Using Global Code Motions to Improve the Quality of Results for High-Level Synthesis

Abstract

Abstract-The quality of synthesis results for most high-level synthesis approaches is strongly affected by the choice of control flow (through conditions and loops) in the input description. This leads to a need for high-level and compiler transformations that overcome the effects of programming style on the quality of generated circuits. To address this issue, we have developed a set of speculative code-motion transformations that enable movement of operations through, beyond, and into conditionals with the objective of maximizing performance. We have implemented these code transformations, along with supporting code-motion techniques and variable renaming techniques, in a high-level synthesis research framework called Spark. Spark takes a behavioral description in ANSI-C as input and generates synthesizable register-transfer level VHDL. We present results for experiments on designs derived from three real-life multimedia and image processing applications, namely, the MPEG-1 and -2 and GNU image manipulation program applications. We find that the speculative-code motions lead to reductions between 36% and 59% in the number of states in the finite-state machine (controller complexity) and the cycles on the longest path (performance) compared with the case when only nonspeculative code motions are employed. Also, logic synthesis results show fairly constant critical path lengths (clock period) and a marginal increase in area.

Μεθοδολογίες Αυτοματοποιημένης Εξερεύνησης Και Σύνθεσης Επαναδιατάξιμων Συνεπεξεργαστών Υλικού

“…Η CompInLoop μεθοδολογία εξερεύνησης εφαρμόζεται στο χώρο λύσεων και βασίζεται στην εισαγωγή δομής στο χώρο αναζήτησης, στο σχηματισμό κατάλληλων άνω οριακών συνθηκών καθώς και σε ένα νέο ευριστικό αλγόριθμο αναζήτησης ο οποίος εκμετ-αλλεύεται την εισηγμένη δομή του χώρου λύσεων, ώστε να μειώσει περαιτέρω το χώρο αναζήτησης, όταν διαπιστώνει πως αξιολογεί λύσεις σε τοπικά μη-βέλτιστες περιοχές κορεσμού. Η προτεινόμενη CompInLoop μεθοδολογία εξερεύνησης υλο-ποιήθηκε επεκτείνοντας το εργαλείου αυτοματοποιημένης αρχιτεκτονικής σύνθεσης SPARK [43].…”

“…The input behavior is actually partitioned in code segments called Basic Blocks (BBs) which capture the control flow (CFG), and their DFGs which handle the data dependencies inside each BB. We adopt an execution model similar to [43]. It is based on the sequential execution of the BBs found in the extracted CDFG.…”

“…4.16). The tool-flow integrates: (i) the academic C-to-RTL SPARK HLS framework [43], [110], (ii) several custom design passes (dark grey boxes in Fig. 4 Specifically, we performed the following seven sets of experimental comparisons.…”

“…4.16 was used in order to schedule and map each of the benchmarks' DSP kernel. For both the CRISP-like and the FCC-based datapaths, the SPARK HLS tool [43] with the default optimization passes was used for scheduling the kernels' DFGs. In case of FCC-based datapaths, SPARK was configured to perform intra-template chaining.…”

“…Gradient-based look-ahead pruning trades accuracy/quality of the derived curve for exploration's runtime and vice versa, by characterizing large or small portions of design solutions inside the already bounded space. An exploration framework implementing the proposed methodology has been build extending the SPARK-HLS tool [43].…”