Running the convert_benchmark.exe program on Windows, I got a Dr. Watson assert in cleanup_converter() since
AFAICS, the uc8_lookup pointer was already freed.
BTW1, I've not built with -DSC16Q11_TABLE_BITS. But assume it has the same issue.
BTW2, here are my results on AMD 3.8GHz (with clang-cl):
Benchmarking: SC16Q11, DC 1451.28M samples in 5.000000 seconds
290.26M samples/second
Benchmarking: SC16Q11, no DC 2352.60M samples in 5.000000 seconds
470.52M samples/second
Benchmarking: UC8, DC 1157.00M samples in 5.000000 seconds
231.40M samples/second
Benchmarking: UC8, no DC 8046.42M samples in 5.000000 seconds
1609.28M samples/second
Benchmarking: SC16, DC 1474.06M samples in 5.000000 seconds
294.81M samples/second
Benchmarking: SC16, no DC 2413.32M samples in 5.000000 seconds
482.66M samples/second
Running the
convert_benchmark.exe
program on Windows, I got a Dr. Watson assert incleanup_converter()
since AFAICS, theuc8_lookup
pointer was already freed.This little patch fixed it:
BTW1, I've not built with
-DSC16Q11_TABLE_BITS
. But assume it has the same issue.BTW2, here are my results on AMD 3.8GHz (with clang-cl):