smyalygames/FiniteVolumeGPU
1
0
Fork 0
mirror of https://github.com/smyalygames/FiniteVolumeGPU.git synced 2026-01-14 15:48:43 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity

Async mem ops

Browse Source
This commit is contained in:
Martin Lilleeng Sætra
2022-04-26 11:34:29 +00:00
parent dcef56a1b9
commit 6c8bac6f7b
6 changed files with 314 additions and 57 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
GPUSimulators/Common.py
View File

@@ -546,9 +546,9 @@ class CudaArray2D:
#self.logger.debug("Downloading [%dx%d] buffer", self.nx, self.ny)
#Allocate host memory
#The following fails, don't know why (crashes python)
#cpu_data = cuda.pagelocked_empty((self.ny, self.nx), np.float32)32)
cpu_data = cuda.pagelocked_empty((int(ny), int(nx)), dtype=np.float32, mem_flags=cuda.host_alloc_flags.PORTABLE)
#Non-pagelocked: cpu_data = np.empty((ny, nx), dtype=np.float32)
cpu_data = self.memorypool.allocate((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]
@@ -759,7 +759,7 @@ class ArakawaA2D:
assert i < len(self.gpu_variables), "Variable {:d} is out of range".format(i)
cpu_variables += [self.gpu_variables[i].download(stream, asynch=True)]
stream.synchronize()
#stream.synchronize()
return cpu_variables
def check(self):