[llvm] [llvm][InstCombine] bitcast bfloat half castpair bug (PR #79832)

[via llvm-commits]

================
@@ -0,0 +1,70 @@
+; RUN: opt -S %s | FileCheck %s
+
+define double @F0([2 x bfloat] %P0) {
----------------
ostannard wrote:

Are the arrays relevant to this test case, or could we use a scalar instead?

https://github.com/llvm/llvm-project/pull/79832