Citation Context

...paration (such as recompiling or relinking) is needed. Also, it gives 100% instrumentation coverage of user-mode code, without requiring source code. Several generic DBI frameworks exist, such as Pin =-=[11]-=-, DynamoRIO [3], and Valgrind [18, 15]. They provide a base system that can instrument and run code, plus an environment for writing tools that plug into the base system. The performance of DBI framew...

Citation Context

...se the Microarchitecture-Independent Workload Characterization(MICA)frameworkdevelopedbyHosteandEeckhout [11] to evaluate the application diversity of the Rodinia benchmark suite. MICA provides a Pin =-=[19]-=- toolkit to collect metrics such as instruction mix, instruction-level parallelism, register traffic, working set, data-stream size and branchpredictability. Each metric also includes several sub-metr...

Citation Context

...pability may also be affected by the larger execution perturbation from AVIO-S. 5. Evaluation Methodology Our software implementation, AVIO-S, is implemented using the PIN binary instrumentation tool =-=[21]-=- and runs on a real machine with four Intel Pentium processors. Our hardware implementation, AVIO-H, is implemented on the Simics [22] whole system simulator, based on the SimFlex timing model [15] be...

Citation Context

...ogy node. We developed a manycore simulation infrastructure where a timing simulator and a functional simulator are decoupled, as in GEMS [28]. We modified a user-level thread library [34] in the Pin =-=[26]-=- binary instrumentation tool to support more pthread APIs, and use it as a functional simulator to run applications. In-order cores, caches, directories, on-chip networks, and memory channels are mode...

Citation Context

... this section we will describe how the BugNet replayer works and our experiences in building the replayer system using Simics. 5.1 Replaying Single Thread We used the Pin Dynamic Instrumentation tool =-=[12]-=- running on Fedora Linux OS to collect the FLLs that were described in Section 4.2 for the programs listed in Table 1. Multiple FLLs were collected per thread leading up to the program crash. The goal...

Citation Context

...on into the appropriate IR statement via our own translation language. Instruction traces can be efficiently generated for most modern architectures including x86 via hardware [5, 49] or via software =-=[1, 37, 42]-=-. An instruction trace contains the instruction address and optionally the value of the operands for each instruction executed. Although the number of instructions executed may be large, the correspon...

Citation Context

...ms. However, it was mainly designed for other applications such as slicing and does not provide some of the capabilities that we need. Dynamic Binary Analysis Platforms. Tools like DynamoRIO [3], Pin =-=[27]-=-, and Valgrind [41] support fine-grained instrumentation of a user-level program. They all provide a well-defined interface for users to implement plugins. However, as they can only instrument a singl...

Citation Context

... configuration, DYTAN instruments the x86 executable on the fly to produce an instrumented executable. To perform instrumentation on the fly, DYTAN leverages the PIN dynamic instrumentation framework =-=[12]-=-, which is well supported and offers a rich APIs for manipulating x86 binaries. Performing controlflow based taint propagation requires DYTAN to compute CFGs and pdom information at code loading time....

Citation Context

...n common between the two executions. Note that this number correctly accounts for extra instructions spent in synchronization, as they will form some communicating pair. We developed a tool using PIN =-=[14]-=- that collects the communication sets for multithreaded program executions. The tool takes a single sample at a specified point in the execution. We cover the whole application by running the tool twi...

Citation Context

...sult in a larger scheduletrace search space. 8.3 TracingBranches qi-dri requires a branch-trace, as discussed in Section 7. ODR captures branches in software using the Pin binary instrumentation tool =-=[15]-=-. Software binary translation incurs some overhead, but it is faster or more flexible than the alternatives (e.g., LibVEX or x86 hardware branch tracing [11]). To obtain low logging overhead, we emplo...