Search

High performance architectures are constantly evolving in order to deliver ever greater compute powers, as well as ever greater energy efficiencies. This applies to multi-core CPUs (with higher core count and wider vector units) as well as to various many-core, possibly heterogeneous, architectures (GPUs, Xeon Phi processors . . . ). Considering performance, power efficiency or performance portability, and relying on new and relevant programming paradigms, we have focused on algorithmic changes allowing to adapt at best specific or key applications in scientific computing to such high performance architectures. Our research work has been structured according to three research directions: (i) designing algorithms for many-core architectures via massive parallelism, for multi-core architectures via task parallelism, or for both via hybrid algorithms; (ii) handling the vector divergence on high performance architectures; and (iii) taking advantage of new heterogeneousarchitectures for scientific applications. We present here our algorithmic contributions, their interdisciplinary context, and the close combination they require between application specificities, algorithmics, programming and architectural features., Les architectures de calcul haute performance évoluent en permanence afin d’offrir des capacités de calcul, et des efficacités énergétiques, toujours plus importantes. Ceci concerne d’une part les CPU multi-cœurs, qui comportent de plus en plus de cœurs et des unités vectorielles toujours plus grandes, et d’autre part les diverses architectures many-core, potentiellement hétérogènes (GPU, processeurs Xeon Phi ...). En prenant en compte la performance, l’efficacité énergétique ou la portabilité des performances, et en nous appuyant sur des paradigmes de programmation récents et appropriés, nous avons apporté des contributions algorithmiques permettant d’adapter au mieux des applications de référence, ou spécifiques, en calcul scientifique à ces architectures haute performance.Nos travaux ont été structurés selon les trois axes de recherche suivants : (i) concevoir des algorithmes pour les architectures many-core via du parallélisme massif, pour les architectures multi-cœurs via du parallélisme de tâches, ou pour les deux via des algorithmes hybrides ; (ii) réduire la divergence au sein des calculs vectoriels sur les architectures haute performance ; et (iii) tirer parti des nouvelles architectures hétérogènes pour des applications encalcul scientifique. Nous présentons ici nos contributions algorithmiques, leur contexte interdisciplinaire, ainsi que les fortes interactions qu’elles requièrent entre les caractéristiques de l’application, l’algorithmique, la programmation et l’architecture matérielle.

We study the effects of the CEGEP innovation of the late 1960s in the province of Quebec on total years of schooling based on a difference-in-difference methodology applied to the microdata from the 20% sample of the 2001 Canadian Census long form. Eight demographic groups are investigated, namely Francophone men and women from Montreal, Quebec City and "Rest-of- Quebec", and Montreal Anglophones of each gender. The average estimated effects are statistically positive and large: 0.33 year for men and 0.41 year for women. Putting these results in context, we further argue that they likely underestimate the true behavioural effects of the reform on educational attainment, particularly in the case of Francophones.

This background paper was prepared for the Bank of Canada conference on the 2021 renewal of the inflation-control agreement between the Bank and the Government of Canada held online in August 2020. The first part of the paper focuses on the fact that the room for conventional monetary stimulus is being limited by the narrow space remaining between the neutral level of the policy interest rate, estimated to be 2.5 percent and expected to remain low for some time, and the effective lower bound on this policy rate, set by the Bank of Canada at 0.25 percent. Two means of getting greater monetary stimulus would be for the Bank to keep on purchasing long-term assets on a large scale, or to increase its inflation target by a couple of percentage points, say from 2 percent to 3 or 4 percent. But the macroeconomic effectiveness of the first option is uncertain, and there would most likely be strong political opposition to increasing the inflation rate to 4 percent, or even only to 3 percent. In the short term, therefore, federal and provincial budgets are the only policy instrument that can make up for the shortcomings of monetary policy - however difficult policy coordination may be - and bring the Canadian economy to fully recover from the current recession without delay. The second part of the paper reviews results that Bank of Canada researchers have obtained in comparing the macroeconomic performance of various monetary policy frameworks with the help of their macro-econometric model of the Canadian economy called ToTEM. I am led to conclude that the 2021 agreement should keep the current flexible inflation targeting framework and continue to have it operated independently by our central bank, but that a somewhat more flexible approach than in the past could be welfare-improving. In particular, it could specify that maximizing employment is a prime concern of the Bank of Canada jointly with keeping inflation low and stable. This dual concern has been the bread and butter of the US Federal Reserve since it was legislated by the Humphrey-Hawkins Act of 1978. It would be beneficial for Canadians if the renewed agreement began to clarify how the two instruments of monetary and fiscal policy will be coordinated to achieve these two interdependent macroeconomic goals of low inflation and maximum employment.