Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Main fix cffi life cycle #28

Merged
merged 2 commits into from
Feb 9, 2025
Merged

Main fix cffi life cycle #28

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ebd18c6
fix cffi life cycle issue
Feb 9, 2025
040bbe1
fix cffi life cycle
Feb 9, 2025
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
16 changes: 9 additions & 7 deletions axengine/_axclrt.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -96,6 +96,8 @@ def __init__(
super().__init__()

self._device_index = 0
self._io = None
self._model_id = None

if provider_options is not None and "device_id" in provider_options[0]:
self._device_index = provider_options[0].get("device_id", 0)
Expand Down Expand Up @@ -214,12 +216,12 @@ def _unload(self):
dev_size = axclrt_cffi.new("uint64_t *")
dev_prt = axclrt_cffi.new("void **")
for i in range(axclrt_lib.axclrtEngineGetNumInputs(self._info[0])):
axclrt_lib.axclrtEngineGetInputBufferByIndex(self._io, i, dev_prt, dev_size)
axclrt_lib.axclrtEngineGetInputBufferByIndex(self._io[0], i, dev_prt, dev_size)
axclrt_lib.axclrtFree(dev_prt[0])
for i in range(axclrt_lib.axclrtEngineGetNumOutputs(self._info[0])):
axclrt_lib.axclrtEngineGetOutputBufferByIndex(self._io, i, dev_prt, dev_size)
axclrt_lib.axclrtEngineGetOutputBufferByIndex(self._io[0], i, dev_prt, dev_size)
axclrt_lib.axclrtFree(dev_prt[0])
axclrt_lib.axclrtEngineDestroyIO(self._io)
axclrt_lib.axclrtEngineDestroyIO(self._io[0])
self._io = None
if self._model_id[0] is not None and self._model_id[0] != 0:
axclrt_lib.axclrtEngineUnload(self._model_id[0])
Expand Down Expand Up @@ -322,7 +324,7 @@ def _prepare_io(self):
ret = axclrt_lib.axclrtEngineSetOutputBufferByIndex(_io[0], i, dev_ptr[0], max_size)
if 0 != ret:
raise RuntimeError(f"axclrtEngineSetOutputBufferByIndex failed 0x{ret:08x} for output {i}.")
return _io[0]
return _io

def run(
self,
Expand Down Expand Up @@ -353,21 +355,21 @@ def run(
if not (npy.flags.c_contiguous or npy.flags.f_contiguous):
npy = np.ascontiguousarray(npy)
npy_ptr = axclrt_cffi.cast("void *", npy.ctypes.data)
ret = axclrt_lib.axclrtEngineGetInputBufferByIndex(self._io, i, dev_prt, dev_size)
ret = axclrt_lib.axclrtEngineGetInputBufferByIndex(self._io[0], i, dev_prt, dev_size)
if 0 != ret:
raise RuntimeError(f"axclrtEngineGetInputBufferByIndex failed for input {i}.")
ret = axclrt_lib.axclrtMemcpy(dev_prt[0], npy_ptr, npy.nbytes, axclrt_lib.AXCL_MEMCPY_HOST_TO_DEVICE)
if 0 != ret:
raise RuntimeError(f"axclrtMemcpy failed for input {i}.")

# execute model
ret = axclrt_lib.axclrtEngineExecute(self._model_id[0], self._context_id[0], 0, self._io)
ret = axclrt_lib.axclrtEngineExecute(self._model_id[0], self._context_id[0], 0, self._io[0])

# get output
outputs = []
if 0 == ret:
for i in range(len(self.get_outputs())):
ret = axclrt_lib.axclrtEngineGetOutputBufferByIndex(self._io, i, dev_prt, dev_size)
ret = axclrt_lib.axclrtEngineGetOutputBufferByIndex(self._io[0], i, dev_prt, dev_size)
if 0 != ret:
raise RuntimeError(f"axclrtEngineGetOutputBufferByIndex failed for output {i}.")
npy = np.zeros(self.get_outputs()[i].shape, dtype=self.get_outputs()[i].dtype)
Expand Down
9 changes: 8 additions & 1 deletion axengine/_axe.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -216,32 +216,39 @@ def __init__(
_inputs= engine_cffi.new(
"AX_ENGINE_IO_BUFFER_T[{}]".format(self._io[0].nInputSize)
)
self._io[0].pInputs = _inputs
_outputs = engine_cffi.new(
"AX_ENGINE_IO_BUFFER_T[{}]".format(self._io[0].nOutputSize)
)
self._io_buffers = (_inputs, _outputs)
self._io[0].pInputs = _inputs
self._io[0].pOutputs = _outputs

self._io_inputs_pool = []
for i in range(len(self.get_inputs())):
max_buf = 0
for j in range(self._shape_count):
max_buf = max(max_buf, self._info[j][0].pInputs[i].nSize)
self._io[0].pInputs[i].nSize = max_buf
phy = engine_cffi.new("AX_U64*")
vir = engine_cffi.new("AX_VOID**")
self._io_inputs_pool.append((phy, vir))
ret = sys_lib.AX_SYS_MemAllocCached(
phy, vir, self._io[0].pInputs[i].nSize, self._align, self._cmm_token
)
if 0 != ret:
raise RuntimeError("Failed to allocate memory for input.")
self._io[0].pInputs[i].phyAddr = phy[0]
self._io[0].pInputs[i].pVirAddr = vir[0]

self._io_outputs_pool = []
for i in range(len(self.get_outputs())):
max_buf = 0
for j in range(self._shape_count):
max_buf = max(max_buf, self._info[j][0].pOutputs[i].nSize)
self._io[0].pOutputs[i].nSize = max_buf
phy = engine_cffi.new("AX_U64*")
vir = engine_cffi.new("AX_VOID**")
self._io_outputs_pool.append((phy, vir))
ret = sys_lib.AX_SYS_MemAllocCached(
phy, vir, self._io[0].pOutputs[i].nSize, self._align, self._cmm_token
)
Expand Down