From 31f43762443d7a6d8b4da347a0f900ca232b16d5 Mon Sep 17 00:00:00 2001
From: Stefan van der Walt <stefan@sun.ac.za>
Date: Sat, 3 Jul 2010 14:15:36 -0500
Subject: [PATCH] Begin working on kernel invocation refactor.
---
pyopencl/array.py | 185 +++++++++++++++++++++++-----------------
pyopencl/elementwise.py | 53 ++++++------
2 files changed, 134 insertions(+), 104 deletions(-)
diff --git a/pyopencl/array.py b/pyopencl/array.py
index c0c74ef6..450aec8e 100644
--- a/pyopencl/array.py
+++ b/pyopencl/array.py
@@ -33,19 +33,19 @@ OTHER DEALINGS IN THE SOFTWARE.
import numpy
import pyopencl.elementwise as elementwise
import pyopencl as cl
+#from pytools import memoize_method
+from decorator import decorator
-def splay(ctx, n):
- max_work_items = max(dev.max_work_group_size for dev in ctx.devices)
- max_work_items = min(128, max_work_items)
+def splay(queue, n):
+ dev = queue.device
+ max_work_items = min(128, dev.max_work_group_size)
min_work_items = min(32, max_work_items)
- max_groups = max(
- 4 * dev.max_compute_units * 8
- # 4 to overfill the device
- # 8 is an Nvidia constant--that's how many
- # groups fit onto one compute device
- for dev in ctx.devices)
+ max_groups = dev.max_compute_units * 4 * 8
+ # 4 to overfill the device
+ # 8 is an Nvidia constant--that's how many
+ # groups fit onto one compute device
if n < min_work_items:
group_count = 1
@@ -74,6 +74,39 @@ def _get_common_dtype(obj1, obj2):
+@decorator
+def elwise_kernel_runner(kernel_getter, *args, **kwargs):
+ """Take a kernel getter of the same signature as the kernel
+ and return a function that invokes that kernel.
+
+ Assumes that the zeroth entry in *args* is an :class:`Array`.
+ """
+
+ knl = kernel_getter(*args)
+
+ repr_ary = args[0]
+ assert isinstance(repr_ary, Array)
+
+ queue = kwargs.pop("queue", None) or repr_ary.queue
+ gs, ls = repr_ary.get_sizes(queue)
+
+ if kwargs:
+ raise TypeError("only the 'queue' keyword argument is supported")
+
+ actual_args = []
+ for arg in args:
+ if isinstance(arg, Array):
+ actual_args.append(arg.data)
+ else:
+ actual_args.append(arg)
+ actual_args.append(repr_ary.mem_size)
+
+ return knl(queue, gs, ls, *actual_args)
+
+
+
+
+
class DefaultAllocator:
def __init__(self, context, flags=cl.mem_flags.READ_WRITE):
self.context = context
@@ -132,7 +165,9 @@ class Array(object):
self.base = base
- self._global_size, self._local_size = splay(context, self.mem_size)
+ #@memoize_method FIXME: reenable
+ def get_sizes(self, queue):
+ return splay(self.queue, self.mem_size)
def set(self, ary, queue=None, async=False):
assert ary.size == self.size
@@ -170,38 +205,28 @@ class Array(object):
raise TypeError("pyopencl arrays are not hashable.")
# kernel invocation wrappers ----------------------------------------------
- def _axpbyz(self, selffac, other, otherfac, out, queue=None):
+ @elwise_kernel_runner
+ @staticmethod
+ def _axpbyz(out, afac, a, bfac, b, queue=None):
"""Compute ``out = selffac * self + otherfac*other``,
where `other` is a vector.."""
assert self.shape == other.shape
- knl = elementwise.get_axpbyz_kernel(
+ return elementwise.get_axpbyz_kernel(
self.context, self.dtype, other.dtype, out.dtype)
- knl(queue or self.queue, self._global_size, self._local_size,
- selffac, self.data, otherfac, other.data,
- out.data, self.mem_size)
-
- return out
-
- def _axpbz(self, selffac, other, out, queue=None):
+ @elwise_kernel_runner
+ def _axpbz(self, out, selffac, other, queue=None):
"""Compute ``out = selffac * self + other``, where `other` is a scalar."""
- knl = elementwise.get_axpbz_kernel(self.context, self.dtype)
- knl(queue or self.queue, self._global_size, self._local_size,
- selffac, self.data, other, out.data, self.mem_size)
+ return elementwise.get_axpbz_kernel(self.context, self.dtype)
- return out
-
- def _elwise_multiply(self, other, out, queue=None):
- knl = elementwise.get_multiply_kernel(
+ @elwise_kernel_runner
+ def _elwise_multiply(self, out, other, queue=None):
+ return elementwise.get_multiply_kernel(
self.context, self.dtype, other.dtype, out.dtype)
- knl(queue or self.queue, self._global_size, self._local_size,
- self.data, other.data,
- out.data, self.mem_size)
-
- return out
- def _rdiv_scalar(self, other, out, queue=None):
+ @elwise_kernel_runner
+ def _rdiv_scalar(self, out, other, queue=None):
"""Divides an array by a scalar::
y = n / self
@@ -209,24 +234,29 @@ class Array(object):
assert self.dtype == numpy.float32
- knl = elementwise.get_rdivide_elwise_kernel(self.context, self.dtype)
- knl(queue or self.queue, self._global_size, self._local_size,
- self.data, other,
- out.data, self.mem_size)
-
- return out
+ return elementwise.get_rdivide_elwise_kernel(self.context, self.dtype)
- def _div(self, other, out, queue=None):
+ @elwise_kernel_runner
+ def _div(self, out, other, queue=None):
"""Divides an array by another array."""
assert self.shape == other.shape
- knl = elementwise.get_divide_kernel(self.context,
+ return elementwise.get_divide_kernel(self.context,
self.dtype, other.dtype, out.dtype)
- knl(queue or self.queue, self._global_size, self._local_size,
- self.data, other.data, out.data, self.mem_size)
- return out
+ @elwise_kernel_runner
+ def _fill(result, scalar):
+ return elementwise.get_fill_kernel(result.context, result.dtype)
+
+ @elwise_kernel_runner
+ def _abs(result, arg):
+ return elementwise.get_unary_func_kernel(
+ arg.context, fname, arg.dtype)
+
+ @elwise_kernel_runner
+ def _pow_scalar(self, ):
+ return elementwise.get_pow_kernel(self.context, self.dtype)
def _new_like_me(self, dtype=None):
if dtype is None:
@@ -236,11 +266,12 @@ class Array(object):
queue=self.queue)
# operators ---------------------------------------------------------------
- def mul_add(self, selffac, other, otherfac, add_timer=None, queue=None):
+ def mul_add(self, selffac, other, otherfac, queue=None):
"""Return `selffac * self + otherfac*other`.
"""
result = self._new_like_me(_get_common_dtype(self, other))
- return self._axpbyz(selffac, other, otherfac, result, add_timer)
+ self._axpbyz(result, selffac, self, other, otherfac)
+ return result
def __add__(self, other):
"""Add an array with an array or an array with a scalar."""
@@ -248,14 +279,16 @@ class Array(object):
if isinstance(other, Array):
# add another vector
result = self._new_like_me(_get_common_dtype(self, other))
- return self._axpbyz(1, other, 1, result)
+ self._axpbyz(result, 1, self, 1, other)
+ return result
else:
# add a scalar
if other == 0:
return self
else:
result = self._new_like_me()
- return self._axpbz(1, other, result)
+ self._axpbz(result, 1, other)
+ return result
__radd__ = __add__
@@ -264,14 +297,15 @@ class Array(object):
if isinstance(other, Array):
result = self._new_like_me(_get_common_dtype(self, other))
- return self._axpbyz(1, other, -1, result)
+ self._axpbyz(result, 1, other, -1)
+ return result
else:
if other == 0:
return self
else:
# create a new array for the result
result = self._new_like_me()
- return self._axpbz(1, -other, result)
+ return self._axpbz(result, 1, -other)
def __rsub__(self,other):
"""Substracts an array by a scalar or an array::
@@ -280,32 +314,37 @@ class Array(object):
"""
# other must be a scalar
result = self._new_like_me()
- return self._axpbz(-1, other, result)
+ return self._axpbz(result, -1, other)
def __iadd__(self, other):
- return self._axpbyz(1, other, 1, self)
+ self._axpbyz(self, 1, self, 1, other)
+ return self
def __isub__(self, other):
- return self._axpbyz(1, other, -1, self)
+ return self._axpbyz(self, 1, self, -1, other)
def __neg__(self):
result = self._new_like_me()
- return self._axpbz(-1, 0, result)
+ return self._axpbz(result, -1, self)
def __mul__(self, other):
if isinstance(other, Array):
result = self._new_like_me(_get_common_dtype(self, other))
- return self._elwise_multiply(other, result)
+ self._elwise_multiply(result, other)
+ return result
else:
result = self._new_like_me()
- return self._axpbz(other, 0, result)
+ self._axpbz(result, other, self, 0)
+ return result
def __rmul__(self, scalar):
result = self._new_like_me()
- return self._axpbz(scalar, 0, result)
+ self._axpbz(result, scalar, self, 0)
+ return result
def __imul__(self, scalar):
- return self._axpbz(scalar, 0, self)
+ self._axpbz(self, scalar, 0)
+ return self
def __div__(self, other):
"""Divides an array by an array or a scalar::
@@ -314,14 +353,15 @@ class Array(object):
"""
if isinstance(other, Array):
result = self._new_like_me(_get_common_dtype(self, other))
- return self._div(other, result)
+ return self._div(result, other)
else:
if other == 1:
return self
else:
# create a new array for the result
result = self._new_like_me()
- return self._axpbz(1/other, 0, result)
+ self._axpbz(result, 1/other, 0)
+ return result
__truediv__ = __div__
@@ -333,29 +373,22 @@ class Array(object):
if isinstance(other, Array):
result = self._new_like_me(_get_common_dtype(self, other))
-
- knl = elementwise.get_divide_kernel(self.context)
- knl(self.queue, self._global_size, self._local_size,
- other.data, self.data, result.data,
- self.mem_size)
-
- return result
+ other._div(result, self)
else:
if other == 1:
return self
else:
# create a new array for the result
result = self._new_like_me()
- return self._rdiv_scalar(other, result)
+ self._rdiv_scalar(result, other)
+
+ return result
__rtruediv__ = __div__
def fill(self, value, queue=None):
"""fills the array with the specified value"""
- knl = elementwise.get_fill_kernel(self.context, self.dtype)
- knl(queue or self.queue, self._global_size, self._local_size,
- value, self.data, self.mem_size)
-
+ self._fill(self, value, queue=queue)
return self
def __len__(self):
@@ -370,8 +403,6 @@ class Array(object):
of `self`.
"""
- result = self._new_like_me()
-
if self.dtype.kind == "f":
fname = "fabs"
elif self.dtype.kind in ["u", "i"]:
@@ -379,11 +410,8 @@ class Array(object):
else:
raise TypeError("unsupported dtype in __abs__()")
- knl = elementwise.get_unary_func_kernel(
- self.context, fname, self.dtype)
- knl(self.queue, self._global_size, self._local_size,
- self.data,result.data, self.mem_size)
-
+ result = self._new_like_me()
+ self._abs(result, self)
return result
def __pow__(self, other):
@@ -406,7 +434,6 @@ class Array(object):
return result
else:
result = self._new_like_me()
- knl = elementwise.get_pow_kernel(self.context, self.dtype)
knl(self.queue, self._global_size, self._local_size,
other, self.data, result.data,
self.mem_size)
diff --git a/pyopencl/elementwise.py b/pyopencl/elementwise.py
index 0fc9aaed..4ec64b73 100644
--- a/pyopencl/elementwise.py
+++ b/pyopencl/elementwise.py
@@ -152,9 +152,9 @@ class ElementwiseKernel:
queue = repr_vec.queue
invocation_args.append(repr_vec.mem_size)
+ gs, ls = repr_vec.get_sizes(queue)
self.kernel.set_args(*invocation_args)
- return cl.enqueue_nd_range_kernel(queue, self.kernel,
- repr_vec._global_size, repr_vec._local_size)
+ return cl.enqueue_nd_range_kernel(queue, self.kernel, gs, ls)
@@ -166,9 +166,9 @@ def get_take_kernel(context, dtype, idx_dtype, vec_count=1):
"tp": dtype_to_ctype(dtype),
}
- args = ([VectorArg(idx_dtype, "idx")]
- + [VectorArg(dtype, "src"+str(i))for i in range(vec_count)]
- + [VectorArg(dtype, "dest"+str(i))for i in range(vec_count)])
+ args = ([VectorArg(dtype, "dest"+str(i))for i in range(vec_count)]
+ + [VectorArg(idx_dtype, "idx")]
+ + [VectorArg(dtype, "src"+str(i))for i in range(vec_count)] )
body = (
("%(idx_tp)s src_idx = idx[i];\n" % ctx)
+ "\n".join(
@@ -188,15 +188,15 @@ def get_take_put_kernel(context, dtype, idx_dtype, with_offsets, vec_count=1):
}
args = [
+ VectorArg(dtype, "dest%d" % i)
+ for i in range(vec_count)
+ ] + [
VectorArg(idx_dtype, "gmem_dest_idx"),
VectorArg(idx_dtype, "gmem_src_idx"),
] + [
VectorArg(dtype, "src%d" % i)
for i in range(vec_count)
] + [
- VectorArg(dtype, "dest%d" % i)
- for i in range(vec_count)
- ] + [
ScalarArg(idx_dtype, "offset%d" % i)
for i in range(vec_count) if with_offsets
]
@@ -228,11 +228,11 @@ def get_put_kernel(context, dtype, idx_dtype, vec_count=1):
}
args = [
- VectorArg(idx_dtype, "gmem_dest_idx"),
- ] + [
VectorArg(dtype, "dest%d" % i)
for i in range(vec_count)
] + [
+ VectorArg(idx_dtype, "gmem_dest_idx"),
+ ] + [
VectorArg(dtype, "src%d" % i)
for i in range(vec_count)
]
@@ -262,6 +262,9 @@ def get_copy_kernel(context, dtype_dest, dtype_src):
@context_dependent_memoize
def get_linear_combination_kernel(summand_descriptors,
dtype_z):
+ # TODO: Port this!
+ raise NotImplementedError
+
from pycuda.tools import dtype_to_ctype
from pycuda.elementwise import \
VectorArg, ScalarArg, get_elwise_module
@@ -311,7 +314,7 @@ def get_linear_combination_kernel(summand_descriptors,
@context_dependent_memoize
def get_axpbyz_kernel(context, dtype_x, dtype_y, dtype_z):
return get_elwise_kernel(context,
- "%(tp_x)s a, %(tp_x)s *x, %(tp_y)s b, %(tp_y)s *y, %(tp_z)s *z" % {
+ "%(tp_z)s *z, %(tp_x)s a, %(tp_x)s *x, %(tp_y)s b, %(tp_y)s *y" % {
"tp_x": dtype_to_ctype(dtype_x),
"tp_y": dtype_to_ctype(dtype_y),
"tp_z": dtype_to_ctype(dtype_z),
@@ -322,7 +325,7 @@ def get_axpbyz_kernel(context, dtype_x, dtype_y, dtype_z):
@context_dependent_memoize
def get_axpbz_kernel(context, dtype):
return get_elwise_kernel(context,
- "%(tp)s a, %(tp)s *x,%(tp)s b, %(tp)s *z" % {
+ "%(tp)s *z, %(tp)s a, %(tp)s *x,%(tp)s b" % {
"tp": dtype_to_ctype(dtype)},
"z[i] = a * x[i] + b",
"axpb")
@@ -330,7 +333,7 @@ def get_axpbz_kernel(context, dtype):
@context_dependent_memoize
def get_multiply_kernel(context, dtype_x, dtype_y, dtype_z):
return get_elwise_kernel(context,
- "%(tp_x)s *x, %(tp_y)s *y, %(tp_z)s *z" % {
+ "%(tp_z)s *z, %(tp_x)s *x, %(tp_y)s *y" % {
"tp_x": dtype_to_ctype(dtype_x),
"tp_y": dtype_to_ctype(dtype_y),
"tp_z": dtype_to_ctype(dtype_z),
@@ -341,7 +344,7 @@ def get_multiply_kernel(context, dtype_x, dtype_y, dtype_z):
@context_dependent_memoize
def get_divide_kernel(context, dtype_x, dtype_y, dtype_z):
return get_elwise_kernel(context,
- "%(tp_x)s *x, %(tp_y)s *y, %(tp_z)s *z" % {
+ "%(tp_z)s *z, %(tp_x)s *x, %(tp_y)s *y, " % {
"tp_x": dtype_to_ctype(dtype_x),
"tp_y": dtype_to_ctype(dtype_y),
"tp_z": dtype_to_ctype(dtype_z),
@@ -352,7 +355,7 @@ def get_divide_kernel(context, dtype_x, dtype_y, dtype_z):
@context_dependent_memoize
def get_rdivide_elwise_kernel(context, dtype):
return get_elwise_kernel(context,
- "%(tp)s *x, %(tp)s y, %(tp)s *z" % {
+ "%(tp)s *z, %(tp)s *x, %(tp)s y" % {
"tp": dtype_to_ctype(dtype),
},
"z[i] = y / x[i]",
@@ -361,7 +364,7 @@ def get_rdivide_elwise_kernel(context, dtype):
@context_dependent_memoize
def get_fill_kernel(context, dtype):
return get_elwise_kernel(context,
- "%(tp)s a, %(tp)s *z" % {
+ "%(tp)s *z, %(tp)s a" % {
"tp": dtype_to_ctype(dtype),
},
"z[i] = a",
@@ -370,7 +373,7 @@ def get_fill_kernel(context, dtype):
@context_dependent_memoize
def get_reverse_kernel(context, dtype):
return get_elwise_kernel(context,
- "%(tp)s *y, %(tp)s *z" % {
+ "%(tp)s *z, %(tp)s *y" % {
"tp": dtype_to_ctype(dtype),
},
"z[i] = y[n-1-i]",
@@ -389,7 +392,7 @@ def get_arange_kernel(context, dtype):
@context_dependent_memoize
def get_pow_kernel(context, dtype):
return get_elwise_kernel(context,
- "%(tp)s value, %(tp)s *y, %(tp)s *z" % {
+ "%(tp)s *z, %(tp)s value, %(tp)s *y, " % {
"tp": dtype_to_ctype(dtype),
},
"z[i] = pow(y[i], value)",
@@ -398,7 +401,7 @@ def get_pow_kernel(context, dtype):
@context_dependent_memoize
def get_pow_array_kernel(context, dtype_x, dtype_y, dtype_z):
return get_elwise_kernel(context,
- "%(tp_x)s *x, %(tp_y)s *y, %(tp_z)s *z" % {
+ "%(tp_z)s *z, %(tp_x)s *x, %(tp_y)s *y" % {
"tp_x": dtype_to_ctype(dtype_x),
"tp_y": dtype_to_ctype(dtype_y),
"tp_z": dtype_to_ctype(dtype_z),
@@ -409,21 +412,21 @@ def get_pow_array_kernel(context, dtype_x, dtype_y, dtype_z):
@context_dependent_memoize
def get_fmod_kernel(context):
return get_elwise_kernel(context,
- "float *arg, float *mod, float *z",
+ "float *z, float *arg, float *mod",
"z[i] = fmod(arg[i], mod[i])",
"fmod_kernel")
@context_dependent_memoize
def get_modf_kernel(context):
return get_elwise_kernel(context,
- "float *x, float *intpart ,float *fracpart",
+ "float *intpart ,float *fracpart, float *x",
"fracpart[i] = modf(x[i], &intpart[i])",
"modf_kernel")
@context_dependent_memoize
def get_frexp_kernel(context):
return get_elwise_kernel(context,
- "float *x, float *significand, float *exponent",
+ "float *significand, float *exponent, float *x",
"""
int expt = 0;
significand[i] = frexp(x[i], &expt);
@@ -434,7 +437,7 @@ def get_frexp_kernel(context):
@context_dependent_memoize
def get_ldexp_kernel(context):
return get_elwise_kernel(context,
- "float *sig, float *expt, float *z",
+ "float *z, float *sig, float *expt",
"z[i] = ldexp(sig[i], (int) expt[i])",
"ldexp_kernel")
@@ -444,7 +447,7 @@ def get_unary_func_kernel(context, func_name, in_dtype, out_dtype=None):
out_dtype = in_dtype
return get_elwise_kernel(context,
- "%(tp_in)s *y, %(tp_out)s *z" % {
+ "%(tp_out)s *z, %(tp_in)s *y" % {
"tp_in": dtype_to_ctype(in_dtype),
"tp_out": dtype_to_ctype(out_dtype),
},
@@ -457,10 +460,10 @@ def get_unary_func_kernel(context, func_name, in_dtype, out_dtype=None):
@context_dependent_memoize
def get_if_positive_kernel(context, crit_dtype, dtype):
return get_elwise_kernel(context, [
+ VectorArg(dtype, "result"),
VectorArg(crit_dtype, "crit"),
VectorArg(dtype, "then_"),
VectorArg(dtype, "else_"),
- VectorArg(dtype, "result"),
],
"result[i] = crit[i] > 0 ? then_[i] : else_[i]",
"if_positive")
--
GitLab