Hera devel

This has MB with 100ms timeout

Update gbe_rx.v

fixed 1 GbE throughput  problem

Casper astro soak test

So now the GPIO yellow block can be used to connect
to arbitrary pins listed in a platform's yaml file,
without having to mess with the GPIO mask to change
the drop down options.

This was implemented so that SNAP can use "rpi_header"
GPIOs to control the HERA phase switches.

Casper astro soak test

Copied from other functions. Not sure if this is correct,
but if it is, the variable being returned should probably have
its name changed. Right now it sounds like it's supposed
to be the number of bytes written.

This is long enough that a few retries can be had during a PROM
burning procedure, where the time between tries has to be ~1s to give
time for the flash erasing

the LWIP implementation of TFTP didn't
have any logic to ignore duplicate write packets
(i.e., a write packet resent by a client because
it failed to receive an ack on a previous packet).
This meant that duplicate packets were written twice,
and the write address was erroneously incremented.

This commit adds logic to track the expected TFTP
blknum (i.e., packet number) and re-ack, but not re-write
any duplicated packets.

words_left actually counted bytes, so not it's called
bytes_left

This is probably too long, but it takes several 100ms for a
flash page erase, so the client usually runs with a relatively
long timeout. If the server-side timeout is too short then
if a packet is lost by the time the client resends the server
will have given up on the connection.

Casper astro soak test

Merge branch 'casper-astro-soak-test' of https://github.com/jack-h/mlib_devel into casper-astro-soak-test

The fft-wideband-real block is based on a complex, biplex
fft core, which requires inputs to be multiples of 4 parallel real streams
to FFT.
The block used to check that this requirement was met by demanding that
the demux-factor -- "simultaneous samples per FPGA clock -- was a multiple
of 4. In reality, if multiple independent ADC streams are being transformed,
the requirement need only be N_ADC_STREAMS * DEMUX_FACTOR is a multiple of 4.
This commit implements this more relaxed requirement

This change courtesy Ezgi Korkmaz

The fft-wideband-real block is based on a complex, biplex
fft core, which requires inputs to be multiples of 4 parallel real streams
to FFT.
The block used to check that this requirement was met by demanding that
the demux-factor -- "simultaneous samples per FPGA clock -- was a multiple
of 4. In reality, if multiple independent ADC streams are being transformed,
the requirement need only be N_ADC_STREAMS * DEMUX_FACTOR is a multiple of 4.
This commit implements this more relaxed requirement

Remove duplicated code

Casper astro soak test

Include Tian's latest adc fixes

Casper astro soak test

Conflicts:
	casper_library/casper_library_accumulators.mdl
	casper_library/casper_library_communications.mdl
	casper_library/casper_library_correlator.mdl
	casper_library/casper_library_delays.mdl
	casper_library/casper_library_ffts.mdl
	casper_library/casper_library_flow_control.mdl
	casper_library/casper_library_forwarding_table.m
	casper_library/casper_library_misc.mdl
	casper_library/casper_library_multipliers.mdl
	casper_library/casper_library_pfbs.mdl
	casper_library/casper_library_reorder.mdl
	casper_library/casper_library_scopes.mdl
	casper_library/casper_library_sources.mdl
	casper_library/update_casper_block.m
	startsg
	startup.m
	xps_library/xps_library.mdl

changed clear_path.m to account for both windows and linux systems by…

changed clear_path.m to account for both windows and linux systems by adding the slash parameter used for the rest of gen_xps_files.m

High resolution adc

Add a testbench for wb_bram
Fix a bug of the MMCM clock generator for adc16_interface

The matlab perfile and frontend python arguments did not work. This has been fixed.

Added `override` to the `CFLAGS += ...` line in Makefile.  This allows
the user to provide some CFLAGS on the make command line without losing
the essential benefit of the CFLAGS defined in the Makefile.