Description

Version Found: v7.1 Memory IP

Version Resolved: See (Xilinx Answer 58435)

Vivado and the UltraScale Memory IP use an absolute path to point to the custom imported CSV file.

This causes problems if the project is moved from its original location.

Solution

To resolve this problem, the Memory IP should only be moved using the "Project Archive" option in Vivado. The output products and synthesized DCP should then be regenerated.

If "Project Archive" cannot be used, then the IP must be regenerated from scratch, or the following steps can be followed:

Option 1:

1. Hand modify the CSV path in the following files:
   
• ./<project>.srcs/sources_1/ip/<ip_name>/<ip_name>.xci\
• ./<project>.srcs/sources_1/ip/<ip_name>/<ip_name>.xcml
• ./<project>.srcs/sources_1/ip/<ip_name>/ip_0/<ip_name>_phy.xci
• ./<project>.srcs/sources_1/ip/<ip_name>/ip_0/<ip_name>_phy.xcml


2. Find each instance of the CSV file path and correct it to the new location (i.e. *_customPartAbsPath)


3. Regenerate the Synthesized DCP if using Out-of-Context mode.
   
   Note: Opt_design uses the absolute CSV file path inside the DCP

Option 2 :

Use the following Tcl commands after synthesis on the synthesized DCP:

set_property CustomPartAbsPath [file normalize ../../<relative path to CSV file>.csv] [get_cells -hierarchical -filter {REF_NAME == <MIG top-level name>}]

write_checkpoint -force [lindex [find_top] 0].dcp

For future designs, starting in Vivado 2016.1, the "Project Archive" option should be used.

Older IP brought into Vivado 2016.1 should be upgraded as the CSV file path is also removed from the DCP to resolve any opt_design errors.

Revision History:

Initial Release - 02/22/2016

Linked Answer Records

Master Answer Records