[clang] [clang] Implement constexpr bit_cast for vectors (PR #66894)

[via cfe-commits]

================
@@ -7304,6 +7382,21 @@ class BufferToAPValueConverter {
     return ArrayValue;
   }
 
+  std::optional<APValue> visit(const VectorType *Ty, CharUnits Offset) {
+    SmallVector<uint8_t, 8> Bytes;
+    if (!Buffer.readObject(Offset, Info.Ctx.getTypeSizeInChars(Ty), Bytes))
+      return std::nullopt;
----------------
DaMatrix wrote:

I've nearly got it working, except now we have another fun thing: pointing the `CK_BitCast` at my new implementation which tracks uninitialized padding bits prevents that test case in `const-init.c` from being constant-evaluated, so it proceeds to generate bitcode for it instead. Unfortunately, this doesn't work either because some bit casting code in LLVM itself seems to assume that a vector of `long double` is tightly packed, triggering this assertion:  
https://github.com/llvm/llvm-project/blob/8768741800aae37a825864e2ee782484ed073ce9/llvm/lib/IR/Instructions.cpp#L3299-L3302

This can actually be reproduced anywhere a vector is bit_casted to or from a vector of long double. [Example](https://godbolt.org/#g:!((g:!((g:!((h:codeEditor,i:(filename:'1',fontScale:14,fontUsePx:'0',j:1,lang:c%2B%2B,selection:(endColumn:1,endLineNumber:3,positionColumn:1,positionLineNumber:3,selectionStartColumn:1,selectionStartLineNumber:3,startColumn:1,startLineNumber:3),source:'using+fp80_vec+%3D+long+double+__attribute__((vector_size(32)))%3B%0Ausing+int16_vec+%3D+short+__attribute__((vector_size(32)))%3B%0A%0Afp80_vec+bit_cast_vec(int16_vec+vec)+%7B%0A++++return+(fp80_vec)vec%3B%0A%7D%0A'),l:'5',n:'0',o:'C%2B%2B+source+%231',t:'0')),k:37.545691906005224,l:'4',n:'0',o:'',s:0,t:'0'),(g:!((g:!((h:compiler,i:(compiler:clang170assert,deviceViewOpen:'1',filters:(b:'0',binary:'1',binaryObject:'1',commentOnly:'0',debugCalls:'1',demangle:'0',directives:'0',execute:'1',intel:'0',libraryCode:'0',trim:'1'),flagsViewOpen:'1',fontScale:14,fontUsePx:'0',j:1,lang:c%2B%2B,libs:!(),options:'-emit-llvm+-march%3Dskylake',overrides:!(),selection:(endColumn:1,endLineNumber:1,positionColumn:1,positionLineNumber:1,selectionStartColumn:1,selectionStartLineNumber:1,startColumn:1,startLineNumber:1),source:1),l:'5',n:'0',o:'+x86-64+clang+17.0.1+(assertions)+(Editor+%231)',t:'0')),k:50,l:'4',m:31.542857142857144,n:'0',o:'',s:0,t:'0'),(g:!((h:output,i:(compilerName:'x86-64+clang+17.0.1+(assertions)',editorid:1,fontScale:14,fontUsePx:'0',j:1,wrap:'1'),l:'5',n:'0',o:'Output+of+x86-64+clang+17.0.1+(assertions)+(Compiler+%231)',t:'0')),header:(),l:'4',m:68.45714285714286,n:'0',o:'',s:0,t:'0')),k:62.45430809399477,l:'3',n:'0',o:'',t:'0')),l:'2',n:'0',o:'',t:'0')),version:4)

This obviously needs to be fixed as well; should I do that here or make a separate PR for that?

https://github.com/llvm/llvm-project/pull/66894