The names RISC-V, SiFive, PIC, AVR, Arm, Kinetis, Freedom Studio, Freedom E310 (FE310), Sparkfun Electronics, RedBoard, RED-V, ARM, and Microsoft, as well as related names, marks, emblems, and images are trademarks of their respective owners.Įxternal Links: Links to external web pages have been provided as a convenience and for informational purposes only. Notice of Non-Affiliation and Disclaimer: As of the publication date, we are not affiliated with, associated with, authorized by, endorsed by, compensated by, or in any way officially connected with the RISC-V Foundation, SiFive, Sparkfun Electronics, ARM Holdings, Microchip Technology Inc., NXP, TSMC, or Microsoft, or their owners, subsidiaries or affiliates. This guide will focus on the FE310-G002 microcontroller created by SiFive, Inc. As a result, there are numerous RISC-V devices in circulation created by universities, startup companies, and national laboratories. The RISC-V name is short for “Reduced Instruction Set Computer Five.” Without going into great historical detail, the project began as an open alternative to traditional ISAs. Figure 1: Two Development Kits for RISC-V Microcontrollers Now it’s time to revisit the RISC-V development kits and try our hand at driving external components. Thanks to new development kits like the SiFive Revision B board, future developers can get started with minimal resources and training. Tensorflow/lite/micro/examples/hello_world/main.The RISC-V instruction set architecture (ISA) is rapidly gaining traction with electronics and robotics enthusiasts. This one is generic, it does not have mbed-specific code: Still not clear for me the: there is the board-specific int main() ?
lib files: BSP_DISCO_F746NG.lib LCD_DISCO_F746NG.lib mbed-os.libįor my board I can probably manually put the. This tag probably leads to appearing in mbed project folder the 2. The original example for DISCO-stm32f746 board uses in the step 1) TAGS="CMSIS disco_f746ng" I think I need to run the following steps:ġ) make -f tensorflow/lite/micro/tools/make/Makefile TARGET=mbed TAGS="cmsis-nn"Ģ) cd tensorflow/lite/micro/tools/make/gen/mbed_cortex-m4/prj/hello_world/mbedĥ) mbed compile -m NUCLEO_H743ZI2 -t GCC_ARMīut I cannot figure out from the original example where to put the board-specific main.cpp and the board-specific *.lib files.
#Importing segger embedded studio project to source insight how to#
I cannot figure out how to modify the original code to compile for my. I do not have this exact board, I have this board: ST-Nucleo-H743ZI2 There is an official "hello world" TF micro example (sine prediction): These types of functions are use the get the needed memory: It's doable since the size is deterministic when the model is known, but in practice it may need some hacking. To run this on x86 can be tricky, perhaps we need a call on that. The best method (I can think of right away) is to check the 'high water mark' in the arena after greedy planner has done its thing. It's not straight fwd to derive how much/if the arena size increases.
Scratch tensors is a general concept and not specific for CMSIS-NN. It may increase the required arena size due to the fact it uses scratch tensors.
0 kommentar(er)
