- 11 Jul, 2014 2 commits
-
-
Lucas Russo authored
This finishes the quick fix for github issue #21.
-
Lucas Russo authored
As of now, we don't have a way of checking which channel is tha valid one. So, we hardcode this limit in software, for now. This is a quick fix for github issue #21.
-
- 10 Jul, 2014 9 commits
-
-
Lucas Russo authored
This avoid buffering the zmq message and hiding our debug
-
Lucas Russo authored
-
Lucas Russo authored
As we have added new logging functions in new files, we need to add them to libclient too
-
Lucas Russo authored
-
Lucas Russo authored
Now, we can pass the log directory we want for dev_mngr and dev_io instances
-
Lucas Russo authored
Now, we can pass the log directory and the dev_mngr puts together the dev_mngr and dev_io log files based on the number of devices currently running. For now, the design is very simple and subject to errors, such as when a dev_io exits. In this case, the variable __dev_nums would be decremented, but we wouldn't know which of the dev_ios instances exited. The next time a dev_io is started, the last dev_io logfile and device name "/dev/fpga%d" would be choose for it, causing the logs to be corrupted as well as PCIe access.
-
Lucas Russo authored
In this way we can redirect zeroMQ message prints to a logfile
-
Lucas Russo authored
-
Lucas Russo authored
-
- 08 Jul, 2014 4 commits
-
-
Lucas Russo authored
For now, we default it to "dev_mngr.log", but it can be any name. This completely fixes #9 github issue.
-
Lucas Russo authored
Added message log initialization with command-line support. It is possible to output messages to stdout as before. Just pass the "stdout" string (without the quotation marks) to the command-line. This partly fixes github issue #9
-
Lucas Russo authored
Added message log initialization with command-line support. It is possible to output messages to stdout as before. Just pass the "stdout" string (without the quotation marks) to the command-line. This partly fixes github issue #9
-
Lucas Russo authored
Now, we can log all messages directly into a file of our choice. Also, it was added the error level and the current timestamp for identification purposes. This partly fixes github issue #9
-
- 07 Jul, 2014 4 commits
-
-
Lucas Russo authored
Now, the addresses are realtive to PCIe BARs and we don't need to know whether we are talking to a PCIe device or an Ethernet one
-
Lucas Russo authored
As we need to switch between BARs, we fake our real addresses by concatenating the BAR number in the 4 MSB. In this way, the PCIe LLIO can differentiate between BARs ans select the correct one
-
Lucas Russo authored
This was clearly a typo, as we handle the error inside the SMIO macro.
-
Lucas Russo authored
-
- 01 Jul, 2014 1 commit
-
-
Lucas Russo authored
-
- 30 Jun, 2014 5 commits
-
-
Lucas Russo authored
Now, whenever possible we use the new error handling, in which we can set an error variable inside the ASSERT macros. This should easy the code maintainability. This, together with commit f1a961df fixes issue #15.
-
Lucas Russo authored
This is bad! I had to add the variable arguement list to all files that used the ASSERT macros in rder to support the ioptional variable setting. There is a need to fix this properly. This is the first commit towards fixing issue #15.
-
Lucas Russo authored
Now, we can set an optional error (already declared) variable. This is useful when return errors through multiple subsystems, where error conversion is necessary.
-
Lucas Russo authored
They are now in a separated file located in "hal/include/hal_varg.h"
-
Lucas Russo authored
This will allow for ease reuse of the variable argument macros expansion
-
- 27 Jun, 2014 6 commits
-
-
Lucas Russo authored
The new poll API have the advantage that it is balanced. So, even if one socket gets more mesages (requests) than the other, they both get served in a fair round robin algorithm. This commit together with 3759861f fixes issue #13.
-
Lucas Russo authored
There are 2 new methods for doing the polling on dev_io sockets. They are named devio_poll2_all_sm () and devio_init_poller2_sm () to differentiate from the old methods and it performs a balanced polling. For now, both of them are going to coexist. In the future the old methods might be removed.
-
Lucas Russo authored
-
Lucas Russo authored
Using the derefenrced pointer in a sizeof is more generic and avoids a mismatch problem if we change the type of the pointer
-
Lucas Russo authored
-
Lucas Russo authored
For better visualization and standardizing, we prefer lowercase names for module names.
-
- 26 Jun, 2014 4 commits
-
-
Lucas Russo authored
Now, all paths, (direct and inverted) are written in this example.
-
Lucas Russo authored
The previous behavior still continued in the loop but didn't executed anything. Now, if zctx_interrupted becomes active the current loop exits appropriately
-
Lucas Russo authored
Now, if the zctx_interrupted becomes active, a BPM_CLIENT_INT error is issued and it exits the current task properly
-
Lucas Russo authored
When acquiring data from the "acq" module, the number of samples has to be multiple of 4 (for ADC channel) or 2 (for other channels). The previous calculation was wrong as it didn't take into account for the alignment number. This fixes issue #10
-
- 24 Jun, 2014 2 commits
-
-
Lucas Russo authored
-
Lucas Russo authored
-
- 16 Jun, 2014 3 commits
-
-
Lucas Russo authored
-
Lucas Russo authored
This closes issue #2
-
Lucas Russo authored
- Fix wrong comment - Fix indentation throughout the file
-