Merge branch 'develop' into groupnorm-mix-precision

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

@@ -266,8 +266,8 @@ struct GridwiseGemm_k0mk1_k0nk1_mn_wmma
         // TODO: also check validity of all components (blockwise-copy, threadwise-copy, etc)
         constexpr long_index_t TwoGB = (long_index_t{1} << 31);
 
-        if(!(a_grid_desc_k0_m_k1.GetElementSpaceSize() * sizeof(ADataType) <= TwoGB &&
-             b_grid_desc_k0_n_k1.GetElementSpaceSize() * sizeof(BDataType) <= TwoGB))
+        if(!(a_grid_desc_k0_m_k1.GetElementSpaceSize() * sizeof(FloatA) <= TwoGB &&
+             b_grid_desc_k0_n_k1.GetElementSpaceSize() * sizeof(FloatB) <= TwoGB))
         {
             return false;
         }

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 defined(__gfx90a__)
+#if CK_WORKAROUND_DENORM_FIX && defined(__gfx90a__)
     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 defined(__gfx90a__)
+#if CK_WORKAROUND_DENORM_FIX && defined(__gfx90a__)
     using FloatABAdjusted = conditional_t<is_same_v<FloatAB, ck::half_t>, ck::bhalf_t, FloatAB>;
 #else
     using FloatABAdjusted = FloatAB;