From 61f5245b020f9a35a06d50a36c47f15f5da37f0a Mon Sep 17 00:00:00 2001
From: Andreas Kloeckner <inform@tiker.net>
Date: Wed, 12 Jun 2013 00:38:57 -0400
Subject: [PATCH] Fix lots of broken device type queries
---
pyopencl/characterize/__init__.py | 14 +++++++-------
pyopencl/clrandom.py | 4 ++--
pyopencl/scan.py | 4 ++--
test/test_algorithm.py | 2 +-
test/test_wrapper.py | 4 ++--
5 files changed, 14 insertions(+), 14 deletions(-)
diff --git a/pyopencl/characterize/__init__.py b/pyopencl/characterize/__init__.py
index eb6c9d01..8fd33a86 100644
--- a/pyopencl/characterize/__init__.py
+++ b/pyopencl/characterize/__init__.py
@@ -119,13 +119,13 @@ def simultaneous_work_items_on_local_access(dev):
return 32
- if dev.type == cl.device_type.GPU:
+ if dev.type & cl.device_type.GPU:
from warnings import warn
warn("wildly guessing conflicting local access size on '%s'"
% dev,
CLCharacterizationWarning)
return 16
- elif dev.type == cl.device_type.CPU:
+ elif dev.type & cl.device_type.CPU:
return 1
else:
from warnings import warn
@@ -157,13 +157,13 @@ def local_memory_bank_count(dev):
return 32
- if dev.type == cl.device_type.GPU:
+ if dev.type & cl.device_type.GPU:
from warnings import warn
warn("wildly guessing local memory bank count on '%s'"
% dev,
CLCharacterizationWarning)
return 16
- elif dev.type == cl.device_type.CPU:
+ elif dev.type & cl.device_type.CPU:
if dev.local_mem_type == cl.device_local_mem_type.GLOBAL:
raise RuntimeError("asking for a bank count is "
"meaningless for cache-based lmem")
@@ -284,7 +284,7 @@ def get_simd_group_size(dev, type_size):
return 32
if ("advanced micro" in lc_vendor or "ati" in lc_vendor):
- if dev.type == cl.device_type.GPU:
+ if dev.type & cl.device_type.GPU:
# Tomasz Rybak says, in response to reduction mishbehaving on the AMD
# 'Loveland' APU:
#
@@ -303,12 +303,12 @@ def get_simd_group_size(dev, type_size):
# This is therefore our best guess as to the SIMD group size.
return reasonable_work_group_size_multiple(dev)
- elif dev.type == cl.device_type.CPU:
+ elif dev.type & cl.device_type.CPU:
return 1
else:
raise RuntimeError("unexpected AMD device type")
- if dev.type == cl.device_type.CPU:
+ if dev.type & cl.device_type.CPU:
# implicit assumption: Impl. will vectorize
if type_size == 1:
diff --git a/pyopencl/clrandom.py b/pyopencl/clrandom.py
index bae170f5..92a5c2de 100644
--- a/pyopencl/clrandom.py
+++ b/pyopencl/clrandom.py
@@ -105,7 +105,7 @@ class RanluxGenerator(object):
luxury = 4
if num_work_items is None:
- if queue.device.type == cl.device_type.CPU:
+ if queue.device.type & cl.device_type.CPU:
num_work_items = 8 * queue.device.max_compute_units
else:
num_work_items = 64 * queue.device.max_compute_units
@@ -151,7 +151,7 @@ class RanluxGenerator(object):
if ("darwin" in sys.platform
and "Apple" in queue.device.platform.vendor
and platform.mac_ver()[0].startswith("10.7")
- and queue.device.type == cl.device_type.CPU):
+ and queue.device.type & cl.device_type.CPU):
wg_size = (1,)
self.wg_size = wg_size
diff --git a/pyopencl/scan.py b/pyopencl/scan.py
index 4d1513bb..7ee47e35 100644
--- a/pyopencl/scan.py
+++ b/pyopencl/scan.py
@@ -1021,7 +1021,7 @@ class _GenericScanKernelBase(object):
arg_ctypes=arg_ctypes,
scan_expr=_process_code_for_macro(scan_expr),
neutral=_process_code_for_macro(neutral),
- is_gpu=self.devices[0].type == cl.device_type.GPU,
+ is_gpu=bool(self.devices[0].type & cl.device_type.GPU),
double_support=all(
has_double_support(dev) for dev in devices),
)
@@ -1086,7 +1086,7 @@ class GenericScanKernel(_GenericScanKernelBase):
if lmem_use + 256 <= avail_local_mem:
solutions.append((wg_size*k_group_size, k_group_size, wg_size))
- if self.devices[0].type == cl.device_type.GPU:
+ if self.devices[0].type & cl.device_type.GPU:
from pytools import any
for wg_size_floor in [256, 192, 128]:
have_sol_above_floor = any(wg_size >= wg_size_floor
diff --git a/test/test_algorithm.py b/test/test_algorithm.py
index ef28a015..fc31e8c5 100644
--- a/test/test_algorithm.py
+++ b/test/test_algorithm.py
@@ -628,7 +628,7 @@ def test_index_preservation(ctx_factory):
classes = [GenericScanKernel]
dev = context.devices[0]
- if dev.type == cl.device_type.CPU:
+ if dev.type & cl.device_type.CPU:
classes.append(GenericDebugScanKernel)
for cls in classes:
diff --git a/test/test_wrapper.py b/test/test_wrapper.py
index 4f6a1526..6c04d7d9 100644
--- a/test/test_wrapper.py
+++ b/test/test_wrapper.py
@@ -341,7 +341,7 @@ def test_image_2d(ctx_factory):
good = la.norm(a_result - a) == 0
if not good:
- if queue.device.type == cl.device_type.CPU:
+ if queue.device.type & cl.device_type.CPU:
assert good, ("The image implementation on your CPU CL platform '%s' "
"returned bad values. This is bad, but common."
% queue.device.platform)
@@ -417,7 +417,7 @@ def test_image_3d(ctx_factory):
good = la.norm(a_result - a) == 0
if not good:
- if queue.device.type == cl.device_type.CPU:
+ if queue.device.type & cl.device_type.CPU:
assert good, ("The image implementation on your CPU CL platform '%s' "
"returned bad values. This is bad, but common."
% queue.device.platform)
--
GitLab