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
207
208
209
210
211
212
213
214
215
216
217
|
/*
* Copyright © 2017 Advanced Micro Devices, Inc.
*
* 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.
*/
#ifndef AC_GPU_INFO_H
#define AC_GPU_INFO_H
#include <stddef.h>
#include <stdint.h>
#include <stdbool.h>
#include "amd_family.h"
#ifdef __cplusplus
extern "C" {
#endif
struct amdgpu_gpu_info;
struct radeon_info {
/* PCI info: domain:bus:dev:func */
uint32_t pci_domain;
uint32_t pci_bus;
uint32_t pci_dev;
uint32_t pci_func;
/* Device info. */
const char *name;
const char *marketing_name;
bool is_pro_graphics;
uint32_t pci_id;
enum radeon_family family;
enum chip_class chip_class;
uint32_t family_id;
uint32_t chip_external_rev;
bool has_graphics; /* false if the chip is compute-only */
uint32_t num_compute_rings;
uint32_t num_sdma_rings;
uint32_t clock_crystal_freq;
uint32_t tcc_cache_line_size;
/* There are 2 display DCC codepaths, because display expects unaligned DCC. */
/* Disable RB and pipe alignment to skip the retile blit. (1 RB chips only) */
bool use_display_dcc_unaligned;
/* Allocate both aligned and unaligned DCC and use the retile blit. */
bool use_display_dcc_with_retile_blit;
/* Memory info. */
uint32_t pte_fragment_size;
uint32_t gart_page_size;
uint64_t gart_size;
uint64_t vram_size;
uint64_t vram_vis_size;
unsigned gds_size;
unsigned gds_gfx_partition_size;
uint64_t max_alloc_size;
uint32_t min_alloc_size;
uint32_t address32_hi;
bool has_dedicated_vram;
bool r600_has_virtual_memory;
/* CP info. */
bool gfx_ib_pad_with_type2;
unsigned ib_start_alignment;
uint32_t me_fw_version;
uint32_t me_fw_feature;
uint32_t pfp_fw_version;
uint32_t pfp_fw_feature;
uint32_t ce_fw_version;
uint32_t ce_fw_feature;
/* Multimedia info. */
bool has_hw_decode;
bool uvd_enc_supported;
uint32_t uvd_fw_version;
uint32_t vce_fw_version;
uint32_t vce_harvest_config;
/* Kernel & winsys capabilities. */
uint32_t drm_major; /* version */
uint32_t drm_minor;
uint32_t drm_patchlevel;
bool is_amdgpu;
bool has_userptr;
bool has_syncobj;
bool has_syncobj_wait_for_submit;
bool has_fence_to_handle;
bool has_ctx_priority;
bool has_local_buffers;
bool kernel_flushes_hdp_before_ib;
bool htile_cmask_support_1d_tiling;
bool si_TA_CS_BC_BASE_ADDR_allowed;
bool has_bo_metadata;
bool has_gpu_reset_status_query;
bool has_eqaa_surface_allocator;
bool has_format_bc1_through_bc7;
bool kernel_flushes_tc_l2_after_ib;
bool has_indirect_compute_dispatch;
bool has_unaligned_shader_loads;
bool has_sparse_vm_mappings;
bool has_2d_tiling;
bool has_read_registers_query;
bool has_gds_ordered_append;
bool has_scheduled_fence_dependency;
/* Shader cores. */
uint32_t r600_max_quad_pipes; /* wave size / 16 */
uint32_t max_shader_clock;
uint32_t num_good_compute_units;
uint32_t num_good_cu_per_sh;
uint32_t num_tcc_blocks;
uint32_t max_se; /* shader engines */
uint32_t max_sh_per_se; /* shader arrays per shader engine */
/* Render backends (color + depth blocks). */
uint32_t r300_num_gb_pipes;
uint32_t r300_num_z_pipes;
uint32_t r600_gb_backend_map; /* R600 harvest config */
bool r600_gb_backend_map_valid;
uint32_t r600_num_banks;
uint32_t gb_addr_config;
uint32_t pa_sc_tile_steering_override; /* CLEAR_STATE also sets this */
uint32_t num_render_backends;
uint32_t num_tile_pipes; /* pipe count from PIPE_CONFIG */
uint32_t pipe_interleave_bytes;
uint32_t enabled_rb_mask; /* GCN harvest config */
uint64_t max_alignment; /* from addrlib */
/* Tile modes. */
uint32_t si_tile_mode_array[32];
uint32_t cik_macrotile_mode_array[16];
};
bool ac_query_gpu_info(int fd, void *dev_p,
struct radeon_info *info,
struct amdgpu_gpu_info *amdinfo);
void ac_compute_driver_uuid(char *uuid, size_t size);
void ac_compute_device_uuid(struct radeon_info *info, char *uuid, size_t size);
void ac_print_gpu_info(struct radeon_info *info);
int ac_get_gs_table_depth(enum chip_class chip_class, enum radeon_family family);
void ac_get_raster_config(struct radeon_info *info,
uint32_t *raster_config_p,
uint32_t *raster_config_1_p,
uint32_t *se_tile_repeat_p);
void ac_get_harvested_configs(struct radeon_info *info,
unsigned raster_config,
unsigned *cik_raster_config_1_p,
unsigned *raster_config_se);
unsigned ac_get_compute_resource_limits(struct radeon_info *info,
unsigned waves_per_threadgroup,
unsigned max_waves_per_sh,
unsigned threadgroups_per_cu);
static inline unsigned ac_get_max_wave64_per_simd(enum radeon_family family)
{
switch (family) {
/* These always have 8 waves: */
case CHIP_POLARIS10:
case CHIP_POLARIS11:
case CHIP_POLARIS12:
case CHIP_VEGAM:
return 8;
default:
return 10;
}
}
static inline unsigned ac_get_num_physical_vgprs(enum chip_class chip_class,
unsigned wave_size)
{
/* The number is per SIMD. */
if (chip_class >= GFX10)
return wave_size == 32 ? 1024 : 512;
else
return 256;
}
static inline uint32_t
ac_get_num_physical_sgprs(const struct radeon_info *info)
{
/* The number is per SIMD. There is enough SGPRs for the maximum number
* of Wave32, which is double the number for Wave64.
*/
if (info->chip_class >= GFX10)
return 128 * ac_get_max_wave64_per_simd(info->family) * 2;
return info->chip_class >= GFX8 ? 800 : 512;
}
#ifdef __cplusplus
}
#endif
#endif /* AC_GPU_INFO_H */
|
