Compiler-Driven Dependence Profiling to Guide Program Parallelization

2010 43rd Annual IEEE/ACM International Symposium on Microarchitecture

Luk

2010

As multicore processors are deployed in mainstream computing, the need for software tools to help parallelize programs is increasing dramatically. Data-dependence profiling is an important technique to exploit parallelism in programs. More specifically, manual or automatic parallelization can use the outcomes of data-dependence profiling to guide where to parallelize in a program.However, state-of-the-art data-dependence profiling techniques are not scalable as they suffer from two major issues when profiling large and long-running applications: (1) runtime overhead and (2) memory overhead. Existing data-dependence profilers are either unable to profile large-scale applications or only report very limited information.In this paper, we propose a scalable approach to datadependence profiling that addresses both runtime and memory overhead in a single framework. Our technique, called SD 3 , reduces the runtime overhead by parallelizing the dependence profiling step itself. To reduce the memory overhead, we compress memory accesses that exhibit stride patterns and compute data dependences directly in a compressed format. We demonstrate that SD 3 reduces the runtime overhead when profiling SPEC 2006 by a factor of 4.1× and 9.7× on eight cores and 32 cores, respectively. For the memory overhead, we successfully profile SPEC 2006 with the reference input, while the previous approaches fail even with the train input. In some cases, we observe more than a 20× improvement in memory consumption and a 16× speedup in profiling time when 32 cores are used.

Section: B Dependence Profiling For Speculative Parallelizationmentioning

confidence: 99%

“…Data-dependence profiling has already been used in parallelization efforts like speculative multithreading [4,7,17,23,29] and finding parallelism [15,22,25,27,31]. It is also being employed in the commercial tools we mentioned.…”

Section: Introductionmentioning

confidence: 98%

SD3: A Scalable Approach to Dynamic Data-Dependence Profiling

2010 43rd Annual IEEE/ACM International Symposium on Microarchitecture

Luk

2010

“…The annotation is currently a manual process. However, this step can be made fully or semi-automatic by several techniques: (1) traditional static analyses from compilers, (2) dynamic dependence analyses [20,21,24,25,27], and (3) an extended critical path analysis [11].…”

Section: Lock_end(int Lock_id)mentioning

confidence: 99%

Predicting Potential Speedup of Serial Code via Lightweight Profiling and Emulations with Memory Performance Model

Kumar

2012 IEEE 26th International Parallel and Distributed Processing Symposium

et al. 2012

We present Parallel Prophet, which projects potential parallel speedup from an annotated serial program before actual parallelization. Programmers want to see how much speedup could be obtained prior to investing time and effort to write parallel code. With Parallel Prophet, programmers simply insert annotations that describe the parallel behavior of the serial program. Parallel Prophet then uses lightweight interval profiling and dynamic emulations to predict potential performance benefit. Parallel Prophet models many realistic features of parallel programs: unbalanced workload, multiple critical sections, nested and recursive parallelism, and specific thread schedulings and paradigms, which are hard to model in previous approaches. Furthermore, Parallel Prophet predicts speedup saturation resulting from memory and caches by monitoring cache hit ratio and bandwidth consumption in a serial program.We achieve very small runtime overhead: approximately a 1.2-10 times slowdown and moderate memory consumption. We demonstrate the effectiveness of Parallel Prophet in eight benchmarks in the OmpSCR and NAS Parallel benchmarks by comparing our predictions with actual parallelized code. Our simple memory model also identifies performance limitations resulting from memory system contention.

2012 IEEE International Symposium on Workload Characterization (IISWC)

“…Therefore, several works have been proposed to uncover the dependencies between different regions. In [6] and [12], they introduced the dependence density metric and the conflict probability to examine specific dependencies will occur, respectively. However, these just estimate probability of expected dependencies and cannot extract precise data dependencies enough to uncover structural parallelism across whole program execution accurately.…”

Section: Introductionmentioning

confidence: 99%

“…So, we can identify these nested loops potentially have loop-parallelism. On the other hand, there are loops that depend on themselves(loop-11, 12, 13, 25) or loops that are dependent on other nodes among the group of nested loops (loop-11,12,13,14,15), (loop-26, 29).…”

mentioning

confidence: 98%

Whole program data dependence profiling to unveil parallel regions in the dynamic execution

Sato

Inoguchi

Nakamura

2012

As application programs become large and complex year by year, mapping these programs into highly parallelized hardware resources by understanding all of the program structures across an application is becoming harder and harder for programmers. In this paper, we present an advanced technique that can extract dynamic memory dataflow across whole program execution to provide programmers with hints for program parallelization and acceleration. Using a precompiled executable binary code as the input, we monitor data dependencies via memory references together with dynamic loop-and call-contexts. We implement our mechanism on a dynamic binary translation system and evaluate it using productive benchmark suites. From these results, we confirm that we can successfully keep track of data dependencies among function calls, loops and their iterations using LCCT+M (LoopCall Context Tree with Memory Dataflow) representation within reasonable time and memory space overheads. We also demonstrate our profiling can help programmers to identify loop-, task-and pipeline-parallelisms in its actual dynamic execution.