I am wondering if 64-bit Matlab in 64-bit OS compiles faster things? Than 32-bit Matlab in 32-bit OS if the machine is the same one!!! If no... what are the advantages that 64-bit Matlab with 64-bit OS brings us???

0 |

12/6/2010 7:53:06 PM

"Sofia Häggberg" <hd.int.assNOSPAM@gmail.com> wrote in message <idjev2$sp6$1@fred.mathworks.com>... > I am wondering if 64-bit Matlab in 64-bit OS compiles faster things? Than 32-bit Matlab in 32-bit OS if the machine is the same one!!! Generally 64-bit is slower. >If no... what are the advantages that 64-bit Matlab with 64-bit OS brings us??? Capability to address larger memory than the limit of 2-4 Gbytes, thus solving very big problem. Bruno

0 |

12/6/2010 8:05:09 PM

On 10-12-06 02:05 PM, Bruno Luong wrote: > "Sofia H=C3=A4ggberg" <hd.int.assNOSPAM@gmail.com> wrote in message > <idjev2$sp6$1@fred.mathworks.com>... >> I am wondering if 64-bit Matlab in 64-bit OS compiles faster things? >> Than 32-bit Matlab in 32-bit OS if the machine is the same one!!! > > Generally 64-bit is slower. > >> If no... what are the advantages that 64-bit Matlab with 64-bit OS >> brings us??? > > Capability to address larger memory than the limit of 2-4 Gbytes, thus > solving very big problem. It will depend upon the hardware details -- though it would not surprise = me if=20 these days the only supported hardware that differs is the 32 bit MacOS P= PC=20 version. When 64 bit code is used, 64 bit pointers have to be loaded and stored. I= t is=20 not uncommon on processors for that to take longer than saving and loadin= g 32=20 bit pointers, because more data has to be transferred across the bus.=20 Depending on the instruction set, 64 bit code might also lead to longer=20 instructions in some cases, instructions that can deal with offsets large= r=20 than 32 bits. Thus, it is not unreasonable for 64 bit code to end up being larger and s= lower=20 than 32 bit code. But as Bruno indicates, if you have the memory then 64 = bit=20 code can be used to solve larger problems. It doesn't take much before yo= u run=20 in to the 32 bit limits -- on Windows, with the way that Matlab 32 bit us= es=20 memory, it is common to start out with a maximum array size less than 800= =20 megabytes. High performance processors tend to be tuned for 64 bit (or larger) loads= and=20 stores, and on them 32 bit loads and stores can end up taking more time. = You=20 have to know the details of your particular processor model (not just fam= ily)=20 in order to figure out which would be best for a particular purpose.

0 |

12/6/2010 8:23:43 PM