epics/pcas
0
0
Fork 0
Code Issues Pull Requests Projects Wiki Activity

made hashing routine portable

Browse Source
This commit is contained in:
Jeff Hill
1998-05-29 20:22:44 +00:00
parent 8f2a7b538c
commit f30571ed6d
2 changed files with 40 additions and 70 deletions
Download Patch File Download Diff File Expand all files Collapse all files

55
src/libCom/fdManager.cc
View File

@@ -4,6 +4,9 @@
//
//
// $Log$
// Revision 1.12 1998/02/05 21:12:09 jhill
// removed questionable inline
//
// Revision 1.11 1997/08/05 00:37:00 jhill
// removed warnings
//
@@ -277,45 +280,27 @@ void fdRegId::show(unsigned level) const
//
// fdRegId::resourceHash()
//
resTableIndex fdRegId::resourceHash (unsigned nBitsId) const
resTableIndex fdRegId::resourceHash (unsigned) const
{
//
// 0.5 uS pentium 200
//
//unsigned src = (unsigned) this->fd;
//resTableIndex hashid;
//
//hashid = src;
//while (src = src >> nBitsId) {
// hashid = hashid ^ src;
//}
//hashid = hashid ^ this->type;
resTableIndex hashid = (unsigned) this->fd;
//
// 0.32 uS Pent 200 MHz
//
// faster because it does not
// check for the hash id size
// (assumes worst case hash id
// of 1 bit)
// This assumes worst case hash table index width of 1 bit.
// We will iterate this loop 5 times on a 32 bit architecture.
//
// A good optimizer will unroll this loop?
// Experiments using the microsoft compiler show that this isnt
// slower than switching on the architecture size and urolling the
// loop explicitly (that solution has resulted in portability
// problems in the past).
//
for (unsigned i=(CHAR_BIT*sizeof(unsigned))/2u; i>0u; i >>= 1u) {
hashid ^= (hashid>>i);
}
//
// evenly distribute based on the type of interest also
//
resTableIndex hashid = (unsigned) this->fd;
#if UINT_MAX >> 128u
hashid ^= (hashid>>128u);
#endif
#if UINT_MAX >> 64u
hashid ^= (hashid>>64u);
#endif
#if UINT_MAX >> 32u
hashid ^= (hashid>>32u);
#endif
#if UINTMAX >> 16u
hashid ^= (hashid>>16u);
#endif
hashid ^= (hashid>>8u);
hashid ^= (hashid>>4u);
hashid ^= (hashid>>2u);
hashid ^= (hashid>>1u);
hashid ^= this->type;
//

55
src/libCom/fdmgr/fdManager.cpp
View File

@@ -4,6 +4,9 @@
//
//
// $Log$
// Revision 1.12 1998/02/05 21:12:09 jhill
// removed questionable inline
//
// Revision 1.11 1997/08/05 00:37:00 jhill
// removed warnings
//
@@ -277,45 +280,27 @@ void fdRegId::show(unsigned level) const
//
// fdRegId::resourceHash()
//
resTableIndex fdRegId::resourceHash (unsigned nBitsId) const
resTableIndex fdRegId::resourceHash (unsigned) const
{
//
// 0.5 uS pentium 200
//
//unsigned src = (unsigned) this->fd;
//resTableIndex hashid;
//
//hashid = src;
//while (src = src >> nBitsId) {
// hashid = hashid ^ src;
//}
//hashid = hashid ^ this->type;
resTableIndex hashid = (unsigned) this->fd;
//
// 0.32 uS Pent 200 MHz
//
// faster because it does not
// check for the hash id size
// (assumes worst case hash id
// of 1 bit)
// This assumes worst case hash table index width of 1 bit.
// We will iterate this loop 5 times on a 32 bit architecture.
//
// A good optimizer will unroll this loop?
// Experiments using the microsoft compiler show that this isnt
// slower than switching on the architecture size and urolling the
// loop explicitly (that solution has resulted in portability
// problems in the past).
//
for (unsigned i=(CHAR_BIT*sizeof(unsigned))/2u; i>0u; i >>= 1u) {
hashid ^= (hashid>>i);
}
//
// evenly distribute based on the type of interest also
//
resTableIndex hashid = (unsigned) this->fd;
#if UINT_MAX >> 128u
hashid ^= (hashid>>128u);
#endif
#if UINT_MAX >> 64u
hashid ^= (hashid>>64u);
#endif
#if UINT_MAX >> 32u
hashid ^= (hashid>>32u);
#endif
#if UINTMAX >> 16u
hashid ^= (hashid>>16u);
#endif
hashid ^= (hashid>>8u);
hashid ^= (hashid>>4u);
hashid ^= (hashid>>2u);
hashid ^= (hashid>>1u);
hashid ^= this->type;
//