aboutsummaryrefslogtreecommitdiffstats
path: root/src/intel/perf/gen_perf.py
blob: caee7725b63cbd5e6327d6d0ef4d02cac51b795a (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
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
# Copyright (c) 2015-2017 Intel Corporation
#
# 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.

import argparse
import os
import sys
import textwrap

import xml.etree.cElementTree as et

hashed_funcs = {}

c_file = None
_c_indent = 0

def c(*args):
    code = ' '.join(map(str,args))
    for line in code.splitlines():
        text = ''.rjust(_c_indent) + line
        c_file.write(text.rstrip() + "\n")

# indented, but no trailing newline...
def c_line_start(code):
    c_file.write(''.rjust(_c_indent) + code)
def c_raw(code):
    c_file.write(code)

def c_indent(n):
    global _c_indent
    _c_indent = _c_indent + n
def c_outdent(n):
    global _c_indent
    _c_indent = _c_indent - n

header_file = None
_h_indent = 0

def h(*args):
    code = ' '.join(map(str,args))
    for line in code.splitlines():
        text = ''.rjust(_h_indent) + line
        header_file.write(text.rstrip() + "\n")

def h_indent(n):
    global _c_indent
    _h_indent = _h_indent + n
def h_outdent(n):
    global _c_indent
    _h_indent = _h_indent - n


def emit_fadd(tmp_id, args):
    c("double tmp{0} = {1} + {2};".format(tmp_id, args[1], args[0]))
    return tmp_id + 1

# Be careful to check for divide by zero...
def emit_fdiv(tmp_id, args):
    c("double tmp{0} = {1};".format(tmp_id, args[1]))
    c("double tmp{0} = {1};".format(tmp_id + 1, args[0]))
    c("double tmp{0} = tmp{1} ? tmp{2} / tmp{1} : 0;".format(tmp_id + 2, tmp_id + 1, tmp_id))
    return tmp_id + 3

def emit_fmax(tmp_id, args):
    c("double tmp{0} = {1};".format(tmp_id, args[1]))
    c("double tmp{0} = {1};".format(tmp_id + 1, args[0]))
    c("double tmp{0} = MAX(tmp{1}, tmp{2});".format(tmp_id + 2, tmp_id, tmp_id + 1))
    return tmp_id + 3

def emit_fmul(tmp_id, args):
    c("double tmp{0} = {1} * {2};".format(tmp_id, args[1], args[0]))
    return tmp_id + 1

def emit_fsub(tmp_id, args):
    c("double tmp{0} = {1} - {2};".format(tmp_id, args[1], args[0]))
    return tmp_id + 1

def emit_read(tmp_id, args):
    type = args[1].lower()
    c("uint64_t tmp{0} = accumulator[query->{1}_offset + {2}];".format(tmp_id, type, args[0]))
    return tmp_id + 1

def emit_uadd(tmp_id, args):
    c("uint64_t tmp{0} = {1} + {2};".format(tmp_id, args[1], args[0]))
    return tmp_id + 1

# Be careful to check for divide by zero...
def emit_udiv(tmp_id, args):
    c("uint64_t tmp{0} = {1};".format(tmp_id, args[1]))
    c("uint64_t tmp{0} = {1};".format(tmp_id + 1, args[0]))
    c("uint64_t tmp{0} = tmp{1} ? tmp{2} / tmp{1} : 0;".format(tmp_id + 2, tmp_id + 1, tmp_id))
    return tmp_id + 3

def emit_umul(tmp_id, args):
    c("uint64_t tmp{0} = {1} * {2};".format(tmp_id, args[1], args[0]))
    return tmp_id + 1

def emit_usub(tmp_id, args):
    c("uint64_t tmp{0} = {1} - {2};".format(tmp_id, args[1], args[0]))
    return tmp_id + 1

def emit_umin(tmp_id, args):
    c("uint64_t tmp{0} = MIN({1}, {2});".format(tmp_id, args[1], args[0]))
    return tmp_id + 1

def emit_lshft(tmp_id, args):
    c("uint64_t tmp{0} = {1} << {2};".format(tmp_id, args[1], args[0]))
    return tmp_id + 1

def emit_rshft(tmp_id, args):
    c("uint64_t tmp{0} = {1} >> {2};".format(tmp_id, args[1], args[0]))
    return tmp_id + 1

def emit_and(tmp_id, args):
    c("uint64_t tmp{0} = {1} & {2};".format(tmp_id, args[1], args[0]))
    return tmp_id + 1

ops = {}
#             (n operands, emitter)
ops["FADD"] = (2, emit_fadd)
ops["FDIV"] = (2, emit_fdiv)
ops["FMAX"] = (2, emit_fmax)
ops["FMUL"] = (2, emit_fmul)
ops["FSUB"] = (2, emit_fsub)
ops["READ"] = (2, emit_read)
ops["UADD"] = (2, emit_uadd)
ops["UDIV"] = (2, emit_udiv)
ops["UMUL"] = (2, emit_umul)
ops["USUB"] = (2, emit_usub)
ops["UMIN"] = (2, emit_umin)
ops["<<"]   = (2, emit_lshft)
ops[">>"]   = (2, emit_rshft)
ops["AND"]  = (2, emit_and)

def brkt(subexp):
    if " " in subexp:
        return "(" + subexp + ")"
    else:
        return subexp

def splice_bitwise_and(args):
    return brkt(args[1]) + " & " + brkt(args[0])

def splice_logical_and(args):
    return brkt(args[1]) + " && " + brkt(args[0])

def splice_ult(args):
    return brkt(args[1]) + " < " + brkt(args[0])

def splice_ugte(args):
    return brkt(args[1]) + " >= " + brkt(args[0])

exp_ops = {}
#                 (n operands, splicer)
exp_ops["AND"]  = (2, splice_bitwise_and)
exp_ops["UGTE"] = (2, splice_ugte)
exp_ops["ULT"]  = (2, splice_ult)
exp_ops["&&"]   = (2, splice_logical_and)


hw_vars = {}
hw_vars["$EuCoresTotalCount"] = "perf->sys_vars.n_eus"
hw_vars["$EuSlicesTotalCount"] = "perf->sys_vars.n_eu_slices"
hw_vars["$EuSubslicesTotalCount"] = "perf->sys_vars.n_eu_sub_slices"
hw_vars["$EuThreadsCount"] = "perf->sys_vars.eu_threads_count"
hw_vars["$SliceMask"] = "perf->sys_vars.slice_mask"
# subslice_mask is interchangeable with subslice/dual-subslice since Gen12+
# only has dual subslices which can be assimilated with 16EUs subslices.
hw_vars["$SubsliceMask"] = "perf->sys_vars.subslice_mask"
hw_vars["$DualSubsliceMask"] = "perf->sys_vars.subslice_mask"
hw_vars["$GpuTimestampFrequency"] = "perf->sys_vars.timestamp_frequency"
hw_vars["$GpuMinFrequency"] = "perf->sys_vars.gt_min_freq"
hw_vars["$GpuMaxFrequency"] = "perf->sys_vars.gt_max_freq"
hw_vars["$SkuRevisionId"] = "perf->sys_vars.revision"

def output_rpn_equation_code(set, counter, equation):
    c("/* RPN equation: " + equation + " */")
    tokens = equation.split()
    stack = []
    tmp_id = 0
    tmp = None

    for token in tokens:
        stack.append(token)
        while stack and stack[-1] in ops:
            op = stack.pop()
            argc, callback = ops[op]
            args = []
            for i in range(0, argc):
                operand = stack.pop()
                if operand[0] == "$":
                    if operand in hw_vars:
                        operand = hw_vars[operand]
                    elif operand in set.counter_vars:
                        reference = set.counter_vars[operand]
                        operand = set.read_funcs[operand[1:]] + "(perf, query, accumulator)"
                    else:
                        raise Exception("Failed to resolve variable " + operand + " in equation " + equation + " for " + set.name + " :: " + counter.get('name'));
                args.append(operand)

            tmp_id = callback(tmp_id, args)

            tmp = "tmp{0}".format(tmp_id - 1)
            stack.append(tmp)

    if len(stack) != 1:
        raise Exception("Spurious empty rpn code for " + set.name + " :: " +
                counter.get('name') + ".\nThis is probably due to some unhandled RPN function, in the equation \"" +
                equation + "\"")

    value = stack[-1]

    if value in hw_vars:
        value = hw_vars[value]
    if value in set.counter_vars:
        value = set.read_funcs[value[1:]] + "(perf, query, accumulator)"

    c("\nreturn " + value + ";")

def splice_rpn_expression(set, counter, expression):
    tokens = expression.split()
    stack = []

    for token in tokens:
        stack.append(token)
        while stack and stack[-1] in exp_ops:
            op = stack.pop()
            argc, callback = exp_ops[op]
            args = []
            for i in range(0, argc):
                operand = stack.pop()
                if operand[0] == "$":
                    if operand in hw_vars:
                        operand = hw_vars[operand]
                    else:
                        raise Exception("Failed to resolve variable " + operand + " in expression " + expression + " for " + set.name + " :: " + counter.get('name'));
                args.append(operand)

            subexp = callback(args)

            stack.append(subexp)

    if len(stack) != 1:
        raise Exception("Spurious empty rpn expression for " + set.name + " :: " +
                counter.get('name') + ".\nThis is probably due to some unhandled RPN operation, in the expression \"" +
                expression + "\"")

    return stack[-1]

def output_counter_read(gen, set, counter):
    c("\n")
    c("/* {0} :: {1} */".format(set.name, counter.get('name')))

    if counter.read_hash in hashed_funcs:
        c("#define %s \\" % counter.read_sym)
        c_indent(3)
        c("%s" % hashed_funcs[counter.read_hash])
        c_outdent(3)
    else:
        ret_type = counter.get('data_type')
        if ret_type == "uint64":
            ret_type = "uint64_t"

        read_eq = counter.get('equation')

        c("static " + ret_type)
        c(counter.read_sym + "(UNUSED struct gen_perf_config *perf,\n")
        c_indent(len(counter.read_sym) + 1)
        c("const struct gen_perf_query_info *query,\n")
        c("const uint64_t *accumulator)\n")
        c_outdent(len(counter.read_sym) + 1)

        c("{")
        c_indent(3)
        output_rpn_equation_code(set, counter, read_eq)
        c_outdent(3)
        c("}")

        hashed_funcs[counter.read_hash] = counter.read_sym


def output_counter_max(gen, set, counter):
    max_eq = counter.get('max_equation')

    if not counter.has_max_func():
        return

    c("\n")
    c("/* {0} :: {1} */".format(set.name, counter.get('name')))

    if counter.max_hash in hashed_funcs:
        c("#define %s \\" % counter.max_sym())
        c_indent(3)
        c("%s" % hashed_funcs[counter.max_hash])
        c_outdent(3)
    else:
        ret_type = counter.get('data_type')
        if ret_type == "uint64":
            ret_type = "uint64_t"

        c("static " + ret_type)
        c(counter.max_sym() + "(struct gen_perf_config *perf)\n")
        c("{")
        c_indent(3)
        output_rpn_equation_code(set, counter, max_eq)
        c_outdent(3)
        c("}")

        hashed_funcs[counter.max_hash] = counter.max_sym()


c_type_sizes = { "uint32_t": 4, "uint64_t": 8, "float": 4, "double": 8, "bool": 4 }
def sizeof(c_type):
    return c_type_sizes[c_type]

def pot_align(base, pot_alignment):
    return (base + pot_alignment - 1) & ~(pot_alignment - 1);

semantic_type_map = {
    "duration": "raw",
    "ratio": "event"
    }

def output_availability(set, availability, counter_name):
    expression = splice_rpn_expression(set, counter_name, availability)
    lines = expression.split(' && ')
    n_lines = len(lines)
    if n_lines == 1:
        c("if (" + lines[0] + ") {")
    else:
        c("if (" + lines[0] + " &&")
        c_indent(4)
        for i in range(1, (n_lines - 1)):
            c(lines[i] + " &&")
        c(lines[(n_lines - 1)] + ") {")
        c_outdent(4)


def output_counter_report(set, counter, current_offset):
    data_type = counter.get('data_type')
    data_type_uc = data_type.upper()
    c_type = data_type

    if "uint" in c_type:
        c_type = c_type + "_t"

    semantic_type = counter.get('semantic_type')
    if semantic_type in semantic_type_map:
        semantic_type = semantic_type_map[semantic_type]

    semantic_type_uc = semantic_type.upper()

    c("\n")

    availability = counter.get('availability')
    if availability:
        output_availability(set, availability, counter.get('name'))
        c_indent(3)

    c("counter = &query->counters[query->n_counters++];\n")
    c("counter->oa_counter_read_" + data_type + " = " + set.read_funcs[counter.get('symbol_name')] + ";\n")
    c("counter->name = \"" + counter.get('name') + "\";\n")
    c("counter->desc = \"" + counter.get('description') + "\";\n")
    c("counter->type = GEN_PERF_COUNTER_TYPE_" + semantic_type_uc + ";\n")
    c("counter->data_type = GEN_PERF_COUNTER_DATA_TYPE_" + data_type_uc + ";\n")
    c("counter->raw_max = " + set.max_values[counter.get('symbol_name')] + ";\n")

    current_offset = pot_align(current_offset, sizeof(c_type))
    c("counter->offset = " + str(current_offset) + ";\n")

    if availability:
        c_outdent(3);
        c("}")

    return current_offset + sizeof(c_type)


register_types = {
    'FLEX': 'flex_regs',
    'NOA': 'mux_regs',
    'OA': 'b_counter_regs',
}

def compute_register_lengths(set):
    register_lengths = {}
    register_configs = set.findall('register_config')
    for register_config in register_configs:
        t = register_types[register_config.get('type')]
        if t not in register_lengths:
            register_lengths[t] = len(register_config.findall('register'))
        else:
            register_lengths[t] += len(register_config.findall('register'))

    return register_lengths


def generate_register_configs(set):
    register_configs = set.findall('register_config')
    for register_config in register_configs:
        t = register_types[register_config.get('type')]

        availability = register_config.get('availability')
        if availability:
            output_availability(set, availability, register_config.get('type') + ' register config')
            c_indent(3)

        for register in register_config.findall('register'):
            c("query->config.%s[query->config.n_%s++] = (struct gen_perf_query_register_prog) { .reg = %s, .val = %s };" %
              (t, t, register.get('address'), register.get('value')))

        if availability:
            c_outdent(3)
            c("}")
        c("\n")


# Wraps a <counter> element from the oa-*.xml files.
class Counter:
    def __init__(self, set, xml):
        self.xml = xml
        self.set = set
        self.read_hash = None
        self.max_hash = None

        self.read_sym = "{0}__{1}__{2}__read".format(self.set.gen.chipset,
                                                     self.set.underscore_name,
                                                     self.xml.get('underscore_name'))

    def get(self, prop):
        return self.xml.get(prop)

    # Compute the hash of a counter's equation by expanding (including all the
    # sub-equations it depends on)
    def compute_hashes(self):
        if self.read_hash is not None:
            return

        def replace_token(token):
            if token[0] != "$":
                return token
            if token not in self.set.counter_vars:
                return token
            self.set.counter_vars[token].compute_hashes()
            return self.set.counter_vars[token].read_hash

        read_eq = self.xml.get('equation')
        self.read_hash = ' '.join(map(replace_token, read_eq.split()))

        max_eq = self.xml.get('max_equation')
        if max_eq:
            self.max_hash = ' '.join(map(replace_token, max_eq.split()))

    def has_max_func(self):
        max_eq = self.xml.get('max_equation')
        if not max_eq:
            return False

        try:
            val = float(max_eq)
            return False
        except ValueError:
            pass

        for token in max_eq.split():
            if token[0] == '$' and token not in hw_vars:
                return False
        return True

    def max_sym(self):
        assert self.has_max_func()
        return "{0}__{1}__{2}__max".format(self.set.gen.chipset,
                                           self.set.underscore_name,
                                           self.xml.get('underscore_name'))

    def max_value(self):
        max_eq = self.xml.get('max_equation')
        if not max_eq:
            return "0 /* undefined */"

        try:
            return "{0}".format(float(max_eq))
        except ValueError:
            pass

        for token in max_eq.split():
            if token[0] == '$' and token not in hw_vars:
                return "0 /* unsupported (varies over time) */"

        return "{0}__{1}__{2}__max(perf)".format(self.set.gen.chipset,
                                                 self.set.underscore_name,
                                                 self.xml.get('underscore_name'))

# Wraps a <set> element from the oa-*.xml files.
class Set:
    def __init__(self, gen, xml):
        self.gen = gen
        self.xml = xml

        self.counter_vars = {}
        self.max_values = {}
        self.read_funcs = {}

        xml_counters = self.xml.findall("counter")
        self.counters = []
        for xml_counter in xml_counters:
            counter = Counter(self, xml_counter)
            self.counters.append(counter)
            self.counter_vars["$" + counter.get('symbol_name')] = counter
            self.read_funcs[counter.get('symbol_name')] = counter.read_sym
            self.max_values[counter.get('symbol_name')] = counter.max_value()

        for counter in self.counters:
            counter.compute_hashes()

    @property
    def hw_config_guid(self):
        return self.xml.get('hw_config_guid')

    @property
    def name(self):
        return self.xml.get('name')

    @property
    def symbol_name(self):
        return self.xml.get('symbol_name')

    @property
    def underscore_name(self):
        return self.xml.get('underscore_name')

    def findall(self, path):
        return self.xml.findall(path)

    def find(self, path):
        return self.xml.find(path)


# Wraps an entire oa-*.xml file.
class Gen:
    def __init__(self, filename):
        self.filename = filename
        self.xml = et.parse(self.filename)
        self.chipset = self.xml.find('.//set').get('chipset').lower()
        self.sets = []

        for xml_set in self.xml.findall(".//set"):
            self.sets.append(Set(self, xml_set))


def main():
    global c_file
    global header_file

    parser = argparse.ArgumentParser()
    parser.add_argument("--header", help="Header file to write", required=True)
    parser.add_argument("--code", help="C file to write", required=True)
    parser.add_argument("xml_files", nargs='+', help="List of xml metrics files to process")

    args = parser.parse_args()

    c_file = open(args.code, 'w')
    header_file = open(args.header, 'w')

    gens = []
    for xml_file in args.xml_files:
        gens.append(Gen(xml_file))


    copyright = textwrap.dedent("""\
        /* Autogenerated file, DO NOT EDIT manually! generated by {}
         *
         * Copyright (c) 2015 Intel Corporation
         *
         * 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.
         */

        """).format(os.path.basename(__file__))

    h(copyright)
    h(textwrap.dedent("""\
        #pragma once

        struct gen_perf_config;

        """))

    c(copyright)
    c(textwrap.dedent("""\
        #include <stdint.h>
        #include <stdbool.h>

        #include <drm-uapi/i915_drm.h>

        #include "util/hash_table.h"

        """))

    c("#include \"" + os.path.basename(args.header) + "\"")

    c(textwrap.dedent("""\
        #include "perf/gen_perf.h"


        #define MIN(a, b) ((a < b) ? (a) : (b))
        #define MAX(a, b) ((a > b) ? (a) : (b))


        """))

    # Print out all equation functions.
    for gen in gens:
        for set in gen.sets:
            for counter in set.counters:
                output_counter_read(gen, set, counter)
                output_counter_max(gen, set, counter)

    # Print out all metric sets registration functions for each set in each
    # generation.
    for gen in gens:
        for set in gen.sets:
            counters = set.counters

            c("\n")
            register_lengths = compute_register_lengths(set);
            for reg_type, reg_length in register_lengths.items():
                c("static struct gen_perf_query_register_prog {0}_{1}_{2}[{3}];".format(gen.chipset,
                                                                                        set.underscore_name,
                                                                                        reg_type, reg_length))

            c("\nstatic struct gen_perf_query_counter {0}_{1}_query_counters[{2}];\n".format(gen.chipset, set.underscore_name, len(counters)))
            c("static struct gen_perf_query_info " + gen.chipset + "_" + set.underscore_name + "_query = {\n")
            c_indent(3)

            c(".kind = GEN_PERF_QUERY_TYPE_OA,\n")
            c(".name = \"" + set.name + "\",\n")
            c(".guid = \"" + set.hw_config_guid + "\",\n")

            c(".counters = {0}_{1}_query_counters,".format(gen.chipset, set.underscore_name))
            c(".n_counters = 0,")
            c(".oa_metrics_set_id = 0, /* determined at runtime, via sysfs */")

            if gen.chipset == "hsw":
                c(textwrap.dedent("""\
                    .oa_format = I915_OA_FORMAT_A45_B8_C8,

                    /* Accumulation buffer offsets... */
                    .gpu_time_offset = 0,
                    .a_offset = 1,
                    .b_offset = 46,
                    .c_offset = 54,
                """))
            else:
                c(textwrap.dedent("""\
                    .oa_format = I915_OA_FORMAT_A32u40_A4u32_B8_C8,

                    /* Accumulation buffer offsets... */
                    .gpu_time_offset = 0,
                    .gpu_clock_offset = 1,
                    .a_offset = 2,
                    .b_offset = 38,
                    .c_offset = 46,
                """))

            c(".config = {")
            c_indent(3)
            for reg_type, reg_length in register_lengths.items():
                c(".{0} = {1}_{2}_{3},".format(reg_type, gen.chipset, set.underscore_name, reg_type))
                c(".n_{0} = 0, /* Determined at runtime */".format(reg_type))
            c_outdent(3)
            c("},")

            c_outdent(3)
            c("};\n")

            c("\nstatic void\n")
            c("{0}_register_{1}_counter_query(struct gen_perf_config *perf)\n".format(gen.chipset, set.underscore_name))
            c("{\n")
            c_indent(3)

            c("static struct gen_perf_query_info *query = &" + gen.chipset + "_" + set.underscore_name + "_query;\n")
            c("struct gen_perf_query_counter *counter;\n")

            c("\n")
            c("/* Note: we're assuming there can't be any variation in the definition ")
            c(" * of a query between contexts so it's ok to describe a query within a ")
            c(" * global variable which only needs to be initialized once... */")
            c("\nif (!query->data_size) {")
            c_indent(3)

            generate_register_configs(set)

            offset = 0
            for counter in counters:
                offset = output_counter_report(set, counter, offset)


            c("\nquery->data_size = counter->offset + gen_perf_query_counter_get_size(counter);\n")

            c_outdent(3)
            c("}");

            c("\n_mesa_hash_table_insert(perf->oa_metrics_table, query->guid, query);")

            c_outdent(3)
            c("}\n")

        h("void gen_oa_register_queries_" + gen.chipset + "(struct gen_perf_config *perf);\n")

        c("\nvoid")
        c("gen_oa_register_queries_" + gen.chipset + "(struct gen_perf_config *perf)")
        c("{")
        c_indent(3)

        for set in gen.sets:
            c("{0}_register_{1}_counter_query(perf);".format(gen.chipset, set.underscore_name))

        c_outdent(3)
        c("}")


if __name__ == '__main__':
    main()