source['arg_index'] = int(
re.search('arg_index(\s*)=(\s*)(?P<arg_index>\d+)',
source['text']).group('arg_index'))
source['arg_value_as_str'] = re.search(
'arg_value\[\](\s*)=(\s*)\"(?P<arg_value>.+)\"',
source['text']).group('arg_value')
source['arg_size'] = int(
re.search('arg_size(\s*)=(\s*)(?P<arg_size>\d+)',
source['text']).group('arg_size'))
# Read from stdin (via pipe).
output = sys.stdin.read()
print 'OUTPUT'
print output
result = prof_parse(output)[0]
result['arg_value_as_str'] = hex_str_as_str(result['arg_value'])
print 'RESULT'
print result
print
status = True
status &= (source['kernel'] == result['kernel'])
status &= (source['arg_index'] == result['arg_index'])
status &= (source['arg_size'] == result['arg_size'])
status &= (source['arg_value_as_str'] == result['arg_value_as_str'])
status &= (0 == result['errcode'])
print '%s%s: %s' % (call, _id, 'PASSED' if status else 'FAILED')
print
def process(i):
"""
Input: {
file_in - stdout to parse
file_out - file with raw dvdt prof data
data - dict with CK data to be updated
env - CK env to customize output if needed
CK_ADD_RAW_DVDT_PROF (if yes then add all raw data to CK pipeline)
deps - CK deps to get info about dvdt_prof
}
Output: {
return - return code = 0, if successful
> 0, if error
(error) - error text if return > 0
}
"""
ck=i['ck_kernel']
env=i['env']
d=i['data']
file_in=i['file_in']
dvdt_prof_file=i['file_out']
deps=i['deps']
# Check if DVDT prof if actually used
dvdt_prof_dep=deps.get('dvdt_prof',{})
if dvdt_prof_dep.get('uoa','')!='':
r=ck.save_json_to_file({'json_file':'tmp-dvdt-prof-deps.json', 'dict':dvdt_prof_dep, 'sort_keys':'yes'})
if r['return']>0: return r
# Load output.
r=ck.load_text_file({
'text_file':file_in,
'split_to_list':'no'
})
if r['return']>0: return r
# Locate profiler parser.
dvdt_prof_dir=dvdt_prof_dep['dict']['env']['CK_ENV_TOOL_DVDT_PROF']
dvdt_prof_src_python=os.path.join(dvdt_prof_dir,'src','python')
sys.path.append(dvdt_prof_src_python)
from prof_parser import prof_parse
# Parse profiler output.
dvdt_prof=prof_parse(r['string'])
r=ck.save_json_to_file({'json_file':dvdt_prof_file, 'dict':dvdt_prof, 'sort_keys':'yes'})
if r['return']>0: return r
if env.get('CK_ADD_RAW_DVDT_PROF','').lower()=='yes':
d['dvdt_prof']=dvdt_prof
nqs=[ call for call in dvdt_prof if call['call'] in ['clEnqueueNDRangeKernel'] ]
# d['execution_time_opencl_us']={ nq['name'] : (nq['profiling']['end']-nq['profiling']['start'])*1e-3 for nq in nqs }
# d['execution_time_opencl_ms']={ nq['name'] : (nq['profiling']['end']-nq['profiling']['start'])*1e-6 for nq in nqs }
# d['execution_time_opencl_s' ]={ nq['name'] : (nq['profiling']['end']-nq['profiling']['start'])*1e-9 for nq in nqs }
d['execution_time_opencl_us']={}
d['execution_time_opencl_ms']={}
d['execution_time_opencl_s' ]={}
d['execution_time_list_opencl']=[]
seq=0
for nq in nqs:
kernel_name=nq['name']
kernel_time=nq['profiling']['end']-nq['profiling']['start']
d['execution_time_list_opencl'].append({'kernel_name':kernel_name,
'kernel_time':kernel_time,
'sequence':seq,
'lws':nq.get('lws',[]),
'gws':nq.get('gws',[])})
if kernel_name not in d['execution_time_opencl_us']:
d['execution_time_opencl_us'][kernel_name]=0.0
d['execution_time_opencl_ms'][kernel_name]=0.0
d['execution_time_opencl_s' ][kernel_name]=0.0
d['execution_time_opencl_us'][kernel_name]+=kernel_time*1e-3
d['execution_time_opencl_ms'][kernel_name]+=kernel_time*1e-6
d['execution_time_opencl_s' ][kernel_name]+=kernel_time*1e-9
seq+=1
return {'return':0}
def get_result(output):
result = prof_parse(output)[0]
return result
