This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 04/14] S390: Optimize 8bit-generic iconv modules.
- From: Florian Weimer <fweimer at redhat dot com>
- To: Stefan Liebler <stli at linux dot vnet dot ibm dot com>
- Cc: libc-alpha at sourceware dot org
- Date: Fri, 15 Apr 2016 15:05:31 +0200
- Subject: Re: [PATCH 04/14] S390: Optimize 8bit-generic iconv modules.
- Authentication-results: sourceware.org; auth=none
- References: <1456219278-5258-1-git-send-email-stli at linux dot vnet dot ibm dot com> <1456219278-5258-5-git-send-email-stli at linux dot vnet dot ibm dot com>
On 02/23/2016 10:21 AM, Stefan Liebler wrote:
+ to the 1 byte generic character. If this table contains only up
+ to 256 entry, then the highest UCS4 value can be stored in 1 byte
“256 entries”? (spelling)
Why don't you compute the required table at compile time? Then it can
live in .rodata and does not have to end up in .bss.
In the inline assembly, I would suggest to out-dent the labels. There
is a typo in a comment, “blcocks”. You could reduce the amount of
inline assembly by falling back on the C code for error handling, I think.
I can't comment on the technical accuracy of the inline assembly.
Florian