smyalygames/FiniteVolumeGPU
1
0
Fork 0
mirror of https://github.com/smyalygames/FiniteVolumeGPU.git synced 2025-05-18 14:34:13 +02:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #6 from babrodtk/autotuning

Browse Source 
Updated domain size benchmark in autotuning
This commit is contained in:
André R. Brodtkorb 2018-08-23 16:06:35 +02:00 committed by GitHub
commit 1ff3485918
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
5 changed files with 303 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

281
Autotuning.ipynb
View File

File diff suppressed because one or more lines are too long

2
GPUSimulators/Autotuner.py
View File

@ -29,7 +29,7 @@ from socket import gethostname
import pycuda.driver as cuda import pycuda.driver as cuda
from GPUSimulators import Common, LxF, FORCE, HLL, HLL2, KP07, KP07_dimsplit, WAF from GPUSimulators import Common, Simulator
class Autotuner: class Autotuner:
def __init__(self, def __init__(self,

3
GPUSimulators/Common.py
View File

@ -100,7 +100,7 @@ class CudaContext(object):
self.cache_path = os.path.join(self.module_path, "cuda_cache") self.cache_path = os.path.join(self.module_path, "cuda_cache")
if not os.path.isdir(self.cache_path): if not os.path.isdir(self.cache_path):
os.mkdir(self.cache_path) os.mkdir(self.cache_path)
self.logger.debug("Using CUDA cache dir %s", self.cache_path) self.logger.info("Using CUDA cache dir %s", self.cache_path)
self.autotuner = None self.autotuner = None
if (autotuning): if (autotuning):
@ -395,6 +395,7 @@ class SWEDataArakawaA:
Uploads initial data to the CL device Uploads initial data to the CL device
""" """
def __init__(self, stream, nx, ny, halo_x, halo_y, h0, hu0, hv0): def __init__(self, stream, nx, ny, halo_x, halo_y, h0, hu0, hv0):
self.logger = logging.getLogger(__name__)
self.h0 = CudaArray2D(stream, nx, ny, halo_x, halo_y, h0) self.h0 = CudaArray2D(stream, nx, ny, halo_x, halo_y, h0)
self.hu0 = CudaArray2D(stream, nx, ny, halo_x, halo_y, hu0) self.hu0 = CudaArray2D(stream, nx, ny, halo_x, halo_y, hu0)
self.hv0 = CudaArray2D(stream, nx, ny, halo_x, halo_y, hv0) self.hv0 = CudaArray2D(stream, nx, ny, halo_x, halo_y, hv0)

32
GPUSimulators/IPythonMagic.py
View File

@ -31,33 +31,43 @@ from GPUSimulators import Common
@magics_class @magics_class
class MyIPythonMagic(Magics): class MyIPythonMagic(Magics):
@line_magic @line_magic
def cuda_context_handler(self, context_name): @magic_arguments.magic_arguments()
@magic_arguments.argument(
'name', type=str, help='Name of context to create')
@magic_arguments.argument(
'--blocking', '-b', action="store_true", help='Enable blocking context')
@magic_arguments.argument(
'--no_cache', '-nc', action="store_true", help='Disable caching of kernels')
@magic_arguments.argument(
'--no_autotuning', '-na', action="store_true", help='Disable autotuning of kernels')
def cuda_context_handler(self, line):
args = magic_arguments.parse_argstring(self.cuda_context_handler, line)
self.logger = logging.getLogger(__name__) self.logger = logging.getLogger(__name__)
self.logger.debug("Registering %s as a global context", context_name) self.logger.info("Registering %s in user workspace", args.name)
if context_name in self.shell.user_ns.keys(): if args.name in self.shell.user_ns.keys():
self.logger.debug("Context already registered! Ignoring") self.logger.debug("Context already registered! Ignoring")
return return
else: else:
self.logger.debug("Creating context") self.logger.debug("Creating context")
#self.shell.ex(context_name + " = Common.CudaContext(blocking=False)") use_cache = False if args.no_cache else True
self.shell.user_ns[context_name] = Common.CudaContext(blocking=False) use_autotuning = False if args.no_autotuning else True
self.shell.user_ns[args.name] = Common.CudaContext(blocking=args.blocking, use_cache=use_cache, autotuning=use_autotuning)
# this function will be called on exceptions in any cell # this function will be called on exceptions in any cell
def custom_exc(shell, etype, evalue, tb, tb_offset=None): def custom_exc(shell, etype, evalue, tb, tb_offset=None):
self.logger.exception("Exception caught: Resetting to CUDA context %s", context_name) self.logger.exception("Exception caught: Resetting to CUDA context %s", args.name)
while (cuda.Context.get_current() != None): while (cuda.Context.get_current() != None):
context = cuda.Context.get_current() context = cuda.Context.get_current()
self.logger.info("Popping <%s>", str(context.handle)) self.logger.info("Popping <%s>", str(context.handle))
cuda.Context.pop() cuda.Context.pop()
if context_name in self.shell.user_ns.keys(): if args.name in self.shell.user_ns.keys():
self.logger.info("Pushing <%s>", str(self.shell.user_ns[context_name].cuda_context.handle)) self.logger.info("Pushing <%s>", str(self.shell.user_ns[args.name].cuda_context.handle))
#self.shell.ex(context_name + ".cuda_context.push()") self.shell.user_ns[args.name].cuda_context.push()
self.shell.user_ns[context_name].cuda_context.push()
else: else:
self.logger.error("No CUDA context called %s found (something is wrong)", context_name) self.logger.error("No CUDA context called %s found (something is wrong)", args.name)
self.logger.error("CUDA will not work now") self.logger.error("CUDA will not work now")
self.logger.debug("==================================================================") self.logger.debug("==================================================================")

2
GPUSimulators/Simulator.py
View File

@ -69,6 +69,8 @@ class BaseSimulator:
self.stream = cuda.Stream() self.stream = cuda.Stream()
#Create data by uploading to device #Create data by uploading to device
free, total = cuda.mem_get_info()
self.logger.debug("GPU memory: %d / %d MB available", int(free/(1024*1024)), int(total/(1024*1024)))
self.data = Common.SWEDataArakawaA(self.stream, \ self.data = Common.SWEDataArakawaA(self.stream, \
nx, ny, \ nx, ny, \
ghost_cells_x, ghost_cells_y, \ ghost_cells_x, ghost_cells_y, \