6022a199065a6018e18bc719ba1c5b63bd31115c
[mesa.git] / src / amd / common / ac_gpu_info.h
1 /*
2 * Copyright © 2017 Advanced Micro Devices, Inc.
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining
5 * a copy of this software and associated documentation files (the
6 * "Software"), to deal in the Software without restriction, including
7 * without limitation the rights to use, copy, modify, merge, publish,
8 * distribute, sub license, and/or sell copies of the Software, and to
9 * permit persons to whom the Software is furnished to do so, subject to
10 * the following conditions:
11 *
12 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
13 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
14 * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
15 * NON-INFRINGEMENT. IN NO EVENT SHALL THE COPYRIGHT HOLDERS, AUTHORS
16 * AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
17 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
18 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
19 * USE OR OTHER DEALINGS IN THE SOFTWARE.
20 *
21 * The above copyright notice and this permission notice (including the
22 * next paragraph) shall be included in all copies or substantial portions
23 * of the Software.
24 */
25
26 #ifndef AC_GPU_INFO_H
27 #define AC_GPU_INFO_H
28
29 #include <stddef.h>
30 #include <stdint.h>
31 #include <stdbool.h>
32 #include "amd_family.h"
33
34 #ifdef __cplusplus
35 extern "C" {
36 #endif
37
38 struct amdgpu_gpu_info;
39
40 struct radeon_info {
41 /* PCI info: domain:bus:dev:func */
42 uint32_t pci_domain;
43 uint32_t pci_bus;
44 uint32_t pci_dev;
45 uint32_t pci_func;
46
47 /* Device info. */
48 const char *name;
49 const char *marketing_name;
50 bool is_pro_graphics;
51 uint32_t pci_id;
52 uint32_t pci_rev_id;
53 enum radeon_family family;
54 enum chip_class chip_class;
55 uint32_t family_id;
56 uint32_t chip_external_rev;
57 uint32_t clock_crystal_freq;
58
59 /* Features. */
60 bool has_graphics; /* false if the chip is compute-only */
61 uint32_t num_rings[NUM_RING_TYPES];
62 uint32_t ib_pad_dw_mask[NUM_RING_TYPES];
63 bool has_clear_state;
64 bool has_distributed_tess;
65 bool has_dcc_constant_encode;
66 bool has_rbplus; /* if RB+ registers exist */
67 bool rbplus_allowed; /* if RB+ is allowed */
68 bool has_load_ctx_reg_pkt;
69 bool has_out_of_order_rast;
70 bool has_packed_math_16bit;
71 bool cpdma_prefetch_writes_memory;
72 bool has_gfx9_scissor_bug;
73 bool has_tc_compat_zrange_bug;
74 bool has_msaa_sample_loc_bug;
75 bool has_ls_vgpr_init_bug;
76
77 /* Display features. */
78 /* There are 2 display DCC codepaths, because display expects unaligned DCC. */
79 /* Disable RB and pipe alignment to skip the retile blit. (1 RB chips only) */
80 bool use_display_dcc_unaligned;
81 /* Allocate both aligned and unaligned DCC and use the retile blit. */
82 bool use_display_dcc_with_retile_blit;
83
84 /* Memory info. */
85 uint32_t pte_fragment_size;
86 uint32_t gart_page_size;
87 uint64_t gart_size;
88 uint64_t vram_size;
89 uint64_t vram_vis_size;
90 uint32_t vram_bit_width;
91 uint32_t vram_type;
92 unsigned gds_size;
93 unsigned gds_gfx_partition_size;
94 uint64_t max_alloc_size;
95 uint32_t min_alloc_size;
96 uint32_t address32_hi;
97 bool has_dedicated_vram;
98 bool has_l2_uncached;
99 bool r600_has_virtual_memory;
100 uint32_t num_sdp_interfaces;
101 uint32_t num_tcc_blocks;
102 uint32_t tcc_cache_line_size;
103 bool tcc_harvested;
104 unsigned pc_lines;
105 uint32_t lds_size_per_workgroup;
106 uint32_t lds_granularity;
107 uint32_t max_memory_clock;
108 uint32_t ce_ram_size;
109 uint32_t l1_cache_size;
110 uint32_t l2_cache_size;
111
112 /* CP info. */
113 bool gfx_ib_pad_with_type2;
114 unsigned ib_alignment; /* both start and size alignment */
115 uint32_t me_fw_version;
116 uint32_t me_fw_feature;
117 uint32_t pfp_fw_version;
118 uint32_t pfp_fw_feature;
119 uint32_t ce_fw_version;
120 uint32_t ce_fw_feature;
121
122 /* Multimedia info. */
123 bool has_hw_decode;
124 bool uvd_enc_supported;
125 uint32_t uvd_fw_version;
126 uint32_t vce_fw_version;
127 uint32_t vce_harvest_config;
128
129 /* Kernel & winsys capabilities. */
130 uint32_t drm_major; /* version */
131 uint32_t drm_minor;
132 uint32_t drm_patchlevel;
133 bool is_amdgpu;
134 bool has_userptr;
135 bool has_syncobj;
136 bool has_syncobj_wait_for_submit;
137 bool has_fence_to_handle;
138 bool has_ctx_priority;
139 bool has_local_buffers;
140 bool kernel_flushes_hdp_before_ib;
141 bool htile_cmask_support_1d_tiling;
142 bool si_TA_CS_BC_BASE_ADDR_allowed;
143 bool has_bo_metadata;
144 bool has_gpu_reset_status_query;
145 bool has_eqaa_surface_allocator;
146 bool has_format_bc1_through_bc7;
147 bool kernel_flushes_tc_l2_after_ib;
148 bool has_indirect_compute_dispatch;
149 bool has_unaligned_shader_loads;
150 bool has_sparse_vm_mappings;
151 bool has_2d_tiling;
152 bool has_read_registers_query;
153 bool has_gds_ordered_append;
154 bool has_scheduled_fence_dependency;
155
156 /* Shader cores. */
157 uint32_t cu_mask[4][2];
158 uint32_t r600_max_quad_pipes; /* wave size / 16 */
159 uint32_t max_shader_clock;
160 uint32_t num_good_compute_units;
161 uint32_t max_good_cu_per_sa;
162 uint32_t min_good_cu_per_sa; /* min != max if SAs have different # of CUs */
163 uint32_t max_se; /* shader engines */
164 uint32_t max_sh_per_se; /* shader arrays per shader engine */
165 uint32_t max_wave64_per_simd;
166 uint32_t num_physical_sgprs_per_simd;
167 uint32_t num_physical_wave64_vgprs_per_simd;
168 uint32_t num_simd_per_compute_unit;
169 uint32_t min_sgpr_alloc;
170 uint32_t max_sgpr_alloc;
171 uint32_t sgpr_alloc_granularity;
172 uint32_t min_wave64_vgpr_alloc;
173 uint32_t max_vgpr_alloc;
174 uint32_t wave64_vgpr_alloc_granularity;
175 bool use_late_alloc; /* VS and GS: late pos/param allocation */
176
177 /* Render backends (color + depth blocks). */
178 uint32_t r300_num_gb_pipes;
179 uint32_t r300_num_z_pipes;
180 uint32_t r600_gb_backend_map; /* R600 harvest config */
181 bool r600_gb_backend_map_valid;
182 uint32_t r600_num_banks;
183 uint32_t gb_addr_config;
184 uint32_t pa_sc_tile_steering_override; /* CLEAR_STATE also sets this */
185 uint32_t num_render_backends;
186 uint32_t num_tile_pipes; /* pipe count from PIPE_CONFIG */
187 uint32_t pipe_interleave_bytes;
188 uint32_t enabled_rb_mask; /* GCN harvest config */
189 uint64_t max_alignment; /* from addrlib */
190 uint32_t pbb_max_alloc_count;
191
192 /* Tile modes. */
193 uint32_t si_tile_mode_array[32];
194 uint32_t cik_macrotile_mode_array[16];
195 };
196
197 bool ac_query_gpu_info(int fd, void *dev_p,
198 struct radeon_info *info,
199 struct amdgpu_gpu_info *amdinfo);
200
201 void ac_compute_driver_uuid(char *uuid, size_t size);
202
203 void ac_compute_device_uuid(struct radeon_info *info, char *uuid, size_t size);
204 void ac_print_gpu_info(struct radeon_info *info);
205 int ac_get_gs_table_depth(enum chip_class chip_class, enum radeon_family family);
206 void ac_get_raster_config(struct radeon_info *info,
207 uint32_t *raster_config_p,
208 uint32_t *raster_config_1_p,
209 uint32_t *se_tile_repeat_p);
210 void ac_get_harvested_configs(struct radeon_info *info,
211 unsigned raster_config,
212 unsigned *cik_raster_config_1_p,
213 unsigned *raster_config_se);
214 unsigned ac_get_compute_resource_limits(struct radeon_info *info,
215 unsigned waves_per_threadgroup,
216 unsigned max_waves_per_sh,
217 unsigned threadgroups_per_cu);
218
219 #ifdef __cplusplus
220 }
221 #endif
222
223 #endif /* AC_GPU_INFO_H */