[Lldb-commits] [lldb] [lldb] Add support for displaying `__float128` variables (PR #98369)

[via lldb-commits]

================
@@ -4737,19 +4750,23 @@ CompilerType TypeSystemClang::CreateGenericFunctionPrototype() {
 // Exploring the type
 
 const llvm::fltSemantics &
-TypeSystemClang::GetFloatTypeSemantics(size_t byte_size) {
+TypeSystemClang::GetFloatTypeSemantics(size_t byte_size, bool prefer_float128) {
   clang::ASTContext &ast = getASTContext();
   const size_t bit_size = byte_size * 8;
   if (bit_size == ast.getTypeSize(ast.FloatTy))
     return ast.getFloatTypeSemantics(ast.FloatTy);
   else if (bit_size == ast.getTypeSize(ast.DoubleTy))
     return ast.getFloatTypeSemantics(ast.DoubleTy);
+  else if (prefer_float128 && bit_size == ast.getTypeSize(ast.Float128Ty))
+    return ast.getFloatTypeSemantics(ast.Float128Ty);
   else if (bit_size == ast.getTypeSize(ast.LongDoubleTy) ||
            bit_size == llvm::APFloat::semanticsSizeInBits(
                            ast.getFloatTypeSemantics(ast.LongDoubleTy)))
     return ast.getFloatTypeSemantics(ast.LongDoubleTy);
   else if (bit_size == ast.getTypeSize(ast.HalfTy))
     return ast.getFloatTypeSemantics(ast.HalfTy);
+  else if (bit_size == ast.getTypeSize(ast.Float128Ty))
----------------
beetrees wrote:

If the user formats a 128-bit value using `eFormatFloat` on a target where the only 128-bit float is float128 (a.k.a. `long double` is some other size), then the value should be formatted as a float128. The `eFormatFloat128` format is only needed to disambiguate when `long double` and float128 are both 128 bits.

https://github.com/llvm/llvm-project/pull/98369