Skip to content
This repository was archived by the owner on Jan 25, 2023. It is now read-only.
Merged
Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
23 commits
Select commit Hold shift + click to select a range
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
Fix for dpctl update
  • Loading branch information
reazulhoque committed Oct 6, 2020
commit 249ad65c087b6e302e6491e1ad170a6c97006a12
20 changes: 14 additions & 6 deletions numba/dppl/compiler.py
Original file line number Diff line number Diff line change
Expand Up @@ -388,9 +388,9 @@ def _unpack_device_array_argument(self, val, kernelargs):
void_ptr_arg = True

# meminfo
kernelargs.append(ctypes.c_long(0))
kernelargs.append(ctypes.c_size_t(0))
# parent
kernelargs.append(ctypes.c_long(0))
kernelargs.append(ctypes.c_size_t(0))

kernelargs.append(ctypes.c_long(val.size))
kernelargs.append(ctypes.c_long(val.dtype.itemsize))
Expand Down Expand Up @@ -505,13 +505,21 @@ def __call__(self, *args, **kwargs):
def specialize(self, *args):
argtypes = tuple([self.typingctx.resolve_argument_type(a)
for a in args])
q = None
kernel = None
# we were previously using the _env_ptr of the device_env, the sycl_queue
# should be sufficient to cache the compiled kernel for now, but we should
# use the device type to cache such kernels
key_definitions = (self.sycl_queue, argtypes)
kernel = self.definitions.get(key_definitions)
if kernel is None:
#key_definitions = (self.sycl_queue, argtypes)
key_definitions = (argtypes)
result = self.definitions.get(key_definitions)
if result:
q, kernel = result

if q and self.sycl_queue.equals(q):
return kernel
else:
kernel = compile_kernel(self.sycl_queue, self.py_func, argtypes,
self.access_types)
self.definitions[key_definitions] = kernel
self.definitions[key_definitions] = (self.sycl_queue, kernel)
return kernel
15 changes: 8 additions & 7 deletions numba/dppl/dppl_host_fn_call_gen.py
Original file line number Diff line number Diff line change
Expand Up @@ -16,8 +16,7 @@ def __init__(self, lowerer, cres, num_inputs):
self.builder = self.lowerer.builder

self.kernel = cres.kernel
self.kernel_addr = self.kernel.get_kernel_ref_address()

self.kernel_addr = self.kernel.addressof_ref()

# Initialize commonly used LLVM types and constant
self._init_llvm_types_and_constants()
Expand Down Expand Up @@ -119,16 +118,18 @@ def resolve_and_return_dpctl_type(self, ty):
val = self.context.get_constant(types.int32, 4)
elif ty == types.uint32:
val = self.context.get_constant(types.int32, 5)
elif ty == types.boolean:
val = self.context.get_constant(types.int32, 5)
elif ty == types.int64:
val = self.context.get_constant(types.int32, 6)
elif ty == types.uint64:
val = self.context.get_constant(types.int32, 7)
elif ty == types.uint64:
val = self.context.get_constant(types.int32, 8)
elif ty == types.float32:
val = self.context.get_constant(types.int32, 11)
elif ty == types.float64:
val = self.context.get_constant(types.int32, 12)
elif ty == types.float64:
val = self.context.get_constant(types.int32, 13)
elif ty == types.voidptr:
val = self.context.get_constant(types.int32, 14)
val = self.context.get_constant(types.int32, 15)
else:
raise NotImplementedError

Expand Down