RMM
E890461
RMM is the RAPIDS Memory Manager, a GPU-accelerated memory management library used within NVIDIA RAPIDS to efficiently allocate and manage device memory for data science and analytics workloads.
Statements (49)
| Predicate | Object |
|---|---|
| instanceOf |
GPU memory management library
ⓘ
RAPIDS component ⓘ open-source software library ⓘ |
| aimsTo |
improve GPU memory utilization
ⓘ
minimize GPU memory fragmentation ⓘ provide consistent memory behavior across RAPIDS libraries ⓘ reduce GPU memory allocation overhead ⓘ |
| designedFor |
ETL workloads
ⓘ
GPU-accelerated data science workloads ⓘ analytics workloads ⓘ machine learning workloads ⓘ |
| developedBy | NVIDIA NERFINISHED ⓘ |
| feature |
arena allocator
ⓘ
asynchronous memory allocation support ⓘ configurable memory resource hierarchy ⓘ device memory resource interface ⓘ host memory resource interface ⓘ managed memory resource interface ⓘ out-of-memory error handling ⓘ pool allocator ⓘ statistics collection for allocations ⓘ stream-ordered memory allocation ⓘ |
| fullName | RAPIDS Memory Manager NERFINISHED ⓘ |
| integratesWith |
CUDA
NERFINISHED
ⓘ
RAPIDS cuDF NERFINISHED ⓘ RAPIDS cuGraph NERFINISHED ⓘ RAPIDS cuML NERFINISHED ⓘ |
| license | Apache License 2.0 ⓘ |
| maintainedBy | RAPIDS AI team NERFINISHED ⓘ |
| partOf |
RAPIDS
NERFINISHED
ⓘ
RAPIDS AI ecosystem NERFINISHED ⓘ |
| programmingLanguage |
C++
ⓘ
CUDA C++ NERFINISHED ⓘ Python (bindings) ⓘ |
| provides |
GPU device memory allocation
ⓘ
GPU device memory deallocation ⓘ logging of memory allocations ⓘ memory pooling ⓘ memory resource abstractions ⓘ memory tracking ⓘ |
| repository | https://github.com/rapidsai/rmm ⓘ |
| supportsLanguageBinding |
C++
ⓘ
Python ⓘ |
| targetHardware | NVIDIA GPU NERFINISHED ⓘ |
| targetPlatform | GPU ⓘ |
| usedBy |
cuDF
NERFINISHED
ⓘ
cuGraph NERFINISHED ⓘ cuML NERFINISHED ⓘ other RAPIDS libraries ⓘ |
Referenced by (1)
Full triples — surface form annotated when it differs from this entity's canonical label.