This is the mail archive of the
mailing list for the glibc project.
Re: [PATCH] powerpc: New feature - HWCAP/HWCAP2 bits in the TCB
- From: David Edelsohn <dje dot gcc at gmail dot com>
- To: Torvald Riegel <triegel at redhat dot com>
- Cc: Richard Henderson <rth at twiddle dot net>, munroesj at us dot ibm dot com, szabolcs dot nagy at arm dot com, Carlos Eduardo Seo <cseo at linux dot vnet dot ibm dot com>, GLIBC Devel <libc-alpha at sourceware dot org>, Steve Munroe <sjmunroe at us dot ibm dot com>
- Date: Wed, 1 Jul 2015 14:17:01 -0400
- Subject: Re: [PATCH] powerpc: New feature - HWCAP/HWCAP2 bits in the TCB
- Authentication-results: sourceware.org; auth=none
- References: <CAGWvnynvibieMA_7D3-DnNG-BFRLQsn4OeOv_=r1gKyDpMgRXw at mail dot gmail dot com> <1435701696 dot 4216 dot 57 dot camel at localhost dot localdomain>
On Tue, Jun 30, 2015 at 6:01 PM, Torvald Riegel <firstname.lastname@example.org> wrote:
> The technical concerns that Richard voiced are still there if we assume
> that Intel systems don't exist at all. Therefore, I don't see how he or
> the glibc community are picking favorites here. Could you point to some
> more evidence that he or the glibc community (ie, backed with community
> consensus, not opinions of individuals) are doing it?
No one has disagreed with Richard's technical concern and
recommendation. If the customer could / would redesign their code,
that would be great. Some customers are open to critiques of their
software design and others are not.
A lot of the feedback in this thread is predicated on the assumption
that a vendor can dictate to a customer how to design and implement
their code. The customer is the one paying and ultimately is the one
setting the rules. I doubt that Jim Whitehurst or Jeff Law go to the
C-suite of Red Hat's biggest customers and tell them that Red Hat's
engineers think their coding style is stupid. They may say that
jokingly, but they are not going to tell the customer "change your
code or we won't sell you our product and won't try to help you."
It's not really productive to tell a customer that you don't care
about their business. At the end of the day, Red Hat (or IBM) need to
try to accommodate the customer's needs.
Steve has been working with IBM processor architects on different
instruction sequences based on the customer coding style. He proposed
the best option for the constraints and current processor reality.
All of the mailing list feedback is appreciated, although the tone for
some of the comments was unconstructive. A number of GLIBC developers
have been shooting from the hip with suggestions about an ISA and
micro-architecture for which they don't have good instincts and then
demanding that Steve prove why each and every suggestion is wrong.
It's easy to pontificate or make demands of others when one has little
or no skin in the game. If anyone wants to come work on challenging
problems, let Steve, Carlos or me know. ;-)