| Urs H olzle and David Ungar. Do object-oriented languages need special hardware support? In Nierstras [Nierstras1995]. |
.... and a small number of megamorphic call sites from which methods were invoked on many different kinds of object [Hlzle94a] Hlzle and Ungar argue that the implementation of self invalidates earlier claims that object oriented programming languages required special hardware or microcode support [Hlzle95] They claim that the optimizations performed at run time in the self vm enable the generation of code which is comparable with that generated from the C programming language. For example, concrete type analysis is effective at identifying where native arithmetic operations or formats may be used ....
Urs Hlzle and David Ungar. Do Object-Oriented Languages Need Special Hardware Support? In Walter G. Olthoff, editor, European Conference on Object-Oriented Programming (ECOOP '95), volume 952 of Lecture Notes in Computer Science, pages 283--302. Springer-Verlag, August 1995. (p 16) 184
....counters on the UltraSPARC II processor. Factors including CPI, instruction and data cache misses, processor stalls due to instruction cache misses and branch misprediction, from real execution of several programs were measured and presented [59, 60] In another study, Holzle and Ungar [33] look at the trade offs between hardware and compiler optimizations for several large SELF applications. The need for specialized hardware for eliminating message dispatch overhead and tagged arithmetic in the SPARC architecture was seen to be less, when compiler optimizations were used. The ....
Urs Holzle and David Ungar. Do object-oriented languages need special hardware support? In ECOOP-95 Conference Proceedings, August 1995.
....totalling over 90,000 lines of code (Table 13) In general, we tried to obtain large, realistic applications rather than small, artificial benchmarks. Two of the benchmarks (deltablue and richards) are much smaller than the others; they are included for comparison with earlier studies (e.g. [72][62] Richards is the only synthetic benchmark in our suite (i.e. the program was never used to solve any real problem) We did not yet test any programs for which only the executables were available. For every program except porky 2 we also tested an all virtual version (indicated by av ....
....still is a lot of time. The introduction of inline caching [35,126] dramatically diminished this overhead. Polymorphic inline caches (PICs) extend the technique to cache multiple targets per call site. For SELF 93 which uses inline caching and polymorphic inline caching [68] Hlzle and Ungar [72] report an average dispatch overhead of 10 15 on a scalar SPARCstation 2 processor, almost half of which (6.4 ) is for inlined tag tests implementing generic integer arithmetic. This figure also includes other inlined type tests, not just dispatched calls. Given the large differences in ....
Urs Hlzle and David Ungar. Do Object-Oriented Languages Need Special Hardware Support? ECOOP `95 Conference Proceedings, rhus, Denmark, August 1995.
....in a high level language can provide valuable opportunities for improving run time performance. Examples of this are pervasive. For instance, information gleaned from type analysis allows natural unboxed representations of primitive quantities like integers, even in the presence of polymorphism [Mor95, HU95]. The precision of pointer aliasing analysis in an imperative language can be improved by considering the types of the pointers involved [App98] Run time feedback can help reduce the overhead of virtual method lookup in an object oriented language [Hol94] This article describes a further ....
Urs Holzle and David Ungar. Do object-oriented languages need special hardware support? In Walter G. Olthoff, editor, ECOOP'95---ObjectOriented Programming, 9th European Conference, volume 952 of Lecture Notes in Computer Science, pages 283--302, Aarhus, Denmark, 7-- 11 August 1995. Springer.
....[CG94b] neither intra nor interprocedural) Using CPA in SmallEioeel would allow us to eliminate even more dispatch call sites, thus further increasing the speed of the generated code. Further studies are needed to precisely estimate the resulting speedup and extra cost. H#lzle and Ungar [HU95] raise an interesting question: whether object oriented languages need special hardware. They conclude that dispatch tests cannot easily be improved with special purpose hardware, and that the most promising way to reduce dispatch overhead is via compiler optimizations. As an exemple, ....
Urs H#lzle and David Ungar. Do ObjectOriented Languages Need Special Hardware Support ? In Proceedings of the 9th European Conference on Object-Oriented Programming (ECOOP'95), volume 952 of Lecture Notes in Computer Sciences, pages 283302. SpringerVerlag, 1995.
No context found.
Urs H olzle and David Ungar. Do object-oriented languages need special hardware support? In Nierstras [Nierstras1995].
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC