Segment registers
in [ASM]
|
From: James Daughtry on 17 Mar 2006 17:14 How and why does one use the segment registers (CS, DS, SS, and ES)? Actually, I'm looking more for the why than the how. ;-) Are they mostly geared toward 16-bit programming and I simply haven't encountered a need for them? Cheers!
From: randyhyde@earthlink.net on 17 Mar 2006 17:51 James Daughtry wrote: > How and why does one use the segment registers (CS, DS, SS, and ES)? > Actually, I'm looking more for the why than the how. ;-) Are they > mostly geared toward 16-bit programming and I simply haven't > encountered a need for them? If you're programming in a typical 32-bit environment (Win32, Linux), you'll not need to use the segment register very often. FS gets used for structured exception handling. (and you don't actually modify FS, the value is given to you, you just reference the segment pointed at by FS). GS also has a special purpose under Win32, but I've forgotten at the moment. I don't know if Linux has similar uses for the segment registers. By and large, however, you are correct. DOS is the main reason people use segmentation. In general, segmentation could be use for some *very* powerful purposes. For example, the Pentia have a 36-bit address space and you're limited to 32-bits in flat model. A decent 32-bit OS that used segmentation would allow you to do some really clever things, such as independent address spaces for memory-mapped files, and things like that. If you can dig up the April 1988 issue of Byte Magazine, you can read about segmentation's benefits in the article I wrote "Memory Management Primer". Cheers, Randy Hyde
From: santosh on 18 Mar 2006 02:58 James Daughtry wrote: > How and why does one use the segment registers (CS, DS, SS, and ES)? > Actually, I'm looking more for the why than the how. ;-) Are they > mostly geared toward 16-bit programming and I simply haven't > encountered a need for them? Well, segmentation exists even in 32 bit protected mode, though most flat memory model operating systems like Windows and Linux effectively hide it by setting the segment registers to hold selectors pointing to the same start address. Under real model, segment registers are used to construct the physical address on every memory reference. The hold a segment address that starts on every paragraph boundary, (i.e. every 10h bytes), in the CPU's address space (a megabyte). The segment address is used as a base address pointing at the start of a "segment". A segment is a region of memory, whose start is specified in a segment register/selector and which extends for some number of bytes. Under real mode the "range" of a segment is always 65536 bytes. In protected mode, the operating system can set the range for each segment. In real mode, to construct the actual physical address, the segment address in the implicit or explicit segment register is multiplied by 16 and the offset address is added to it. The CPU does this automatically and transparently. You simply have to specify the locations that contain the segment and offset addresses. For the former, it is either a segment register assumed from a built-in set of implicit rules or one specified by yourself as a segment override. The later is of course a label, or a register value, (for which only BX, SI, DI, SP may hold offsets), or a literal constant. Segmentation came about because of certain decisions taken by Intel for porting CP/M programs from 8080 to the 8086/8088 under DOS. See the Intel manuals for definitive guide to segmentation and it's history.
|
Pages: 1 Prev: My final hope!! you lovely people. Next: Formatting in assembly |