summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/radeonsi/si_query.h
blob: aaf0bd03aca76a376a28190adc02cde40a3e5cdf (plain)
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
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
/*
 * Copyright 2015 Advanced Micro Devices, Inc.
 * All Rights Reserved.
 *
 * 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, sublicense,
 * 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 above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * 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 NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS 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.
 */

#ifndef SI_QUERY_H
#define SI_QUERY_H

#include "util/u_threaded_context.h"

struct pipe_context;
struct pipe_query;
struct pipe_resource;

struct si_screen;
struct si_context;
struct si_query;
struct si_query_buffer;
struct si_query_hw;
struct si_resource;

enum {
	SI_QUERY_DRAW_CALLS = PIPE_QUERY_DRIVER_SPECIFIC,
	SI_QUERY_DECOMPRESS_CALLS,
	SI_QUERY_MRT_DRAW_CALLS,
	SI_QUERY_PRIM_RESTART_CALLS,
	SI_QUERY_SPILL_DRAW_CALLS,
	SI_QUERY_COMPUTE_CALLS,
	SI_QUERY_SPILL_COMPUTE_CALLS,
	SI_QUERY_DMA_CALLS,
	SI_QUERY_CP_DMA_CALLS,
	SI_QUERY_NUM_VS_FLUSHES,
	SI_QUERY_NUM_PS_FLUSHES,
	SI_QUERY_NUM_CS_FLUSHES,
	SI_QUERY_NUM_CB_CACHE_FLUSHES,
	SI_QUERY_NUM_DB_CACHE_FLUSHES,
	SI_QUERY_NUM_L2_INVALIDATES,
	SI_QUERY_NUM_L2_WRITEBACKS,
	SI_QUERY_NUM_RESIDENT_HANDLES,
	SI_QUERY_TC_OFFLOADED_SLOTS,
	SI_QUERY_TC_DIRECT_SLOTS,
	SI_QUERY_TC_NUM_SYNCS,
	SI_QUERY_CS_THREAD_BUSY,
	SI_QUERY_GALLIUM_THREAD_BUSY,
	SI_QUERY_REQUESTED_VRAM,
	SI_QUERY_REQUESTED_GTT,
	SI_QUERY_MAPPED_VRAM,
	SI_QUERY_MAPPED_GTT,
	SI_QUERY_BUFFER_WAIT_TIME,
	SI_QUERY_NUM_MAPPED_BUFFERS,
	SI_QUERY_NUM_GFX_IBS,
	SI_QUERY_NUM_SDMA_IBS,
	SI_QUERY_GFX_BO_LIST_SIZE,
	SI_QUERY_GFX_IB_SIZE,
	SI_QUERY_NUM_BYTES_MOVED,
	SI_QUERY_NUM_EVICTIONS,
	SI_QUERY_NUM_VRAM_CPU_PAGE_FAULTS,
	SI_QUERY_VRAM_USAGE,
	SI_QUERY_VRAM_VIS_USAGE,
	SI_QUERY_GTT_USAGE,
	SI_QUERY_GPU_TEMPERATURE,
	SI_QUERY_CURRENT_GPU_SCLK,
	SI_QUERY_CURRENT_GPU_MCLK,
	SI_QUERY_GPU_LOAD,
	SI_QUERY_GPU_SHADERS_BUSY,
	SI_QUERY_GPU_TA_BUSY,
	SI_QUERY_GPU_GDS_BUSY,
	SI_QUERY_GPU_VGT_BUSY,
	SI_QUERY_GPU_IA_BUSY,
	SI_QUERY_GPU_SX_BUSY,
	SI_QUERY_GPU_WD_BUSY,
	SI_QUERY_GPU_BCI_BUSY,
	SI_QUERY_GPU_SC_BUSY,
	SI_QUERY_GPU_PA_BUSY,
	SI_QUERY_GPU_DB_BUSY,
	SI_QUERY_GPU_CP_BUSY,
	SI_QUERY_GPU_CB_BUSY,
	SI_QUERY_GPU_SDMA_BUSY,
	SI_QUERY_GPU_PFP_BUSY,
	SI_QUERY_GPU_MEQ_BUSY,
	SI_QUERY_GPU_ME_BUSY,
	SI_QUERY_GPU_SURF_SYNC_BUSY,
	SI_QUERY_GPU_CP_DMA_BUSY,
	SI_QUERY_GPU_SCRATCH_RAM_BUSY,
	SI_QUERY_NUM_COMPILATIONS,
	SI_QUERY_NUM_SHADERS_CREATED,
	SI_QUERY_BACK_BUFFER_PS_DRAW_RATIO,
	SI_QUERY_NUM_SHADER_CACHE_HITS,
	SI_QUERY_GPIN_ASIC_ID,
	SI_QUERY_GPIN_NUM_SIMD,
	SI_QUERY_GPIN_NUM_RB,
	SI_QUERY_GPIN_NUM_SPI,
	SI_QUERY_GPIN_NUM_SE,
	SI_QUERY_TIME_ELAPSED_SDMA,
	SI_QUERY_TIME_ELAPSED_SDMA_SI, /* emulated, measured on the CPU */

	SI_QUERY_FIRST_PERFCOUNTER = PIPE_QUERY_DRIVER_SPECIFIC + 100,
};

enum {
	SI_QUERY_GROUP_GPIN = 0,
	SI_NUM_SW_QUERY_GROUPS
};

struct si_query_ops {
	void (*destroy)(struct si_screen *, struct si_query *);
	bool (*begin)(struct si_context *, struct si_query *);
	bool (*end)(struct si_context *, struct si_query *);
	bool (*get_result)(struct si_context *,
			   struct si_query *, bool wait,
			   union pipe_query_result *result);
	void (*get_result_resource)(struct si_context *,
				    struct si_query *, bool wait,
				    enum pipe_query_value_type result_type,
				    int index,
				    struct pipe_resource *resource,
				    unsigned offset);

	void (*suspend)(struct si_context *, struct si_query *);
	void (*resume)(struct si_context *, struct si_query *);
};

struct si_query {
	struct threaded_query b;
	const struct si_query_ops *ops;

	/* The PIPE_QUERY_xxx type of query */
	unsigned type;

	/* The number of dwords for suspend. */
	unsigned num_cs_dw_suspend;

	/* Linked list of queries that must be suspended at end of CS. */
	struct list_head active_list;
};

enum {
	SI_QUERY_HW_FLAG_NO_START = (1 << 0),
	/* gap */
	/* whether begin_query doesn't clear the result */
	SI_QUERY_HW_FLAG_BEGIN_RESUMES = (1 << 2),
};

struct si_query_hw_ops {
	bool (*prepare_buffer)(struct si_context *, struct si_query_buffer *);
	void (*emit_start)(struct si_context *,
			   struct si_query_hw *,
			   struct si_resource *buffer, uint64_t va);
	void (*emit_stop)(struct si_context *,
			  struct si_query_hw *,
			  struct si_resource *buffer, uint64_t va);
	void (*clear_result)(struct si_query_hw *, union pipe_query_result *);
	void (*add_result)(struct si_screen *screen,
			   struct si_query_hw *, void *buffer,
			   union pipe_query_result *result);
};

struct si_query_buffer {
	/* The buffer where query results are stored. */
	struct si_resource		*buf;
	/* Offset of the next free result after current query data */
	unsigned			results_end;
	/* If a query buffer is full, a new buffer is created and the old one
	 * is put in here. When we calculate the result, we sum up the samples
	 * from all buffers. */
	struct si_query_buffer	*previous;
};

void si_query_buffer_destroy(struct si_screen *sctx, struct si_query_buffer *buffer);
void si_query_buffer_reset(struct si_context *sctx, struct si_query_buffer *buffer);
bool si_query_buffer_alloc(struct si_context *sctx, struct si_query_buffer *buffer,
			   bool (*prepare_buffer)(struct si_context *, struct si_query_buffer*),
			   unsigned size);


struct si_query_hw {
	struct si_query b;
	struct si_query_hw_ops *ops;
	unsigned flags;

	/* The query buffer and how many results are in it. */
	struct si_query_buffer buffer;
	/* Size of the result in memory for both begin_query and end_query,
	 * this can be one or two numbers, or it could even be a size of a structure. */
	unsigned result_size;
	/* For transform feedback: which stream the query is for */
	unsigned stream;

	/* Workaround via compute shader */
	struct si_resource *workaround_buf;
	unsigned workaround_offset;
};

void si_query_hw_destroy(struct si_screen *sscreen,
			 struct si_query *squery);
bool si_query_hw_begin(struct si_context *sctx,
		       struct si_query *squery);
bool si_query_hw_end(struct si_context *sctx,
		     struct si_query *squery);
bool si_query_hw_get_result(struct si_context *sctx,
			    struct si_query *squery,
			    bool wait,
			    union pipe_query_result *result);
void si_query_hw_suspend(struct si_context *sctx, struct si_query *query);
void si_query_hw_resume(struct si_context *sctx, struct si_query *query);


/* Performance counters */
struct si_perfcounters {
	unsigned num_groups;
	unsigned num_blocks;
	struct si_pc_block *blocks;

	unsigned num_stop_cs_dwords;
	unsigned num_instance_cs_dwords;

	bool separate_se;
	bool separate_instance;
};

struct pipe_query *si_create_batch_query(struct pipe_context *ctx,
					 unsigned num_queries,
					 unsigned *query_types);

int si_get_perfcounter_info(struct si_screen *,
			    unsigned index,
			    struct pipe_driver_query_info *info);
int si_get_perfcounter_group_info(struct si_screen *,
				  unsigned index,
				  struct pipe_driver_query_group_info *info);

struct si_qbo_state {
	void *saved_compute;
	struct pipe_constant_buffer saved_const0;
	struct pipe_shader_buffer saved_ssbo[3];
};

#endif /* SI_QUERY_H */