Simplify handling of Liberty file related variables #2619
Conversation
…SC_LIB_FILE Signed-off-by: Emil J. Tywoniak <[email protected]>
widlarizer
force-pushed
the
emil/dont-dont-use
branch
from
e7da6d6 to
d0dee53
Compare
|
This can be followed up by integrating YosysHQ/yosys#4797 and redoing #2139 if possible |
| @@ -23,7 +23,7 @@ proc write_keep_hierarchy {} { | |||
| if {[env_var_exists_and_non_empty DFF_LIB_FILE]} { | |||
| dfflibmap -liberty $::env(DFF_LIB_FILE) | |||
| } else { | |||
| dfflibmap -liberty $::env(DONT_USE_SC_LIB) | |||
| dfflibmap -liberty $::env(SC_LIB_FILE) | |||
There was a problem hiding this comment.
The only reason for this if/else is because dfflibmap used to only take one liberty file. If it accepts multiple now, then you can simply pass in LIB_FILES_EXTRACTED and eliminate the DFF_LIB_FILE variable entirely. In fact, DFF_LIB_FILE was added exclusively to support ASAP7, no other PDK uses it.
| @@ -67,7 +67,7 @@ foreach cell $::env(DONT_USE_CELLS) { | |||
| if {[env_var_exists_and_non_empty DFF_LIB_FILE]} { | |||
| dfflibmap -liberty $::env(DFF_LIB_FILE) {*}$dfflibmap_args | |||
| } else { | |||
| dfflibmap -liberty $::env(DONT_USE_SC_LIB) {*}$dfflibmap_args | |||
| dfflibmap -liberty $::env(SC_LIB_FILE) {*}$dfflibmap_args | |||
| $(LIB_FILES_EXTRACTED): $$(filter %$$(@F) %$$(@F).gz,$(LIB_FILES)) | ||
| @mkdir -p $(OBJECTS_DIR)/lib | ||
| $(UTILS_DIR)/preprocessLib.py -i $^ -o $@ | ||
| @if echo "$<" | grep -q "\.gz$$"; then \ | ||
| gunzip -c $< > $@; \ | ||
| else \ | ||
| cp $< $@; \ | ||
| fi |
There was a problem hiding this comment.
Does Yosys support reading in compressed .gz files and decompressing on-the-fly? This step could be entirely eliminated if so.
There was a problem hiding this comment.
It does not, but it could. It requires zlib by default to support .fst wave files. I'll see what I can do
There was a problem hiding this comment.
I think it does when you supply the file to a frontend (e.g. read_liberty) but not e.g. in dfflibmap -liberty
original_pinused in Liberty file pin declarations of cells it actually uses to map to even iforiginal_pinis not defined as user-defined. Probably from Synopsys PrimeTime folklore? Can't find much of a trace. Let me know if private PDKs break with thisdont_useattributes to some of the cells in those filesobjects/I tried this out with asap7 and nangate and it seems to work fine
cc @povik