Inter-thread Data Dependences Research Articles

Exploring speculative procedure and loop level parallelism in SPLASH2

How to make use of multicore computing resources to accelerate high performance computing HPC applications has become a common concern problem. However, HPC applications have not yet been explored in thread level speculation TLS thoroughly, especially in the procedure level. This paper proposes a procedure and loop level speculation architecture model for speeding up HPC applications, including its speculative mechanism, analysis method, etc. It also takes several applications from SPLASH2 to analyse their speculative parallel potential together with performance impacting factors. The experimental results show that: 1 the best Barnes application can get a 90.9× speedup in loop level speculation while Lu application can get 40.2× speedup in procedure level speculation; 2 limited parallelism coverage and severe inter-thread data dependence violations badly affect both loop and procedure level speculative parallelism in some HPC applications; 3 It is found that although loop structure is the main source of speculative parallelism, procedure structure can be treated as its important supplement. Especially in applications that their 'hot' iteration body concludes multiple procedure calls, higher speculative procedure level speedup can be achieved than that in loop level speculation.

Disjoint out-of-order execution processor

High-performance superscalar architectures used to exploit instruction level parallelism in single-thread applications have become too complex and power hungry for the multicore processors era. We propose a new architecture that uses multiple small latency-tolerant out-of-order cores to improve single-thread performance. Improving single-thread performance with multiple small out-of-order cores allows designers to place more of these cores on the same die. Consequently, emerging highly parallel applications can take full advantage of the multicore parallel hardware without sacrificing performance of inherently serial and hard to parallelize applications. Our architecture combines speculative multithreading (SpMT) with checkpoint recovery and continual flow pipeline architectures. It splits single-thread program execution into disjoint control and data threads that execute concurrently on multiple cooperating small and latency-tolerant out-of-order cores. Hence we call this style of execution Disjoint Out-of-Order Execution (DOE). DOE uses latency tolerance to overcome performance issues of SpMT caused by interthread data dependences. To evaluate this architecture, we have developed a microarchitecture performance model of DOE based on PTLSim, a simulation infrastructure of the x86 instruction set architecture. We evaluate the potential performance of DOE processor architecture using a simple heuristic to fork control independent threads in hardware at the target addresses of future procedure return instructions. Using applications from SpecInt 2000, we study DOE under ideal as well as realistic architectural constraints. We discuss the performance impact of key DOE architecture and application variables such as number of cores, interthread data dependences, intercore data communication delay, buffers capacity, and branch mispredictions. Without any DOE specific compiler optimizations, our results show that DOE outperforms conventional SpMT architectures by 15%, on average. We also show that DOE with four small cores can perform on average equally well to a large superscalar core, consuming about the same power. Most importantly, DOE improves throughput performance by a significant amount over a large superscalar core, up to 2.5 times, when running multitasking applications.

Selective restart of threads for efficient thread-level speculation on multicore architecture

An efficient recovery method for thread-level speculation (TLS) is proposed. The method tracks the inter-thread data dependence as a method for identifying those threads that are obviously unaffected by a data dependence violation. The method is simple to implement. Still, the simulation results using benchmark applications show that the method can significantly reduce the number of unnecessary thread restarts and consequently improve the performance of TLS. Specifically, when compared with the baseline TLS, TLS with the proposed method is 2.3 times faster for IS, 1.7 times faster for equake, and 3.5 times faster for mcf with the use of 64 cores. With the method, the performance of TLS increases steadily up to 64 cores for IS, equake, and mcf, while the speedup of the baseline TLS starts to saturate at 8 or 16 cores.

ParaLog

Instruction-grain lifeguards monitor the events of a running application at the level of individual instructions in order to identify and help mitigate application bugs and security exploits. Because such lifeguards impose a 10-100X slowdown on existing platforms, previous studies have proposed hardware designs to accelerate lifeguard processing. However, these accelerators are either tailored to a specific class of lifeguards or suitable only for monitoring singlethreaded programs. We present ParaLog, the first design of a system enabling fast online parallel monitoring of multithreaded parallel applications. ParaLog supports a broad class of software-defined lifeguards. We show how three existing accelerators can be enhanced to support online multithreaded monitoring, dramatically reducing lifeguard overheads. We identify and solve several challenges in monitoring parallel applications and/or parallelizing these accelerators, including (i) enforcing inter-thread data dependences, (ii) dealing with inter-thread effects that are not reflected in coherence traffic, (iii) dealing with unmonitored operating system activity, and (iv) ensuring lifeguards can access shared metadata with negligible synchronization overheads. We present our system design for both Sequentially Consistent and Total Store Ordering processors. We implement and evaluate our design on a 16 core simulated CMP, using benchmarks from SPLASH-2 and PARSEC and two lifeguards: a data-flow tracking lifeguard and a memory-access checker lifeguard. Our results show that (i) our parallel accelerators improve performance by 2-9X and 1.13-3.4X for our two lifeguards, respectively, (ii) we are 5-126X faster than the time-slicing approach required by existing techniques, and (iii) our average overheads for applications with eight threads are 51% and 28% for the two lifeguards, respectively.

ParaLog

Instruction-grain lifeguards monitor the events of a running application at the level of individual instructions in order to identify and help mitigate application bugs and security exploits. Because such lifeguards impose a 10-100X slowdown on existing platforms, previous studies have proposed hardware designs to accelerate lifeguard processing. However, these accelerators are either tailored to a specific class of lifeguards or suitable only for monitoring singlethreaded programs. We present ParaLog, the first design of a system enabling fast online parallel monitoring of multithreaded parallel applications. ParaLog supports a broad class of software-defined lifeguards. We show how three existing accelerators can be enhanced to support online multithreaded monitoring, dramatically reducing lifeguard overheads. We identify and solve several challenges in monitoring parallel applications and/or parallelizing these accelerators, including (i) enforcing inter-thread data dependences, (ii) dealing with inter-thread effects that are not reflected in coherence traffic, (iii) dealing with unmonitored operating system activity, and (iv) ensuring lifeguards can access shared metadata with negligible synchronization overheads. We present our system design for both Sequentially Consistent and Total Store Ordering processors. We implement and evaluate our design on a 16 core simulated CMP, using benchmarks from SPLASH-2 and PARSEC and two lifeguards: a data-flow tracking lifeguard and a memory-access checker lifeguard. Our results show that (i) our parallel accelerators improve performance by 2-9X and 1.13-3.4X for our two lifeguards, respectively, (ii) we are 5-126X faster than the time-slicing approach required by existing techniques, and (iii) our average overheads for applications with eight threads are 51% and 28% for the two lifeguards, respectively.

Mitosis compiler

Speculative parallelization can provide significant sources of additional thread-level parallelism, especially for irregular applications that are hard to parallelize by conventional approaches. In this paper, we present the Mitosis compiler, which partitions applications into speculative threads, with special emphasis on applications for which conventional parallelizing approaches fail.The management of inter-thread data dependences is crucial for the performance of the system. The Mitosis framework uses a pure software approach to predict/compute the thread's input values. This software approach is based on the use of pre-computation slices (p-slices), which are built by the Mitosis compiler and added at the beginning of the speculative thread. P-slices must compute thread input values accurately but they do not need to guarantee correctness, since the underlying architecture can detect and recover from misspeculations. This allows the compiler to use aggressive/unsafe optimizations to significantly reduce their overhead. The most important optimizations included in the Mitosis compiler and presented in this paper are branch pruning, memory and register dependence speculation, and early thread squashing.Performance evaluation of Mitosis compiler/architecture shows an average speedup of 2.2.