================
@@ -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

Reply via email to