Floating point and fixed point

doonny / PipeCNN

An OpenCL-based FPGA Accelerator for Convolutional Neural Networks

Apache License 2.0

1.22k stars 370 forks source link

Floating point and fixed point #38

Closed zhao-lun closed 6 years ago

zhao-lun commented 6 years ago

hi Prof @doonny , I noticed the current version is in fixed point mac. And how was the resource utilization compared to Floating point mac? What is the Pipe depth ? #define PIPE_DEPTH 6

doonny commented 6 years ago

There is no floating point version.

PIPE_DEPTH refers to the depth of the accumulate buffers.

zhao-lun commented 6 years ago

[History:] -v1.0 Initial release https://github.com/doonny/PipeCNN/blob/01799e66f69f4ce7dc123795717b1f3e0ab8aeb0/project/device/conv_pipe.cl Here you uses float for both data and weight. float mac(lane_data input, lane_data weights) I read your paper, you got 33.9GOPS when running on Stratix-V (floating point) Prof @doonny , Why is cov split used? conv_split 0,1,0,1,1,0,0,0 //layer 1 -> layer 8

doonny commented 6 years ago

Floating-point format is no longer supported.