https://issues.dlang.org/show_bug.cgi?id=17229
--- Comment #1 from Jon Degenhardt <[email protected]> --- (In reply to Jon Degenhardt from comment #0) > Results for the count 9's program, against the 2.7, 14 million line file: > byLine: 8.98 seconds > byChunk: 1.64 seconds Update: The byLine Count 9s program given above is being affected by autodecode. Changing counting line as follows: from: chunkedStream.each!(x => count += x.count('9')); to: chunkedStream.each!(x => count += (cast(ubyte[])x).count('9')); Gives updated times: byLine: 3.13 byChunk: 1.64 This is more consistent. byChunk is faster than byLine in the Count 9s task (read and access data), by not by 5x. The 15x performance deficit of byChunk relative to byLine on the file copy task remains. --
