mirror-github-bspeice/cuburn
1
0
Fork 0
mirror of https://github.com/stevenrobertson/cuburn.git synced 2025-02-05 11:40:04 -05:00
Code Issues Actions Packages Projects Releases Wiki Activity

Changes to CUDA module loading

Browse Source 
Modules may (once again) be compiled and loaded in separate stages,
including compiling without having a CUDA context on hand. Also, modules
will be reused if they are already loaded.
This commit is contained in:
Steven Robertson 2012-05-20 13:05:28 -07:00
parent 5083fefad7
commit 8c7db9d0fc
1 changed files with 18 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
cuburn/render.py
View File

@ -200,17 +200,26 @@ class Renderer(object):
# asynchronous, and avoid expensive CPU polling, this hangs on to
# a number of (relatively small) CUDA modules and flushes them together.
MAX_MODREFS = 20
_modrefs = []
_modrefs = {}
@classmethod
def compile(cls, gnm, arch=None):
packer, lib = iter.mkiterlib(gnm)
cubin = util.compile('iter', assemble_code(lib), arch=arch)
return packer, lib, cubin
def load(self, cubin):
if cubin in self._modrefs:
return self._modrefs[cubin]
mod = cuda.module_from_buffer(self.cubin)
if len(self._modrefs) > self.MAX_MODREFS:
self._modrefs.clear()
self._modrefs[cubin] = mod
return mod
def __init__(self, gnm, gprof):
self.packer, self.lib = iter.mkiterlib(gnm)
cubin = util.compile('iter', assemble_code(self.lib))
self.mod = cuda.module_from_buffer(cubin)
if len(self._modrefs) > self.MAX_MODREFS:
del self._modrefs[:]
self._modrefs.append(self.mod)
self.packer, self.lib, self.cubin = self.compile(gnm)
self.mod = self.load(self.cubin)
self.filts = filters.create(gprof)
self.out = output.PILOutput()