| Electronic Components Datasheet Search |
|
STM32MP255CAL3 Datasheet(PDF) 15 Page - STMicroelectronics |
|
|
|||||||||||||||||||||||||||||
STM32MP255CAL3 Datasheet(HTML) 15 Page - STMicroelectronics |
|
15 / 241 page ![]() 3.4 Graphic processing unit (GPU) Note: Features may be limited or absent in some devices or packages (see Section 2 for details). STM32MP25xC/F devices include a 3D graphics engine (VeriSilicon GC8000UL). The GPU is a dedicated graphics processing unit accelerating numerous 3D graphics applications such as graphical user interface (GUI), menu display or animations. It works together with an optimized software stack design for industry-standard APIs with support for Android™ and Linux® embedded development platforms. The GPU is used to accelerate parallel computing (GPGPU), via the typical OpenCL or Vulkan API, or more image-based API like OpenVX or OpenCV. This wide support guarantees to be able to accelerate any application up to the most recent ones, with graphic performances reaching 25.6 GFlops. The GPU is built in a separate power domain, which allows the GPU to be switched off when not used in the long- term, or even to play with dynamic voltage frequency scaling (DVFS). The GPU graphic hardware acceleration is exposed through the following API: • OpenVG 1.2 for 2D or curve drawing • OpenGL/ES 3.1 for 3D apps (backward compatible: OES2.1 and OES1.1) • Vulkan 1.3 for modern 3D apps • OpenCL 3.0 for parallel programming • OpenVX 1.3 for acceleration of computer vision applications The GPU provides the following graphic theoretical performance (values for 800 MHz): • Vertex: 200 MVtx/s • Triangle: 133 MTrg/s • Texel: 800 MTex/s • Pixel: 800 MPix/s • Float 16bit: 25.6 GFlops • Float 32bit: 12.8 GFlops 3.5 Neural processor unit (NPU) Note: Features may be limited or absent in some devices or packages (see Section 2 for details). The NPU provides powerful hardware acceleration for neural network, to allow efficient artificial intelligence (AI) applications. The NPU acceleration is implemented by hardware neural operator inserted into the GPU. As such, it benefits of both optimized hardware (the neural operator), and of the flexibility and efficiency of the existing GPU shaders. The NPU neural hardware acceleration is exposed through the following API: • TensorFlowLite-API • ONNX • Linux NN-API-Adapter The NPU flexibility is used to optimally accelerate the following frameworks (nonexhaustive list) • TensorFlow non-exhaustive hardware support • TensorFlowLite full hardware support (including its SoftMax subset) • Caffe, Caffe2 • CNTK, Torch, Theano, Darknet The NPU provides the following neural theoretical performance, below values for 800 MHz (values for 900 MHz overdrive inside parenthesis): • Integer operations: 1.2 (1.35) TOPS (8-bit integer) To reduce the required DDR bandwidth during neural computations, the NPU embeds natively 128 Kbytes of memory. 3.6 Memories 3.6.1 External SDRAM STM32MP25xC/F devices embed a controller for the external SDRAM which supports the following devices STM32MP251C/F STM32MP253C/F STM32MP255C/F STM32MP257C/F Functional overview DS14284 - Rev 3 page 15/241 |
|
Link URL |
| Does ALLDATASHEET help your business so far? [ DONATE ] |
About Alldatasheet | Advertisement | Datasheet Upload | Contact us | Privacy Policy | Link to Datasheet | Link Exchange | Manufacturer List All Rights Reserved©Alldatasheet.com |
| Russian : Alldatasheetru.com | Korean : Alldatasheet.co.kr | Spanish : Alldatasheet.es | French : Alldatasheet.fr | Italian : Alldatasheetit.com Portuguese : Alldatasheetpt.com | Polish : Alldatasheet.pl | Vietnamese : Alldatasheet.vn Indian : Alldatasheet.in | Mexican : Alldatasheet.com.mx | British : Alldatasheet.co.uk | New Zealand : Alldatasheet.co.nz |
|
Family Site : ic2ic.com |
icmetro.com |