On 26 February 2014 14:44, Robert Lytton <robert at xmos.com> wrote:> This is related to a patch I submitted a little while ago (still pending): > http://llvm-reviews.chandlerc.com/D2760 > > If accepted, would it make this patch (and a others) unnecessary?Hi Robert, It is, but hijacking your patch a little, why not use __ORDER_LITTLE_ENDIAN__? Why do we need to create __LITTLE_ENDIAN__ ? cheers, --renato
Hi Renato, You are quite right. The question is are __LITTLE_ENDIAN__ & __BIG_ENDIAN__ defines becoming 'standard'? Do we want all targets to have one or the other? Still waiting for feedback... Robert ________________________________________ From: Renato Golin [renato.golin at linaro.org] Sent: 26 February 2014 17:11 To: Robert Lytton Cc: Daniel Dunbar; LLVM Dev Subject: Re: [LLVMdev] test-suite wrongly using big-endian results On 26 February 2014 14:44, Robert Lytton <robert at xmos.com> wrote:> This is related to a patch I submitted a little while ago (still pending): > http://llvm-reviews.chandlerc.com/D2760 > > If accepted, would it make this patch (and a others) unnecessary?Hi Robert, It is, but hijacking your patch a little, why not use __ORDER_LITTLE_ENDIAN__? Why do we need to create __LITTLE_ENDIAN__ ? cheers, --renato
On 26 February 2014 18:00, Robert Lytton <robert at xmos.com> wrote:> The question is are __LITTLE_ENDIAN__ & __BIG_ENDIAN__ defines becoming 'standard'?They're redundant with __ORDER_*_ENDIAN__, so I personally wouldn't even emit them. Even though code might have start depending on them and it could be hard to deprecate it, I'd still go through. It's not like people would have to re-implement anything, this is a simple regex's job.> Do we want all targets to have one or the other?I'd say every target needs one of the __ORDER_*_ENDIAN__ options, yes, and also __BYTE_ORDER__ set. cheers, --renato
Joerg Sonnenberger
2014-Feb-26 21:46 UTC
[LLVMdev] test-suite wrongly using big-endian results
On Wed, Feb 26, 2014 at 06:00:26PM +0000, Robert Lytton wrote:> The question is are __LITTLE_ENDIAN__ & __BIG_ENDIAN__ defines becoming 'standard'? > Do we want all targets to have one or the other? > Still waiting for feedback...At the very least NetBSD always wants to have __LITTLE_ENDIAN__ / __BIG_ENDIAN__ as they are much older than __ORDER_*. Joerg