Page 356 - FUNDAMENTALS OF COMPUTER

NPP

356 Fundamentals of Computers NPP

8. H$_ _mÌm _| Cn`moJ _| bmB© OmVr h¡Ÿ& A{YH$ _mÌm _| Cn`moJ _| bmB© OmVr h¡Ÿ&
9. H$moB© `m§{ÌH$ {hñgo Zht hmoVo h¢Ÿ& Bg_| `m§{ÌH$ {hñgo hmoVo h¢Ÿ&
10. g_mZm§Va S>mQ>m Q´>m§g\$a hmoVm h¡Ÿ& loUr _| S>mQ>m Q´>m§g\$a hmoVm h¡&
11. h_oem aoÝS>_ EŠgog hmoVr h¡Ÿ& gr[a`b `m AY© aoÝS>_ EŠgog hmoVr h¡Ÿ&
4.10 Cache Memory 4.10 Ho$e _o_moar
“A cache memory is a fast and very ""`h EH$ CÀM J{V go H$m`© H$aZo dmbr, bo{H$Z
expensive memory placed in between CPU and AË`{YH$ _h±Jr, AY©MmbH$ _o_moar h¡, Omo {H$ CPU Am¡a
main memory to increase the overall speed of _oZ _o_moar Ho$ _Ü` bJr hmoVr h¡ Ÿ&"" Bggo H$åß`yQ>a Ho$
program execution.” Normally the size of cache àmoJ«m_ {H«$`mÝd`Z H$s J{V ~‹T>mVo h¢ Ÿ& _h±Jr hmoZo Ho$ H$maU
memory is very small in a computer system,
because of its cost factor. The access time of Bgo H$_ _mÌm _| bJmVo h¢& BgH$m EŠgog Q>mB©_ ~hþV H$_
cache is very low. The aim of any caching hmoVm h¡Ÿ& Bg àH$ma go Omo {gñQ>_ ~ZVm h¡, Cgo Ho$qeJ
system is to achieve highest speed with the {gñQ>_ H$hVo h¢Ÿ& Bg {gñQ>_ H$m CÔoí` hmoVm h¡, H$_-
lowest cost. Large amount of main memory is go-H$_ IM© _| A{YH$ go A{YH$ J{V àmßV H$aZmŸ& _oZ
used because it has a moderate price. By using _o_moar H$s H$s_V H$_ hmoZo go BgH$s _mÌm Vmo A{YH$ aIr
small amount of cache and certain algorithm OmVr h¡Ÿ& H$_ _mÌm _| Ho$e H$m Cn`moJ H$aHo$ Bg Vah
an engineer wants to achieve the speed of cache. go J{V àmßV H$aZo H$s H$mo{ee H$aVo h¢ O¡go gånyU© _o_moar
This memory is based on the principle of hr Ho$e hmoŸ& `h bmoHo${bQ>r (ñWmZr`Vm) Am°\$ [a\$a|g
Locality of Reference, which tells that in a Ho$ {gÕm§V na H$m`© H$aVr h¡Ÿ& `h {gÕm§V ~VmVm h¡ {H$
short interval of time CPU tends to reference a
small portion of memory area. {H$gr g_` _| CPU, _o_moar Ho$ EH$ N>moQ>o go joÌ H$mo hr
àmßV H$a nmVm h¡&
Principle of Locality of Reference: It has bmoHo${bQ>r Am°\$ [a\$a|g H$m {gÕm§V: Eogm XoIm
been seen that in a large program CPU J`m h¡ {H$ {H$VZm ^r ~‹S>m àmoJ«m_ hmo, EH$ {ZpíMV g_`
references are confined in a small portion of
memory area. This principle is called locality _| CPU Bg ~‹S>o àmoJ«m_ Ho$ EH$ N>moQ>o go joÌ na hr AnZm
of reference. For example, consider a simple H$m`© H$a nmVm h¡& Bgr VÏ` H$mo bmoHo${bQ>r Am°\$ [a\$a|g
program with no loops. This program is H$m {gÕm§V H$hVo h¢Ÿ& O¡go `{X EH$ Eogm àmoJ«m_ h¡, {Og_|
executed sequentially. If line number A con-
taining an instruction is referenced, then the H$moB© byn Zht h¡, Vmo `h H«$_~Õ gånÞ hmoJmŸ& `{X EH$
next reference will be to line number A + 1. H$WZ Omo {H$ bmB©Z Zå~a A na h¡, EŠgrŠ`yQ> hmo ahm
Now consider the programs containing loops h¡, Vmo AJbm H$WZ A + 1 dmbm hr gånÞ hmoJmŸ& `{X
and procedures. If CPU is entered to execute a
loop then the references made by CPU are àmoJ«m_ _| byn ^r h¡Ÿ& O¡go \$m°a byn h¡ Vmo N>moQ>o g_m`m§Va
confined in a memory area where the loop _| CPU byn Ho$ AÝXa Ho$ H$WZ hr àmßV H$aVm ahoJmŸ&
resides. Similar is the case with the procedures. `hr ~mV àmo{gOa d \§$ŠeZ Ho$ {bE ^r gË` h¡Ÿ&
The programs and data which are needed Eogo hr àmoJ«m_ d S>mQ>m {OZH$m CPU H$mo ~ma-~ma
frequently, are kept in a cache memory. The H$m_ n‹S>Vm h¡, Ho$e _o_moar _| aI {XE OmVo h¢ & _oZ _o_moar
access time of memory is lower by a factor of 5 Ho$ gmnoj Ho$e _o_moar H$m EŠgog Q>mB©_ 5 go 10 JwZm
to 10 as compared to main memory. If a very