[oharboe]

Draft: waits for #4751 and #4810. This branch merges both so it can be built and tested; review only the last commit, 371ea26. It is rebased onto master once they merge.

Before: a variant that wanted another variant's synthesis or placement had to copy that variant's results into its own directory first.

Now: FLOW_INPUT_VARIANT names the variant a stage reads the previous stage's result from. One synthesis, floorplan or CTS is shared by many experiments, each forked off it with nothing copied:

make DESIGN_CONFIG=... FLOW_VARIANT=base cts
make DESIGN_CONFIG=... FLOW_INPUT_VARIANT=base FLOW_VARIANT=high_effort do-grt
make DESIGN_CONFIG=... FLOW_VARIANT=high_effort do-route do-final

(do-route running only route is #4811.)

This redoes #4475, which was closed after review: it touched 13 files to resolve every read against a search path (this variant first, then the other), and every reader had to remember to use the helper. #4751 and #4810 remove what made that necessary:

• flow: delete find_sdc_file and the per-step .sdc rules; every .odb has its .sdc #4751 pairs every .odb with its .sdc, so a stage's input is one stem read in one place, load_design, and find_sdc_file's ranking across two directories is gone.
• AUTO_MEMORIES cleanup: read the files blackboxes.txt names, no globs or existence checks #4810 makes AUTO_MEMORIES read the files its lists name, not what memories/ holds.

What is left is one explicit read per stage, and no search.

How

• variables.mk: FLOW_INPUT_VARIANT ?= $(FLOW_VARIANT) and INPUT_RESULTS_DIR. With the default every path is what it was.
• load.tcl: load_stage_input reads from INPUT_RESULTS_DIR, load_design from RESULTS_DIR; both are load_design_from <dir> <file>.
• The six reads of a previous stage's result call load_stage_input (floorplan.tcl 1_synth, global_place_skip_io.tcl 2_floorplan, cts.tcl 3_place, global_route.tcl 4_cts, detail_route.tcl 5_1_grt, density_fill.tcl 5_route), and their Makefile prerequisites name $(INPUT_RESULTS_DIR). Every other read, a stage's own earlier steps among them, stays in RESULTS_DIR.
• variables.yaml/.json, FlowVariables.md: the variable, its stages and its rules.

Fails loudly

When FLOW_INPUT_VARIANT names another variant, load_stage_input stops:

• The input is in both variants. One invocation that runs two stages, make FLOW_INPUT_VARIANT=base do-grt do-route say, would read route's input, 5_1_grt.odb, from base, not the one grt just wrote; a copy left in the fork from an earlier run is the same ambiguity. The fork reads another variant for its first stage only, and the stages after it with FLOW_INPUT_VARIANT unset.
• AUTO_MEMORIES=1. The memories are read from the variant that synthesised them, a second location this pull request does not add.

Known limit

After a fork, a file target such as make FLOW_VARIANT=high_effort route follows its prerequisites back through the fork's own directory and reruns the flow from synthesis there: correct, and the cost of the whole flow. do- targets run on the files in the tree and do not. #4811 makes do-route run only route, as the other do-<stage> targets do; until then, a fork at grt runs do-5_2_route do-5_3_fillcell do-5_route after do-grt.

make metadata in a fork reads the fork's own LOG_DIR and REPORTS_DIR: it reports the stages the fork ran, and the shared stages read N/A/ERR, as for any stage that did not run; their metrics are in the input variant's metadata.

copy_endpoint_slack reads the input step's endpoint-slack report from REPORTS_DIR and skips it when it is not there, so a fork whose first step only copies its input (GPL_PLACE_IOS=1 at place) has no endpoint-slack report for that step. That existence check predates this pull request.

Testing

tclfmt/tclint 0.7.0 clean. asap7/gcd with ORFS's make on this branch:

• base through cts, then a fork: do-5_1_grt with FLOW_INPUT_VARIANT=base, then do-5_2_route do-5_3_fillcell do-5_route do-6_1_fill in the fork. Against base run straight on, the .odb and .sdc of 5_1_grt, 5_2_route, 5_3_fillcell, 5_route and 6_1_fill are byte-identical. The fork's directory holds only those; it read base/4_cts.odb once.
• do-5_1_grt do-5_2_route in one fork invocation, with base already holding 5_1_grt.odb: route stops at 5_1_grt.odb is in both FLOW_VARIANT=span and FLOW_INPUT_VARIANT=base. .... Without the check it would read base's global route.
• A stale 4_cts.odb in the fork: do-5_1_grt stops at the same error for 4_cts.odb.
• grt route file targets in one fork invocation before base has grt: make stops at No rule to make target '.../base/5_1_grt.odb'.
• AUTO_MEMORIES=1 in a fork: stops at AUTO_MEMORIES=1 reads the memories synthesised in this variant, ....

No CI design sets FLOW_INPUT_VARIANT. With the default, every CI run takes load_stage_input with INPUT_RESULTS_DIR equal to RESULTS_DIR; a fork exercised in CI would need a step in the CI pipeline, which lives outside this repository.

🤖 Generated with Claude Code

[gemini-code-assist]

Code Review

This pull request introduces the FLOW_INPUT_VARIANT mechanism, allowing a flow run to read inputs from a shared upstream variant directory while writing outputs to its own variant directory. This is supported by helper functions in Tcl and updated Makefile variables. However, the feedback highlights a critical issue in flow/util/utils.mk where passing only INPUT_*_DIR to genMetrics.py in the metadata-generate and update_metadata_autotuner targets will result in incorrect or incomplete metadata, as metrics for stages run in the current variant will be read from the upstream variant instead.

[openroad-ci]

🔍 QoR check

Metrics reflect the PR merge build — i.e. what will land on the target branch.

Commit 84d336e · Jenkins build #1 · Baseline: build · View build on dashboard

62 design(s) checked — 0 with regression(s), 0 without a comparable baseline.