Author: JT Smith
Brian writes: “There’s a new article up at Ace’s Hardware which takes a look at the floating-point
performance of the Athlon, Pentium 4, and Pentium III, as influenced by various compilers.
In particular, the article compares the performance of MSVC 5.0, 6.0, and 7.0 to each other
as well as to Intel C/C++ 5.0.1. As you may know, Intel’s latest compiler can automatically
vectorize loops with SSE-2 when the /QaxW flag is used. This results in a dramatic increase
in performance in some instances, as is visible in the benchmarks reported in the article.
The question remains as to how applicable this functionality is in real-life, especially
considering the unoptimized code-path produced very unimpressive results on non-SSE-2
microprocessors. However, it may offer some interesting possibilities to developers. Floating-Point Compiler Performance Analysis“
performance of the Athlon, Pentium 4, and Pentium III, as influenced by various compilers.
In particular, the article compares the performance of MSVC 5.0, 6.0, and 7.0 to each other
as well as to Intel C/C++ 5.0.1. As you may know, Intel’s latest compiler can automatically
vectorize loops with SSE-2 when the /QaxW flag is used. This results in a dramatic increase
in performance in some instances, as is visible in the benchmarks reported in the article.
The question remains as to how applicable this functionality is in real-life, especially
considering the unoptimized code-path produced very unimpressive results on non-SSE-2
microprocessors. However, it may offer some interesting possibilities to developers. Floating-Point Compiler Performance Analysis“
