TEST RESULTS FOR THE 5 COMPUTATION METHODS

For positive inputs...
   Benchmark using floating-point double      +13485.75
   1) int  products with  int accumulator       +189  
   2) int  products with long accumulator      -8321  
   3) long products with long accumulator       -132  
   4) long products with 2 long accumulators  +13485  
   5) In-line assembler version of the above  +13485

For negative inputs...
   Benchmark using floating-point double      -13485.75
   1) int  products with  int accumulator       -189  
   2) int  products with long accumulator      +8321  
   3) long products with long accumulator       +132  
   4) long products with 2 long accumulators  -13485  
   5) In-line assembler version of the above  -13485


TIME TRIALS AGAINST FLOATING-POINT DOUBLE BENCHMARK:
   Floating-point double  345.15 secs  (Benchmark)          
   4) 'C' using *(I+i)      9.34 secs  ( 36.95 times as fast)
   4) 'C' using *(pi+i)     9.39 secs  ( 36.76 times as fast)
   4) 'C' using *pi++       9.72 secs  ( 35.51 times as fast)
   4) 'C' for(..., pw++)    9.61 secs  ( 35.92 times as fast)
   5) asm using *(I+i)     15.99 secs  ( 21.59 times as fast)
Note: time trials were only done on the methods that worked.


TESTING WITH MAXIMUM INPUT AND WEIGHT
   For max positive inputs  +32766       7FFE
   For zero inputs...           +0          0
   For max negative inputs  -32766   FFFF8002


TEST RESULTS FOR THE 5 COMPUTATION METHODS

For positive inputs...
   Benchmark using floating-point double      +13485.75
   1) int  products with  int accumulator       +189  
   2) int  products with long accumulator      -8321  
   3) long products with long accumulator       -132  
   4) long products with 2 long accumulators  +13485  
   5) In-line assembler version of the above  +13485

For negative inputs...
   Benchmark using floating-point double      -13485.75
   1) int  products with  int accumulator       -189  
   2) int  products with long accumulator      +8321  
   3) long products with long accumulator       +132  
   4) long products with 2 long accumulators  -13485  
   5) In-line assembler version of the above  -13485


TIME TRIALS AGAINST FLOATING-POINT DOUBLE BENCHMARK:
   Floating-point double    0.05 secs  (Benchmark)          
   4) 'C' using *(I+i)      0.11 secs  (  0.45 times as fast)
   4) 'C' using *(pi+i)     0.11 secs  (  0.45 times as fast)
   4) 'C' using *pi++       0.11 secs  (  0.45 times as fast)
   4) 'C' for(..., pw++)    0.17 secs  (  0.29 times as fast)
   5) asm using *(I+i)      0.16 secs  (  0.31 times as fast)
Note: time trials were only done on the methods that worked.


TESTING WITH MAXIMUM INPUT AND WEIGHT
   For max positive inputs  +32766       7FFE
   For zero inputs...           +0          0
   For max negative inputs  -32766   FFFF8002