Auto-parallelization: Threshold Control and Diagnostics

Threshold Control

The -par-threshold{n} (Linux*) or /Qpar-threshold[:n] (Windows*) option sets a threshold for auto-parallelization of loops based on the probability of profitable execution of the loop in parallel. The value of n can be from 0 to 100.

Diagnostics

The -par-report (Linux) or /Qpar-report (Windows) option controls the auto-parallelizer's diagnostic levels 0, 1, 2, or 3.

Example of Parallelization Diagnostics Report

The following examples shows an output generated by -par-report3 (Linux) or /Qpar-report3 (Windows):

Example 1: Output
program MYPROG procedure: myprog serial loop: line 5: not a parallel candidate due to statement at line 6 serial loop: line 9 flow data dependence from line 10 to line 10, due to "a" 12 Lines Compiled

The sample output resulted from the command:

Platform	Commands
Linux	ifort -c -parallel -par-report3 myprog.f90
Windows	ifort -c /Qparallel /Qpar-report3 myprog.f90

where the program code for myprog.f90 was:

Example 2: Sample code
program myprog integer a(10000), q C Assumed side effects do i=1,10000 a(i) = foo(i) enddo C Actual dependency do i=1,10000 a(i) = a(i-1) + i enddo end

Example 2: Sample code

program myprog

integer a(10000), q

C Assumed side effects

do i=1,10000

a(i) = foo(i)

enddo

C Actual dependency

do i=1,10000

a(i) = a(i-1) + i

enddo

end

Troubleshooting Tips

Use -par-threshold0 (Linux) or /Qpar-threshold:0 (Windows) to auto-parallelize loops regardless of computational work.
Use -par-report3 (Linux) or /Qpar-report3 (Windows) to view diagnostics (see example above).
Use -ipo[value] (Linux) or /Qipo (Windows) to eliminate assumed side-effects done to function calls.
Use the !DEC$ PARALLEL directive to eliminate assumed data dependency.

Auto-parallelization: Threshold Control and Diagnostics

Threshold Control

Diagnostics

Example of Parallelization Diagnostics Report

Example 1: Output

Platform

Commands

Example 2: Sample code

Troubleshooting Tips