def _generate_kernel_bin(self, k, ctx):
gk = gpuarray.GpuKernel(k.code, k.name, k.params, context=ctx,
**k.flags)
bin = gk._binary
bcode = ','.join(hex(ord(c)) for c in bin)
return ("""static const char %(bname)s[] = { %(bcode)s };""" %
dict(bname=k.binvar, bcode=bcode))
def c_support_code_apply(self, node, name):
nb = 0
if self.kshp is not None:
nb = self.kshp[1]
kernels = self.gpu_kernels(node, name)
k = kernels[0]
code = """
#define THEANO_KERN_WID %(nb)d
""" % locals()
code += "\n".join([
open(os.path.join(os.path.split(__file__)[0], f)).read()
for f in ["conv_kernel.cu", "conv_full_kernel.cu"]
])
kname = "conv_full_load_everything"
gk = gpuarray.GpuKernel(code, k.name, k.params, **k.flags)
bin = gk._binary
bcode = ','.join(hex(ord(c)) for c in bin)
code = code.replace('\\', '\\\\')
code = code.replace('"', '\\"')
code = code.replace('\n', '\\n')
mod = """
static const char conv_bcode[] = {%(bcode)s};
static const char *conv_code = "%(code)s";
""" % locals()
for k in kernels:
mod += "static GpuKernel " + k.name + '_' + name + ";\n"
mod += open(os.path.join(os.path.split(__file__)[0], "conv.cu")).read()
return mod
