Yup, GCC is "fast enough" again. Thanks :)
On Tue, Oct 11, 2016 at 9:13 PM, Mehdi Amini wrote:
> Reverted in r283920, can you check if it is enough to “fix” the GCC issue?
>
>> On Oct 11, 2016, at 12:04 PM, Benjamin Kramer wrote:
>>
>> Committing this patch before the constexpr change seems back
ber 11, 2016 12:14
> To: Benjamin Kramer
> Cc: cfe-commits
> Subject: Re: r283802 - Change Builtins name to be stored as StringRef
> instead of raw pointers (NFC)
>
> Reverted in r283920, can you check if it is enough to “fix” the GCC
> issue?
>
> > On Oct 11, 2
Reverted in r283920, can you check if it is enough to “fix” the GCC issue?
> On Oct 11, 2016, at 12:04 PM, Benjamin Kramer wrote:
>
> Committing this patch before the constexpr change seems backwards
> then? The static initializers are already breaking stuff because it
> takes GCC with optimizat
Yes, do you have a specific concern for this table size in particular?
> On Oct 11, 2016, at 12:07 PM, Craig Topper wrote:
>
> But this also increases the size of the builtin table too right? Since
> StringRef is twice the size of a pointer.
>
> ~Craig
>
> On Tue, Oct 11, 2016 at 11:40 AM, Me
> On Oct 11, 2016, at 12:04 PM, Benjamin Kramer wrote:
>
> Committing this patch before the constexpr change seems backwards
> then?
Well not really because I need to make StringRef(const char *) explicit, so all
the others have to go first.
> The static initializers are already breaking stuf
But this also increases the size of the builtin table too right? Since
StringRef is twice the size of a pointer.
~Craig
On Tue, Oct 11, 2016 at 11:40 AM, Mehdi Amini via cfe-commits <
cfe-commits@lists.llvm.org> wrote:
> This is temporary: the last patch of my series of patches adds the
> conste
Committing this patch before the constexpr change seems backwards
then? The static initializers are already breaking stuff because it
takes GCC with optimization and debug info takes 10+ minutes to
generate megabytes of static initializer code in Targets.cpp. Can you
please revert this until the co
This is temporary: the last patch of my series of patches adds the constexpr
ctor and remove all these static initializers.
> On Oct 11, 2016, at 11:26 AM, Benjamin Kramer wrote:
>
> I don't think this change is worth it. We create huge static arrays
> with Builtin::Info in Builtins.cpp and Tar
I don't think this change is worth it. We create huge static arrays
with Builtin::Info in Builtins.cpp and Targets.cpp, StringRef(const
char*) is not constexpr (because of strlen). This means you'll get a
huge generated initialization function for it. We want to reduce the
number of global initiali
