Fixed "minimum revisions" */
34 #define NO_ANISO  (~0)
35 #define HAS_ANISO (0)
36 
37 #define MODEL(gpu_id_, shortname, counters_, min_rev_anisotropic_, tib_size_,  \
38               quirks_)                                                         \
39    {                                                                           \
40       .gpu_id = gpu_id_, .name = "Mali-" shortname " (Panfrost)",              \
41       .performance_counters = counters_,                                       \
42       .min_rev_anisotropic = min_rev_anisotropic_,                             \
43       .tilebuffer_size = tib_size_, .quirks = quirks_,                         \
44    }
45 
46 /* Table of supported Mali GPUs */
47 /* clang-format off */
48 const struct panfrost_model panfrost_model_list[] = {
49         MODEL(0x620, "T620",    "T62x", NO_ANISO,          8192, {}),
50         MODEL(0x720, "T720",    "T72x", NO_ANISO,          8192, { .no_hierarchical_tiling = true }),
51         MODEL(0x750, "T760",    "T76x", NO_ANISO,          8192, {}),
52         MODEL(0x820, "T820",    "T82x", NO_ANISO,          8192, { .no_hierarchical_tiling = true }),
53         MODEL(0x830, "T830",    "T83x", NO_ANISO,          8192, { .no_hierarchical_tiling = true }),
54         MODEL(0x860, "T860",    "T86x", NO_ANISO,          8192, {}),
55         MODEL(0x880, "T880",    "T88x", NO_ANISO,          8192, {}),
56 
57         MODEL(0x6000, "G71",    "TMIx", NO_ANISO,          8192, {}),
58         MODEL(0x6221, "G72",    "THEx", 0x0030 /* r0p3 */, 16384, {}),
59         MODEL(0x7090, "G51",    "TSIx", 0x1010 /* r1p1 */, 16384, {}),
60         MODEL(0x7093, "G31",    "TDVx", HAS_ANISO,         16384, {}),
61         MODEL(0x7211, "G76",    "TNOx", HAS_ANISO,         16384, {}),
62         MODEL(0x7212, "G52",    "TGOx", HAS_ANISO,         16384, {}),
63         MODEL(0x7402, "G52 r1", "TGOx", HAS_ANISO,         16384, {}),
64         MODEL(0x9091, "G57",    "TNAx", HAS_ANISO,         16384, {}),
65         MODEL(0x9093, "G57",    "TNAx", HAS_ANISO,         16384, {}),
66 };
67 /* clang-format on */
68 
69 #undef NO_ANISO
70 #undef HAS_ANISO
71 #undef MODEL
72 
73 /*
74  * Look up a supported model by its GPU ID, or return NULL if the model is not
75  * supported at this time.
76  */
77 const struct panfrost_model *
panfrost_get_model(uint32_t gpu_id)78 panfrost_get_model(uint32_t gpu_id)
79 {
80    for (unsigned i = 0; i < ARRAY_SIZE(panfrost_model_list); ++i) {
81       if (panfrost_model_list[i].gpu_id == gpu_id)
82          return &panfrost_model_list[i];
83    }
84 
85    return NULL;
86 }
87 
88 unsigned
panfrost_query_l2_slices(const struct pan_kmod_dev_props * props)89 panfrost_query_l2_slices(const struct pan_kmod_dev_props *props)
90 {
91    /* L2_SLICES is MEM_FEATURES[11:8] minus(1) */
92    return ((props->mem_features >> 8) & 0xF) + 1;
93 }
94 
95 struct panfrost_tiler_features
panfrost_query_tiler_features(const struct pan_kmod_dev_props * props)96 panfrost_query_tiler_features(const struct pan_kmod_dev_props *props)
97 {
98    /* Default value (2^9 bytes and 8 levels) to match old behaviour */
99    uint32_t raw = props->tiler_features;
100 
101    /* Bin size is log2 in the first byte, max levels in the second byte */
102    return (struct panfrost_tiler_features){
103       .bin_size = (1 << (raw & BITFIELD_MASK(5))),
104       .max_levels = (raw >> 8) & BITFIELD_MASK(4),
105    };
106 }
107 
108 unsigned
panfrost_query_core_count(const struct pan_kmod_dev_props * props,unsigned * core_id_range)109 panfrost_query_core_count(const struct pan_kmod_dev_props *props,
110                           unsigned *core_id_range)
111 {
112    /* On older kernels, worst-case to 16 cores */
113 
114    unsigned mask = props->shader_present;
115 
116    /* Some cores might be absent. In some cases, we care
117     * about the range of core IDs (that is, the greatest core ID + 1). If
118     * the core mask is contiguous, this equals the core count.
119     */
120    *core_id_range = util_last_bit(mask);
121 
122    /* The actual core count skips overs the gaps */
123    return util_bitcount(mask);
124 }
125 
126 unsigned
panfrost_query_thread_tls_alloc(const struct pan_kmod_dev_props * props)127 panfrost_query_thread_tls_alloc(const struct pan_kmod_dev_props *props)
128 {
129    unsigned tls = props->thread_tls_alloc;
130 
131    return (tls > 0)
132              ? tls
133              : panfrost_max_thread_count(pan_arch(props->gpu_prod_id), 0);
134 }
135 
136 uint32_t
panfrost_query_compressed_formats(const struct pan_kmod_dev_props * props)137 panfrost_query_compressed_formats(const struct pan_kmod_dev_props *props)
138 {
139    return props->texture_features[0];
140 }
141 
142 /* Check for AFBC hardware support. AFBC is introduced in v5. Implementations
143  * may omit it, signaled as a nonzero value in the AFBC_FEATURES property. */
144 
145 bool
panfrost_query_afbc(const struct pan_kmod_dev_props * props)146 panfrost_query_afbc(const struct pan_kmod_dev_props *props)
147 {
148    unsigned reg = props->afbc_features;
149 
150    return (pan_arch(props->gpu_prod_id) >= 5) && (reg == 0);
151 }
152 
153 /*
154  * To pipeline multiple tiles, a given tile may use at most half of the tile
155  * buffer. This function returns the optimal size (assuming pipelining).
156  *
157  * For Mali-G510 and Mali-G310, we will need extra logic to query the tilebuffer
158  * size for the particular variant. The CORE_FEATURES register might help.
159  */
160 unsigned
panfrost_query_optimal_tib_size(const struct panfrost_model * model)161 panfrost_query_optimal_tib_size(const struct panfrost_model *model)
162 {
163    /* Preconditions ensure the returned value is a multiple of 1 KiB, the
164     * granularity of the colour buffer allocation field.
165     */
166    assert(model->tilebuffer_size >= 2048);
167    assert(util_is_power_of_two_nonzero(model->tilebuffer_size));
168 
169    return model->tilebuffer_size / 2;
170 }
171 
172 uint64_t
panfrost_clamp_to_usable_va_range(const struct pan_kmod_dev * dev,uint64_t va)173 panfrost_clamp_to_usable_va_range(const struct pan_kmod_dev *dev, uint64_t va)
174 {
175    struct pan_kmod_va_range user_va_range =
176       pan_kmod_dev_query_user_va_range(dev);
177 
178    if (va < user_va_range.start)
179       return user_va_range.start;
180    else if (va > user_va_range.start + user_va_range.size)
181       return user_va_range.start + user_va_range.size;
182 
183    return va;
184 }
185