Solving hard instances of floorplacement

Abstract

Physical Design of modern systems on chip is extremely challenging. Such digital integrated circuits often contain tens of millions of logic gates, intellectual property blocks, embedded memories and custom RTL blocks. At current and future technology nodes, their power and performance are impacted, more than ever, by the placement of their modules. However, our experiments show that traditional techniques for placement and floorplanning, and existing academic tools cannot reliably solve the placement task.To study this problem, we identify particularly difficult industrial instances and reproduce the failures of existing tools by modifying public-domain netlists. Furthermore, we propose algorithms that facilitate floorplacement of these difficult instances. Empirically, our techniques consistently produced legal placements, and on instances where comparison is possible, reduced wirelength by 3.5% over Capo 9.4 and 14.5% over PATOMA 1.0 -the pre-existing tools that most frequently produced legal placements in our experiments.

Capo: Congestion-Driven Placement for Standard-cell and RTL Netlists with Incremental Capability

“…For example, a partitioner may misapproximate the area required by a set of macros and incorrectly allocate whitespace. To address these issues, we have integrated into Capo 10 the SCAMPI (SCalable Advanced Macro Placement Improvements) work [31]. The top-down partitioning flow is modified to selectively place large macros, while smaller macros are clustered into soft modules that will be placed later.…”

“…One case that is not considered by either the original floorplacement techniques [34] or those introduced in SCAMPI [31] is where there are an extreme number of movable modules and an extreme ratio between the largest and smallest macro. An example of this is the newblue1 benchmark from the ISPD'06 placement contest suite.…”

“…The main considerations include robustness, the handling of movable macros and fixed obstacles, as well as consistent routability of placements and the handling of density constraints. Based on recent empirical evidence [31,35,38], the top-down framework appears a somewhat better choice. Indeed the 2 out of 9 contestants in the ISPD 2006 Competition that satisfied density constraints were top-down placers.…”

“…With the addition of macros, the flow of top-down placement becomes more complex. We adopt the technique of "floorplacement" which proceeds as traditional placement until a bin satisfies criteria for block-packing [31,34]. If the criteria suggest that the bin should be packed rather than partitioned, a fixed-outline floorplanning instance is induced from the bin where macros are treated as hard blocks and standard cells are clustered into soft blocks.…”

“…Mixed-size Benchmarks. To show Capo's performance on difficult mixed-size placement instances, we show results on difficult floorplanning instances identified by the authors of [31]. Comparisons of Capo with other tools on two difficult benchmark suites are shown in Tables 5 and 6.…”