[PATCH] D48937: [TableGen] Increase the number of supported decoder fix-ups.

[Dave Green via Phabricator via llvm-commits]

dmgreen accepted this revision.
dmgreen added a comment.
This revision is now accepted and ready to land.

Thanks Sander!

For context, we (in a slightly different part of Arm), started hitting this with the latest SVE patches combined with our downstream changes. I believe upstream llvm will start hitting the same problem once more 8.X/SVE patches are in.

I tested this out and it fixes our build, but will increase the size of the decode table. Looking through, I think this looks OK. Just some comments to update.



================
Comment at: utils/TableGen/FixedLenDecoderEmitter.cpp:751
 
       // 16-bit numtoskip value.
       uint8_t Byte = *I++;
----------------
24-bit!


================
Comment at: utils/TableGen/FixedLenDecoderEmitter.cpp:774
       OS << (unsigned)*I++ << ", ";
       // 16-bit numtoskip value.
       uint8_t Byte = *I++;
----------------
24-bit!


================
Comment at: utils/TableGen/FixedLenDecoderEmitter.cpp:794
 
       // 16-bit numtoskip value.
       uint8_t Byte = *I++;
----------------
24-bit!


================
Comment at: utils/TableGen/FixedLenDecoderEmitter.cpp:837
 
       // 16-bit numtoskip value.
       uint8_t Byte = *I++;
----------------
24-bit!


================
Comment at: utils/TableGen/FixedLenDecoderEmitter.cpp:2209
      << "      Ptr += Len;\n"
      << "      // NumToSkip is a plain 16-bit integer.\n"
      << "      unsigned NumToSkip = *Ptr++;\n"
----------------
24-bit!


https://reviews.llvm.org/D48937