search

doonny / PipeCNN

An OpenCL-based FPGA Accelerator for Convolutional Neural Networks
Apache License 2.0
1.23k stars 370 forks source link

Can't fit design on DE1-SoC Board #86

Closed KamelAbdelouahab closed 6 years ago

KamelAbdelouahab commented 6 years ago

Thanks for the great work guys ! I'm trying to reproduce your results on a DE1-SoC device. However, it seems Quartus can't fit the the design on the board. C:\Users\Kamel\dev\PipeCNN\project> aoc -v -report -board=de1soc_sharedonly -I device/RTL -L device/RTL -l rtl_lib.aoclib ./device/conv_pipe.cl -o bin/pipe_cnn.aocx aoc: Selected target board de1soc_sharedonly aoc: Running OpenCL parser.... aoc: OpenCL parser completed successfully. aoc: Optimizing and doing static analysis of code... aoc: Linking with IP library ... Checking if memory usage is larger than 100%

!=========================================================================== ! The report below may be inaccurate. A more comprehensive ! resource usage report can be found at conv_pipe/reports/report.html !===========================================================================``

+--------------------------------------------------------------------+ ; Estimated Resource Usage Summary ; +----------------------------------------+---------------------------+ ; Resource + Usage ; +----------------------------------------+---------------------------+ ; Logic utilization ; 100% `; ; ALUTs ; 64% ; ; Dedicated logic registers ; 42% ; ; Memory blocks ; 110% ; ; DSP blocks ; 31% ; +----------------------------------------+---------------------------; aoc: First stage compilation completed successfully. Compiling for FPGA. This process may take a long time, please be patient. Error (170012): Fitter requires 4594 LABs to implement the design, but the device contains only 3207 LABs Error: Cannot fit kernel(s) on device Anyone of you tried with the DE1-SoC board ? or is it related to Quartus 18?

manojrohit commented 6 years ago

Try decreasing the LANE_NUM in project/device/hw_param.cl

KamelAbdelouahab commented 6 years ago

Thanks for the help, I was able to compile the Kernel for DE1-SoC with LANE_NUM = 4 Could tweak this value to have better throughput I guess. I will investigate that 👍

crazylazydboy commented 6 years ago

@doonny @manojrohit @KamelAbdelouahab Thanks for the help.. But now I'm trying to figure out how to implement Kernal including OpenCV part on DE1SoC.. Getting errors..