# 08 иЃ°зЃЧжЬЇзїДжИР--ж±ЗзЉЦз®ЛеЇП

жЬђзѓЗжЦЗж°£дЄїи¶БдїЛзїНдЇЖж±ЗзЉЦз®ЛеЇПе¶ВдљХеЈ•дљЬпЉМдї•еК†ж≥ХжМЗдї§дЄЇдЊЛпЉМзіѓеК†еЩ®еЬ®жУНдљЬдєЛеЙНе≠ШеВ®жУНдљЬжХ∞пЉМеЬ®жУНдљЬдєЛеРОе≠ШеВ®зїУжЮЬгАВдїЛзїНдЇЖMIPSжШѓдЄАдЄ™иіЯиљље≠ШеВ®жЮґжЮДгАБзЃАеМЦжМЗдї§йЫЖиЃ°зЃЧжЬЇ(RISC)гАБж†ИзїУжЮДзЪДдЊЛе≠РгАБиЃ°зЃЧжЬЇињРзЃЧгАБжЇҐеЗЇз≠ЙзЯ•иѓЖгАВ
е±ХеЉАжЯ•зЬЛиѓ¶жГЕ

1.How does an assembler work? In a two-pass assembler PASS 1: Symbol table generation PASS 2: Code generation Illustration of the two passes (follow the class lecture) .data L1: .word 0x2345 # some arbitrary value L2: .word 0x3366 # some arbitrary value Res: .space 4 .text .globl main main: lw \$t0, L1(\$0) #load the first value lw \$t1, L2(\$0) # load the second value and \$t2, \$t0, \$t1 # compute the bit-by-bit AND or \$t3, \$t0, \$t1 # compute the bit-by-bit OR sw \$t3, Res(\$0) # store result at location in memory li \$v0, 10 # code for program end syscall

2.Other architectures Not all processors are like MIPS. Example. Accumulator-based machines A single register, called the accumulator, stores the operand before the operation, and stores the result after the operation. Load x # into accumulator from memory Add y # add y from memory to the acc Store z # store acc to memory as z Can we have an instruction like add z, x, y # z:= x + y, (x, y, z in memory) ? For some machines, YES, but not in MIPS! What are the advantages and disadvantages of such an instruction?

3.Load-store machines MIPS is a load-store architecture. Only load and store instructions access the memory, all other instructions use registers as operands. What is the motivation? Register access is much faster than memory access, so the program will run faster. Reduced Instruction Set Computers (RISC) вАҐ The instruction set has only a small number of frequently used instructions. This lowers processor cost, without much impact on performance. вАҐ All instructions have the same length. вАҐ Load-store architecture. Non-RISC machines are called CISC (Complex Instruction Set Computer). Example: Pentium

4.Another classification 3-address add r1, r2, r3 (r1 вЖР r2 + r3) 2-address add r1, r2 (r1 вЖР r1 + r2) 1-address add r1 (to the accumulator) 0-address or stack machines (see below) Example of stack architecture Consider evaluating z = x * (y + z) Push x x x x x * (y+z) Push y y y+z Push z z Add Multiply Pop z

5.Computer Arithmetic How to represent negative integers? The most widely used convention is 2вАЩs complement representation. +14 = 0, 1 1 1 0 -14 = 1, 0 0 1 0 Largest integer represented using n-bits is + (2n-1 вАУ 1) Smallest integer represented using n-bits is - 2n-1 So, using 4-bits (that includes 1 sign bit), the largest integer is 0,111 (=7), and the smallest integer is 1,000 (= -8) Review binary-to decimal and binary-to-hex conversions. Review BCD (Binary Coded Decimal) and ASCII codes. How to represent fractions?

6.Overflow +12 = 0, 1 1 0 0 +12 = 0, 1 1 0 0 +2 = 0, 0 0 1 0 +7 = 0, 0 1 1 1 add add +14 = 0, 1 1 1 0 ? = 1, 0 0 1 1 (WRONG) Addition of a positive and a negative number does not lead to overflow. How to detect overflow? Here is a clue. 0 0 0 вКХ 0 = 0(OK ) 0 1 0 вКХ1 = 1¬†(NOT ¬†OK) +12 = 0, 1 1 0 0 +12 = 0, 1 1 0 0 +2 = 0, 0 0 1 0 +7 = 0, 0 1 1 1 add add +14 = 0, 1 1 1 0 ? = 1, 0 0 1 1 (WRONG)

7.The following sequence of MIPS instructions can detect overflow in signed addition of \$t1 and \$t2: addu \$t0, \$t1, \$t2 # add unsigned xor \$t3, \$t1, \$t2 # check if signs differ slt \$t3, \$t3, \$zero # \$t3=1 if signs differ bne \$t3, \$zero, no_overflow xor \$t3, \$t0, \$t1 # sum sign = operand sign? slt \$t3, \$t3, \$zero # if not, then \$t3=1 bne \$t3, \$zero, overflow no_overflow: ... ... overflow: <Do something to handle overflow>

жИСе∞±жШѓжИСпЉБ
еЈ≤е∞ЖйУЊжО•е§НеИґиЗ≥еЙ™иііжЭњ