Update the CK_WORKAROUND_DENORM_FIX flag handling

include/ck/ck.hpp

@@ -171,7 +171,10 @@
 // denorm test fix, required to work around dissue
 #ifndef CK_WORKAROUND_DENORM_FIX
 #define CK_WORKAROUND_DENORM_FIX 0
-#endif
+#elif
+// enable only on MI200
+#define CK_WORKAROUND_DENORM_FIX = CK_WORKAROUND_DENORM_FIX && defined(__gfx90a__)
+#endif // CK_WORKAROUND_DENORM_FIX
 
 namespace ck {
 

include/ck/tensor_operation/gpu/grid/gridwise_gemm_multiple_d_xdl_cshuffle.hpp

@@ -96,7 +96,7 @@ struct GridwiseGemmMultipleD_xdl_cshuffle
     // we convert fp16->fp32->bf16 and execute bf16 mfma instruction
     // when mfma if fixed, remove this section and update
     // ABDataTypeAdjusted -> ABDataType throughout this file
-#if CK_WORKAROUND_DENORM_FIX && defined(__gfx90a__)
+#if CK_WORKAROUND_DENORM_FIX
     using ABDataTypeAdjusted =
         conditional_t<is_same_v<ABDataType, ck::half_t>, ck::bhalf_t, ABDataType>;
 #else

include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_bwd_weight.hpp

@@ -265,7 +265,7 @@ struct GridwiseGemm_bk0mk1_bk0nk1_mn_xdlops_bwd_weight
     // we convert fp16->fp32->bf16 and execute bf16 mfma instruction
     // when mfma if fixed, remove this section and update
     // FloatABAdjusted -> FloatAB throughout this file
-#if CK_WORKAROUND_DENORM_FIX && defined(__gfx90a__)
+#if CK_WORKAROUND_DENORM_FIX
     using FloatABAdjusted = conditional_t<is_same_v<FloatAB, ck::half_t>, ck::bhalf_t, FloatAB>;
 #else
     using FloatABAdjusted = FloatAB;

include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r3.hpp

@@ -135,7 +135,7 @@ struct GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3
     // we convert fp16->fp32->bf16 and execute bf16 mfma instruction
     // when mfma if fixed, remove this section and update
     // FloatABAdjusted -> FloatAB throughout this file
-#if CK_WORKAROUND_DENORM_FIX && defined(__gfx90a__)
+#if CK_WORKAROUND_DENORM_FIX
     using FloatABAdjusted = conditional_t<is_same_v<FloatAB, ck::half_t>, ck::bhalf_t, FloatAB>;
 #else
     using FloatABAdjusted = FloatAB;