Want a faster ML processor? Do it yourself! -- A framework for playing with custom opcodes to accelerate TensorFlow Lite for Microcontrollers (TFLM). . . . . . Online tutorial: https://google.github.io/CFU-Playground/ For reference docs, see the link below.
Hi @tcal-x and @alanvgreen ,
Im trying to synthesize and run a large cfu design with multiple compute units and am having an issue where it is hanging at a very basic cfu store command. So, this is the interface Im using:
//Multi cycle Interface handshaking
if(func_id_full[2:0] == 3'b111) begin
rsp_valid = rsp_valid_sig;
cmd_ready = ~rsp_valid_sig;
end
else begin
rsp_valid = rsp_valid_sig;
cmd_ready = 0;
end
Im seeing that the whole thing hangs at line 289 in conv.h, corresponding to the first cfu_op0 command. Ive looked at the trace and seen that the rsp_valid_sig register (combinationally connected to rsp_valid) is
not being enabled on the cfu's end however, Ive confirmed that in the cfu.v file rsp_valid_sig is being flipped to 1 when it receives the cfu_op0 command as you can see in line 356 in cfu.v. This is baffling and Im not sure how to write it correctly for synthesis. Could you give me suggestions on the cfu.v file?
I've attached a debug dir with all relevant files including the cfu as well.
Please let me know about this as soon as you can, I have to do this within a deadline. Thanks for the help in advance.
Thanks,
Bala.
Hi @tcal-x and @alanvgreen , Im trying to synthesize and run a large cfu design with multiple compute units and am having an issue where it is hanging at a very basic cfu store command. So, this is the interface Im using:
Im seeing that the whole thing hangs at line 289 in conv.h, corresponding to the first cfu_op0 command. Ive looked at the trace and seen that the rsp_valid_sig register (combinationally connected to rsp_valid) is
not being enabled on the cfu's end however, Ive confirmed that in the cfu.v file rsp_valid_sig is being flipped to 1 when it receives the cfu_op0 command as you can see in line 356 in cfu.v. This is baffling and Im not sure how to write it correctly for synthesis. Could you give me suggestions on the cfu.v file?
I've attached a debug dir with all relevant files including the cfu as well. Please let me know about this as soon as you can, I have to do this within a deadline. Thanks for the help in advance. Thanks, Bala.