Skip to content

cuda.core.system: More device-related APIs #1465

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

Merged
mdboom merged 13 commits into from
Jan 15, 2026
+663 −28
Merged

cuda.core.system: More device-related APIs #1465

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
13 commits
Select commit Hold shift + click to select a range
14332cd
cuda.core.system: More device-related APIs
mdboom Jan 12, 2026
421ebda
Fix line wrapping
mdboom Jan 12, 2026
16e71e0
Hide handle as an implementation detail
mdboom Jan 12, 2026
aafd1e9
Update cuda_core/cuda/core/system/_device.pyx
mdboom Jan 12, 2026
adad250
Address comments from Copilot
mdboom Jan 12, 2026
eb08820
Working on tests
mdboom Jan 13, 2026
51496f7
Fix tests
mdboom Jan 13, 2026
47a83ce
Fix creating new device
mdboom Jan 13, 2026
3cfdabb
Fix iterator
mdboom Jan 13, 2026
dd3118a
Fix affinity test
mdboom Jan 13, 2026
2293660
Fix nearest GPUs
mdboom Jan 13, 2026
7a9783a
Merge remote-tracking branch 'upstream/main' into cuda.core.system-sy…
mdboom Jan 15, 2026
db0f621
Merge remote-tracking branch 'upstream/main' into cuda.core.system-sy…
mdboom Jan 15, 2026
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
2 changes: 1 addition & 1 deletion cuda_bindings/cuda/bindings/_nvml.pxd
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -173,7 +173,7 @@ cpdef str device_get_inforom_version(intptr_t device, int object)
cpdef str device_get_inforom_image_version(intptr_t device)
cpdef unsigned int device_get_inforom_configuration_checksum(intptr_t device) except? 0
cpdef device_validate_inforom(intptr_t device)
cpdef unsigned long device_get_last_bbx_flush_time(intptr_t device, intptr_t timestamp) except? 0
cpdef tuple device_get_last_bbx_flush_time(intptr_t device)
Copy link
Contributor

@cpcloud cpcloud Jan 15, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This seems like a weird signature. Does this function really return a tuple?

Copy link
Contributor Author

@mdboom mdboom Jan 15, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yep, it does. It returns two numbers.

cpdef int device_get_display_mode(intptr_t device) except? -1
cpdef int device_get_display_active(intptr_t device) except? -1
cpdef int device_get_persistence_mode(intptr_t device) except? -1
Expand Down
33 changes: 18 additions & 15 deletions cuda_bindings/cuda/bindings/_nvml.pyx
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22228,23 +22228,26 @@ cpdef device_validate_inforom(intptr_t device):
check_status(__status__)


cpdef unsigned long device_get_last_bbx_flush_time(intptr_t device, intptr_t timestamp) except? 0:
cpdef tuple device_get_last_bbx_flush_time(intptr_t device):
"""Retrieves the timestamp and the duration of the last flush of the BBX (blackbox) infoROM object during the current run.

Args:
device (intptr_t): The identifier of the target device.
timestamp (intptr_t): The start timestamp of the last BBX Flush.

Returns:
unsigned long: The duration (us) of the last BBX Flush.
A 2-tuple containing:

- unsigned long long: The start timestamp of the last BBX Flush.
- unsigned long: The duration (us) of the last BBX Flush.

.. seealso:: `nvmlDeviceGetLastBBXFlushTime`
"""
cdef unsigned long long timestamp
cdef unsigned long duration_us
with nogil:
__status__ = nvmlDeviceGetLastBBXFlushTime(<Device>device, <unsigned long long*>timestamp, &duration_us)
__status__ = nvmlDeviceGetLastBBXFlushTime(<Device>device, &timestamp, &duration_us)
check_status(__status__)
return duration_us
return (timestamp, duration_us)


cpdef int device_get_display_mode(intptr_t device) except? -1:
Expand Down Expand Up @@ -27101,8 +27104,8 @@ cpdef object system_get_topology_gpu_set(unsigned int cpuNumber):
__status__ = nvmlSystemGetTopologyGpuSet(cpuNumber, <unsigned int*>count, NULL)
check_status_size(__status__)
if count[0] == 0:
return view.array(shape=(1,), itemsize=sizeof(intptr_t), format="i", mode="c")[:0]
cdef view.array deviceArray = view.array(shape=(count[0],), itemsize=sizeof(intptr_t), format="i", mode="c")
return view.array(shape=(1,), itemsize=sizeof(intptr_t), format="P", mode="c")[:0]
cdef view.array deviceArray = view.array(shape=(count[0],), itemsize=sizeof(intptr_t), format="P", mode="c")
with nogil:
__status__ = nvmlSystemGetTopologyGpuSet(cpuNumber, <unsigned int*>count, <nvmlDevice_t *>deviceArray.data)
check_status(__status__)
Expand Down Expand Up @@ -27141,8 +27144,8 @@ cpdef object unit_get_devices(intptr_t unit):
__status__ = nvmlUnitGetDevices(<nvmlUnit_t *>unit, <unsigned int*>deviceCount, NULL)
check_status_size(__status__)
if deviceCount[0] == 0:
return view.array(shape=(1,), itemsize=sizeof(intptr_t), format="i", mode="c")[:0]
cdef view.array deviceArray = view.array(shape=(deviceCount[0],), itemsize=sizeof(intptr_t), format="i", mode="c")
return view.array(shape=(1,), itemsize=sizeof(intptr_t), format="P", mode="c")[:0]
cdef view.array deviceArray = view.array(shape=(deviceCount[0],), itemsize=sizeof(intptr_t), format="P", mode="c")
with nogil:
__status__ = nvmlUnitGetDevices(<nvmlUnit_t *>unit, <unsigned int*>deviceCount, <nvmlDevice_t *>deviceArray.data)
check_status(__status__)
Expand All @@ -27169,8 +27172,8 @@ cpdef object device_get_topology_nearest_gpus(intptr_t device, unsigned int leve
)
check_status_size(__status__)
if count[0] == 0:
return view.array(shape=(1,), itemsize=sizeof(intptr_t), format="i", mode="c")[:0]
cdef view.array deviceArray = view.array(shape=(deviceCount[0],), itemsize=sizeof(intptr_t), format="i", mode="c")
return view.array(shape=(1,), itemsize=sizeof(intptr_t), format="P", mode="c")[:0]
cdef view.array deviceArray = view.array(shape=(deviceCount[0],), itemsize=sizeof(intptr_t), format="P", mode="c")
with nogil:
__status__ = nvmlDeviceGetTopologyNearestGpus(
<Device>device,
Expand Down Expand Up @@ -27834,9 +27837,9 @@ cpdef object device_get_gpu_instances(intptr_t device, unsigned int profile_id):
check_status_size(__status__)

if count[0] == 0:
view.array(shape=(1,), itemsize=sizeof(intptr_t), format="i", mode="c")[:0]
view.array(shape=(1,), itemsize=sizeof(intptr_t), format="P", mode="c")[:0]

cdef view.array gpuInstances = view.array(shape=(count[0],), itemsize=sizeof(intptr_t), format="i", mode="c")
cdef view.array gpuInstances = view.array(shape=(count[0],), itemsize=sizeof(intptr_t), format="P", mode="c")
with nogil:
__status__ = nvmlDeviceGetGpuInstances(<Device>device, profile_id, <nvmlGpuInstance_t *>gpuInstances.data, count)
check_status(__status__)
Expand All @@ -27860,9 +27863,9 @@ cpdef object gpu_instance_get_compute_instances(intptr_t gpu_instance, unsigned
check_status_size(__status__)

if count[0] == 0:
view.array(shape=(1,), itemsize=sizeof(intptr_t), format="i", mode="c")[:0]
view.array(shape=(1,), itemsize=sizeof(intptr_t), format="P", mode="c")[:0]

cdef view.array computeInstances = view.array(shape=(count[0],), itemsize=sizeof(intptr_t), format="i", mode="c")
cdef view.array computeInstances = view.array(shape=(count[0],), itemsize=sizeof(intptr_t), format="P", mode="c")
with nogil:
__status__ = nvmlGpuInstanceGetComputeInstances(<GpuInstance>gpu_instance, profile_id, <nvmlComputeInstance_t *>computeInstances.data, count)
check_status(__status__)
Expand Down
Loading
Loading