clang-format

include/ck/tensor_operation/gpu/block/blockwise_gemm_wmma.hpp

@@ -216,22 +216,15 @@ struct BlockwiseGemmWMMA
         return make_naive_tensor_descriptor(
             //        |MRepeat           |MWave |MSubGroup |NRepeat           |NWave
             //        |NThreadPerSubGroup |MAccVgprs
-            make_tuple(Number<MRepeat>{}, 
-                       I1,
-                       I1,
-                       Number<NRepeat>{},
-                       I1,
-                       I1,
-                       MAccVgprs),
+            make_tuple(Number<MRepeat>{}, I1, I1, Number<NRepeat>{}, I1, I1, MAccVgprs),
             make_tuple(Number<NRepeat>{} * MAccVgprs * AccStride,
                        Number<NRepeat>{} * MAccVgprs * AccStride,
                        Number<NRepeat>{} * MAccVgprs * AccStride,
                        MAccVgprs * AccStride,
                        MAccVgprs * AccStride,
                        MAccVgprs * AccStride,
-                       AccStride)
-            );
-        #if 0
+                       AccStride));
+#if 0
         return make_naive_tensor_descriptor_packed(
             //        |MRepeat           |MWave |MSubGroup |NRepeat           |NWave
             //        |NThreadPerSubGroup |MAccVgprs
@@ -242,7 +235,7 @@ struct BlockwiseGemmWMMA
                        I1,
                        NThreadPerSubGroup,
                        MAccVgprs));
-        #endif
+#endif
     }
 
     template <typename CGridDesc_M_N>

include/ck/tensor_operation/gpu/device/impl/device_batched_gemm_softmax_gemm_permute_wmma_cshuffle.hpp

@@ -151,9 +151,9 @@ struct DeviceBatchedGemmSoftmaxGemmPermute_Wmma_CShuffle
     static constexpr auto B0EnableLds_manu = true;
     static constexpr auto B1EnableLds_manu = true;
 
-    static constexpr auto AEnableLds  = AEnableLds_auto || AEnableLds_manu || (NumPrefetch >1);
-    static constexpr auto B0EnableLds = B0EnableLds_auto || B0EnableLds_manu || (NumPrefetch >1);
-    static constexpr auto B1EnableLds = B1EnableLds_auto || B1EnableLds_manu || (NumPrefetch >1);
+    static constexpr auto AEnableLds  = AEnableLds_auto || AEnableLds_manu || (NumPrefetch > 1);
+    static constexpr auto B0EnableLds = B0EnableLds_auto || B0EnableLds_manu || (NumPrefetch > 1);
+    static constexpr auto B1EnableLds = B1EnableLds_auto || B1EnableLds_manu || (NumPrefetch > 1);
 
     using Transform = TransformBatchedContractionContractionToBatchedGemmGemm<
         Sequence<NumDimG, NumDimM, NumDimL, NumDimK, NumDimN>,

include/ck/tensor_operation/gpu/device/impl/device_gemm_wmma.hpp

@@ -94,8 +94,8 @@ struct DeviceGemmWmma_CShuffle : public DeviceGemm<ALayout,
     static constexpr auto AEnableLds_manu = false;
     static constexpr auto BEnableLds_manu = false;
 
-    static constexpr auto AEnableLds = AEnableLds_auto || AEnableLds_manu || (NumPrefetch>1);
-    static constexpr auto BEnableLds = BEnableLds_auto || BEnableLds_manu || (NumPrefetch>1);
+    static constexpr auto AEnableLds = AEnableLds_auto || AEnableLds_manu || (NumPrefetch > 1);
+    static constexpr auto BEnableLds = BEnableLds_auto || BEnableLds_manu || (NumPrefetch > 1);
 
     static constexpr auto matrix_padder =
         MatrixPadder<GemmSpec, index_t, index_t, index_t>{MPerBlock, NPerBlock, KPerBlock};
@@ -467,7 +467,8 @@ struct DeviceGemmWmma_CShuffle : public DeviceGemm<ALayout,
         if(ck::get_device_name() == "gfx1100" || ck::get_device_name() == "gfx1101" ||
            ck::get_device_name() == "gfx1102")
         {
-            if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, ck::half_t> || is_same_v<AccDataType, int32_t>))
+            if constexpr(!(is_same_v<AccDataType, float> || is_same_v<AccDataType, ck::half_t> ||
+                           is_same_v<AccDataType, int32_t>))
             {
                 printf("DeviceOp err: AccDataType");
                 return false;

include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_d_wmma_cshuffle.hpp

@@ -177,8 +177,10 @@ struct DeviceGroupedConvFwdMultipleD_Wmma_CShuffle
     static constexpr auto AEnableLds_manu = false;
     static constexpr auto BEnableLds_manu = false;
 
-    static constexpr auto AEnableLds = AEnableLds_auto || AEnableLds_manu || (NumGemmKPrefetchStage > 1);
-    static constexpr auto BEnableLds = BEnableLds_auto || BEnableLds_manu || (NumGemmKPrefetchStage > 1);
+    static constexpr auto AEnableLds =
+        AEnableLds_auto || AEnableLds_manu || (NumGemmKPrefetchStage > 1);
+    static constexpr auto BEnableLds =
+        BEnableLds_auto || BEnableLds_manu || (NumGemmKPrefetchStage > 1);
 
     static constexpr auto conv_to_gemm_transformer =
         TransformConvFwdToGemm<NDimSpatial, ConvForwardSpecialization>{};

include/ck/tensor_operation/gpu/warp/wmma_gemm.hpp

@@ -104,11 +104,7 @@ struct wmma_type<WmmaInstr::wmma_f32_16x16x16_f16,
         m_per_wmma * n_per_wmma * acc_data_size * acc_pack_number / wave_size / 4;
     static constexpr index_t num_subgroups = wave_size / num_thread_per_subgroups;
 
-    template <index_t MPerWmma,
-              index_t NPerWmma,
-              class FloatA,
-              class FloatB,
-              class FloatC>
+    template <index_t MPerWmma, index_t NPerWmma, class FloatA, class FloatB, class FloatC>
     __device__ void run(const FloatA& a, const FloatB& b, FloatC& reg_c) const
     {
         if constexpr(wave_size == 32)
@@ -142,7 +138,7 @@ struct wmma_type<WmmaInstr::wmma_f32_16x16x16_bf16,
     static constexpr index_t num_src_a_vgprs_per_wave = m_per_wmma * src_a_data_size / 4;
     static constexpr index_t num_src_b_vgprs_per_wave = n_per_wmma * src_b_data_size / 4;
     static constexpr index_t num_acc_vgprs_per_wave =
-        m_per_wmma * n_per_wmma * acc_data_size *acc_pack_number/ wave_size / 4;
+        m_per_wmma * n_per_wmma * acc_data_size * acc_pack_number / wave_size / 4;
     static constexpr index_t num_subgroups = wave_size / num_thread_per_subgroups;
 
     template <index_t MPerWmma, index_t NPerWmma, class FloatA, class FloatB, class FloatC>
@@ -182,11 +178,7 @@ struct wmma_type<WmmaInstr::wmma_f16_16x16x16_f16,
         m_per_wmma * n_per_wmma * acc_data_size * acc_pack_number / wave_size / 4;
     static constexpr index_t num_subgroups = wave_size / num_thread_per_subgroups;
 
-    template <index_t MPerWmma,
-              index_t NPerWmma,
-              class FloatA,
-              class FloatB,
-              class FloatC>
+    template <index_t MPerWmma, index_t NPerWmma, class FloatA, class FloatB, class FloatC>
     __device__ void run(const FloatA& a, const FloatB& b, FloatC& reg_c) const
     {
         if constexpr(wave_size == 32)
@@ -261,7 +253,7 @@ struct wmma_type<WmmaInstr::wmma_i32_16x16x16_iu8,
     static constexpr index_t num_src_a_vgprs_per_wave = m_per_wmma * src_a_data_size / 4;
     static constexpr index_t num_src_b_vgprs_per_wave = n_per_wmma * src_b_data_size / 4;
     static constexpr index_t num_acc_vgprs_per_wave =
-        m_per_wmma * n_per_wmma * acc_data_size *acc_pack_number / wave_size / 4;
+        m_per_wmma * n_per_wmma * acc_data_size * acc_pack_number / wave_size / 4;
     static constexpr index_t num_subgroups = wave_size / num_thread_per_subgroups;
 
     template <index_t MPerWmma,
@@ -496,13 +488,11 @@ struct WmmaGemm
             "(int8, int32) or (int4, int32)!");
         if constexpr(!TransposeC)
         {
-            wmma_instr.template run<MPerWmma, NPerWmma>(
-                p_a_wave, p_b_wave, p_c_thread);
+            wmma_instr.template run<MPerWmma, NPerWmma>(p_a_wave, p_b_wave, p_c_thread);
         }
         else
         {
-            wmma_instr.template run<MPerWmma, NPerWmma>(
-                p_b_wave, p_a_wave, p_c_thread);
+            wmma_instr.template run<MPerWmma, NPerWmma>(p_b_wave, p_a_wave, p_c_thread);
         }
     }
 
@@ -555,7 +545,10 @@ struct WmmaGemm
     __host__ __device__ static constexpr auto
     GetCMSubGroupNThreadPerSubGroupMAccVgprsThreadBlkLengths()
     {
-        return make_tuple(I1, I1, Number<wmma_instr.num_acc_vgprs_per_wave>{}, Number<wmma_instr.acc_pack_number>{});
+        return make_tuple(I1,
+                          I1,
+                          Number<wmma_instr.num_acc_vgprs_per_wave>{},
+                          Number<wmma_instr.acc_pack_number>{});
     }
 };