[PATCH] Tablegen: Extend target register information by giving the users access to call preserved register mask names.

[Alex L]

This makes sense, I squashed this patch together with D10673 (MIR
Serialization: Serialize register mask machine operands.).

Cheers,
Alex

2015-06-23 12:12 GMT-07:00 Duncan P. N. Exon Smith <dexonsmith at apple.com>:

>
> > On 2015-Jun-22, at 14:30, Alex Lorenz <arphaman at gmail.com> wrote:
> >
> > This is what tablegen would produce for the X86 TRI:
> >
> >  ArrayRef<const uint32_t *> X86GenRegisterInfo::getRegMasks() const {
> >    static const uint32_t *Masks[] = {
> >      CSR_32_RegMask,
> >      CSR_32EHRet_RegMask,
> >      CSR_64_RegMask,
> >      CSR_64EHRet_RegMask,
> >      CSR_64_AllRegs_RegMask,
> >      CSR_64_AllRegs_AVX_RegMask,
> >      CSR_64_Intel_OCL_BI_RegMask,
> >      CSR_64_Intel_OCL_BI_AVX_RegMask,
> >      CSR_64_Intel_OCL_BI_AVX512_RegMask,
> >      CSR_64_MostRegs_RegMask,
> >      CSR_64_RT_AllRegs_RegMask,
> >      CSR_64_RT_AllRegs_AVX_RegMask,
> >      CSR_64_RT_MostRegs_RegMask,
> >      CSR_NoRegs_RegMask,
> >      CSR_Win64_RegMask,
> >      CSR_Win64_Intel_OCL_BI_AVX_RegMask,
> >      CSR_Win64_Intel_OCL_BI_AVX512_RegMask,
> >      nullptr
> >    };
> >    return ArrayRef<const uint32_t *>(Masks, (size_t)17);
> >  }
> >
> >  ArrayRef<const char *> X86GenRegisterInfo::getRegMaskNames() const {
> >    static const char *Names[] = {
> >      "CSR_32",
> >      "CSR_32EHRet",
> >      "CSR_64",
> >      "CSR_64EHRet",
> >      "CSR_64_AllRegs",
> >      "CSR_64_AllRegs_AVX",
> >      "CSR_64_Intel_OCL_BI",
> >      "CSR_64_Intel_OCL_BI_AVX",
> >      "CSR_64_Intel_OCL_BI_AVX512",
> >      "CSR_64_MostRegs",
> >      "CSR_64_RT_AllRegs",
> >      "CSR_64_RT_AllRegs_AVX",
> >      "CSR_64_RT_MostRegs",
> >      "CSR_NoRegs",
> >      "CSR_Win64",
> >      "CSR_Win64_Intel_OCL_BI_AVX",
> >      "CSR_Win64_Intel_OCL_BI_AVX512",
> >      nullptr
> >    };
> >    return ArrayRef<const char *>(Names, (size_t)17);
> >  }
> >
> >
> > REPOSITORY
> >  rL LLVM
> >
> > http://reviews.llvm.org/D10615
>
> This introduces code without using or testing it.  I think it'd be
> better to squash it with its first use.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150623/d96d2524/attachment.html>