2 * Copyright © 2018 Intel Corporation
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:
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
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
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
31 #include <sys/sysmacros.h>
33 #include "util/hash_table.h"
34 #include "util/ralloc.h"
36 struct gen_device_info
;
39 struct gen_perf_query_info
;
41 enum gen_perf_counter_type
{
42 GEN_PERF_COUNTER_TYPE_EVENT
,
43 GEN_PERF_COUNTER_TYPE_DURATION_NORM
,
44 GEN_PERF_COUNTER_TYPE_DURATION_RAW
,
45 GEN_PERF_COUNTER_TYPE_THROUGHPUT
,
46 GEN_PERF_COUNTER_TYPE_RAW
,
47 GEN_PERF_COUNTER_TYPE_TIMESTAMP
,
50 enum gen_perf_counter_data_type
{
51 GEN_PERF_COUNTER_DATA_TYPE_BOOL32
,
52 GEN_PERF_COUNTER_DATA_TYPE_UINT32
,
53 GEN_PERF_COUNTER_DATA_TYPE_UINT64
,
54 GEN_PERF_COUNTER_DATA_TYPE_FLOAT
,
55 GEN_PERF_COUNTER_DATA_TYPE_DOUBLE
,
58 struct gen_pipeline_stat
{
65 * The largest OA formats we can use include:
67 * 1 timestamp, 45 A counters, 8 B counters and 8 C counters.
69 * 1 timestamp, 1 clock, 36 A counters, 8 B counters and 8 C counters
71 #define MAX_OA_REPORT_COUNTERS 62
73 struct gen_perf_query_result
{
75 * Storage for the final accumulated OA counters.
77 uint64_t accumulator
[MAX_OA_REPORT_COUNTERS
];
80 * Hw ID used by the context on which the query was running.
85 * Number of reports accumulated to produce the results.
87 uint32_t reports_accumulated
;
90 * Frequency in the slices of the GT at the begin and end of the
93 uint64_t slice_frequency
[2];
96 * Frequency in the unslice of the GT at the begin and end of the
99 uint64_t unslice_frequency
[2];
102 struct gen_perf_query_counter
{
105 enum gen_perf_counter_type type
;
106 enum gen_perf_counter_data_type data_type
;
112 uint64_t (*oa_counter_read_uint64
)(struct gen_perf
*perf
,
113 const struct gen_perf_query_info
*query
,
114 uint64_t *accumulator
);
115 float (*oa_counter_read_float
)(struct gen_perf
*perf
,
116 const struct gen_perf_query_info
*query
,
117 uint64_t *accumulator
);
118 struct gen_pipeline_stat pipeline_stat
;
122 struct gen_perf_query_register_prog
{
127 struct gen_perf_query_info
{
128 enum gen_perf_query_type
{
129 GEN_PERF_QUERY_TYPE_OA
,
130 GEN_PERF_QUERY_TYPE_RAW
,
131 GEN_PERF_QUERY_TYPE_PIPELINE
,
135 struct gen_perf_query_counter
*counters
;
141 uint64_t oa_metrics_set_id
;
144 /* For indexing into the accumulator[] ... */
146 int gpu_clock_offset
;
151 /* Register programming for a given query */
152 struct gen_perf_query_register_prog
*flex_regs
;
153 uint32_t n_flex_regs
;
155 struct gen_perf_query_register_prog
*mux_regs
;
158 struct gen_perf_query_register_prog
*b_counter_regs
;
159 uint32_t n_b_counter_regs
;
163 struct gen_perf_query_info
*queries
;
166 /* Variables referenced in the XML meta data for OA performance
167 * counters, e.g in the normalization equations.
169 * All uint64_t for consistent operand types in generated code
172 uint64_t timestamp_frequency
; /** $GpuTimestampFrequency */
173 uint64_t n_eus
; /** $EuCoresTotalCount */
174 uint64_t n_eu_slices
; /** $EuSlicesTotalCount */
175 uint64_t n_eu_sub_slices
; /** $EuSubslicesTotalCount */
176 uint64_t eu_threads_count
; /** $EuThreadsCount */
177 uint64_t slice_mask
; /** $SliceMask */
178 uint64_t subslice_mask
; /** $SubsliceMask */
179 uint64_t gt_min_freq
; /** $GpuMinFrequency */
180 uint64_t gt_max_freq
; /** $GpuMaxFrequency */
181 uint64_t revision
; /** $SkuRevisionId */
184 /* OA metric sets, indexed by GUID, as know by Mesa at build time, to
185 * cross-reference with the GUIDs of configs advertised by the kernel at
188 struct hash_table
*oa_metrics_table
;
190 /* Location of the device's sysfs entry. */
191 char sysfs_dev_dir
[256];
193 int (*ioctl
)(int, unsigned long, void *);
196 static inline struct gen_perf_query_info
*
197 gen_perf_query_append_query_info(struct gen_perf
*perf
, int max_counters
)
199 struct gen_perf_query_info
*query
;
201 perf
->queries
= reralloc(perf
, perf
->queries
,
202 struct gen_perf_query_info
,
204 query
= &perf
->queries
[perf
->n_queries
- 1];
205 memset(query
, 0, sizeof(*query
));
207 if (max_counters
> 0) {
208 query
->max_counters
= max_counters
;
210 rzalloc_array(perf
, struct gen_perf_query_counter
, max_counters
);
217 gen_perf_query_info_add_stat_reg(struct gen_perf_query_info
*query
,
220 uint32_t denominator
,
222 const char *description
)
224 struct gen_perf_query_counter
*counter
;
226 assert(query
->n_counters
< query
->max_counters
);
228 counter
= &query
->counters
[query
->n_counters
];
229 counter
->name
= name
;
230 counter
->desc
= description
;
231 counter
->type
= GEN_PERF_COUNTER_TYPE_RAW
;
232 counter
->data_type
= GEN_PERF_COUNTER_DATA_TYPE_UINT64
;
233 counter
->size
= sizeof(uint64_t);
234 counter
->offset
= sizeof(uint64_t) * query
->n_counters
;
235 counter
->pipeline_stat
.reg
= reg
;
236 counter
->pipeline_stat
.numerator
= numerator
;
237 counter
->pipeline_stat
.denominator
= denominator
;
243 gen_perf_query_info_add_basic_stat_reg(struct gen_perf_query_info
*query
,
244 uint32_t reg
, const char *name
)
246 gen_perf_query_info_add_stat_reg(query
, reg
, 1, 1, name
, name
);
249 static inline struct gen_perf
*
250 gen_perf_new(void *ctx
, int (*ioctl_cb
)(int, unsigned long, void *))
252 struct gen_perf
*perf
= rzalloc(ctx
, struct gen_perf
);
254 perf
->ioctl
= ioctl_cb
;
259 bool gen_perf_load_oa_metrics(struct gen_perf
*perf
, int fd
,
260 const struct gen_device_info
*devinfo
);
261 bool gen_perf_load_metric_id(struct gen_perf
*perf
, const char *guid
,
262 uint64_t *metric_id
);
264 void gen_perf_query_result_read_frequencies(struct gen_perf_query_result
*result
,
265 const struct gen_device_info
*devinfo
,
266 const uint32_t *start
,
267 const uint32_t *end
);
268 void gen_perf_query_result_accumulate(struct gen_perf_query_result
*result
,
269 const struct gen_perf_query_info
*query
,
270 const uint32_t *start
,
271 const uint32_t *end
);
272 void gen_perf_query_result_clear(struct gen_perf_query_result
*result
);
275 #endif /* GEN_PERF_H */