aboutsummaryrefslogtreecommitdiffstats
path: root/contrib/TODO.template
diff options
context:
space:
mode:
authorbryan newbold <bnewbold@leaflabs.com>2013-11-12 12:54:30 -0500
committerbryan newbold <bnewbold@leaflabs.com>2013-11-12 12:54:30 -0500
commit92ed49e97997d769f41f42260e56a41f8391559d (patch)
treeedfa72346aea5ebbbbcd550d52ae8f299cf12e5c /contrib/TODO.template
parent79815d744079daadb5a3808b03ed4d45ffa76365 (diff)
downloadbasic-hdl-template-92ed49e97997d769f41f42260e56a41f8391559d.tar.gz
basic-hdl-template-92ed49e97997d769f41f42260e56a41f8391559d.zip
Part 1 of refactoring template files into ./contrib
BROKEN without later parts (documentation and Makefile updates) This commit moves and deletes a lot of Xula2 and SP605 files around. It also includes a large cleanup of xilinx.mk
Diffstat (limited to 'contrib/TODO.template')
-rw-r--r--contrib/TODO.template45
1 files changed, 45 insertions, 0 deletions
diff --git a/contrib/TODO.template b/contrib/TODO.template
new file mode 100644
index 0000000..cc47521
--- /dev/null
+++ b/contrib/TODO.template
@@ -0,0 +1,45 @@
+
+switch to .EXPORT_ALL_VARIABLES and/or .ONESHELL (as a refactor/cleanup)?
+ or is that too gmake specific...
+
+BUG: synth still seems to continue even if first build (verilog compile)
+ fails
+
+add .PRECIOUS for intermediate files we don't want to get deleted
+
+'lint' should use vfiles, not -I./hdl
+
+for fpga_editor:
+ DISPLAY=`echo $DISPLAY |sed s/'\.0'//` fpga_editor <.ncd file>
+
+effort levels seem high by default:
+ Overall effort level (-ol): High
+ Router effort level (-rl): High
+
+impact:
+ impact -mode bscan -b build/sp605.bit -port auto -autoassign (needs testing)
+
+requests from AJ:
+ anything related to not rebuilding all the coregen when not necessary.
+
+ e.g. want a make clean equivalent to removing the build dir but not the
+ cores. This should actually be the default behavior, with different
+ operator for nixing the cores.
+
+ not auto rebuilding the cores when switching branches/commits if not
+ strictly necessary. Because git touches all the files, this may be
+ difficult.
+
+ make isim/simulate will run to completion even if there were errors on
+ the build.
+
+ In the case of make isim, isim will load and run the previously valid
+ simulation. Unless you happen to see the error go by in the build log,
+ you will unknowingly be simulating your previous build, whereas your
+ current build failed to compile.
+
+ The solution is to have make simulate begin by deleting the previous
+ simulation executable, so that it must successfully create a new one
+ before loading isim.
+
+Add linting to tests (aka, tb/*_tb.v)