240 likes | 344 Views
Pertemuan 13 Mesin IA-32 Pentium: I. Matakuliah : T0324 / Arsitektur dan Organisasi Komputer Tahun : 2005 Versi : 1. Learning Outcomes. Pada akhir pertemuan ini, diharapkan mahasiswa akan mampu :
E N D
Pertemuan 13Mesin IA-32 Pentium: I Matakuliah : T0324 / Arsitektur dan Organisasi Komputer Tahun : 2005 Versi : 1
Learning Outcomes Pada akhir pertemuan ini, diharapkan mahasiswa akan mampu : • Membandingkan implementasi instruksi mesin untuk Arsitektur Komputer Intel Pentium ( C4 ) ( No TIK : 6 )
Chapter 3. The IA-32 Pentium Instruction Set: I (OFC3)
Figure 3.38. Compatibility of the IA-32 register structure with earlier Intel processor register structures.
Main memory address doubleword 1000 Base register EBP 1000 * * * * 60 = displacement * * Operand 1060 Operand address (EA) = [EBP] + 60 (a) Base with displacement mode, expressed as [EBP + 60] 1000 Base register EBP 1000 40 * * Index register ESI * * * * 200 = displacement 1200 scale factor = 4 List of 4-byte * * ´ (double w ord) 160 = [Index register] 4 * * data items * * Operand 1360 * * * * * * * * * * * * ´ Operand address (EA) = [EBP] + [ESI] 4 + 200 (b) Base with displacement and index mode, expressed as [EBP + ESI * 4 + 200] Figure 3.39. Examples of addressing modes in the IA-32 architecture.
LEA EBX,NUM1 Initialize base (EBX) and MO V ECX,N coun ter (ECX) registers. MO V EAX,0 Clear accum ulator (EAX) MO V EDI,0 and index (EDI) registers. * ST AR T ADD: ADD EAX,[EBX + EDI 4] Add next n um b er in to EAX. INC EDI Incremen t index register. DEC ECX Decremen t coun ter register. JG ST AR T ADD Branc h bac k if [ECX] > 0. MO V SUM,EAX Store sum in memory . (a) Straightforward approach LEA EBX,NUM1 Load base register EBX and – SUB EBX,4 adjust to hold NUM1 4. MO V ECX,N Initialize coun ter/index (ECX). MO V EAX,0 Clear the accum ulator (EAX). * ST AR T ADD: ADD EAX,[EBX + ECX 4] Add next n um b er in to EAX. LOOP ST AR T ADD Decremen t ECX and branc h bac k if [ECX] > 0. MO V SUM,EAX Store sum in memory . (b) More compact program Figure 3.40. IA-32 program for adding numbers.
.data NUM1 DD 17 , 3 , 51 , 242 , 113 N DD 5 Assem bler directiv es SUM DD 0 .co de MAIN : LEA EBX , NUM1 SUB EBX , 4 MO V ECX , N Statemen ts that generate MO V EAX , 0 mac hine instructions * STARTADD : ADD EAX , [EBX +ECX 4] LOOP ST AR T ADD MO V SUM , EAX Assem bler directiv e END MAIN Figure 3.42. Complete IA-32 assembly language representation for the program in Figure 3.40b.
LEA EBP ,LOC EBP p oin ts to first b yte. MO V AL,[EBP] Load first b yte in to AL. SHL AL,4 Shift left b y 4 bit p ositions. MO V BL,[EBP+1] Load second b yte in to BL. AND BL,0FH Clear high-order 4 bits to zero. OR AL,BL Concatenate the BCD digits. MO V P A CKED,AL Store the result. Figure 3.43. An IA-32 routine to pack two BCD digits into a byte.
LEA EBP ,LOC EBP p oin ts to memory area. READ: BT INST A TUS,3 W ait for c haracter to b e JNC READ en tered in to D A T AIN. MO V AL,DATAIN T ransfer c haracter in to AL. MO V [EBP],AL Store the c haracter in memory INC EBP and incremen t pointer. ECHO: BT OUTST A TUS,3 W ait for displa y to JNC ECHO b e ready . MO V D A T A OUT,AL Send c haracter to displa y . CMP AL,CR If not carriage return, JNE READ read more c haracters. Figure 3.44. An IA-32 program that reads a line of characters and displays it.
Pertemuan 14Mesin IA-32 Pentium: II Matakuliah : T0324 / Arsitektur dan Organisasi Komputer Tahun : 2005 Versi : 1
Learning Outcomes Pada akhir pertemuan ini, diharapkan mahasiswa akan mampu : • Membandingkan implementasi instruksi mesin untuk Arsitektur Komputer Intel Pentium ( C4 ) ( No TIK : 6 )
Chapter 3. The IA-32 Pentium Instruction Set: II (OFC4)
Calling program . . . LEA EBX,NUM1 Load parameters MO V ECX,N in to EBX,ECX. CALL LIST ADD Branc h to subroutine. MO V SUM,EAX Store sum in to memory . . . . Subroutine LIST ADD: PUSH EDI Sa v e EDI. MO V EDI,0 Use EDI as index register. MO V EAX,0 Use EAX as accum ulator register. * ST AR T ADD: ADD EAX, [EBX + EDI 4] Add next n um b er. INC EDI Incremen t index. DEC ECX Decremen t coun ter. JG STARTADD Branc h bac k if [ECX] > 0. POP EDI Restore EDI. RET Branc h bac k to Calling program. (a) Calling program and subroutine [EDI] ESP Return Address Old TOS (b) Stack contents after saving EDI in subroutine Figure 3.45. Program of Figure 3.40a written as an IA-32 subroutine; parameters passed through registers.
(Assume top of stac k is at lev el 1 b elo w.) Calling program PUSH OFFSET NUM1 Push parameters on to the stack. PUSH N CALL LIST ADD Branc h to the subroutine. ADD ESP ,4 Remo v e n from the stack. POP SUM P op the sum in to SUM. . . . Subroutine LIST ADD: PUSH EDI Sa v e EDI and use MO V EDI,0 as index register. PUSH EAX Sa v e EAX and use as MO V EAX,0 accummulator register. PUSH EBX Sa v e EBX and load MO V EBX,[ESP+20] address NUM1. PUSH ECX Sa v e ECX and MO V ECX,[ESP+20] load count n . * ST AR T ADD: ADD EAX,[EBX+EDI 4] Add next n umber. INC EDI Incremen t index. DEC ECX Decremen t coun ter. JG ST AR T ADD Branc h bac k if not done. MO V [ESP+24],EAX Ov erwrite NUM1 in stac k with sum. POP ECX Restore registers. POP EBX POP EAX POP EDI RET Return. (a) Calling program and subroutine [ECX] Lev el 3 [EBX] [EAX] [EDI] Return Address Lev el 2 n NUM1 Lev el 1 (b) Stack contents at different times Figure 3.46. Program of Figure 3.40a written as an IA-32 subroutine; parameters passed on the stack.
Address Instructions Commen ts Calling program . . . 2000 PUSH P ARAM2 Place parameters 2006 PUSH P ARAM1 on stac k. 2012 CALL SUB1 2017 POP RESUL T Store result. ADD ESP ,4 Restore stac k lev el. . . . First subroutine 2100 SUB1: PUSH EBP Sa v e frame p oin ter register. MO V EBP ,ESP Load frame p oin ter. PUSH EAX Sa v e registers. PUSH EBX PUSH ECX PUSH ED X MO V EAX,[EBP+8] Get first parameter. MO V EBX,[EBP+12] Get second parameter. . . . PUSH P ARAM3 Place parameter on stac k. 2160 CALL SUB2 2165 POP ECX P op SUB2 result in to ECX. . . . MO V [EBP+8],ED X Place answ er on stac k. POP ED X Restore registers. POP ECX POP EBX POP EAX POP EBP Restore frame p oin ter register. RET Return to Main program. Second subroutine 3000 SUB2: PUSH EBP Sa v e frame p oin ter register. MO V EBP ,ESP Load frame p oin ter. PUSH EAX Sa v e registers. PUSH EBX MO V EAX,[EBP+8] Get parameter. . . . MO V [EBP+8],EBX Place SUB2 result on stac k. POP EBX Restore registers. POP EAX POP EBP Restore frame p oin ter register. RET Return to first subroutine. Figure 3.47. Nested subroutines in IA-32 assembly language.
LEA EBP ,A VEC EBP p oin ts to v ector A. LEA EBX,BVEC EBX p oin ts to v ector B. MO V ECX,N ECX is the lo op coun ter. MO V EAX,0 EAX accum ulates the dot pro duct. MO V EDI,0 EDI is an index register. * LOOPST AR T: MO V ED X,[EBP+EDI 4] Compute the pro duct * IMUL ED X,[EBX+EDI 4] of next comp onen ts. INC EDI Incremen t index. ADD EAX,ED X Add to previous sum. LOOP LOOPST AR T Branc h bac k if not done. MO V DOTPR OD,EAX Store dot pro duct in memory . Figure 3.49. IA-32 dot product program.
– – (j = n 1; j > 0; j = j 1) for – – { ( k = j 1; k > = 0; k = k 1 ) for { (LIST[ k ] > LIST[ j ]) if { TEMP = LIST[ k ]; LIST[ k ] = LIST[ j ]; LIST[ j ] = TEMP; } } } (a) C-language program for sorting LEA EAX,LIST Load list p oin ter base MO V EDI,N register (EAX), and initialize DEC EDI outer lo op index register – (EDI) to j = n 1. OUTER: MO V ECX,EDI Initialize inner lo op index – DEC ECX register (ECX) to k = j 1. MO V DL,[EAX + EDI] Load LIST(j) in to register DL. INNER: CMP [EAX + ECX],DL Compare LIST(k) to LIST(j). JLE NEXT If LIST(k) LIST(j), go to next lo w er k index en try; X CHG [EAX + ECX],DL Otherwise, in terc hange LIST(k) and LIST(j), lea ving MO V [EAX + EDI],DL new LIST(j) in DL. NEXT: DEC ECX Decremen t inner lo op index k. JGE INNER Rep eat or terminate inner lo op. DEC EDI Decremen t outer lo op index j. JG OUTER Rep eat or terminate outer lo op. (b) IA-32 program implementation Figure 3.50. An IA-32 byte-sorting program using straight-selection sort.
Subroutine INSER TION: MO V RNEWID,[RNEWREC] CMP RHEAD,0 Chec k if list empt y . JG HEAD MO V RHEAD,RNEWREC If y es, new record b ecomes RET one-en try list. HEAD: CMP RNEWID,[RHEAD] Chec k if new record b ecomes head. JG SEAR CH MO V [RNEWREC+4],RHEAD If y es, mak e new record MO V RHEAD,RNEWREC the head. RET SEAR CH: MO V R CURRENT,RHEAD Otherwise, use LOOPST AR T: MO V RNEXT,[R CURRENT+4] R CURRENT CMP RNEXT,0 and RNEXT JE T AIL to mo v e through CMP RNEWID,[RNEXT] the list to find JL INSER T the insertion p oin t. MO V R CURRENT,RNEXT JMP LOOPST AR T INSER T: MO V [RNEWREC+4],RNEXT T AIL: MO V [R CURRENT+4],RNEWREC RET Figure 3.51. An IA-32 subroutine for inserting a new record into a linked list.
Subroutine DELETION: CMP RIDNUM,[RHEAD] Chec k if head. JGT SEAR CH MO V RHEAD,[RHEAD+4] If y es, remo v e. RET SEAR CH: MO V R CURRENT,RHEAD Otherwise, LOOPST AR T: MO V RNEXT,[R CURRENT+4] use R CURRENT CMP RIDNUM,[RNEXT] and RNEXT JEQ DELETE to mo v e through MO V R CURRENT,RNEXT the list to JMP LOOPST AR T find the record. DELETE: MO V R TEMP ,[RNEXT+4] MO V [R CURRENT+4],R TEMP RET Figure 3.52. An IA-32 subroutine for deleting a record from a linked list.
Program 1 Program 2 CLR.L D0 MO VE.W #$FFFF,D0 MO VEA.L #LIST,A0 MO VEA.L #LIST,A0 LOOP MO VE.W (A0)+,D1 LOOP LSL.W (A0)+ BGE LOOP BCC LOOP ADDQ.L #1,D0 LSL.W #1,D0 CMPI #17,D0 BCS LOOP – BL T LOOP MO VE.W 2(A0),RSL T – MO VE.W 2(A0),RSL T Figure P3.1. Two 68000 programs for Problem 3.29.
T able 3.3 IA-32 addressing modes Name Assem bler syn tax Addressing function Immediate V alue Op erand = V alue Direct Lo cation EA = Lo cation Register Reg EA = Reg that is, Op erand = [Reg] Register indirect [Reg] EA = [Reg] Base with [Reg + Disp] EA = [Reg] + Disp displacemen t * Index with [Reg EA = [Reg] S + Disp S + Disp] displacemen t * Base with index [Reg1 + Reg2 S] EA = [Reg1] + [Reg2] S * Base with index [Reg1 + Reg2 S + Disp] EA = [Reg1] + [Reg2] S + Disp and displacemen t V alue = an 8- or 32-bit signed n um b er Lo cation = a 32-bit address Reg, Reg1, Reg2 = one of the general purp ose registers EAX, EBX, ECX, ED X, ESP , EBP , ESI, EDI, with the exception that ESP cannot be used as an index register Disp = an 8- or 32-bit signed n um b er, except that in the Index with displacemen t mode it can only b e 32 bits. S = a scale factor of 1, 2, 4, or 8