Skip to content

[DO NOT MERGE] test: Jenkins QoR stage - #4640

Closed
vvbandeira wants to merge 56 commits into
The-OpenROAD-Project:masterfrom
vvbandeira:qor-red-stage-test
Closed

vvbandeira wants to merge 56 commits into
The-OpenROAD-Project:masterfrom
vvbandeira:qor-red-stage-test

Conversation

@vvbandeira

Copy link
Copy Markdown
Member

Do not merge. Test only.

Runs nangate45/gcd and sky130hd/gcd with a jenkins-ci branch that changes how a QoR regression shows in Jenkins. nangate45/gcd has a tighter clock so it fails the QoR check.

I will close this PR after the check.

migueldalberto and others added 23 commits September 28, 2026 20:15
Adds --buildUserEmail and --changeAuthorEmail to uploadMetadata.py, sent
as the top-level build_user_email and change_author_email keys on every
schema version, v1 per-design fallback included. Blank values are left
out. The dashboard stores them on the build and attributes the build to
the user who started it, so a "my branches" view can include builds a
user started by hand.

They describe the build rather than the payload's shape, so they do not
bump payload_schema_version. The backend must already register them as
metadata before this lands: a v1 message carries metrics at its root, and
an older backend would ingest the keys as metrics.

Signed-off-by: Miguel Dalberto Pedro <miguel.pedro@precisioninno.com>
A new test design with one concern: an asynchronous clock crossing,
constrained without exceptions that hide a path, and implemented by the
flow as constrained. It is XiangShan's CHI bridge between its tile and
its network-on-chip, both halves in one block: six flit channels, each
a rocket-chip async queue, with their link credits, handshakes and a
reset synchroniser per domain, as readable parameterised SystemVerilog
read by slang.

constraint.sdc bounds every crossing with set_max_delay
-ignore_clock_latency, budgets the ports per domain as the platform's
constraints.sdc budgets a macro's, keeps synchroniser hops short, and
checks the names it constrains against the synthesised netlist,
stopping the flow if they are wrong.

The core period, 300 ps, is where XiangShan's own bridge, elaborated
from its Chisel and hardened alone through this flow, just fails at
global route; this design fails there on the same path, within 1.5 %.
README.md has the numbers. base is XiangShan's bridge as it is;
FLOW_VARIANT=small|medium are faster rungs.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The dashboard resolves inline checks in two ways. A request that names a
pipeline (JobName or PipelineId) is a build of that pipeline and follows
its CI branch rules. Any other request is a checkout: the dashboard
searches its default baseline pipelines and takes a master build that
has the design, at the commit's merge base or else the latest one
before the commit.

So the script no longer sends IsLocal or a default job name. Without
--job-name it sends just the commit metadata.json records (--commit /
DASHBOARD_COMMIT overrides it, and must be a full 40-character SHA:
the dashboard matches it exactly and falls back to the latest master
build on a miss). With --job-name it sends that name and
the branch, from --branch / DASHBOARD_BRANCH or git, as CI does. A
--base-commit pin no longer needs a job name. A run whose metadata.json
records no commit is compared against the latest master build, and the
report now says so.

--api-url / DASHBOARD_API_URL picks the deployment and takes precedence
over --beta. The report names the pipeline that provided the baseline.

Also drop non-finite metric values before sending: json.dumps renders
them as bare Infinity/NaN, and the dashboard rejects the whole request.

Signed-off-by: Miguel Dalberto Pedro <miguel.pedro@precisioninno.com>
Signed-off-by: Miguel Dalberto Pedro <miguel.pedro@precisioninno.com>

# Conflicts:
#	flow/util/uploadMetadata.py
A memory inferred inside a larger module was being converted. The macro
was generated, the module name written into blackboxes.txt, and
`blackbox <name>` then matched nothing -- because the name is a memory
inside a module, not a module. The design synthesized to flip-flops
while memories.json reported the memory as converted.

That is worse than not converting it: the flow completes, the inventory
and the netlist disagree, and the area and power look plausible. On
picorv32 the register file -- an inline `reg [31:0] cpuregs [0:31]`
array -- was recorded as

    "behavioral_model": {"module": "$paramod$...\\picorv32"},
    "idiomatic": true, "reason": "meets ASAP7 macro floors"

and synthesized as 1024 flip-flops.

An inline array is the design asking for flip-flops, so leaving it alone
is the right answer, and the reason says how to get the other one:
instantiate the memory as its own module. The gate is whether the memory
is its own behavioral model, which is what "memories at the module
boundary" already means everywhere else in this code.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
…efuse a combinational read

Chisel designs reach the flow through firtool, which emits every
inferred memory as its own module -- ram_<depth>x<width>,
array_<depth>x<width> -- whose ports follow one convention,
<R|W|RW><n>_<addr|en|clk|data|wdata|rdata|wmode|wmask>, and whose
register array is always called `Memory`. The detector got three things
wrong about them, and together they meant that no firtool memory was
ever converted, silently, while the inventory said two entries and
nothing failed.

Naming. A memory was named after its module only when the module held
exactly one cell of any kind. firtool's modules carry glue -- the
`'bx` read mux, the mask ANDs -- so every one of them was named `Memory`
instead, and the inventory is keyed by name, so forty memories collapsed into
one. On XiangShan's XSCore: 41 memory modules in the design, two
entries in memories.json, both declined. A firtool memory module -- one
memory, every port in the convention below -- now *is* that memory and
takes the module's name, whatever glue surrounds it. Any other module
keeps the old rule: named for the module only when the memory is its
only cell, else <module>.<cell>, so an inline array inside a module
that does something else still fails idiomatic.judge's own-module
gate.

Pins. The pin list was synthesized from yosys's port counts as
R0_addr, W0_data and so on. The generated liberty view has to carry the
ports of the module the flow blackboxes, or the module never links to
it; and a read-write port, which yosys reports as a read plus a write,
came out as two ports with a per-bit write enable where the module has
one RW0_wmask of a few lanes. When every port of the module fits the
convention, the pins are now the module's own, RW ports are counted as
RW, and the mask width is the port's. `port_convention` in the schema
says which happened; synthesized pins keep the old behaviour.

Combinational reads. $mem_v2 RD_CLK_ENABLE says whether a read port is
clocked. An SRAM macro reads on a clock edge, so a memory the design
reads combinationally -- XiangShan's difftest RAT state banks, 8 x
704x776 -- cannot become one without moving every read a cycle later.
Such a memory is now counted (`comb_read_ports`) and refused with a
reason; flops are the only faithful implementation.

The FakeRAM asap7 backend emits a single read-write port today, so a
memory declared with separate read and write ports under the convention
is refused with a reason rather than handed a macro whose pins do not
match.

Verified on XiangShan's own firtool output: array_64x114 detects as one
RW port, 2 mask lanes, clocked read, idiomatic; the 704x776 difftest
bank as one combinational read port, refused.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
yosys writes RD_CLK_ENABLE as a binary string of exactly RD_PORTS
digits ('01' for a clocked and a combinational read). The helper padded
nothing and truncated a long value, so anything else -- a short string,
a decimal one -- miscounted combinational reads silently and could let
one through as a macro. It now raises, naming the value.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
`memory -nomap` in extract_memories.tcl walks every module of the design
in each of its passes, and memory_dff builds its per-bit driver and
consumer index (ModWalker::setup, memory_dff.cc:225) for a module before
run() asks whether the module has a memory; opt_mem_priority and
opt_mem_feedback scan every module's cells the same way. On a design that
keeps its memories in small generated modules and its logic in large
ones -- a firtool core has every memory in its own ram_* module -- the
index over the memory-less modules is nearly the whole cost of the step.

Measured on XiangShan's flat XSCore (1870 modules, 407 with memory cells,
read through slang, yosys 0.68), the memory passes alone:

    unscoped    721 s   memory_dff 462, opt_mem_priority 67,
                        opt_mem_feedback 52, two opt_clean 99, share 37
    scoped       64 s   memory_dff 24, opt_mem_priority 27

Both produce the same 407 $mem_v2 cells. The extraction step as a whole
went from about fifteen minutes to under four, and the four remaining
minutes are reading and proc.

memory_bmux2rom still runs first and unscoped, since it is what turns a
module's constant muxes into a memory. The module list comes from the
memory cells' names (the module before the first slash) rather than the
selection's `%m` expansion: `%m` leaves a module partially selected when
its name carries a `$`, as slang's uniquified names do, and every pass
then skips the module with "Ignoring partially selected module" and no
memory is collected. write_json -selected writes the memory modules
only; gen_memories.py's detector reads each module's $mem_v2 cells and
nothing across modules.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
read_verilog leaves memory reads as $memrd until memory_collect, so the
_v2-only list found a memory's module through its write or init cells
alone. t:$mem* takes every memory cell type.

The module name is the cell path up to its first slash. An escaped name
holding a slash is cut short, select warns that it matched nothing, and
that module's memories would be skipped without an error. The selection
must now hold every memory cell, or the step stops.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
Every OpenROAD submodule bump recompiled yosys and kepler-formal in the
builder image. They were built in the same RUN as OpenROAD, in an
intermediate stage that inline cache does not export.

Dockerfile.builder gains an orfs-tools target holding only the
tools/install of every tool except OpenROAD, and a toolsImage build arg
to take that install from a prebuilt image. Without the arg it builds
the orfs-tools stage itself, so existing callers are unchanged.

etc/DockerTag.sh -tools hashes the inputs of that image: the dev tag,
the build scripts and the checked-out tools other than OpenROAD and
AutoTuner. CI can pull an image with that tag and skip the build.

build_openroad.sh gains --openroad_only to build OpenROAD without
reconfiguring Yosys, whose Verific sources are gone by then.

Signed-off-by: Matt Liberty <mliberty@precisioninno.com>
The FAx1 and HAxp5 cells in the asap7sc7p5t_SIMPLE libraries have
intentional CON->SN output-to-output timing arcs (mirror-adder topology,
CON is reused internally by the sum stage). Explain this in
liberty_suppressions.tcl so the arcs are not mistakenly removed from the
Liberty files, and note that the suppression also hides STA-1212 for
user-added libraries.

Signed-off-by: dsengupta0628 <dsengupta@precisioninno.com>
The module list was cut out of `select -list` lines at the first slash
and handed back to `select`. An escaped module name may itself hold a
slash, as in `\ram/one`: the cut name matched nothing and the step
stopped. A memory's own name may hold one too, so no cut point is right.

`t:$mem* %m` selects every module that holds a memory cell, whole, and
never goes through a name. Checked with slang on yosys 0.68 against
master's unscoped run: slang's uniquified `ram$top.x.a` and the escaped
`ram/one$top.z` give the same $mem_v2 cells, and no module is left
partially selected. Tested with slang only: yosys's own Verilog
frontend is deprecated.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
uploadMetadata.py used to skip the publish when no design had a
metadata.json, so builds that generated no metrics never reached the QoR
dashboard. It now publishes the pipeline payload with an empty designs
list through the normal path, and a failed publish still exits 1.

Three optional arguments describe the build on the pipeline payload:
--buildTime (build_time, defaulting to now in genMetrics.py's format and
emitted on every pipeline payload), --expectedDesigns (design_count) and
--noMetricsReason (no_metrics_reason, only on an empty payload). v1
fallback messages are unchanged, since the backend reads their unknown
root keys as metrics.

Part of The-OpenROAD-Project-private/lembas#209.

Signed-off-by: Miguel Dalberto Pedro <miguel.pedro@precisioninno.com>
… platform's

The comment and README said the ports are budgeted with
$PLATFORM_DIR/constraints.sdc's ratio. That file has no ratio: it gives
a single-clock macro a fixed 80 ps per port path. The budgets here are
80 % of each domain's period, a choice of this file; say so. The
constraints themselves are unchanged, so the README's results stand.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
noc_tx_{req,rsp,dat}_ready reach noc_tx_*_flitv through one gate
(chi_async_bridge_sink.sv: tx_flitv = valid && tx_ready), and nothing
constrained them: on the small variant's synthesised netlist they were
the only port-to-port paths, required INF. Give each domain an in2out
budget as the platform's constraints.sdc does, at the same 80 % of that
domain's period as its other port paths, and an in2out path group.

Now required 816.8 ps, met by 804 ps; every other path is unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
1021 ps against 300 ps is 0.29 times the frequency, not half. Only the
comment changes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
The synchroniser check parsed DEPTH from VERILOG_TOP_PARAMS but fixed
the stage count at 3, so a flow that set SYNC stopped at synthesis:

  constraint.sdc: synchronisers: sink.asyncQSink_dat_flit.widx_gray.bit_[0] has 4 stages, not 3

Parse SYNC the same way, with the RTL's default. The small variant with
SYNC 2, 3 and 4 now passes the check; with SYNC 4 it stopped before.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
constraint.sdc is read once, by synth_odb.tcl against yosys's netlist;
later stages read what write_sdc made of it, one set_max_delay per hop
by cell name (172 on the small variant). So the check catches what
synthesis does to a synchroniser, not a buffer that resizing or hold
repair puts inside a hop afterwards, which the hop's budget still
bounds. The README now says so.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
The SystemVerilog is rewritten from XiangShan's Chisel rather than
generated from it, but it is still derived from that source and keeps
its licenses. The async queue and synchroniser come from rocket-chip
(Apache-2.0, LICENSE.SiFive); the bridge halves, shadow buffer, reset
synchroniser and top-level wiring from XSCache, Utility and XiangShan
(Mulan PSL v2).

As picorv32 does: LICENSES/ holds both texts verbatim, SOURCE.md names
each file's source, commit and license, and each file starts with its
SPDX identifier, the upstream copyright lines and a note of what
changed. The README points at them. No RTL changes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
OPENROAD_HIERARCHICAL=1, the -hier mode that is to become the default.
A flat netlist gives -hier nothing to link, so SYNTH_KEEP_MODULES keeps
what a clock-crossing bridge is usually kept as: the two clock-domain
halves, their asynchronous queues and their shadow buffers.

The synchroniser check in constraint.sdc now finds the chains with
get_cells -hierarchical; their names are the kept instances joined
with '/', e.g. sink/asyncQSink_dat_flit/widx_gray.bit_[0].sync[1]. On
the small variant all 172 hops are found and written back by name.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
…n does

An integration test: a timed asynchronous crossing, a self-checking
constraint file, slang with SYNTH_KEEP_MODULES under -hier, instance
names resolved through the hierarchy, three threshold voltages on a
multi-clock design, and the top's parameters set from the flow. Checked
against every config.mk and SDC under flow/designs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
… grt

At 300 ps the worst slack reached +0.001 ps at global route (flat
netlist), so repair stopped there. At 290 ps, linked hierarchically,
the worst slack is -35.8 ps at CTS, -35.5 ps at global route and
-30.1 ps at final, on the core half's async queue sink read, as in
XiangShan. README.md's Results say so; the XiangShan calibration there
is marked as measured on the flat netlist.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>
Do not merge. Runs only nangate45/gcd and sky130hd/gcd on a jenkins-ci branch.

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>

@gemini-code-assist gemini-code-assist Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request updates the Jenkinsfile to use the 'utils@orfs-qor-red-stage' library and adds a filter parameter to the pipelineORFS execution. Additionally, it reduces the clock period (clk_period) from 0.25 to 0.15 in the SDC constraints for the 'nangate45/gcd' design. There are no review comments, and I have no feedback to provide.

@openroad-ci

openroad-ci commented Oct 6, 2026 •

Copy link
Copy Markdown
Member

🔍 QoR check

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

Commit b83a02c · Jenkins build #7 · Baseline: build · View build on dashboard

3 design(s) checked — 1 with regression(s), 1 without a comparable baseline.

❌ nangate45/gcd base — 6 failing metric(s)
Metric target base delta limit band
cts__timing__setup__tns -12.2456 -6.95522 -76.06344587230886% -8.346264 PeriodPadding 20%
cts__timing__setup__ws -0.261172 -0.156758 -66.60840276094362% -0.169258 PeriodPadding 5.0%
finish__timing__setup__tns -12.4146 -7.19445 -72.55801346871547% -8.63334 PeriodPadding 20%
finish__timing__setup__ws -0.271991 -0.163541 -66.31364611932176% -0.176041 PeriodPadding 5.0%
globalroute__timing__setup__tns -12.7365 -7.5407 -68.90341745461296% -9.04884 PeriodPadding 20%
globalroute__timing__setup__ws -0.275208 -0.171167 -60.783328562164435% -0.183667 PeriodPadding 5.0%

vvbandeira and others added 5 commits October 6, 2026 07:41
Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
palantir reads build_time as UTC, but the default came from local time, so
a build published from a machine outside UTC was dated off by its offset.
A beta test from a UTC-3 machine showed the build three hours in the past.

Signed-off-by: Miguel Dalberto Pedro <miguel.pedro@precisioninno.com>
Add the CPU time (user + sys) of each step and its ratio to the
elapsed time, a rough measure of thread utilization. The ratio uses
the fractional elapsed seconds so short steps are not skewed, and the
total ratio only counts steps that report CPU time.

To keep the table narrow the column headers are split over two lines,
the Log column fits the longest step name, and the result sha1 is
shortened from 20 to 10 hex characters, plenty to tell a pair of
result files apart.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Matt Liberty <mliberty@precisioninno.com>
Wrap kepler-formal the way the flow wraps yosys and openroad so each
LEC/SEC run gets a ${step}_lec.log / ${step}_sec.log with the tool's
stdout, previously discarded, and the elapsed time, CPU time and peak
memory. kepler-formal still writes its detailed log to
${step}_lec_check.log, which the pass/fail check reads.

genElapsedTime now skips every _lec_check/_sec_check log, not just
4_rsz's, fixing "No elapsed time found in .../6_final_lec_check.log".
The new LEC/SEC rows are left out of the total as they run inside
another step (4_1_cts, 6_report) that already counts their time.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Matt Liberty <mliberty@precisioninno.com>
- Exclude tools/install from the orfs-tools build context so stale
  local install output cannot leak into the image.
- Hash the effective Docker ignore file in the -tools tag.
- Use git submodule status for the yosys/kepler-formal commits so an
  uninitialized submodule does not error.
- Reject --openroad_only without --local or with --skip_openroad.

Signed-off-by: Matt Liberty <mliberty@precisioninno.com>
maliberty and others added 28 commits October 7, 2026 04:31
…t-staging/elapsed-cpu

genElapsedTime: report CPU time; time kepler-formal LEC/SEC
…t-staging/docker-builder-yosys-layer

feat(docker): build yosys into a reusable tools image
Ubuntu 22.04's Ruby 3.0 bundles net-imap 0.1.1 as a default gem. The flow
never uses it, but CVE scanners flag it on downstream images. Remove the
stale files at image build time.

Fixes The-OpenROAD-Project#4376

Signed-off-by: noahwithoutark22 <yash.dubey.ece24@iitbhu.ac.in>
Address review feedback: a separate RUN leaves the gem in the earlier
layer, so remove it in the same RUN as the dependency install.

Signed-off-by: noahwithoutark22 <yash.dubey.ece24@iitbhu.ac.in>
Add REPORT_ENDPOINT_SLACK: when set, every flow step writes the setup and
hold slack of every timing endpoint to
$REPORTS_DIR/<step>_endpoint_slack.json, timed with parasitics chosen by
step as in open.tcl. Timing is updated before reading the endpoint set,
which otherwise misses gated clock enable checks.

flow/util/endpointSlack.py lists, exports, tracks and compares snapshots.
Comparisons report endpoint movement and agreement metrics (Spearman, top
agreement, bias/MAE/RMSE, violator recall/precision) on the critical
endpoints, and write interactive HTML slack scatters; track --html writes
one per step transition plus an index whose rows are grey when nothing
changed and green when everything improved. Endpoints with positive slack
in both steps are not considered, and changes below the displayed
resolution (--tolerance) count as unchanged. Numbers use a fixed-width
font with a fixed number of decimals (--digits).

Signed-off-by: Matt Liberty <mliberty@precisioninno.com>
open.tcl already applies WIRE_LOAD_MODEL; do the same in load_design so
stages that load the design through load.tcl get wire delay on nets
without parasitics too.

Signed-off-by: Matt Liberty <mliberty@precisioninno.com>
The 6_final snapshot was written by orfs_write_db before parasitic
extraction, so it was timed with global routing estimates. Those
overestimate short nets and were about 0.25 ns pessimistic on
sky130hd/riscv32i compared with the extracted SPEF.

Write the 6_final snapshot in final_outputs.tcl after extraction instead,
using the loaded parasitics (rcx, or global_routing on platforms without
RCX_RULES) so it matches 6_finish.rpt. write_endpoint_slack takes an
optional parasitics name to skip its own estimation, and orfs_write_db
can skip the snapshot.

Signed-off-by: Matt Liberty <mliberty@precisioninno.com>
…t-staging/load-wire-load-model

load.tcl: apply WIRE_LOAD_MODEL in load_design
…t-staging/endpoint-slack-tracking

Track endpoint slack across flow steps
…es-firtool-modules

AUTO_MEMORIES: convert only a memory that is its own module; read firtool memory modules
…ories-scoped

memories: run the memory passes on the modules that hold memory cells
The Jenkins jobs use the root Jenkinsfile as their script path. The
symlink only pointed to that file.

The symlink also broke builds of untrusted fork pull requests. For such
a build, Jenkins reads the pipeline script from the base branch through
the git plugin. The git plugin returns the link text "../Jenkinsfile"
and does not follow the link, so the script does not compile.

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
…ile-no-symlink

ci: remove the jenkins/public_tests_all.Jenkinsfile symlink
…x/4376-remove-stale-net-imap

fix(docker): remove unused net-imap default gem
…sync-bridge

asap7/chi_async_bridge: XiangShan's asynchronous CHI bridge, to replace mock-cpu
…t-staging/asap7-document-sta1212-suppression

asap7: document why STA-1212 is suppressed
* commit 'refs/stacktest/qor-check-ci-mode':
  Check QoR as a checkout of the recorded commit

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
* commit 'refs/stacktest/upload-no-metrics-builds':
  Default the reported build time to UTC
  Publish a metrics payload even when no design produced metrics

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
* commit 'refs/stacktest/upload-build-requester':
  Send who requested the build in the metrics payload

Conflicts in flow/util/uploadMetadata.py were additive: the new
argparse arguments and resolve_requester next to resolve_build_time.
Both sides are kept in full.

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
Throwaway test branch. Do not merge.

This branch shows how the jenkins-ci stack reports failed designs in a
real ORFS build. The Jenkinsfile loads utils@upload-build-requester and
runs only nangate45/gcd, sky130hd/gcd and ihp-sg13g2/gcd.

- nangate45/gcd: clk_period goes from 0.25 to 0.15. The flow completes,
  but the QoR check fails.
- sky130hd/gcd: VERILOG_FILES points to missing.v, so the flow fails.
- ihp-sg13g2/gcd: no change. It is the control.

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
Throwaway test branch. Do not merge.

This branch shows how the jenkins-ci stack uploads a build in which no
design writes metrics. The Jenkinsfile loads utils@upload-build-requester
and runs only nangate45/gcd. Its VERILOG_FILES points to missing.v, so
the flow fails before it writes metrics.

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
…rwards

* commit '78a3b6f9fe56cc6bbed85067f421addae5612112':
  test: rerun with the updated jenkins-ci branch
  test: force a QoR regression to check the Jenkins QoR stage

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
* commit 'fc0c63ac05c8085767649376bba074cb24d59091':
  test: run three gcd designs on the jenkins-ci stack, two broken
  test: rerun with the updated jenkins-ci branch
  test: force a QoR regression to check the Jenkins QoR stage

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
This is a throwaway test branch. Do not merge it.
The library branch adds one follow-up PR on top of the stack.

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
* commit 'b18925d142283a7180ef8734ac233689851f7878':
  test: load the jenkins-ci library from orfs-qor-one-stage

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
This is a throwaway test branch. Do not merge it.

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
* commit '2eb6616e7489483843bf027a21079a0f5b60703e':
  test: load the jenkins-ci library from orfs-qor-one-stage
  test: run nangate45/gcd broken so no design writes metrics

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
This is a throwaway test branch. It is not for merge.

Signed-off-by: Vitor Bandeira <vvbandeira@precisioninno.com>
@maliberty

Copy link
Copy Markdown
Member

@vvbandeira close?

@vvbandeira vvbandeira closed this Oct 13, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

7 participants