nec_09mar2004_training_2
Seite
Seite
Vectorization examples
ex. 1: v = s + v (cont.)
Seite
Chaining
ex. 1: v = s + v (cont.)
ex. 1: v = s + v (cont.)
ex. 1: v = s + v
ex. 2: v = v + v
Seite
ex. 3: v = v + s * v
Seite
ex. 3: v = v + s * v
Seite
ex. 4: s = s + v * v
ex. 4: s = s + v * v (cont.)
ex. 5: matrix multiply
ex. s1: matrix multiply
Levels of Parallelism
Seite
Basic Rules to achieve Performance
Basic Rules for Performance
Basic Rules for Performance
Basic Rules for Performance
Basic Rules for Performance
Basic Rules for Performance
Basic Rules for Performance
Basic Rules for Performance
Basic Rules for Performance
Basic Rules for Performance
Basic Rules for Performance
Basic Rules for Performance
Basic Rules for Performance: Concurrency
Basic Rules for Performance
Basic Rules for Performance
Basic Rules for Performance
Basic Rules for Performance: Pointers
Basic Rules for Performance
vectorization of if-blocks
vectorization of if-blocks
vectorization of if-blocks
vectorization of if-blocks
vectorization of if-blocks
vectorization of if-blocks
vectorization of if-blocks
vectorization of if-blocks
vectorization of if-blocks
vectorization of if-blocks
vectorization of if-blocks
constant if
boundary if
ex. s1: matrix multiply
IO optimization
Basic Rules for Performance
IO optimization
Basic Rules for Performance
Seite
FORTRAN 90 !cdir
FORTRAN 90 !cdir
FORTRAN 90 !cdir
FORTRAN 90 !cdir
FORTRAN 90 !cdir
Stride for 1d-Arrays
Bank busy time measured
Stride for 1d-Arrays
Bank busy time measured
Stride for 1d-Arrays
FORTRAN 90 !cdir
Stride for 1d-Arrays
Bank busy time measured
Seite
loop interchange
loop interchange
loop expansion
loop expansion
loop expansion
loop expansion
loop expansion
loop division
loop expansion
loop division
loop division
loop division
call to function
vectorization example
vectorization example (2)
vectorization example (3)
vectorization example (4)
Seite
2D recursion
2D recursion (2)
Seite
Seite
Basic Rules for Performance
Seite