Update handling of -cl-denorms-are-zero

Replaced "-cl-denorms-are-zero" with "-fdenormal-fp-math=preserve-sign"
in accordance with llvm/llvm-project@a4451d88A "Consolidate internal denormal
flushing controls"

Signed-off-by: Alexey Sachkov <[email protected]>

Author: AlexeySachkov

options_compile.cpp

@@ -75,14 +75,16 @@ std::string EffectiveOptionsFilter::processOptions(const OpenCLArgList &args,
     case OPT_COMPILE_I:
     case OPT_COMPILE_Werror:
     case OPT_COMPILE_cl_single_precision_constant:
-    case OPT_COMPILE_cl_denorms_are_zero:
     case OPT_COMPILE_cl_fp32_correctly_rounded_divide_sqrt:
     case OPT_COMPILE_cl_opt_disable:
     case OPT_COMPILE_cl_mad_enable:
     case OPT_COMPILE_cl_no_signed_zeros:
     case OPT_COMPILE_cl_unsafe_math_optimizations:
       effectiveArgs.push_back((*it)->getAsString(args));
       break;
+    case OPT_COMPILE_cl_denorms_are_zero:
+      effectiveArgs.push_back("-fdenormal-fp-math=preserve-sign");
+      break;
     case OPT_COMPILE_s: {
       std::string newSourceName = (*it)->getValue();
       if (!newSourceName.empty()) {