[clang] [llvm] [ARM64EC] Implement `-mlong-double-{64,80,128}` options (PR #237057)
LIU Hao via cfe-commits
cfe-commits at lists.llvm.org
Sun Nov 1 11:33:42 PST 2026
lhmouse wrote:
The changes to ARM64EC backend have been pushed.
This effects a change to an x86-32 test: The 10-byte `long double` is stored in a 12-byte slot, which takes up 2.5 integer registers. Previously, the compiler generated `movzx` for the last load, to zero out the padding bytes. `DAGTypeLegalizer::SoftenFloatRes_BITCAST()` now uses `DAG.getAnyExtOrTrunc()` so the zero-extension is eliminated; the compiler now generates three full-integer loads.
https://github.com/llvm/llvm-project/pull/237057
More information about the cfe-commits
mailing list