1 / 15

Multiplication

Multiplication. CPSC 252 Computer Organization Ellen Walker, Hiram College. Multiplication. Multiplication result needs twice as many bits E.g. 7*7 = 49 (7 is 3 bits, 49 is 6 bits) Multiplication is more complex than addition/subtraction Develop algorithm, implement using simpler hardware.

aleta
Download Presentation

Multiplication

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Multiplication CPSC 252 Computer Organization Ellen Walker, Hiram College

  2. Multiplication • Multiplication result needs twice as many bits • E.g. 7*7 = 49 (7 is 3 bits, 49 is 6 bits) • Multiplication is more complex than addition/subtraction • Develop algorithm, implement using simpler hardware

  3. Multiplication Algorithms • Repeated addition • Easy to implement in software • Can only multiply positive numbers directly • Time depends on magnitude of operands • Shift-and add • Efficient use of hardware • Time depends only on width of operands • No special case for negative numbers

  4. Shift and Add Algorithm • For each bit in multiplier • If bit = 1, add multiplicand to result • Shift multiplicand left by 1 • This is the “usual” multiplication algorithm you learned many years ago (but simpler in binary)

  5. Shift and Add example 10101 21 X 101 5 --------- 10101 000000 (included for completeness) + 1010100 ------------------ 1101001 105 (1+8+32+64)

  6. Hardware to Multiply • 64-bit register for multiplicand • 32 bits for original value • Option to shift left 32 times • 64-bit ALU • Add shifted multiplicand to product • 32-bit register for multiplier • Shift right after each step (low bits fall off) • Control hardware

  7. Multiplication: Implementation Datapath Control

  8. How Long Does it Take? • 32 shift/add cycles • Each cycle has 3 steps (add, shift left, shift right) • Without parallelism, that’s 96 cycles per multiply

  9. Improvements • Save hardware by using 1 64-bit register for the product and multiplier • Left part is product (so far) • Right part is unused multiplier • Add a bit to the product and lose a bit from the multiplier each cycle • Shift product right & add instead of shift multiplicand left & add! • Shift and add in the same cycle (in parallel) • This revision requires only 32 cycles

  10. Final Version • Multiplier starts in right half of product What goes here?

  11. Parallel Multiplication • Instead of 32 cycles, use 32 adders • Each adds 0 or multiplicand • Arrange in tree so results cascade properly • Result pulls each bit from the appropriate adder

  12. Parallel Multiplication (cont)

  13. 010101 x 000101 (6 bits) 000000 000101 (initial product) +010101 Rightmost bit is 1, add multiplicand 010101 000101 0010101 00010 Shift product right 00010101 0001 Shift product right +010101 Rightmost bit is 1, add multiplicand 01101001 0001 001101001 000 Shift product right 0001101001 00 Shift product right 00001101001 0 Shift product right 000001101001 Shift product right, answer is 105

  14. Negative numbers • Convert to positive, multiply, then negate if initial signs were different • Or, consider 3 cases: • Both negative: convert to positive and multiply • One negative: make multiplier positive, multiplicand negative • Negative multiplicand: when high bit is 1, shift in 1’s (shift right arithmetic)

  15. MIPS multiplication • 2 32-bit registers: Hi and Lo mflo $s0 move from lo (to $s0) mfhi $s0 move from hi (to $s0) • Multiplication Operations mult $s0, $s1 {Hi,Lo}= $s0 x $s1 multu $s0, $s1 unsigned version

More Related