Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!uunet!pilchuck!dataio!bright
From: bright@Data-IO.COM (Walter Bright)
Newsgroups: comp.lang.c++
Subject: Re: inline and vectorization
Keywords: C++, vector syntax, vectorization.
Message-ID: <2239@dataio.Data-IO.COM>
Date: 8 Dec 89 19:39:48 GMT
References: <sZTKVau00VoLMDkUw5@andrew.cmu.edu> <40827@lll-winken.LLNL.GOV>
Reply-To: bright@dataio.Data-IO.COM (Walter Bright)
Organization: Data I/O Corporation; Redmond, WA
Lines: 26

In article <40827@lll-winken.LLNL.GOV> jac@muslix.UUCP (James Crotinger) writes:
<Suppose I have a vector type and the following code:
<   vector A, B, C, D, E
<   E = A + B*C     // meaning elementwise multiplication
<   D = A - B*C 
<will the compilers be able to do "loop jamming" on the loops that 
<are implied by the vector syntax.

No compiler I know of can do this (in any language) if the loops are
inserted by inline functions.

The types of optimizations done are driven by the type of code that
programmers tend to write. If a language encourages things to be written
a certain way, but the optimizer generates lousy code for that, then
there is pressure to change the optimizer. I've made a number of
modifications to the optimizer (for ZTC) to recognize certain things
that are common in C++, but very rare in ordinary C.

I don't see a serious technical obstacle to an optimizer doing the loop
jamming you described. The optimization is not done currently because
in C people just don't write code that such an optimization would improve.
C++ changes the way code is written, however, so there will be pressure
on C++ vendors to do those optimizations.

Of course, that doesn't solve your problem today. Perhaps 2-3 years from
now :-).