R devel - Mar 2015 - Notes on building a gcc toolchain for Rtools (but not multilib)

On Tue, Mar 10, 2015 at 3:17 PM, Duncan Murdoch
<murdoch.duncan at gmail.com> wrote:
>
> That's a bug.  I haven't tracked down what's going wrong with
our
> regular code.  If I can't find that and fix it soon, I'll make
Internet2
> the default.  For now, you can do it yourself using the instructions on
> ?setInternet2.
>
> Duncan Murdoch
I successfully rebuilt R-devel_2015-03-09 with the most recent version
of Rtools tonight, building both ICU_531 and this time libcurl (7.39)
as well (and OpenBLAS, of course). The internet bug is still there,
but the rest of make-check all passed with flying colors, as did
building 'microbenchmark' from source (with all the other needed
packages, including Rcpp and Hsiu-Kheurn's change to NM was *not*
used). My non-BLAS test tonight ran faster than last night; maybe 1000
iterations aren't enough or I had something else eating up clock
cycles last night. Either way, outside the internet bug, it's looking
good for Windows 64bit (Win7 at least).

Thanks,

Avi

On Tue, Mar 10, 2015 at 8:54 PM, Avraham Adler <avraham.adler at
gmail.com> wrote:
> On Tue, Mar 10, 2015 at 3:17 PM, Duncan Murdoch
> <murdoch.duncan at gmail.com> wrote:
>>
>> That's a bug.  I haven't tracked down what's going wrong
with our
>> regular code.  If I can't find that and fix it soon, I'll make
Internet2
>> the default.  For now, you can do it yourself using the instructions on
>> ?setInternet2.
>>
>> Duncan Murdoch
>
> I successfully rebuilt R-devel_2015-03-09 with the most recent version
> of Rtools tonight, building both ICU_531 and this time libcurl (7.39)
> as well (and OpenBLAS, of course). The internet bug is still there,
> but the rest of make-check all passed with flying colors, as did
> building 'microbenchmark' from source (with all the other needed
> packages, including Rcpp and Hsiu-Kheurn's change to NM was *not*
> used).
> ...
The NM_FILTER change seems to be needed only when compiling C++ code
with -std=c++11 or -std=gnu++11.

Even though the current CRAN policy doesn't allow the C++11 standard,
it would still be useful to document this, perhaps in the Writing R
Extensions manual.

- Hsiu-Khuern

On Wed, Mar 11, 2015 at 1:40 AM, Hsiu-Khuern Tang <tangoh at gmail.com>
wrote:
> On Tue, Mar 10, 2015 at 8:54 PM, Avraham Adler <avraham.adler at
gmail.com> wrote:
>>
>> I successfully rebuilt R-devel_2015-03-09 with the most recent version
>> of Rtools tonight, building both ICU_531 and this time libcurl (7.39)
>> as well (and OpenBLAS, of course). The internet bug is still there,
>> but the rest of make-check all passed with flying colors, as did
>> building 'microbenchmark' from source (with all the other
needed
>> packages, including Rcpp and Hsiu-Kheurn's change to NM was *not*
>> used).
>> ...
>
> The NM_FILTER change seems to be needed only when compiling C++ code
> with -std=c++11 or -std=gnu++11.
>
> Even though the current CRAN policy doesn't allow the C++11 standard,
> it would still be useful to document this, perhaps in the Writing R
> Extensions manual.
>
> - Hsiu-Khuern
Hello, Hsiu-Khuern.

All the times I have built, I've passed -std-gnu++11. Specifically, I
pass: -march=native -O3 -mfpmath=sse -std=gnu++11 -msse2avx
-mavx256-split-unaligned-load -mavx256-split-unaligned-store
-mvzeroupper --param l1-cache-line-size=64 --param l1-cache-size=64
--param l2-cache-size=256. I live with the repeated warnings that it
isn't needed for gcc, but I know it takes as I see the same string
(gnu++11) passed when I build Rcpp from inside an R session.

Avi