B. Demoen, P.-L. Nguyen. On the impact of argument passing on the performance of the WAM and B-Prolog. CW report 300, September 2000 http://www.cs.kuleuven.ac.be/publicaties/rapporten/cw/CW300.abs.html Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Instruction Merging and Specialization in the SICStus.. - Nässén, Carlsson.. (Correct)
....of many WAM instructions is mode dependent, whereas others change the current mode. To avoid expensive runtime tests, the mode dependent opcodes can be linked to a read mode routine and a write mode routine. In a bytecode interpreter, this can be done either using two switch statements (as in [4]) or e.g. as a common one where an opcode 2 i refers to a read mode routine and 2 i 1 refers to the write mode version. An indirect threaded interpreter can use a similar technique with even and odd offsets. These techniques are used in SICStus Prolog. In a direct threaded interpreter, mode ....
....instructions get changed by the compiler to first value instructions. As a consequence, the frequency of variable pairs decreases, and this either results in not considering these pairs for merging or to obtaining a significantly larger instruction set after merging. The experience reported in [4] is similar. 3.3 Evaluation To evaluate a particular version of the WAM, we ran our benchmark suite on it and measured the execution time. The experiments were conducted both on SPARC (Solaris) and Pentium (Linux) processors, to reduce any architectural bias. We also considered counting machine ....
[Article contains additional citation context not shown here]
B. Demoen and P.-L. Nguyen. On the impact of argument passing on the performance of the WAM and B-Prolog. Technical Report CW 300, K.U. Leuven, 2000.
Argument Registers Or Local Stack for parameter passing in the.. - Demoen, Nguyen Self-citation (Demoen Nguyen) (Correct)
....equal turned out to be too much work. Still, while investigating these issues, we became more and more aware of other di erences between B Prolog and dProlog (as well as other WAM implementations we know o ) which might be responsible for the perceived speed of B Prolog. Indeed, as shown in [7], closer inspection of the implementation technology (generated abstract machine code, emulator structure . shows that mainly the instruction compression employed in B Prolog, its double emulator (for propagation of read write mode) and its built in calling convention are responsible for its ....
....of the implementation technology (generated abstract machine code, emulator structure . shows that mainly the instruction compression employed in B Prolog, its double emulator (for propagation of read write mode) and its built in calling convention are responsible for its performance. In [7], we have introduced in dProlog a series of instruction compressions partly inspired by B Prolog, partly by doing dynamic pro ling as in [10] and a double emulator, thereby making an ANSI C compliant version of dProlog of roughly the same speed as B Prolog. This version of dProlog is later ....
[Article contains additional citation context not shown here]
B. Demoen, P.-L. Nguyen. On the impact of argument passing on the performance of the WAM and B-Prolog. CW report 300, September 2000 http://www.cs.kuleuven.ac.be/publicaties/rapporten/cw/CW300.abs.html
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