================
@@ -424,6 +432,94 @@ void validateAccessChain(const SPIRVSubtarget &STI,
MachineRegisterInfo *MRI,
}
}
+inline void validateVec1Ops(const SPIRVSubtarget &STI, MachineRegisterInfo
*MRI,
+ SPIRVGlobalRegistry &GR, MachineInstr &MI) {
+ // IRTranslator does not believe that rank-1 vectors exist, unlike upstream
+ // LLVM which happily creates <1 x T> vectors. This leads to operations over
+ // <1 x T> vectors getting translated as their scalar counterparts, which is
+ // wrong if we used SPV_EXT_long_vector we to preserve the actual
vector-ness.
----------------
jmmartinez wrote:
It looks like the last part of the comment is missing a verb:
_which is wrong if we used SPV_EXT_long_vector we to preserve the actual
vector-ness._
https://github.com/llvm/llvm-project/pull/210279
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
