Remove Hardcoded register space 64KB in shim and zocl #8300
Conversation
ManojTakasi
requested review from
chvamshi-xilinx,
stsoe,
mamin506 and
rozumx
as code owners
ManojTakasi
changed the title
| try { | ||
| for (auto& xml_inst : xml_kernel) { | ||
| if (xml_inst.first != "instance") | ||
| continue; | ||
| for (auto& xml_remap : xml_inst.second) { | ||
| if (xml_remap.first != "addrRemap") | ||
| continue; | ||
| address_range = convert(xml_remap.second.get<std::string>("<xmlattr>.range")); | ||
| if (address_range != 0) | ||
| return address_range; | ||
| else | ||
| return default_address_range; | ||
| } |
There was a problem hiding this comment.
This is not the same behavior. The value returned is not the same after this change. There are xclbins with kernel attribute "range=0xValue" but same xclbins do not have range attribute. I will vaguer that you cannot make this change.
How can RISK be N/A when you are changing the behavior of an existing function?
There was a problem hiding this comment.
Thank you Soren for your inputs and for pointing that out. I haven't considered the case where xclbins might not have the range attribbute. I will make the necessary adjustments to ensure the behaviour remains consistent.
There was a problem hiding this comment.
Hi Manoj,
As discussed offline, Please remove this change and ask users to have their xclbin's updated with proper range information of slave,
There was a problem hiding this comment.
Hi Vamshi, Soren,
The Changes have been made as suggested. could you pelase review them once again? Please let me know if any further changes are required.
Signed-off-by: Manoj Takasi <[email protected]>
Signed-off-by: Manoj Takasi <[email protected]>
Problem solved by the commit
CR-1185445 XRT has hardcoded the register space for read/write to 64KB.
Bug / issue (if any) fixed, which PR introduced the bug, how it was discovered
In the zocl side, we are hardcoding CU size as 64KB size, mmap is failing for IP's which has more than 64KB size .
How problem was solved, alternative solutions (if any) and why they were rejected
Removed Hardcoded check 0x1000 from the user side and zocl side.
Risks (if any) associated the changes in the commit
NA
What has been tested and how, request additional testing if necessary
Tested on vck190
Documentation impact (if any)
NA