Files
rocm-systems/script/hsaap.py
T
Laurent Morichetti 3adc56bd9b Add the roctracer-tests package
The roctracer-tests package contains all the roctracer test binaries
and scripts needed to run the testsuite outside of the build directory.

Change-Id: Id11f862fb4bdb2425d68f455074172c38814ec92
2022-07-05 17:03:21 -04:00

543 lines
20 KiB
Python
Executable File

#!/usr/bin/env python3
################################################################################
# Copyright (c) 2018-2022 Advanced Micro Devices, Inc.
#
# 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 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.
################################################################################
from __future__ import print_function
import os, sys, re
OUT='hsa_prof_str.h'
API_TABLES_H = 'hsa_api_trace.h'
API_HEADERS_H = (
('CoreApiTable', 'hsa.h'),
('AmdExtTable', 'hsa_ext_amd.h'),
('ImageExtTable', 'hsa_ext_image.h'),
('AmdExtTable', API_TABLES_H),
)
LICENSE = \
'/*\n' + \
'Copyright (c) 2018 Advanced Micro Devices, Inc. All rights reserved.\n' + \
'\n' + \
'Permission is hereby granted, free of charge, to any person obtaining a copy\n' + \
'of this software and associated documentation files (the "Software"), to deal\n' + \
'in the Software without restriction, including without limitation the rights\n' + \
'to use, copy, modify, merge, publish, distribute, sublicense, and/or sell\n' + \
'copies of the Software, and to permit persons to whom the Software is\n' + \
'furnished to do so, subject to the following conditions:\n' + \
'\n' + \
'The above copyright notice and this permission notice shall be included in\n' + \
'all copies or substantial portions of the Software.\n' + \
'\n' + \
'THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\n' + \
'IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\n' + \
'FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE\n' + \
'AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\n' + \
'LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,\n' + \
'OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN\n' + \
'THE SOFTWARE.\n' + \
'*/\n'
#############################################################
# Error handler
def fatal(module, msg):
print (module + ' Error: "' + msg + '"', file = sys.stderr)
sys.exit(1)
# Get next text block
def NextBlock(pos, record):
if len(record) == 0: return pos
space_pattern = re.compile(r'(\s+)')
word_pattern = re.compile(r'([\w\*]+)')
if record[pos] != '(':
m = space_pattern.match(record, pos)
if not m:
m = word_pattern.match(record, pos)
if m:
return pos + len(m.group(1))
else:
fatal('NextBlock', "bad record '" + record + "' pos(" + str(pos) + ")")
else:
count = 0
for index in range(pos, len(record)):
if record[index] == '(':
count = count + 1
elif record[index] == ')':
count = count - 1
if count == 0:
index = index + 1
break
if count != 0:
fatal('NextBlock', "count is not zero (" + str(count) + ")")
if record[index - 1] != ')':
fatal('NextBlock', "last char is not ')' '" + record[index - 1] + "'")
return index
#############################################################
# API table parser class
class API_TableParser:
def fatal(self, msg):
fatal('API_TableParser', msg)
def __init__(self, header, name):
self.name = name
if not os.path.isfile(header):
self.fatal("file '" + header + "' not found")
self.inp = open(header, 'r')
self.beg_pattern = re.compile('^\s*struct\s+' + name + '\s*{\s*$')
self.end_pattern = re.compile('^\s*};\s*$')
self.array = []
self.parse()
# normalizing a line
def norm_line(self, line):
return re.sub(r'^\s+', r' ', line[:-1])
# check for start record
def is_start(self, record):
return self.beg_pattern.match(record)
# check for end record
def is_end(self, record):
return self.end_pattern.match(record)
# check for declaration entry record
def is_entry(self, record):
return re.match(r'^\s*decltype\(([^\)]*)\)', record)
# parse method
def parse(self):
active = 0
for line in self.inp.readlines():
record = self.norm_line(line)
if self.is_start(record): active = 1
if active != 0:
if self.is_end(record): return
m = self.is_entry(record)
if m:
self.array.append(m.group(1))
#############################################################
# API declaration parser class
class API_DeclParser:
def fatal(self, msg):
fatal('API_DeclParser', msg)
def __init__(self, header, array, data):
if not os.path.isfile(header):
self.fatal("file '" + header + "' not found")
self.inp = open(header, 'r')
self.end_pattern = re.compile('\);\s*$')
self.data = data
for call in array:
if call in data:
self.fatal(call + ' is already found')
self.parse(call)
# api record filter
def api_filter(self, record):
record = re.sub(r'\sHSA_API\s', r' ', record)
record = re.sub(r'\sHSA_DEPRECATED\s', r' ', record)
return record
# check for start record
def is_start(self, call, record):
return re.search('\s' + call + '\s*\(', record)
# check for API method record
def is_api(self, call, record):
record = self.api_filter(record)
return re.match('\s+\S+\s+' + call + '\s*\(', record)
# check for end record
def is_end(self, record):
return self.end_pattern.search(record)
# parse method args
def get_args(self, record):
struct = {'ret': '', 'args': '', 'astr': {}, 'alst': [], 'tlst': []}
record = re.sub(r'^\s+', r'', record)
record = re.sub(r'\s*(\*+)\s*', r'\1 ', record)
rind = NextBlock(0, record)
struct['ret'] = record[0:rind]
pos = record.find('(')
end = NextBlock(pos, record);
args = record[pos:end]
args = re.sub(r'^\(\s*', r'', args)
args = re.sub(r'\s*\)$', r'', args)
args = re.sub(r'\s*,\s*', r',', args)
struct['args'] = re.sub(r',', r', ', args)
if len(args) == 0: return struct
pos = 0
args = args + ','
while pos < len(args):
ind1 = NextBlock(pos, args) # type
ind2 = NextBlock(ind1, args) # space
if args[ind2] != '(':
while ind2 < len(args):
end = NextBlock(ind2, args)
if args[end] == ',': break
else: ind2 = end
name = args[ind2:end]
else:
ind3 = NextBlock(ind2, args) # field
m = re.match(r'\(\s*\*\s*(\S+)\s*\)', args[ind2:ind3])
if not m:
self.fatal("bad block3 '" + args + "' : '" + args[ind2:ind3] + "'")
name = m.group(1)
end = NextBlock(ind3, args) # the rest
item = args[pos:end]
struct['astr'][name] = item
struct['alst'].append(name)
struct['tlst'].append(item)
if args[end] != ',':
self.fatal("no comma '" + args + "'")
pos = end + 1
return struct
# parse given api
def parse(self, call):
record = ''
active = 0
found = 0
api_name = ''
prev_line = ''
self.inp.seek(0)
for line in self.inp.readlines():
record += ' ' + line[:-1]
record = re.sub(r'^\s*', r' ', record)
if active == 0:
if self.is_start(call, record):
active = 1
m = self.is_api(call, record)
if not m:
record = ' ' + prev_line + ' ' + record
m = self.is_api(call, record)
if not m:
self.fatal("bad api '" + line + "'")
if active == 1:
if self.is_end(record):
self.data[call] = self.get_args(record)
active = 0
found = 0
if active == 0: record = ''
prev_line = line
#############################################################
# API description parser class
class API_DescrParser:
def fatal(self, msg):
fatal('API_DescrParser', msg)
def __init__(self, out_file, hsa_dir, api_table_h, api_headers, license):
out_macro = re.sub(r'[\/\.]', r'_', out_file.upper()) + '_'
self.content = ''
self.api_names = []
self.api_calls = {}
self.api_rettypes = set()
self.api_id = {}
api_data = {}
api_list = []
ns_calls = []
for i in range(0, len(api_headers)):
(name, header) = api_headers[i]
if i < len(api_headers) - 1:
api = API_TableParser(hsa_dir + api_table_h, name)
api_list = api.array
self.api_names.append(name)
self.api_calls[name] = api_list
else:
api_list = ns_calls
ns_calls = []
for call in api_list:
if call in api_data:
self.fatal("call '" + call + "' is already found")
API_DeclParser(hsa_dir + header, api_list, api_data)
for call in api_list:
if not call in api_data:
# Not-supported functions
ns_calls.append(call)
else:
# API ID map
self.api_id[call] = 'HSA_API_ID_' + call
# Return types
self.api_rettypes.add(api_data[call]['ret'])
self.api_rettypes.discard('void')
self.api_data = api_data
self.ns_calls = ns_calls
self.content += "// automatically generated\n\n" + license + '\n'
self.content += "/////////////////////////////////////////////////////////////////////////////\n"
self.content += "//\n"
self.content += "// HSA API tracing primitives\n"
self.content += "//\n"
for (name, header) in api_headers:
self.content += "// '" + name + "', header '" + header + "', " + str(len(self.api_calls[name])) + ' funcs\n'
for call in self.ns_calls:
self.content += '// ' + call + ' was not parsed\n'
self.content += "//\n"
self.content += "/////////////////////////////////////////////////////////////////////////////\n"
self.content += '\n'
self.content += '#ifndef ' + out_macro + '\n'
self.content += '#define ' + out_macro + '\n'
self.add_section('API ID enumeration', ' ', self.gen_id_enum)
self.add_section('API arg structure', ' ', self.gen_arg_struct)
self.content += '\n'
self.content += '#if PROF_API_IMPL\n'
self.content += '#include \"util/callback_table.h\"\n';
self.content += '#include <atomic>\n'
self.content += 'namespace roctracer {\n'
self.content += 'namespace hsa_support {\n'
self.content += 'std::atomic<uint64_t> hsa_counter_{1};\n'
self.content += 'static thread_local uint64_t hsa_correlation_id_tls = 0;\n'
self.add_section('API callback functions', '', self.gen_callbacks)
self.add_section('API intercepting code', '', self.gen_intercept)
self.add_section('API get_name function', ' ', self.gen_get_name)
self.add_section('API get_code function', ' ', self.gen_get_code)
self.content += '\n};};\n'
self.content += '#endif // PROF_API_IMPL\n'
self.add_section('API output stream', ' ', self.gen_out_stream)
self.content += '\n'
self.content += '#endif // ' + out_macro
# add code section
def add_section(self, title, gap, fun):
n = 0
self.content += '\n// section: ' + title + '\n\n'
fun(-1, '-', '-', {})
for index in range(len(self.api_names)):
last = (index == len(self.api_names) - 1)
name = self.api_names[index]
if n != 0:
if gap == '': fun(n, name, '-', {})
self.content += '\n'
self.content += gap + '// block: ' + name + ' API\n'
for call in self.api_calls[name]:
fun(n, name, call, self.api_data[call])
n += 1
fun(n, '-', '-', {})
# generate API ID enumeration
def gen_id_enum(self, n, name, call, data):
if n == -1:
self.content += 'enum hsa_api_id_t {\n'
return
if call != '-':
self.content += ' ' + self.api_id[call] + ' = ' + str(n) + ',\n'
else:
self.content += '\n'
self.content += ' HSA_API_ID_DISPATCH = ' + str(n) + ',\n'
self.content += ' HSA_API_ID_NUMBER = ' + str(n + 1) + ',\n'
self.content += '};\n'
# generate API args structure
def gen_arg_struct(self, n, name, call, struct):
if n == -1:
self.content += 'struct hsa_api_data_t {\n'
self.content += ' uint64_t correlation_id;\n'
self.content += ' uint32_t phase;\n'
self.content += ' union {\n'
for ret_type in self.api_rettypes:
self.content += ' ' + ret_type + ' ' + ret_type + '_retval;\n'
self.content += ' };\n'
self.content += ' union {\n'
return
if call != '-':
self.content += ' struct {\n'
for (var, item) in struct['astr'].items():
self.content += ' ' + item + ';\n'
if call == "hsa_amd_memory_async_copy_rect" and item == "const hsa_dim3_t* range":
self.content += ' hsa_dim3_t range__val;\n'
self.content += ' } ' + call + ';\n'
else:
self.content += ' } args;\n'
self.content += '};\n'
# generate API callbacks
def gen_callbacks(self, n, name, call, struct):
if n == -1:
self.content += 'typedef util::CallbackTable<ACTIVITY_DOMAIN_HSA_API, HSA_API_ID_NUMBER> cb_table_t;\n'
self.content += 'extern cb_table_t cb_table;\n'
self.content += '\n'
if call != '-':
call_id = self.api_id[call];
ret_type = struct['ret']
self.content += 'static ' + ret_type + ' ' + call + '_callback(' + struct['args'] + ') {\n'
self.content += ' hsa_api_data_t api_data{};\n'
for var in struct['alst']:
item = struct['astr'][var];
if re.search(r'char\* ', item):
self.content += ' api_data.args.' + call + '.' + var + ' = ' + '(' + var + ' != NULL) ? strdup(' + var + ')' + ' : NULL;\n'
else:
self.content += ' api_data.args.' + call + '.' + var + ' = ' + var + ';\n'
if call == 'hsa_amd_memory_async_copy_rect' and var == 'range':
self.content += ' api_data.args.' + call + '.' + var + '__val = ' + '*(' + var + ');\n'
self.content += ' auto [ api_callback_fun, api_callback_arg ] = cb_table.Get(' + call_id + ');\n'
self.content += ' api_data.phase = 0;\n'
self.content += ' api_data.correlation_id = hsa_support::hsa_counter_.fetch_add(1, std::memory_order_relaxed);\n'
self.content += ' hsa_correlation_id_tls = api_data.correlation_id;\n'
self.content += ' if (api_callback_fun) api_callback_fun(ACTIVITY_DOMAIN_HSA_API, ' + call_id + ', &api_data, api_callback_arg);\n'
if ret_type != 'void':
self.content += ' ' + ret_type + ' ret ='
self.content += ' ' + name + '_saved.' + call + '_fn(' + ', '.join(struct['alst']) + ');\n'
if ret_type != 'void':
self.content += ' api_data.' + ret_type + '_retval = ret;\n'
self.content += ' api_data.phase = 1;\n'
self.content += ' if (api_callback_fun) api_callback_fun(ACTIVITY_DOMAIN_HSA_API, ' + call_id + ', &api_data, api_callback_arg);\n'
if ret_type != 'void':
self.content += ' return ret;\n'
self.content += '}\n'
# generate API intercepting code
def gen_intercept(self, n, name, call, struct):
if n > 0 and call == '-':
self.content += '};\n'
if n == 0 or (call == '-' and name != '-'):
self.content += 'static void intercept_' + name + '(' + name + '* table) {\n'
self.content += ' ' + name + '_saved = *table;\n'
if call != '-':
if call != 'hsa_shut_down':
self.content += ' table->' + call + '_fn = ' + call + '_callback;\n'
else:
self.content += ' { void* p = (void*)' + call + '_callback; (void)p; }\n'
# generate API name function
def gen_get_name(self, n, name, call, struct):
if n == -1:
self.content += 'static const char* GetApiName(const uint32_t& id) {\n'
self.content += ' switch (id) {\n'
return
if call != '-':
self.content += ' case ' + self.api_id[call] + ': return "' + call + '";\n'
else:
self.content += ' }\n'
self.content += ' return "unknown";\n'
self.content += '}\n'
# generate API code function
def gen_get_code(self, n, name, call, struct):
if n == -1:
self.content += 'static uint32_t GetApiCode(const char* str) {\n'
return
if call != '-':
self.content += ' if (strcmp("' + call + '", str) == 0) return ' + self.api_id[call] + ';\n'
else:
self.content += ' return HSA_API_ID_NUMBER;\n'
self.content += '}\n'
# generate stream operator
def gen_out_stream(self, n, name, call, struct):
if n == -1:
self.content += '#ifdef __cplusplus\n'
self.content += 'typedef std::pair<uint32_t, hsa_api_data_t> hsa_api_data_pair_t;\n'
self.content += 'inline std::ostream& operator<< (std::ostream& out, const hsa_api_data_pair_t& data_pair) {\n'
self.content += ' const uint32_t cid = data_pair.first;\n'
self.content += ' const hsa_api_data_t& api_data = data_pair.second;\n'
self.content += ' switch(cid) {\n'
return
if call != '-':
self.content += ' case ' + self.api_id[call] + ': {\n'
self.content += ' out << "' + call + '(";\n'
arg_list = struct['alst']
if len(arg_list) != 0:
for ind in range(len(arg_list)):
arg_var = arg_list[ind]
arg_val = 'api_data.args.' + call + '.' + arg_var
if re.search(r'char\* ', struct['astr'][arg_var]):
self.content += ' out << "0x" << std::hex << (uint64_t)' + arg_val
else:
self.content += ' out << ' + arg_val
if call == "hsa_amd_memory_async_copy_rect" and arg_var == "range":
self.content += ' << ", ";\n'
self.content += ' out << ' + arg_val + '__val'
'''
arg_item = struct['tlst'][ind]
if re.search(r'\(\* ', arg_item): arg_pref = ''
elif re.search(r'void\* ', arg_item): arg_pref = ''
elif re.search(r'\*\* ', arg_item): arg_pref = '**'
elif re.search(r'\* ', arg_item): arg_pref = '*'
else: arg_pref = ''
if arg_pref != '':
self.content += ' if (' + arg_val + ') out << ' + arg_pref + '(' + arg_val + '); else out << ' + arg_val
else:
self.content += ' out << ' + arg_val
'''
if ind < len(arg_list) - 1: self.content += ' << ", ";\n'
else: self.content += ';\n'
if struct['ret'] != 'void':
self.content += ' out << ") = " << api_data.' + struct['ret'] + '_retval;\n'
else:
self.content += ' out << ") = void";\n'
self.content += ' break;\n'
self.content += ' }\n'
else:
self.content += ' default:\n'
self.content += ' out << "ERROR: unknown API";\n'
self.content += ' abort();\n'
self.content += ' }\n'
self.content += ' return out;\n'
self.content += '}\n'
self.content += '#endif\n'
#############################################################
# main
# Usage
if len(sys.argv) != 3:
print ("Usage:", sys.argv[0], " <OUT prefix> <HSA runtime include path>", file=sys.stderr)
sys.exit(1)
else:
PREFIX = sys.argv[1] + '/'
HSA_DIR = sys.argv[2] + '/'
descr = API_DescrParser(OUT, HSA_DIR, API_TABLES_H, API_HEADERS_H, LICENSE)
out_file = PREFIX + OUT
print ('Generating "' + out_file + '"')
f = open(out_file, 'w')
f.write(descr.content[:-1])
f.close()
#############################################################