================ @@ -319,10 +319,9 @@ define void @sincos_v3f32(<3 x float> %x, ptr addrspace(1) nocapture writeonly % ; CHECK-NEXT: entry: ; CHECK-NEXT: [[__SINCOS_:%.*]] = alloca <3 x float>, align 16, addrspace(5) ; CHECK-NEXT: [[TMP0:%.*]] = call contract <3 x float> @_Z6sincosDv3_fPU3AS5S_(<3 x float> [[X]], ptr addrspace(5) [[__SINCOS_]]) -; CHECK-NEXT: [[TMP1:%.*]] = load <3 x float>, ptr addrspace(5) [[__SINCOS_]], align 16 +; CHECK-NEXT: [[EXTRACTVEC6:%.*]] = load <4 x float>, ptr addrspace(5) [[__SINCOS_]], align 16 ---------------- vtjnash wrote:
n.b. this odd looking optimization is caused by the more accurate computation of dereferenceable bytes in llvm/lib/IR/Value.cpp (which uses the full alloca size, including alignment, instead of just the store size) https://github.com/llvm/llvm-project/pull/176486 _______________________________________________ lldb-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
