Self-Adapting Linear Algebra Algorithms and Software

One of the main obstacles to the efficient solution of scientific problems is the problem of tuning software, both to the available architecture and to the user problem at hand. We describe approaches for obtaining tuned high-performance kernels and for automatically choosing suitable algorithms. Specifically, we describe the generation of dense and sparse Basic Linear Algebra Subprograms (BLAS) kernels, and the selection of linear solver algorithms. However, the ideas presented here extend beyond these areas, which can be considered proof of concept.     

A theoretical study of light emission from nanoscale silicon

In comparing our calculated exciton energies with those obtained from pseudopotential calculations (Ref. 27) and from a previous tight binding calculation (Ref. 30), we stated that the differences between the three semi-empirical calculations arise because of different treatment of the nanocrystal surfaces. This appears not to be correct. Subsequent calculations with variable Si-H parameters have shown that the band gap is actually rather insensitive to the actual value of these. Instead, the important feature appears to be the overall quality of the bulk band structure parameterization. References 27 and 30 use more extensive and higher quality empirical parameterizations for bulk Si than the sp³s∗ model used by us. Repeating our time dependent calculations with an improved sp³d⁵ parameterization results in similar values to those of Refs. 27 and 30 for the exciton energies.¹ The agreement of the sp³s∗ values with experimental photoluminescence energies (Fig. 7) cannot, therefore, be regarded as well understood at this time.^1,2     

A study of the methylolation of phenol using gel permeation chromatography

Gel permeation chromatography (GPC) was used to investigate the course of the sodium hydroxide catalysed methylolation of phenol in dilute aqueous solution. At low concentrations, the initial rate of methylolation was directly proportional to the concentrations of sodium hydroxide and of formaldehyde. By computer integration of the rate equations for the methylolation of phenol, of 2-methylol phenol and of 4-methylol phenol, the experimental rate curves were simulated and optimal values for the rate constants of all the methylolation steps were obtained.     

Achieving accurate and context‐sensitive timing for code optimization

Key computational kernels must run near their peak efficiency for most high‐performance computing (HPC) applications. Getting this level of efficiency has always required extensive tuning of the kernel on a particular platform of interest. The success or failure of an optimization is usually measured by invoking a timer. Understanding how to build reliable and context‐sensitive timers is one of the most neglected areas in HPC, and this results in a host of HPC software that looks good when reported in the papers, but delivers only a fraction of the reported performance when used by actual HPC applications. In this paper, we motivate the importance of timer design and then discuss the techniques and methodologies we have developed in order to accurately time HPC kernel routines for our well‐known empirical tuning framework, ATLAS. Copyright © 2008 John Wiley & Sons, Ltd.     

Minimizing development and maintenance costs in supporting persistently optimized BLAS

The Basic Linear Algebra Subprograms (BLAS) define one of the most heavily used performance‐critical APIs in scientific computing today. It has long been understood that the most important of these routines, the dense Level 3 BLAS, may be written efficiently given a highly optimized general matrix multiply routine. In this paper, however, we show that an even larger set of operations can be efficiently maintained using a much simpler matrix multiply kernel. Indeed, this is how our own project, ATLAS (which provides one of the most widely used BLAS implementations in use today), supports a large variety of performance‐critical routines. Copyright © 2004 John Wiley & Sons, Ltd.     

Oral ondansetron for the control of cisplatin-induced delayed emesis: a large, multicenter, double-blind, randomized comparative trial of ondansetron versus placebo

PURPOSE: To investigate the efficacy and safety of oral ondansetron in the control of cisplatin-induced delayed emesis in patients who do not require rescue antiemetic therapy for acute emesis. PATIENTS AND METHODS: Five hundred thirty-eight chemotherapy-naive patients who received cisplatin chemotherapy (> or = 70 mg/m2), and who were not rescued for acute emesis, were eligible to be randomized to receive one of the three oral regimens to control delayed emesis. Group I received placebo on days 2 to 6; group II received ondansetron 8 mg twice daily on days 2 and 3 and placebo on days 4 to 6; group III received ondansetron 8 mg twice daily on days 2 to 6. All patients received intravenous ondansetron (0.15 mg/kg every 4 hours for three doses) for the control of acute emesis on day 1. The number of emetic episodes on days 2 and 3 combined (days 2/3, when incidence and severity of delayed emesis were expected to be greatest) was considered the primary measure of efficacy. RESULTS: Patients who received odansetron had significantly fewer emetic episodes on days 2/3, 4, and 5 than those who received placebo (P < or = .002 on each day). Additionally, significantly more patients who received ondansetron had a complete plus major response (C+MR; < or = two two emetic episodes) than those who received placebo on days 2/3 (56% v 37%, P = .001), 4 (94% v 85%, P = .005), and 5 (98% v 88%, P = .006). Patients who received ondansetron had significantly less nausea on day 2/3 when day-1 nausea was used as the baseline score (P = .025). Patients who received ondansetron also had significantly less nausea on day 4 (P = .042) and the results approached significance on day 5 (P = .066). CONCLUSION: Oral ondansetron had a significant effect in the control of cisplatin-induced delayed emesis and nausea in patients who had not required rescue antiemetics during the acute emesis period. The control of delayed nausea and vomiting was most notable in the immediate 2 days following cisplatin administration, with the clinical difference narrowing between the two treatment arms on subsequent days.     

Breathing of awake goats during prolonged dysfunction of caudal M ventrolateral medullary neurons

Recent scientific studies have demonstrated the efficacy of various forms of immunotherapy for the treatment of allergic diseases. Traditional subcutaneous immunotherapy, sublingual, oral, and intranasal immunotherapy have been shown to significantly reduce symptoms and favorably modulate the immune response. Outcome studies that use patient response data from standardized surveys represent the next challenge to all practicing allergists.     

Blocked signal transduction to the ERK and JNK protein kinases in anergic CD4+ T cells

T cells activated by antigen receptor stimulation in the absence of accessory cell-derived costimulatory signals lose the capacity to synthesize the growth factor interleukin-2 (IL-2), a state called clonal anergy. An analysis of CD3- and CD28-induced signal transduction revealed reduced ERK and JNK enzyme activities in murine anergic T cells. The amounts of ERK and JNK proteins were unchanged, and the kinases could be fully activated in the presence of phorbol 12-myristate 13-acetate. Dephosphorylation of the calcineurin substrate NFATp (preexisting nuclear factor of activated T cells) also remained inducible. These results suggest that a specific block in the activation of ERK and JNK contributes to defective IL-2 production in clonal anergy.     

Scattering from finite by infinite arrays of slots in a thinconducting wedge

The radar scattering from a finite by infinite array of slots cut into a thin conducting wedge is considered. The wedge is formed by taking a thin ground plane and applying a bend to create a sharp edge which is parallel to the columns of slots in the infinite axis. Results are derived for thin linear slots whose major axes are either parallel or perpendicular to the edge. A hybrid moment method and geometrical theory of diffraction approach is used, with magnetic current expansion functions defined using Floquet's theorem on single columns of slots. Predictions generally agree with scattering measurements of finite by finite array physical models with monostatic patterns taken in a plane orthogonal to the sharp edge     

Modulation of C3b hemolytic activity by a plasma protein distinct from C3b inactivator

A human plasma protein binds to cell-bound C3b, the major cleavage product of the third component of complement. Consequent upon this binding, C3b no longer functions in either the classical or alternative pathways. This C3b inhibitory activity is a property of a protein previously designated beta 1H on the basis of its electrophoretic mobility.