[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
Re: lib/49745: tests/include/sys/t_bitops: improvement
On Mar 14, 9:55am, isaki%pastel-flower.jp@localhost (isaki%pastel-flower.jp@localhost) wrote:
-- Subject: lib/49745: tests/include/sys/t_bitops: improvement
| >Number: 49745
| >Category: lib
| >Synopsis: tests/include/sys/t_bitops: improvement
| >Confidential: no
| >Severity: non-critical
| >Priority: low
| >Responsible: lib-bug-people
| >State: open
| >Class: sw-bug
| >Submitter-Id: net
| >Arrival-Date: Sat Mar 14 09:55:00 +0000 2015
| >Originator: Tetsuya Isaki
| >Release: NetBSD-current (2015/02)
| NetBSD XXXXX 7.99.4 NetBSD 7.99.4 (GENERIC) #1: Wed Feb 11 12:54:54 JST 2015 isaki@XXXXX:/var/obj/current/x68k/obj/sys/arch/x68k/compile/GENERIC x68k
| The current ilog2()'s implementation consists of 32bit variable
| part, 64bit variable part and constant part. But 'ilog2_basic'
| test checks only 64bit variable. 32bit variable and constant
| were not checked.
| 'ilog2_log2' test is fully meaningless. It checks only bit 0
| and bit 16-31 of 64bit variable though it spent 65000 times loop.
| All range of 64bit variable was already checked by 'ilog2_basic'
| In addition, comparing it with math log2() is also really bad
| idea. It should be compared with prepared(=pre-calculated)
| value. The current 'ilog2_log2' test does not seem to work
| on vax and QEMU due to lack/fail of math log2() according to
| source code comment (I don't know details and it seems to work
| on my QEMU though..). However, it's not necessary to use math
| log2() here.
I would prefer if you put the constants in an array and loop instead
of inlining it.
Main Index |
Thread Index |