Rev |
Log message |
Author |
Age |
Path |
23 |
Updated headers |
dgisselq |
1957d 01h |
/qspiflash/trunk |
22 |
Read only is now an option of wbqspiflash
It was a macro before. |
dgisselq |
1957d 01h |
/qspiflash/trunk |
21 |
Added a formal proof of the lower level QSPI driver |
dgisselq |
2430d 14h |
/qspiflash/trunk |
20 |
Changed reference to flash_config to flashconfig.v |
dgisselq |
2430d 14h |
/qspiflash/trunk |
19 |
LLQSPI now passes formal tests |
dgisselq |
2430d 14h |
/qspiflash/trunk |
18 |
Renamed the flash config file |
dgisselq |
2430d 14h |
/qspiflash/trunk |
17 |
Added an option to turn off verbose outputs per tick |
dgisselq |
2430d 14h |
/qspiflash/trunk |
16 |
Added a full blown test bench to the controller
This includes:
- A global make file, and a "make test" which will build the test bench.
- Merging other versions of the QSPI flash driver I had lying around
- Grabbing the best of these, and verifying that they work
- The result is a reduction in overall logic |
dgisselq |
2572d 19h |
/qspiflash/trunk |
15 |
Made the flash simulator's size variable.
Also adjusted the C interface, providing some new functions, so that the flash
simulation can be loaded and queried easier. |
dgisselq |
2594d 01h |
/qspiflash/trunk |
14 |
Updates: little-big endian, various other fixes
1. Made the wbqspiflash.v and llqspi.v files compile with default_nettype none
2. Changed the internal flash representation to big endian. A little-big
endian conversion is now required when writing to the flash from a PC.
3. Simplified the address description via w_wb_addr and w_spif_addr, so that
the core is more flexible when changing sizes.
4. Removed the dependence upon the WB_CYC line ... as part of the WB
simplifications I've been doing.
5. Got XIP working for the EQSPI flash (I guess --- it's been a while since
I made those changes)
6. Adjusted (fixed) sim of read/writes to the volatile config register
(necessary for XIP) |
dgisselq |
2594d 01h |
/qspiflash/trunk |
13 |
Fixed VERILATOR_ROOT static dependency. |
dgisselq |
2796d 15h |
/qspiflash/trunk |
12 |
Adding the sources for the extended Quad SPI flash test bench. |
dgisselq |
2863d 16h |
/qspiflash/trunk |
11 |
This code has been proven, and is currently working within an Arty platform. |
dgisselq |
2863d 16h |
/qspiflash/trunk |
10 |
Switched to Quad Output mode by default. |
dgisselq |
2880d 14h |
/qspiflash/trunk |
9 |
Minor changes to the baseline, FIRST RELEASE OF THE EQSPIFLASH controller!! |
dgisselq |
2880d 16h |
/qspiflash/trunk |
8 |
Minor documentation edits. |
dgisselq |
3111d 13h |
/qspiflash/trunk |
7 |
Added the ability to make the Quad/SPI flash run in read-only mode, in an
effort to reduce it's resource/LUT usage enough to run on a Digilent Cmod-S6. |
dgisselq |
3131d 22h |
/qspiflash/trunk |
6 |
Minor documentation (formatting) changes. |
dgisselq |
3299d 19h |
/qspiflash/trunk |
5 |
Minor changes to the documentation. |
dgisselq |
3306d 23h |
/qspiflash/trunk |
4 |
FIXED: in the previous version, the first read of the device would always fail.
This was due to the fact that the Xilinx loader that read from the device
to load the Xilinx configuration left the Quad SPI flash chip in a high
data rate state. Thus, on reset, the first thing this core does it to
take the device out of the high data rate state.
Also, on a bus action waiting on a write to complete, the timing was
off. This has now been corrected. Reads (in both modes) should now
stall the bus (properly) while a write/erase cycle is ongoing, and
complete when finished.
The bug in high speed writing was traced to a read bug, not a write bug.
High speed (i.e. 4-bit transaction) writing has been re-enabled.
The bug in transitioning from 1-bit mode to 4-bit reads has been fixed.
This was causing the high speed write bug mentioned above.
Read ID was stalling the bus. This was traced to a sign error while
referencing the chip select line, and has been fixed.
The lowerlevel QSPI project was modified to add a holding state. We were
struggling with a bug whereby a late read would hang the device. The upper
level driver would issue it when the lower level driver was busy, and yet
think that it was accomplished. The lower level driver was moving from ready
to idle, so it never saw the read. Now, hold keeps the lower level driver
in the ready state at the end of a read until the bus transaction is complete,
or until it goes on to some other transaction other than reading data.
(This was a big change.)
Finally, unnecessary debugging lines were disabled in the simulator. |
dgisselq |
3311d 20h |
/qspiflash/trunk |