Repair a short path with the least allowed noninverting delay-cell area.
Hardware context: Two-register shift path with delayed capture clock. Repair minimum delay within a cell-area budget.
Five clock buffers delay the capture edge of a two-register shift path. Its minimum delay fails, while its maximum path has room for a small repair.
Choose one supplied noninverting cell or leave the path unchanged. Meet both timing floors with at most 4 µm² of added logical cell area. The clock tree, 1 ns period and register behavior remain fixed.
Requirements
| Requirement | Limit | Unit |
|---|---|---|
| Maximum internal slack | ≥ 0.75 | ns |
| Minimum internal slack | ≥ 0.02 | ns |
Budget: Added logical cell area ≤ 4 µm². Smaller is better after every requirement passes.
Live analysis: TT ideal-wire OpenSTA with the authored clock buffer chain propagated. Only the internal register-to-register path is assessed; external I/O timing and other corners remain open.
Full specification
| JSON field | Allowed values | Meaning |
|---|---|---|
| delay | "none", "clkbuf_1", "dlygate4sd1_1", "dlygate4sd3_1" | One noninverting data-path buffer, or no insertion. |
TT ideal-wire OpenSTA with the authored clock buffer chain propagated. Only the internal register-to-register path is assessed; external I/O timing and other corners remain open.
Numeric checks use inclusive bounds. A missing measurement always fails. Only the documented fields are accepted.
Examples
Example 1
Unrepaired- Input
{ "delay": "none" }- Output
Maximum internal slack: 0.9734 ns Minimum internal slack: -0.0399 ns
Delay can help minimum timing while consuming maximum-delay margin and area.
Example 2
Large Delay Cell- Input
{ "delay": "dlygate4sd3_1" }- Output
Maximum internal slack: 0.4336 ns Minimum internal slack: 0.4835 ns
Delay can help minimum timing while consuming maximum-delay margin and area.
Example 3
Smaller Delay Cell- Input
{ "delay": "dlygate4sd1_1" }- Output
Maximum internal slack: 0.7623 ns Minimum internal slack: 0.1364 ns
Delay can help minimum timing while consuming maximum-delay margin and area.
Submission Rules
- Submit one JSON object with every documented configuration field. The supplied circuit and external environment stay fixed.
- Run measures your configuration. Submit applies the same requirements and records the result.
- The stated stage and cell corner define this exercise; passing does not establish full chip sign-off.
Hints
Hint 1
A buffer used on the clock network can also be a noninverting delay element in this bounded experiment.
Hint 2
Compare the hold margin with the remaining setup margin.
Hint 3
A candidate that passes timing must still meet the added-area budget.
What is checked: TT ideal-wire OpenSTA with the authored clock buffer chain propagated. Only the internal register-to-register path is assessed; external I/O timing and other corners remain open.
Run and Submit use the same measured requirements. Run is a preview; Submit records an attempt. The fixed circuit and allowed settings use the pinned OpenROAD/OpenSTA engine.
More problems on this topic