TALM: A Hybrid Execution Model with Distributed Speculation Support

Abstract

Parallel programming has become mandatory to fully exploit the potential of modern CPUs. The data-flow model provides a natural way to exploit parallelism. However, traditional data-flow programming is not trivial: specifying dependencies and control using fine-grained tasks (such as instructions) can be complex and present unwanted overheads. To address this issue we have built a coarse-grained data-flow model with speculative execution support to be used on top of widespread architectures, implemented as a hybrid Von Neumanm/data-flow execution system. We argue that speculative execution fits naturally with the data-flow model. Using speculative execution liberates the programmer to consider only the main dependencies, and still allows correct data-flow execution of coarse-grained tasks. Moreover, our speculation mechanism does not demand centralised control, which is a key feature for upcoming manycore systems, where scalability has become an important concern. An initial study on a artificial bank server application suggests that there is a wide range of scenarios where speculation can be very effective.

Exploring the Equivalence between Dynamic Dataflow Model and Gamma - General Abstract Model for Multiset mAnipulation

“…This value will be increment, in each iteration, as an effect of the Inctag operation. The concepts of Steer and Inctag were presented in [5]. Now consider an example of Inctag, the vertex R11.…”

“…Management of the loop iterations are supported by a special node inctag which increases the tag at beginning of next iteration. Function calling might be supported by manipulating tags following the similar solutions applied in loops [5].…”

Optimising loops in dynamic dataflow

“…It comprises a set of identical processing elements (PEs), connected in a network, each one being responsible for interpretation and execution of a group of instructions, according to the dataflow principles. Trebuchet implements a work‐stealing mechanism for dynamic load balancing [10], allows speculative (transactional) execution of tasks that might have memory dependencies [11] and provides error detection/recovery [12].…”

Gamma — General Abstract Model for Multiset mAnipulation and dynamic dataflow model: An equivalence study

“…The Trebuchet 19 is a hybrid von Neumann/dataflow execution system that implements the TALM (an Architecture and Language for Multi‐threading). TALM 20 introduces a user‐defined coarse‐grained parallel data‐flow model, where programmers identify code blocks, called super‐instructions, to be run in parallel and connect them in a data‐flow graph. Using the Trebuchet as a TALM's implementation, the Couillard 21 is a full compiler that creates a dataflow graph based on a super‐instructions expressed in an annotated C‐program.…”

“…Management of the loop iterations are supported by a special node inctag which increases the tag at beginning of next iteration. Function calling might be supported by manipulating tags following the similar solutions applied in loops 20 …”