We need to include unistd.h for usleep():
hardware/asix-sigma/asix-sigma.c: In function 'sigma_fpga_init_bitbang':
hardware/asix-sigma/asix-sigma.c:450:3: warning: implicit declaration of function 'usleep'
usleep(10000);
And we need to tweak the type of chunks_per_read to match dl_lines_total:
hardware/asix-sigma/asix-sigma.c: In function 'download_capture':
hardware/asix-sigma/asix-sigma.c:1161:39: warning: comparison between signed and unsigned integer expressions
dl_lines_curr = MIN(chunks_per_read, dl_lines_total);
Init the cur_sample_rate to 200kHz. This is now the default sample
rate for the ASIX Sigma if no other is specified by -c samplerate=N
on the command line. Without this change, the driver would crash on
SIGFPE because a division by zero would happen.
Signed-off-by: Marek Vasut <marex@denx.de>
Due to the recent changes in the sigma driver, the trigger handling got
broken. Fix this by properly calculating and propagating the fact that
the trigger happened.
Signed-off-by: Marek Vasut <marex@denx.de>
Pull out the code for decoding the DRAM clusters into separate function.
This shall improve readability some more.
Signed-off-by: Marek Vasut <marex@denx.de>
These two values can pulled out from devc->state, so there really is
no need to pass them as function argument when we already pass the
devc.
Signed-off-by: Marek Vasut <marex@denx.de>
This function could have never returned correct data. Rework the decoding
logic according to the Sigma documentation. This patch also fixes the
handling of partial DRAM lines and partial DRAM clusters.
While at it, do not allocate megabytes of data on stack, just allocate 2KiB
which is plenty for all the work this function needs to do.
Signed-off-by: Marek Vasut <marex@denx.de>
The support for trailing DRAM lines was broken. This patch starts
rework of support for this, but in order to do that, we need to
rework decode_chunk_ts() a little first.
This patch adjusts the decode_chunk_ts() a little to receive the
total amount of events in DRAM line instead of some nonsense value.
This patch temporarily removes the support for the trailing DRAM
lines until the decode_chunk_ts() is fixed to cope with this, so
yes, this patch introduces breakage!
Signed-off-by: Marek Vasut <marex@denx.de>
Instead of this ad-hoc readout voodoo, use proper function to read
the timestemp. One which will not break even on big-endian machines.
Signed-off-by: Marek Vasut <marex@denx.de>
Use proper structure instead of plain buffer of uint8_t for the contents
of DRAM in download_capture(). This is beneficial as we can interpret the
contents easily.
Signed-off-by: Marek Vasut <marex@denx.de>
Clean up the way decode_chunk_ts() is called a little. Introduce a variable
which will not be -1 only in case a trigger happened on the particular DRAM
line.
Signed-off-by: Marek Vasut <marex@denx.de>
In the current configuration, the download capture used 32KiB buffer
for samples. This was the upper limit this function could download
from the Sigma. Even the sigma_read_dram() was only called once to
read up-to 32 DRAM line from address 0x0 in the DRAM.
This patch reworks the function to call sigma_read_dram() in a loop
in case there is need to download more than 32 DRAM lines of data
from Sigma. The data are then correctly passed for decoding to the
decoding function.
Signed-off-by: Marek Vasut <marex@denx.de>
Avoid allocating 32KiB of data on stack in download_capture(). Instead,
do a glib-variant of calloc(1, ) to allocate the data for the samples
which will be downloaded. This avoids explosions of stack on systems
with tight stack limits.
Furthermore, define structures describing the organisation of Sigma's
DRAM memory and start using those instead of ad-hoc preprocessor macros
defining the sizes of various structures in memory.
Signed-off-by: Marek Vasut <marex@denx.de>
The trigger position, stop position and chunk in which the trigger happened
are no longer needed in the global scope. Make those variables local to the
download_capture() function.
Signed-off-by: Marek Vasut <marex@denx.de>
Move all the register I/O that is necessary to do the download of samples
from Sigma into download_capture() function. This makes the downloading
code contained a bit more again.
Signed-off-by: Marek Vasut <marex@denx.de>
Reorder the sigma_capture_mode() function so that the part which handles
the download of samples from Sigma is clearly separated from the tests if
the download should be started.
Signed-off-by: Marek Vasut <marex@denx.de>
Pull out the code handling the Sigma which is in CAPTURE mode into
a separate function. This is so we can start reworking this entire
code easily soon.
Signed-off-by: Marek Vasut <marex@denx.de>
Check the position of ForceStop and Trigger events only in case we
are in CAPTURE state, it's useless to do this unconditionally when
receive_data() is called.
Signed-off-by: Marek Vasut <marex@denx.de>
Decode the logic mode start sequence into a series of FPGA instructions
instead and get rid of this sequence of magic numbers.
Signed-off-by: Marek Vasut <marex@denx.de>
Just encode the opcodes as hexadecimal values. This makes for better
readability when mapping the communication dump with the sigma to the
code.
Signed-off-by: Marek Vasut <marex@denx.de>
Pull out the code which switches Sigma from the FPGA programming mode
into Logic-Analyzer mode into separate function. Also, given the reply
is only ever 3-byte long, do not allocate 32 byte big buffer, but only
a 3-byte long one.
Signed-off-by: Marek Vasut <marex@denx.de>
Move the sigma_fw_2_bitbang() function closer to the upload_firmware()
function so there's not such a horrible mess in the file.
Signed-off-by: Marek Vasut <marex@denx.de>
Rework the bit2bitbang function. Remarkable changes are:
- The function was renamed to sigma_fw_2_bitbang()
- We use glib function calls to work with the file
- We mmap the file containing firmware instead of opening the file
and then reading it into a buffer.
- The magic firmware transformation is now type-safe.
- Documentation and comments were added where applicable.
Signed-off-by: Marek Vasut <marex@denx.de>
Remove all those if ((ret = foo(bar)) < 0) constructs from upload_firmware()
function. This is just a confusing programming practice, kill it. While at it,
replace all the uses of &devc->ftdic with plain ftdic , which is defined at
the begining.
Signed-off-by: Marek Vasut <marex@denx.de>
The code silently assumed the firmware path can be no longer than 128 bytes.
This doesn't scale. This patch fixes it in such a way that it completely
rips out the run-time computation of firmware path and instead replaces it
with compile-time computation. It's true this makes the library grow by a
couple bytes, but makes the code cleaner.
Signed-off-by: Marek Vasut <marex@denx.de>
Rework the pre-firmware-upload FPGA initialization sequence so it matches
the documentation. Also, since this sequence is documented in a separate
section, wrap it into separate function.
Signed-off-by: Marek Vasut <marex@denx.de>
The NUM_CHANNELS macro is inflexible, since in 100MHz and 200MHz modes
we don't support 16 channels. Moreover, it's only used to limit the size
of array of channel labels, which can be done in much cleaner way.
Signed-off-by: Marek Vasut <marex@denx.de>
Add comments to the sampling rate table explaining how the frequencies
are selected and where do those numbers come from.
Signed-off-by: Marek Vasut <marex@denx.de>
Also, consistently use 'ch' for channel variables. This matches how we
consistently use sdi, devc, and so on all over the code-base.
This fixes parts of bug #259.
The name 'probe' (and thus 'probe group') is a relic from the times when
sigrok was mostly about logic analyzers. Nowadays we support a lot more
device types where 'probe' is not really a good term and 'channel' is
much better suited.
This fixes parts of bug #259.
Instead of >= 44 Makefile.am's we now only have one top-level
Makefile.am, and use the 'subdir-objects' automake option to
handle the build via non-recursive (auto)make.
This has the advantage of fewer (boilerplate or other) files and less
clutter in general, as well as performance advantages since the new
setup can build many files in parallel (with 'make -j'), not only 2 or 3
files within the same (e.g. hardware/xxxx/* subdirectory) and also since
we no longer need to build intermediate libtool helper libs per subdirectory.
A quick, non-scientific test build on a quad-core laptop with 'make -j 4'
yields a build time reduction from 35s to 19s.
All autotools features that worked before are still intact without any
regressions, including the Make targets 'install', 'uninstall', 'check',
'dist', 'clean', 'distclean' and so on, as well as all the usual portability
handling (build works on any OS, with any Make implementation such as
GNU Make or BSD Make, with any shell such as sh/ksh/zsh/bash/dash, etc. etc.)
and features such as out-of-tree build support, cross-compile support,
testsuite support (also with colored output), "silent make rules", etc. etc.
This fixes the following warning:
asix-sigma.c: In function 'receive_data':
asix-sigma.c:1064:4: warning: dereferencing type-punned pointer will break strict-aliasing rules [-Wstrict-aliasing]
devc->state.lastts = *(uint16_t *) buf - 1;
^
The per-driver API calls no longer have a hw_ prefix (e.g. hw_init()
became init() and so on), so drop the 'hw_' from the std versions
for those API callbacks too.
Mike Frysinger
Bert Vermeulen
Marek Vasut
Uwe Hermann
Aurelien Jacobs
Martin Ling