This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [RFC] Enable SSE2 math intrinsics for fpmath=sse and fast-math

On Wed, 8 Jun 2005, Jan Hubicka wrote:

> Hmm, for math libraries we probably won't need much future changes
> ineed.  I am thinking about similar mechanizm for the string functions
> (ie glibc string functions are pretty poor on i386 for quite a while and
> calling them with stack passing conventions is quite a bottleneck).  If
> we get into busyness of having our own library pieces of string
> functions I guess we can handle it independently on this issue anyway...

Before creating yet more versions of string functions I think it would be 
desirable to work on an infrastructure for a string function testsuite 
that can be shared between GCC and glibc.  glibc has thorough testing of 
its string functions; this thoroughness should apply also to testing of 
all GCC's existing built-in string functions (both machine-independent and 
machine-dependent optimizations) and to any new string function versions.  
There is a lot of scope for problems with e.g. reading a few bytes more 
than allowed which may only show up at page boundaries.

Joseph S. Myers      (personal mail) (CodeSourcery mail) (Bugzilla assignments and CCs)

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]