Hi, The attached patches change Clang's CodeGen for ARM NEON intrinsics to use more normal LLVM operations (and modular intrinsics where necessary) instead of a single special-purpose one.
The hope is that if LLVM can understand the operations being performed, it can optimise them, at least in some cases. It will also allow us to tidy up the ARM backend slightly by removing the original intrinsics entirely. All of the features being altered are already tested (albeit under "REQUIRES: long_tests") so I haven't added any tests here. Is it OK to commit? Cheers. Tim.
0001-ARM-use-vqdmull-and-vqadds-vqsubs-to-implement-vqdml.patch
Description: Binary data
0002-ARM-Emit-normal-IR-for-vaddhn-vsubhn-NEON-intrinsics.patch
Description: Binary data
0003-ARM-use-natural-IR-for-vmull-operations.patch
Description: Binary data
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
