Added some flatglob handling to the GDS import script for magic
for the I/O library to get the refgen_new cell to read correctly.
Made a similar modification to the magic tech file as was done
on a recent commit:  That one modified the cifinput rule for
mvpdiff to capture an instance in the pad cell where the PSDM
is in a parent cell.  That correction sufficed for doing FOM fill,
but a similar change needed to be made to correctly read the
contacts over the same mvpdiff.

VERSION

diff --git a/VERSION b/VERSION
index a24cb9b..f7d7753 100644
--- a/VERSION
+++ b/VERSION
@@ -1 +1 @@
-1.0.499
+1.0.500

sky130/custom/scripts/gds_import_io.tcl

diff --git a/sky130/custom/scripts/gds_import_io.tcl b/sky130/custom/scripts/gds_import_io.tcl
index f020fd5..01d3e09 100644
--- a/sky130/custom/scripts/gds_import_io.tcl
+++ b/sky130/custom/scripts/gds_import_io.tcl
@@ -65,3 +65,9 @@
 gds flatglob *amuxsplitv2_switch_*
 gds flatglob *amuxsplitv2_hvsbt_nand*
 gds flatglob *amuxsplitv2_delay*
+
+### flatten parts of refgen_new, where there are subcells that are
+### spread across areas inside and outside of dnwell
+gds flatglob *refgen_ctl*
+gds flatglob *refgen_out*
+gds flatglob *__opamp_*

sky130/magic/sky130.tech

diff --git a/sky130/magic/sky130.tech b/sky130/magic/sky130.tech
index 8ca96e3..0f8519c 100644
--- a/sky130/magic/sky130.tech
+++ b/sky130/magic/sky130.tech
@@ -3436,12 +3436,13 @@
  or mvnscbase
  labels CONT
 
+ # See note above:  Not checking PSDM
  templayer mvpdcbase CONT
  or barecont
  and LI
  or barelicont
  and DIFF
- and PSDM
+ and-not NSDM
  and NWELL,nwelcheck
  and HVI,hvcheck