smyalygames/FiniteVolumeGPU
1
0
Fork 0
mirror of https://github.com/smyalygames/FiniteVolumeGPU.git synced 2025-10-31 13:07:41 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

Pinned memory

Browse Source
This commit is contained in:
André R. Brodtkorb 2018-12-10 12:08:10 +01:00
parent 12174b39db
commit 2c6ecc8d88
2 changed files with 18 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
GPUSimulators/Common.py
View File

@ -38,6 +38,7 @@ import json
import pycuda.compiler as cuda_compiler
import pycuda.gpuarray
import pycuda.driver as cuda
from pycuda.tools import PageLockedMemoryPool
@ -482,6 +483,9 @@ class CudaArray2D:
#Should perhaps use pycuda.driver.mem_alloc_data.pitch() here
self.data = pycuda.gpuarray.zeros((ny_halo, nx_halo), dtype)
#For returning to download
self.memorypool = PageLockedMemoryPool()
#If we don't have any data, just allocate and return
if cpu_data is None:
return
@ -518,8 +522,10 @@ class CudaArray2D:
if (cpu_data is None):
#self.logger.debug("Downloading [%dx%d] buffer", self.nx, self.ny)
#Allocate host memory
#cpu_data = cuda.pagelocked_empty((self.ny, self.nx), np.float32)
cpu_data = np.empty((ny, nx), dtype=np.float32)
#The following fails, don't know why (crashes python)
#cpu_data = cuda.pagelocked_empty((self.ny, self.nx), np.float32)32)
#Non-pagelocked: cpu_data = np.empty((ny, nx), dtype=np.float32)
cpu_data = self.memorypool.allocate((ny, nx), dtype=np.float32)
assert nx == cpu_data.shape[1]
assert ny == cpu_data.shape[0]
@ -610,6 +616,9 @@ class CudaArray3D:
#Should perhaps use pycuda.driver.mem_alloc_data.pitch() here
self.data = pycuda.gpuarray.zeros((nz_halo, ny_halo, nx_halo), dtype)
#For returning to download
self.memorypool = PageLockedMemoryPool()
#If we don't have any data, just allocate and return
if cpu_data is None:
return
@ -662,7 +671,8 @@ class CudaArray3D:
#self.logger.debug("Downloading [%dx%d] buffer", self.nx, self.ny)
#Allocate host memory
#cpu_data = cuda.pagelocked_empty((self.ny, self.nx), np.float32)
cpu_data = np.empty((self.nz, self.ny, self.nx), dtype=np.float32)
#cpu_data = np.empty((self.nz, self.ny, self.nx), dtype=np.float32)
cpu_data = self.memorypool.allocate((self.nz, self.ny, self.nx), dtype=np.float32)
#Create copy object from device to host
copy = cuda.Memcpy2D()

6
GPUSimulators/IPythonMagic.py
View File

@ -47,6 +47,10 @@ class MagicCudaContext(Magics):
self.logger.info("Registering %s in user workspace", args.name)
context_flags = None
if (args.blocking):
context_flags = cuda.ctx_flags.SCHED_BLOCKING_SYNC
if args.name in self.shell.user_ns.keys():
self.logger.debug("Context already registered! Ignoring")
return
@ -54,7 +58,7 @@ class MagicCudaContext(Magics):
self.logger.debug("Creating context")
use_cache = False if args.no_cache else True
use_autotuning = False if args.no_autotuning else True
self.shell.user_ns[args.name] = CudaContext.CudaContext(blocking=args.blocking, use_cache=use_cache, autotuning=use_autotuning)
self.shell.user_ns[args.name] = CudaContext.CudaContext(context_flags=context_flags, use_cache=use_cache, autotuning=use_autotuning)
# this function will be called on exceptions in any cell
def custom_exc(shell, etype, evalue, tb, tb_offset=None):