Fix double => half cast on clang (tinygrad#8265)

test/test_dtype.py

@@ -52,9 +52,6 @@ def _test_cast(a:Tensor, target_dtype:DType):
   if target_dtype == dtypes.half and Device.DEFAULT == "PYTHON":
     # TODO: struct.pack cannot pack value > 65504 (max of half) into e format
     a = (a > 65504).where(65504, a)
-  if CI and Device.DEFAULT == "CLANG" and (target_dtype, a.dtype) in [(dtypes.double, dtypes.half), (dtypes.half, dtypes.double)]:
-    # TODO: cast between double and half are broken https://github.com/tinygrad/tinygrad/issues/4084
-    return
 
   _test_op(lambda: a.cast(target_dtype), target_dtype, list(a.numpy().astype(_to_np_dtype(target_dtype))))
 def _test_bitcast(a:Tensor, target_dtype:DType, target=None):

tinygrad/renderer/cstyle.py

@@ -172,6 +172,9 @@ class ClangRenderer(CStyleLanguage):
   type_map = {dtypes.bool:"_Bool", dtypes.half:"__fp16"}
   code_for_op = {**({k:v for k,v in CStyleLanguage.code_for_op.items() if k not in [Ops.EXP2, Ops.SIN, Ops.LOG2]}),
                  Ops.SQRT: lambda x,dtype: f"__builtin_sqrt({x})" if dtype == dtypes.float64 else f"__builtin_sqrtf({x})"}
+  # LLVM legalizes double => half cast on systems that don't support it natively (like x86 cpus without AVX512-FP16) into a compiler-rt libcall.
+  extra_matcher = PatternMatcher([(UPat.var("x", dtypes.float64).cast(dtypes.float16), lambda x: x.cast(dtypes.float32).cast(dtypes.float16))]) + \
+    CStyleLanguage.extra_matcher
 
   if AMX:
     tensor_cores = [TensorCore(dims=(sz,sz,1), threads=[], reduce_axes=[], upcast_axes=([(1,sz)],[(0,sz)],[(1,sz),(0,sz)]), dtype_in=dt, dtype_out=dt)