Hey David, The control addresses also need changing in ROACH1 core_info.tab
(I'm pretty sure this is my fault) You're right that there's no need to map so much memory, but I had a memory allocation table from the ROACH2 verilog test design. A bundle was allocated to the sniffer, so I just mapped the whole lot. My feeling was, that this would generate EDK warnings if some memory supposedly for (but unused by) the sniffer were used somewhere else, which seemed like a good thing. I've just changed the core_info.tab for both roaches on the casper_vanilla branch of the oxfork github repository. I've left the full 64KB mapped for now. On 31 January 2013 17:24, David MacMahon <dav...@astro.berkeley.edu> wrote: > Hi, Jack, > > Thanks for your helpful reply! > > On Jan 31, 2013, at 2:01 AM, Jack Hickish wrote: > > > However, digging through the library, calibration flags (cal_fail and > phy_ready and a software qdr_reset) for the 4 QDR chips should be available > in 0x7000, 0x8000, 0x9000, 0xA0000 (does one of these appear in the > system.mhs you just compiled?) -- which don't appear to go into > core_info.tab anywhere. > > I do see an "opb_qdr_sniffer" instance in system.mhs that has (among other > details): > > PARAMETER C_CONFIG_BASEADDR = 0x000A_0000 > PARAMETER C_CONFIG_HIGHADDR = 0x000A_FFFF > PARAMETER C_BASEADDR = 0x0380_0000 > PARAMETER C_HIGHADDR = 0x03FF_FFFF > > so it does map a (64 KB!) config space to offset 0xA0000. > > > I expect these should be the addresses in core_info.tab pointed to by > qdr[X]_ctrl. > > I suspect you're right. The core_info.tab has: > > qdr3_memory 3 3800000 800000 > qdr3_ctrl 3 60000 100 > > The qdr3_memory at offset 0x3800000 matches the C_BASEADDR from > system.mhs, but it does seem that qdr3_ctrl in core_info.tab should have an > offset of 0xA0000 rather than 0x60000. Similar changes should probably be > made for all the qdrX_ctrl lines in the ROACH2 core_info.tab file. > > Thanks again, > Dave > >
