gallium/radeon: print more info about HTILE
[mesa.git] / src / gallium / drivers / radeon / r600_query.h
1 /*
2 * Copyright 2015 Advanced Micro Devices, Inc.
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 * Nicolai Hähnle <nicolai.haehnle@amd.com>
25 *
26 */
27
28 #ifndef R600_QUERY_H
29 #define R600_QUERY_H
30
31 #include "pipe/p_defines.h"
32 #include "util/list.h"
33
34 struct pipe_context;
35 struct pipe_query;
36
37 struct r600_common_context;
38 struct r600_common_screen;
39 struct r600_query;
40 struct r600_query_hw;
41 struct r600_resource;
42
43 #define R600_QUERY_DRAW_CALLS (PIPE_QUERY_DRIVER_SPECIFIC + 0)
44 #define R600_QUERY_REQUESTED_VRAM (PIPE_QUERY_DRIVER_SPECIFIC + 1)
45 #define R600_QUERY_REQUESTED_GTT (PIPE_QUERY_DRIVER_SPECIFIC + 2)
46 #define R600_QUERY_BUFFER_WAIT_TIME (PIPE_QUERY_DRIVER_SPECIFIC + 3)
47 #define R600_QUERY_NUM_CS_FLUSHES (PIPE_QUERY_DRIVER_SPECIFIC + 4)
48 #define R600_QUERY_NUM_BYTES_MOVED (PIPE_QUERY_DRIVER_SPECIFIC + 5)
49 #define R600_QUERY_VRAM_USAGE (PIPE_QUERY_DRIVER_SPECIFIC + 6)
50 #define R600_QUERY_GTT_USAGE (PIPE_QUERY_DRIVER_SPECIFIC + 7)
51 #define R600_QUERY_GPU_TEMPERATURE (PIPE_QUERY_DRIVER_SPECIFIC + 8)
52 #define R600_QUERY_CURRENT_GPU_SCLK (PIPE_QUERY_DRIVER_SPECIFIC + 9)
53 #define R600_QUERY_CURRENT_GPU_MCLK (PIPE_QUERY_DRIVER_SPECIFIC + 10)
54 #define R600_QUERY_GPU_LOAD (PIPE_QUERY_DRIVER_SPECIFIC + 11)
55 #define R600_QUERY_NUM_COMPILATIONS (PIPE_QUERY_DRIVER_SPECIFIC + 12)
56 #define R600_QUERY_NUM_SHADERS_CREATED (PIPE_QUERY_DRIVER_SPECIFIC + 13)
57 #define R600_QUERY_FIRST_PERFCOUNTER (PIPE_QUERY_DRIVER_SPECIFIC + 100)
58
59 struct r600_query_ops {
60 void (*destroy)(struct r600_common_context *, struct r600_query *);
61 boolean (*begin)(struct r600_common_context *, struct r600_query *);
62 void (*end)(struct r600_common_context *, struct r600_query *);
63 boolean (*get_result)(struct r600_common_context *,
64 struct r600_query *, boolean wait,
65 union pipe_query_result *result);
66 };
67
68 struct r600_query {
69 struct r600_query_ops *ops;
70
71 /* The type of query */
72 unsigned type;
73 };
74
75 enum {
76 R600_QUERY_HW_FLAG_NO_START = (1 << 0),
77 R600_QUERY_HW_FLAG_TIMER = (1 << 1),
78 R600_QUERY_HW_FLAG_PREDICATE = (1 << 2),
79 };
80
81 struct r600_query_hw_ops {
82 void (*prepare_buffer)(struct r600_common_context *,
83 struct r600_query_hw *,
84 struct r600_resource *);
85 void (*emit_start)(struct r600_common_context *,
86 struct r600_query_hw *,
87 struct r600_resource *buffer, uint64_t va);
88 void (*emit_stop)(struct r600_common_context *,
89 struct r600_query_hw *,
90 struct r600_resource *buffer, uint64_t va);
91 void (*clear_result)(struct r600_query_hw *, union pipe_query_result *);
92 void (*add_result)(struct r600_common_context *ctx,
93 struct r600_query_hw *, void *buffer,
94 union pipe_query_result *result);
95 };
96
97 struct r600_query_buffer {
98 /* The buffer where query results are stored. */
99 struct r600_resource *buf;
100 /* Offset of the next free result after current query data */
101 unsigned results_end;
102 /* If a query buffer is full, a new buffer is created and the old one
103 * is put in here. When we calculate the result, we sum up the samples
104 * from all buffers. */
105 struct r600_query_buffer *previous;
106 };
107
108 struct r600_query_hw {
109 struct r600_query b;
110 struct r600_query_hw_ops *ops;
111 unsigned flags;
112
113 /* The query buffer and how many results are in it. */
114 struct r600_query_buffer buffer;
115 /* Size of the result in memory for both begin_query and end_query,
116 * this can be one or two numbers, or it could even be a size of a structure. */
117 unsigned result_size;
118 /* The number of dwords for begin_query or end_query. */
119 unsigned num_cs_dw_begin;
120 unsigned num_cs_dw_end;
121 /* Linked list of queries */
122 struct list_head list;
123 /* For transform feedback: which stream the query is for */
124 unsigned stream;
125 };
126
127 boolean r600_query_hw_init(struct r600_common_context *rctx,
128 struct r600_query_hw *query);
129 void r600_query_hw_destroy(struct r600_common_context *rctx,
130 struct r600_query *rquery);
131 boolean r600_query_hw_begin(struct r600_common_context *rctx,
132 struct r600_query *rquery);
133 void r600_query_hw_end(struct r600_common_context *rctx,
134 struct r600_query *rquery);
135 boolean r600_query_hw_get_result(struct r600_common_context *rctx,
136 struct r600_query *rquery,
137 boolean wait,
138 union pipe_query_result *result);
139
140 /* Performance counters */
141 enum {
142 /* This block is part of the shader engine */
143 R600_PC_BLOCK_SE = (1 << 0),
144
145 /* Expose per-instance groups instead of summing all instances (within
146 * an SE). */
147 R600_PC_BLOCK_INSTANCE_GROUPS = (1 << 1),
148
149 /* Expose per-SE groups instead of summing instances across SEs. */
150 R600_PC_BLOCK_SE_GROUPS = (1 << 2),
151
152 /* Shader block */
153 R600_PC_BLOCK_SHADER = (1 << 3),
154
155 /* Non-shader block with perfcounters windowed by shaders. */
156 R600_PC_BLOCK_SHADER_WINDOWED = (1 << 4),
157 };
158
159 /* Shader enable bits. Chosen to coincide with SQ_PERFCOUNTER_CTRL values */
160 enum {
161 R600_PC_SHADER_PS = (1 << 0),
162 R600_PC_SHADER_VS = (1 << 1),
163 R600_PC_SHADER_GS = (1 << 2),
164 R600_PC_SHADER_ES = (1 << 3),
165 R600_PC_SHADER_HS = (1 << 4),
166 R600_PC_SHADER_LS = (1 << 5),
167 R600_PC_SHADER_CS = (1 << 6),
168
169 R600_PC_SHADER_ALL = R600_PC_SHADER_PS | R600_PC_SHADER_VS |
170 R600_PC_SHADER_GS | R600_PC_SHADER_ES |
171 R600_PC_SHADER_HS | R600_PC_SHADER_LS |
172 R600_PC_SHADER_CS,
173
174 R600_PC_SHADER_WINDOWING = (1 << 31),
175 };
176
177 /* Describes a hardware block with performance counters. Multiple instances of
178 * each block, possibly per-SE, may exist on the chip. Depending on the block
179 * and on the user's configuration, we either
180 * (a) expose every instance as a performance counter group,
181 * (b) expose a single performance counter group that reports the sum over all
182 * instances, or
183 * (c) expose one performance counter group per instance, but summed over all
184 * shader engines.
185 */
186 struct r600_perfcounter_block {
187 const char *basename;
188 unsigned flags;
189 unsigned num_counters;
190 unsigned num_selectors;
191 unsigned num_instances;
192
193 unsigned num_groups;
194 char *group_names;
195 unsigned group_name_stride;
196
197 char *selector_names;
198 unsigned selector_name_stride;
199
200 void *data;
201 };
202
203 struct r600_perfcounters {
204 unsigned num_groups;
205 unsigned num_blocks;
206 struct r600_perfcounter_block *blocks;
207
208 unsigned num_start_cs_dwords;
209 unsigned num_stop_cs_dwords;
210 unsigned num_instance_cs_dwords;
211 unsigned num_shaders_cs_dwords;
212
213 void (*get_size)(struct r600_perfcounter_block *,
214 unsigned count, unsigned *selectors,
215 unsigned *num_select_dw, unsigned *num_read_dw);
216
217 void (*emit_instance)(struct r600_common_context *,
218 int se, int instance);
219 void (*emit_shaders)(struct r600_common_context *, unsigned shaders);
220 void (*emit_select)(struct r600_common_context *,
221 struct r600_perfcounter_block *,
222 unsigned count, unsigned *selectors);
223 void (*emit_start)(struct r600_common_context *,
224 struct r600_resource *buffer, uint64_t va);
225 void (*emit_stop)(struct r600_common_context *,
226 struct r600_resource *buffer, uint64_t va);
227 void (*emit_read)(struct r600_common_context *,
228 struct r600_perfcounter_block *,
229 unsigned count, unsigned *selectors,
230 struct r600_resource *buffer, uint64_t va);
231
232 void (*cleanup)(struct r600_common_screen *);
233
234 boolean separate_se;
235 boolean separate_instance;
236 };
237
238 struct pipe_query *r600_create_batch_query(struct pipe_context *ctx,
239 unsigned num_queries,
240 unsigned *query_types);
241
242 int r600_get_perfcounter_info(struct r600_common_screen *,
243 unsigned index,
244 struct pipe_driver_query_info *info);
245 int r600_get_perfcounter_group_info(struct r600_common_screen *,
246 unsigned index,
247 struct pipe_driver_query_group_info *info);
248
249 boolean r600_perfcounters_init(struct r600_perfcounters *, unsigned num_blocks);
250 void r600_perfcounters_add_block(struct r600_common_screen *,
251 struct r600_perfcounters *,
252 const char *name, unsigned flags,
253 unsigned counters, unsigned selectors,
254 unsigned instances, void *data);
255 void r600_perfcounters_do_destroy(struct r600_perfcounters *);
256
257 #endif /* R600_QUERY_H */