Chapter 3. Arithmetic for Computers

Transcription

1 Chapter 3 Arithmetic for Computers

2 Arithmetic for Computers Unsigned vs. Signed Integers Operations on integers Addition and subtraction Multiplication and division Dealing with overflow Floating-point real numbers Representation and operations 3.1 Int roduction Chapter 3 Arithmetic for Computers 2

3 Unsigned Binary Integers Given an n-bit number x = x + n 1 n n 12 + xn 22 + L+ x12 x02 Range: 0 to +2 n 1 Example 2.4 Sig gned and Unsigne ed Numb bers = = = Using 32 bits 0 to +4,294,967,295 Chapter 3 Arithmetic for Computers 3

4 Review: 2 s Complement Binary Representation Negate 1011 and add a complement all the bits Note: negate and invert are different! -2 3 = -(2 3-1) = = 2 sc binary decimal Chapter 3 Arithmetic for Computers 4

5 2s-Complement Signed Integers Given an n-bit number x = x + n 1 n n 12 + xn 22 + L+ x12 x02 Range: 2 n 1 to +2 n 1 1 Example = = 2,147,483, ,147,483,644 = 4 10 Using 32 bits 2,147,483,648 to +2,147,483,647 Chapter 3 Arithmetic for Computers 5

6 2s-Complement Signed Integers Bit 31 is sign bit 1f for negative numbers 0 for non-negative numbers ( 2 n 1 ) can t be represented Non-negative numbers have the same unsigned and 2s-complement representation Some specific numbers 0: : Most-negative: Most-positive: Chapter 3 Arithmetic for Computers 6

7 MIPS Number Representations 32-bit signed numbers (2 s complement): two = 0 ten two = + 1 ten... maxint MSB two = + 2,147,483,646,, ten two = + 2,147,483,647 ten two = 2,147,483,648 ten two = 2,147,483,647 ten two = 2 minint ten two = 1 ten LSB Converting <32-bit values into 32-bit values copy the most significant bit (the sign bit) into the empty bits > > sign extend versus zero extend (lb vs. lbu) Chapter 3 Arithmetic for Computers 7

8 Signed Negation Complement and add 1 Complement means 1 0, 0 1 x + x = = 1 x + 1= x Example: negate = = = Chapter 3 Arithmetic for Computers 8

9 Sign Extension Representing a number using more bits Preserve the numeric value In ARM instruction set LDRSB,LDRSH: extend loaded byte/halfword Replicate the sign bit to the left c.f. unsigned values: extend with 0s Examples: 8-bit to 16-bit +2: => : => Chapter 3 Arithmetic for Computers 9

10 Integer Addition Example: Ad dition and Subtraction Overflow if result out of range Adding +ve and ve operands, no overflow Adding two +ve operands Overflow if result sign is 1 Adding two ve operands Overflow if result sign is 0 Chapter 3 Arithmetic for Computers 10

11 Integer Subtraction Add negation of second operand Example: 7 6 = 7 + ( 6) +7: : : Overflow if result out of range Subtracting two +ve or two ve operands, no overflow Subtracting +ve from ve operand Overflow if result sign is 0 Subtracting ve from +ve operand Overflow if result sign is 1 Chapter 3 Arithmetic for Computers 11

12 Dealing with Overflow Some languages (e.g., C) ignore overflow Use ARM ADD,SUB instructions Other languages (e.g., Ada, Fortran) require raising an exception Use ARM ADDS,SUBS instructions On overflow, invoke exception handler Chapter 3 Arithmetic for Computers 12

13 Arithmetic for Multimedia Graphics and media processing operates on vectors of 8-bit and 16-bit data Use 64-bit adder, with partitioned carry chain Operate on 8 8-bit, 4 16-bit, or 2 32-bit vectors SIMD (single-instruction, multiple-data) Saturating operations On overflow, result is largest representable value c.f. 2s-complement modulo arithmetic E.g., clipping in audio, saturation in video Chapter 3 Arithmetic for Computers 13

14 Multiplication Start with long-multiplication approach 3.3 Mu ultiplicatio on multiplicand 1000 multiplier product Length of product is the sum of operand lengths Chapter 3 Arithmetic for Computers 14

15 Multiplication Hardware Initially 0 Chapter 3 Arithmetic for Computers 15

16 Optimized Multiplier Perform steps in parallel: add/shift One cycle per partial-product addition That s ok, if frequency of multiplications is low Chapter 3 Arithmetic for Computers 16

17 Faster Multiplier Uses multiple adders Cost/performance tradeoff Can be pipelined Several multiplication performed in parallel Chapter 3 Arithmetic for Computers 17

18 ARM Multiplication The MUL instruction is used to multiply signed or unsigned variables to produce a 32-bit result Instruction MUL Rd,Rm,Rs ; Rd = Rm * Rs (32 bits) Because the MUL instruction produces only the lower 32 bits of the 64-bit product, MUL gives the same answer for both signed and unsigned numbers Chapter 3 Arithmetic for Computers 18

19 Division divisor quotient dividend remainder n-bit operands yield n-bit quotient and remainder Check for 0 divisor Long division approach If divisor dividend bits 1 bit in quotient, subtract Otherwise 0 bit in quotient, bring down next dividend bit Restoring division Do the subtract, and if remainder goes < 0, add divisor back Signed division Divide using absolute values Adjust sign of quotient and remainder as required 3.4 Div vision Chapter 3 Arithmetic for Computers 19

20 Division Hardware Initially divisor in left half Initially dividend Chapter 3 Arithmetic for Computers 20

21 Optimized Divider One cycle per partial-remainder subtraction Looks a lot like a multiplier! Same hardware can be used for both Chapter 3 Arithmetic for Computers 21

22 Faster Division Can t use parallel hardware as in multiplier Subtraction is conditional on sign of remainder Faster dividers (e.g. SRT devision) generate multiple quotient bits per step Still require multiple steps Chapter 3 Arithmetic for Computers 22

23 ARM Division The classic ARM instruction set does not include divide instruction ARMv7R and ARMv7M include Signed integer divide - SDIV Unsigned integer divide id - UDIV Chapter 3 Arithmetic for Computers 23

24 Floating Point Representation for non-integral numbers Including very small and very large numbers 3.5 Flo oating Po oint Like scientific notation normalized not normalized In binary ±1.xxxxxxx 2 2 yyyy Types float and double in C Chapter 3 Arithmetic for Computers 24

25 Floating Point Standard Defined by IEEE Std Developed in response to divergence of representations Portability issues for scientific code Now almost universally adopted Two representations Single precision (32-bit) Double precision i (64-bit) Chapter 3 Arithmetic for Computers 25

26 IEEE Floating-Point Format single: 8 bits double: 11 bits S Exponent single: 23 bits double: 52 bits Fraction x = ( 1) S (1+ Fraction) 2 (Exponent Bias) S: sign bit (0 non-negative, 1 negative) Normalize significand: 1.0 significand < 2.0 Always has a leading pre-binary-point 1 bit, so no need to represent it explicitly (hidden bit) Significand is Fraction with the 1. restored Exponent: excess representation: actual exponent + Bias Ensures exponent is unsigned Single: Bias = 127; Double: Bias = 1203 Chapter 3 Arithmetic for Computers 26

27 IEEE 754 FP Standard Encoding Most (all?) computers these days conform to the IEEE 754 floating point standard (-1) sign x (1+F) x 2 E-bias Formats for both single and double precision F is stored in normalized form where the msb in the fraction is 1 (so there is no need to store it!) called the hidden bit To simplify sorting FP numbers, E comes before F in the word and E is represented in excess (biased) notation Single Precision Double Precision Object Represented E (8) F (23) E (11) F (52) true zero (0) 0 nonzero 0 nonzero ± denormalized number ± anything ± anything ± floating point number ± ± ± infinity 255 nonzero 2047 nonzero not a number (NaN) Chapter 3 Arithmetic for Computers 27

28 Single-Precision Range Exponents and reserved Smallest value Exponent: actual exponent = = 126 Fraction: significand = 1.0 ± ± Largest value exponent: actual exponent = = +127 Fraction: significand 2.0 ± ± Chapter 3 Arithmetic for Computers 28

29 Double-Precision Range Exponents and reserved Smallest value Exponent: actual exponent = = 1022 Fraction: significand = 1.0 ± ± Largest value Exponent: actual exponent = = Fraction: significand 2.0 ± ± Chapter 3 Arithmetic for Computers 29

30 Floating-Point Precision Relative precision all fraction bits are significant Single: approx 2 23 Equivalent to 23 log decimal digits of precision Double: approx 2 52 Equivalent to 52 log decimal digits of precision Chapter 3 Arithmetic for Computers 30

31 Floating-Point Example Represent = ( 1) S = 1 Fraction = Exponent = 1+Bi Bias Single: = 126 = Double: = 1022 = Single: Double: Chapter 3 Arithmetic for Computers 31

32 Floating-Point Example What number is represented by the single- precision i float S = 1 Fraction = Exponent = = 129 x=( ( 1) 1 ( ) 2 = ( 1) = 5.0 ( ) Chapter 3 Arithmetic for Computers 32

33 Denormal Numbers Exponent = hidden bit is 0 x = ( 1) S (0 + Fraction) 2 Bias Smaller than normal numbers allow for gradual underflow, with diminishing precision Denormal with fraction = x S = ( 1) (0 + 0) 2 Bias Two representations of f00! 0.0! = ±0.00 Chapter 3 Arithmetic for Computers 33

34 Infinities and NaNs Exponent = , Fraction = ±Infinity Can be used in subsequent calculations, avoiding need for overflow check Exponent = , 1 Fraction Not-a-Number (NaN) Indicates illegal or undefined result e.g., 0.0 / 0.0 Can be used in subsequent calculations Chapter 3 Arithmetic for Computers 34

35 Floating-Point Addition Consider a 4-digit decimal example Align decimal points Shift number with smaller exponent Add significands ifi = Normalize result & check for over/underflow Round and renormalize if necessary Chapter 3 Arithmetic for Computers 35

36 Floating-Point Addition Now consider a 4-digit binary example ( ) 1. Align binary points Shift number with smaller exponent Add significands ifi = Normalize result & check for over/underflow , with no over/underflow 4. Round and renormalize if necessary (no change) = Chapter 3 Arithmetic for Computers 36

37 FP Adder Hardware Much more complex than integer adder Doing it in one clock cycle would take too long Much longer than integer operations Slower clock would penalize all instructions FP adder usually takes several cycles Can be pipelined Chapter 3 Arithmetic for Computers 37

38 FP Adder Hardware Step 1 Step 2 Step 3 Step 4 Chapter 3 Arithmetic for Computers 38

39 Floating-Point Multiplication Consider a 4-digit decimal example Add exponents For biased exponents, subtract bias from sum New exponent = = 5 2. Multiply significands = Normalize result & check for over/underflow Round and renormalize if necessary Determine sign of result from signs of operands Chapter 3 Arithmetic for Computers 39

40 Floating-Point Multiplication Now consider a 4-digit binary example ( ) 1. Add exponents Unbiased: = 3 Biased: ( ) + ( ) = = Multiply significands = Normalize result & check for over/underflow (no change) with no over/underflow 4. Round and renormalize if necessary (no change) 5. Determine sign: +ve ve ve = Chapter 3 Arithmetic for Computers 40

41 FP Arithmetic Hardware FP multiplier is of similar complexity to FP adder But uses a multiplier for significands instead of an adder FP arithmetic hardware usually does Addition, subtraction, multiplication, division, reciprocal, square-root FP integer conversion Operations usually takes several cycles Can be pipelined Chapter 3 Arithmetic for Computers 41

42 FP Instructions in ARM FP hardware is coprocessor 1 Adjunct processor that extends the ISA Separate FP registers 32 single-precision: s0, s1, s31 Paired for double-precision: s0/s1, s2/s3, to give 16 double precision registers : d0,d1..d15 FP instructions operate only on FP registers Programs generally don t do integer ops on FP data, or vice versa More registers with minimal code-size impact FP load and store instructions FLDS,FSTS,FLDD,FSTD e.g., FLDS s1, [r1,#100] ; s1 = mem [r1+400] Chapter 3 Arithmetic for Computers 42

43 FP Instructions in ARM Single-precision arithmetic FADDS,FSUBS,FMULS,FDIVSFSUBS FMULS FDIVS e.g., FADDS s2,s4,s6 ; s2 = s4 + s6 Double-precision arithmetic FADDD,FSUBD,FMULD,FDIVD e.g., FSUBD d2,d4,d6 ; d2 = d4 d6 Single- and double-precision comparison FCMPS, FCMPD e.g. FCMPS s2,s4 ; if (s2 s4) Branch on FP condition flag FMSTAT ; cond.flags = FP cond. flags Copy FP condition flags to integer condition flag Chapter 3 Arithmetic for Computers 43

44 Summary of ARM FP instructions discussed thus far Chapter 3 Arithmetic for Computers 44

45 FP Example: F to C C code: float f2c (float fahr) { return ((5.0/9.0)*(fahr )); } fahr in s12, result in s0, literals in global memory space Compiled ARM code: f2c: FLDS s16, [r12, const5] ;s16 = 5.0 FLDS s18, [r12, const9] ;s18 = 9.0 FDIVS s16,s16,s18 ; s16 = 5.0/9.0 FLDS s18, [r12,const32] ; s18 = 32.0 FSUBS s18,s12,s18 s12 ; s18 = fahr 32 FMULS s0,s16,s18 ; s0 = (5/9)* (fahr 32) MOV pc,lr ; return Chapter 3 Arithmetic for Computers 45

46 FP Example: Array Multiplication X = X + Y Z All matrices, 64-bit double-precision i elements C code: void mm (double x[][], double y[][], double z[][]) { int i, j, k; for (i = 0; i! = 32; i = i + 1) for (j = 0; j! = 32; j = j + 1) for (k = 0; k! = 32; k = k + 1) x[i][j] = x[i][j] + y[i][k] * z[k][j]; } Chapter 3 Arithmetic for Computers 46

47 FP Example: Array Multiplication ARM code: Addresses of variable x, y, z in r0, r1, r2, and i,j, j k in r3, r4, r5. The address of x[ i ][ j ] is r6 and address of y[ i ][ j ] or z[ i ][ j ] is r12. Mm: SUB sp,sp,#12 ; make room on stack for 3 regs STR r4,[sp,#8] ; save r4 on stack STR r5,[sp,#4] ; save r5 on stack STR r6,[sp,#0] ; save r6 on stack MOV i, 0 ; i = 0 initialize 1 st for loop L1: MOV j, 0 ; j = 0 restart 2 nd for loop L2: MOV k,0 ; k = 0 restart 3 rd for loop ADD xijaddr,j,i, LSL #5 ; xijaddr = i*size(row) + j ADD xijaddr,x,xijaddr,lsl #3 ; xijaddr = byte address of x[i][j] Chapter 3 Arithmetic for Computers 47

48 FP Example: Array Multiplication FLDD s4, [xijaddr,#0] ; s4 = 8 bytes of x[i][j] L3: ADD tempaddr, j,k,lsl #5 ; tempaddr = k *size(row) + j ADD tempaddr,z,tempaddr,lsl #3 ; tempaddr= byte ;address of z[k][j] FLDD s16,[tempaddr,#0] ; s16 = 8 bytes of z[k][j] ADD tempaddr, k,i,lsl #5 ; tempaddr = i *size(row) + k ADD tempaddr,y,tempaddr,lsl #3 ; tempaddr= byte ;address of y[i][k] FLDD s18,[tempaddr,#0] ; s18 = 8 bytes of y[i][k] FMULD s16,s18,s16 s18 ; s16 = y[i][k] * z[k][j] FADDD s4,s4,s16 ; s4 =x[i][j] + ; y[i][k]*z[k][j] Chapter 3 Arithmetic for Computers 48

49 FP Example: Array Multiplication ADD k,k,#1 ; k = k+1 CMP k, #32 BLT L3 ; if (k<32) go to L3 FSTD s4, [xijaddr,#0] ; x[i][j] = s4 ADD j,j, #1 ; j = j+1 CMP j #32 BLT ls ; if ( j<32) go to L2 ADD i,i, #1 ; i = i + 1 CMP i, #32 BLT L1 ; if ( i<32) go to L1 Chapter 3 Arithmetic for Computers 49

50 Accurate Arithmetic IEEE Std 754 specifies additional rounding control Extra bits of precision (guard, round, sticky) Choice of rounding modes Allows programmer to fine-tune numerical behavior of a computation Not all FP units implement all options Most programming g languages g and FP libraries just use defaults Trade-off between hardware complexity, performance, and market requirements Chapter 3 Arithmetic for Computers 50

51 Interpretation of Data The BIG Picture Bits have no inherent meaning Interpretation depends on the instructions applied Computer representations of numbers Finite range and precision Need to account for this in programs Chapter 3 Arithmetic for Computers 51

52 Associativity Parallel programs may interleave operations in unexpected orders Assumptions of associativity may fail (x+y)+z x+(y+z) x -1.50E E+38 y 1.50E E+00 z E E E+00 Need to validate parallel l programs under varying degrees of parallelism 3.6 Pa arallelism and Com mputer Arithmetic c: Associa ativity Chapter 3 Arithmetic for Computers 52

53 Right Shift and Division Left shift by i places multiplies an integer by 2 i Right shift divides by 2 i? Only for unsigned integers For signed integers Arithmetic right shift: replicate the sign bit eg e.g., 5/ >> 2 = = 2 Rounds toward c.f >>> 2 = = Fa llacies an nd Pitfalls Chapter 3 Arithmetic for Computers 53

54 Who Cares About FP Accuracy? Important for scientific code But for everyday consumer use? My bank balance is out by ! The Intel Pentium FDIV bug The market expects accuracy See Colwell, The Pentium Chronicles Chapter 3 Arithmetic for Computers 54

55 Concluding Remarks ISAs support arithmetic Signed and unsigned integers Floating-point approximation to reals Bounded range and precision Operations can overflow and underflow ARM core instructions dominate integer SPEC2006 execution and integer and arithmetic core dominate SPEC2006 floating point execution. 3.9 Co oncluding Remark ks Chapter 3 Arithmetic for Computers 55