Hi all,
What is the recommended way of supplying run-time configurable parameters to
HLS custom logic in an RFNoC block?
I tried modifying the gain example OOT block by adding a HLS option for the
user logic:
if (IP_OPTION == "HLS") begin: gen_hls
gain_hls gain_hls_i (
.ap_clk (axis_data_clk),
.ap_rst_n (~axis_data_rst),
.in_r_TDATA (m_in_payload_tdata),
.in_r_TVALID (m_in_payload_tvalid),
.in_r_TREADY (m_in_payload_tready),
.in_r_TLAST (m_in_payload_tlast),
.gain_TDATA (reg_gain), // (reg_gain) Real gain, {16'd0,
reg_gain}
.gain_TLAST (m_in_payload_tlast),
.gain_TVALID (m_in_payload_tvalid),
.gain_TREADY (),
.out_r_TDATA (mult_tdata),
.out_r_TVALID (mult_tvalid),
.out_r_TREADY (mult_tready),
.out_r_TLAST (mult_tlast)
So, similar to the HDL_IP option, reg_gain is input as tdata to an AXI stream
input with a name "gain". Tlast and tvalid come from the master payload stream.
In the HLS source I have an extra input port for the gain parameters:
void gain_hls(
hls::stream<axis_cplx>& in,
hls::stream<axis_gain>& gain,
hls::stream<axis_cplx_32>& out)
...
Data type for gain is an AXI stream for short ints.
However, testbed simulation times out when various gain values are tried. If I
completely comment out the gain input stream, testbed works fine, and fails due
to wrong output values (with gain being hard-coded as a constant).
I am puzzled about what could be wrong, since the definition for the HDL_IP
option is defined similarly and works without problems.
BR,
Kalle Hanhijärvi
_______________________________________________
USRP-users mailing list -- usrp-users@lists.ettus.com
To unsubscribe send an email to usrp-users-le...@lists.ettus.com
