Source author record

Steven

Steven appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

gr-qc math.DS Performance

Catalog footprint

What is connected

3works

3topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2022arXiv

CachePerf: A Unified Cache Miss Classifier via Hybrid Hardware Sampling

The cache plays a key role in determining the performance of applications, no matter for sequential or concurrent programs on homogeneous and heterogeneous architecture. Fixing cache misses requires to understand the origin and the type of cache misses. However, this remains to be an unresolved issue even after decades of research. This paper proposes a unified profiling tool--CachePerf--that could correctly identify different types of cache misses, differentiate allocator-induced issues from those of applications, and exclude minor issues without much performance impact. The core idea behind CachePerf is a hybrid sampling scheme: it employs the PMU-based coarse-grained sampling to select very few susceptible instructions (with frequent cache misses) and then employs the breakpoint-based fine-grained sampling to collect the memory access pattern of these instructions. Based on our evaluation, CachePerf only imposes 14% performance overhead and 19% memory overhead (for applications with large footprints), while identifying the types of cache misses correctly. CachePerf detected 9 previous-unknown bugs. Fixing the reported bugs achieves from 3% to 3788% performance speedup. CachePerf will be an indispensable complementary to existing profilers due to its effectiveness and low overhead.

preprint2020arXiv

The Prime-Power Map

We introduce a modification of Pillai's prime map: the prime-power map. This map fixes $1$, divides its argument by $p$ if it is a prime-power $p^k$, otherwise subtracts from its argument the largest prime-power not exceeding it. We study the iteration of this map over the positive integers, developing, firstly, results parallel to those known for the prime map. Subsequently, we compare its dynamical properties to those of a more manageable variant of the map under which any orbit admits an explicit description. Finally, we present some experimental observations, based on which we conjecture that almost every orbit of the prime-power map contains no prime-power.

preprint2016arXiv

Complete Affine Connection in the Causal Boundary: Static, Spherically Symmetric Spacetimes

The boundary at $\Cal I^+$, future null infinity, for a standard static, spherically symmetric spactime is examined for possible linear connections. Two independent methods are employed, one for treating $\Cal I^+$ as the future causal boundary, and one for treating it as a conformal boundary (the latter is subsumed in the former, which is of greater generality). Both methods provide the same result: a constellation of various possible connections, depending on an arbitrary choice of a certain function, a sort of gauge freedom in obtaining a natural connection on $\Cal I^+$; choosing that function to be constant (for instance) results in a complete connection. Treating $\Cal I^+$ as part of the future causal boundary, the method is to impute affine connections on null hypersurfaces going out to $\Cal I^+$, in terms of a transverse vector field on each null hypersurface (there is much gauge freedom on choice of the transverse vector fields). Treating $\Cal I^+$ as part of a conformal boundary, the method is to make a choice of conformal factor that makes the boundary totally geodesic in the enveloping manifold (there is much gauge freedom in choice of that conformal factor). Similar examination is made of other boundaries, such as timelike infinity and timelike and spacelike singularities. These are much simpler, as they admit a unique connection from a similar limiting process (i.e., no gauge freedom); and that connection is complete.