bigloo/qsort.scm in bglstone confuses the number of runs with the size of the vector to be sorted. Hence, runtimes are tiny and not informative. I attach a fixed file that also tries to increase the task difficulty and to avoid too aggressive optimizations (it uses the result of calls). The necessary fix affects only the function doit.
qsort.scm.txt
Thanks I have applied the modifications to the Bigloo, Gambit, and Kawa sort. It will make the comparison with older systems impossible but the benefit of having significant numbers justifies the
modifications.
bigloo/qsort.scm in bglstone confuses the number of runs with the size of the vector to be sorted. Hence, runtimes are tiny and not informative. I attach a fixed file that also tries to increase the task difficulty and to avoid too aggressive optimizations (it uses the result of calls). The necessary fix affects only the function doit. qsort.scm.txt