[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sat, 7 Jul 2012 11:00:49 -0500
From: "jfoug" <jfoug@....net>
To: <john-dev@...ts.openwall.com>
Subject: RE: Build error for generic (was: SHA2 added to bleeding)
>From: Frank Dittrich [mailto:frank_dittrich@...mail.com]
>On 06/30/2012 06:13 PM, jfoug wrote:
>> I have added sha2.c and sha2.h and modified a lot of formats, to
>> remove the oSSL version dependency for SHA224/256/384/512 from JtR.
>
>Now that this has arrived in magnum-jumbo, building generic on my 32bit
>Linux system fails:
I have changed the macro, from using the uintptr_t to using ARCH_WORD. I
am pretty sure this will be adequate, since we are already properly (I hope)
using the proper machine sized int for ARCH_WORD. The proper machine sized
int was what uintptr_t was supposed to be.
I also changed code in the non-aligned CPU part of memory.c (in the
non-aligned CPU's alter endianity), to use the is_aligned macro. This
should help any problems show up faster.
The only 2 usages of this macro, are wrapped with #define
!ARCH_ALLOWS_UNALIGNED. That the generic build of linux uses this SLOWER
code bothers me. I think we should look at adding at least that much smarts
to the generic build. It can easily detect the ARCH_ALLOWS_UNALIGNED by
checking these (likely not all encompassing, but it would catch it MOST of
the time)
#if defined(i386) || defined(__i386__) || defined(__i486__) || \
defined(__i586__) || defined(__i686__) || defined(__pentium__) || \
defined(__pentiumpro__) || defined(__pentium4__) || \
defined(__nocona__) || defined(prescott) || defined(__core2__) || \
defined(__k6__) || defined(__k8__) || defined(__athlon__) || \
defined(__amd64) || defined(__amd64__) || \
defined(__x86_64) || defined(__x86_64__) || defined(_M_IX86) || \
defined(_M_AMD64) || defined(_M_IA64) || defined(_M_X64)
#define ARCH_ALLOWS_UNALIGNED 1
#else
#define ARCH_ALLOWS_UNALIGNED 0
#endif
There is no reason to always generically use ARCH_ALLOWS_UNALIGNED 0 I
can easily 'fix' detect.c to output a 0 or 1, depending upon the above
defines. This should improve the speed of generic builds for any Intel/AMD
CPU listed. Certainly, will improve speed of most any dynamic format, if a
generic build was done, on a CPU allowing non-aligned access.
Jim.
Download attachment "JtR-bleed-align-macro-fix-1.diff" of type "application/octet-stream" (1157 bytes)
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
