Merge branch 'develop' into gemm_layernorm_welford

client_example/04_contraction/contraction_bilinear.cpp

@@ -12,6 +12,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/gpu/contraction_bilinear.hpp"
+#include "ck/library/utility/numeric.hpp"
 
 using F32 = float;
 
@@ -192,20 +193,14 @@ int main(int argc, char* argv[])
         {
             float ave_time = invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, true});
 
-            ck::index_t M = std::accumulate(e_ms_ns_lengths.begin(),
-                                            e_ms_ns_lengths.begin() + NumDimM,
-                                            ck::index_t{1},
-                                            std::multiplies<ck::index_t>{});
+            ck::index_t M = ck::accumulate_n<ck::index_t>(
+                e_ms_ns_lengths.begin(), NumDimM, 1, std::multiplies<>{});
 
-            ck::index_t N = std::accumulate(e_ms_ns_lengths.begin() + NumDimM,
-                                            e_ms_ns_lengths.begin() + NumDimM + NumDimN,
-                                            ck::index_t{1},
-                                            std::multiplies<ck::index_t>{});
+            ck::index_t N = ck::accumulate_n<ck::index_t>(
+                e_ms_ns_lengths.begin() + NumDimM, NumDimN, 1, std::multiplies<>{});
 
-            ck::index_t K = std::accumulate(a_ms_ks_lengths.begin() + NumDimM,
-                                            a_ms_ks_lengths.begin() + NumDimM + NumDimK,
-                                            ck::index_t{1},
-                                            std::multiplies<ck::index_t>{});
+            ck::index_t K = ck::accumulate_n<ck::index_t>(
+                a_ms_ks_lengths.begin() + NumDimM, NumDimK, 1, std::multiplies<>{});
 
             std::size_t flop      = std::size_t(2) * M * N * K;
             std::size_t num_btype = sizeof(ADataType) * M * K + sizeof(BDataType) * K * N +

client_example/04_contraction/contraction_scale.cpp

@@ -12,6 +12,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/tensor_operation_instance/gpu/contraction_scale.hpp"
+#include "ck/library/utility/numeric.hpp"
 
 using F32 = float;
 
@@ -178,20 +179,14 @@ int main(int argc, char* argv[])
         {
             float ave_time = invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, true});
 
-            ck::index_t M = std::accumulate(e_ms_ns_lengths.begin(),
-                                            e_ms_ns_lengths.begin() + NumDimM,
-                                            ck::index_t{1},
-                                            std::multiplies<ck::index_t>{});
+            ck::index_t M = ck::accumulate_n<ck::index_t>(
+                e_ms_ns_lengths.begin(), NumDimM, 1, std::multiplies<>{});
 
-            ck::index_t N = std::accumulate(e_ms_ns_lengths.begin() + NumDimM,
-                                            e_ms_ns_lengths.begin() + NumDimM + NumDimN,
-                                            ck::index_t{1},
-                                            std::multiplies<ck::index_t>{});
+            ck::index_t N = ck::accumulate_n<ck::index_t>(
+                e_ms_ns_lengths.begin() + NumDimM, NumDimN, 1, std::multiplies<>{});
 
-            ck::index_t K = std::accumulate(a_ms_ks_lengths.begin() + NumDimM,
-                                            a_ms_ks_lengths.begin() + NumDimM + NumDimK,
-                                            ck::index_t{1},
-                                            std::multiplies<ck::index_t>{});
+            ck::index_t K = ck::accumulate_n<ck::index_t>(
+                a_ms_ks_lengths.begin() + NumDimM, NumDimK, 1, std::multiplies<>{});
 
             std::size_t flop = std::size_t(2) * M * N * K;
             std::size_t num_btype =

example/09_convnd_fwd/convnd_fwd_dl_common.hpp

@@ -10,6 +10,7 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
+#include "ck/library/utility/algorithm.hpp"
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
@@ -88,7 +89,7 @@ bool run_grouped_conv_fwd_dl(bool do_verification,
     std::array<ck::index_t, NDimSpatial> input_left_pads{};
     std::array<ck::index_t, NDimSpatial> input_right_pads{};
 
-    auto copy = [](auto& x, auto& y) { std::copy(x.begin(), x.end(), y.begin()); };
+    auto copy = [](auto& x, auto& y) { ck::ranges::copy(x, y.begin()); };
 
     copy(in_g_n_c_wis_desc.GetLengths(), a_g_n_c_wis_lengths);
     copy(in_g_n_c_wis_desc.GetStrides(), a_g_n_c_wis_strides);

example/12_reduce/reduce_blockwise.cpp

@@ -142,7 +142,7 @@ bool reduce_blockwise_test(bool do_verification,
 
         std::array<int, ShapeType::NumReduceDim_> arrReduceDims;
 
-        std::copy(reduceDims.begin(), reduceDims.end(), arrReduceDims.begin());
+        ck::ranges::copy(reduceDims, arrReduceDims.begin());
 
         result = reduce_blockwise_impl<InOutDataType,
                                        AccDataType,

example/12_reduce/reduce_blockwise_impl.hpp

@@ -10,6 +10,7 @@
 #include "ck/tensor_operation/gpu/device/reduction_operator_mapping.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_reduce_multiblock.hpp"
 
+#include "ck/library/utility/algorithm.hpp"
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
@@ -263,10 +264,10 @@ int reduce_blockwise_impl(bool do_verification,
     std::array<index_t, NumOutDim> arrOutLengths;
     std::array<index_t, NumOutDim> arrOutStrides;
 
-    std::copy(inLengths.begin(), inLengths.end(), arrInLengths.begin());
-    std::copy(inStrides.begin(), inStrides.end(), arrInStrides.begin());
-    std::copy(outLengths.begin(), outLengths.end(), arrOutLengths.begin());
-    std::copy(outStrides.begin(), outStrides.end(), arrOutStrides.begin());
+    ck::ranges::copy(inLengths, arrInLengths.begin());
+    ck::ranges::copy(inStrides, arrInStrides.begin());
+    ck::ranges::copy(outLengths, arrOutLengths.begin());
+    ck::ranges::copy(outStrides, arrOutStrides.begin());
 
     auto reduce = DeviceReduceInstance{};
 

example/12_reduce/reduce_blockwise_two_call.cpp

@@ -221,12 +221,12 @@ int main(int argc, char* argv[])
     std::array<index_t, 3> arrOutLengths;
     std::array<index_t, 3> arrOutStrides;
 
-    std::copy(inLengths_1.begin(), inLengths_1.end(), arrInLengths_1.begin());
-    std::copy(inStrides_1.begin(), inStrides_1.end(), arrInStrides_1.begin());
-    std::copy(inLengths_2.begin(), inLengths_2.end(), arrInLengths_2.begin());
-    std::copy(inStrides_2.begin(), inStrides_2.end(), arrInStrides_2.begin());
-    std::copy(outLengths.begin(), outLengths.end(), arrOutLengths.begin());
-    std::copy(outStrides.begin(), outStrides.end(), arrOutStrides.begin());
+    ck::ranges::copy(inLengths_1, arrInLengths_1.begin());
+    ck::ranges::copy(inStrides_1, arrInStrides_1.begin());
+    ck::ranges::copy(inLengths_2, arrInLengths_2.begin());
+    ck::ranges::copy(inStrides_2, arrInStrides_2.begin());
+    ck::ranges::copy(outLengths, arrOutLengths.begin());
+    ck::ranges::copy(outStrides, arrOutStrides.begin());
 
     auto reduce_1 = DeviceReduceInstance_1{};
 

example/12_reduce/reduce_multiblock_atomic_add.cpp

@@ -140,7 +140,7 @@ bool reduce_multiblock_atomic_add_test(bool do_verification,
 
         std::array<int, ShapeType::NumReduceDim_> a_reduceDims;
 
-        std::copy(reduceDims.begin(), reduceDims.end(), a_reduceDims.begin());
+        ck::ranges::copy(reduceDims, a_reduceDims.begin());
 
         result = reduce_multiblock_atomic_add_impl<InOutDataType,
                                                    AccDataType,

example/12_reduce/reduce_multiblock_atomic_add_impl.hpp

@@ -10,6 +10,7 @@
 #include "ck/tensor_operation/gpu/device/reduction_operator_mapping.hpp"
 #include "ck/tensor_operation/gpu/device/impl/device_reduce_multiblock.hpp"
 
+#include "ck/library/utility/algorithm.hpp"
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
@@ -176,10 +177,10 @@ int reduce_multiblock_atomic_add_impl(bool do_verification,
     std::array<index_t, NumOutDim> arrOutLengths;
     std::array<index_t, NumOutDim> arrOutStrides;
 
-    std::copy(inLengths.begin(), inLengths.end(), arrInLengths.begin());
-    std::copy(inStrides.begin(), inStrides.end(), arrInStrides.begin());
-    std::copy(outLengths.begin(), outLengths.end(), arrOutLengths.begin());
-    std::copy(outStrides.begin(), outStrides.end(), arrOutStrides.begin());
+    ck::ranges::copy(inLengths, arrInLengths.begin());
+    ck::ranges::copy(inStrides, arrInStrides.begin());
+    ck::ranges::copy(outLengths, arrOutLengths.begin());
+    ck::ranges::copy(outStrides, arrOutStrides.begin());
 
     auto reduce = DeviceReduceInstance{};
 

example/25_gemm_bias_e_permute/gemm_bias_e_permute_g1m2n3k1_xdl_fp16.cpp

@@ -15,6 +15,7 @@
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 #include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/numeric.hpp"
 
 #include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
 
@@ -317,20 +318,14 @@ int main(int argc, char* argv[])
 
     float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
 
-    std::size_t M = std::accumulate(e_gs_ms_ns_lengths.begin() + NumDimG,
-                                    e_gs_ms_ns_lengths.begin() + NumDimG + NumDimM,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    std::size_t M = ck::accumulate_n<ck::index_t>(
+        e_gs_ms_ns_lengths.begin() + NumDimG, NumDimM, 1, std::multiplies<>{});
 
-    std::size_t N = std::accumulate(e_gs_ms_ns_lengths.begin() + NumDimG + NumDimM,
-                                    e_gs_ms_ns_lengths.begin() + NumDimG + NumDimM + NumDimN,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    std::size_t N = ck::accumulate_n<ck::index_t>(
+        e_gs_ms_ns_lengths.begin() + NumDimG + NumDimM, NumDimN, 1, std::multiplies<>{});
 
-    std::size_t K = std::accumulate(a_gs_ms_ks_lengths.begin() + NumDimG + NumDimM,
-                                    a_gs_ms_ks_lengths.begin() + NumDimG + NumDimM + NumDimK,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    std::size_t K = ck::accumulate_n<ck::index_t>(
+        a_gs_ms_ks_lengths.begin() + NumDimG + NumDimM, NumDimK, 1, std::multiplies<>{});
 
     std::size_t flop      = std::size_t(2) * M * N * K;
     std::size_t num_btype = sizeof(ADataType) * M * K + sizeof(BDataType) * K * N +

example/25_gemm_bias_e_permute/gemm_bias_e_permute_g1m3n2k1_xdl_fp16.cpp

@@ -15,6 +15,7 @@
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 #include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/numeric.hpp"
 
 template <ck::index_t... Is>
 using S = ck::Sequence<Is...>;
@@ -317,20 +318,14 @@ int main(int argc, char* argv[])
 
     float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
 
-    ck::index_t M = std::accumulate(e_gs_ms_ns_lengths.begin(),
-                                    e_gs_ms_ns_lengths.begin() + NumDimM,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    ck::index_t M =
+        ck::accumulate_n<ck::index_t>(e_gs_ms_ns_lengths.begin(), NumDimM, 1, std::multiplies<>{});
 
-    ck::index_t N = std::accumulate(e_gs_ms_ns_lengths.begin() + NumDimM,
-                                    e_gs_ms_ns_lengths.begin() + NumDimM + NumDimN,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    ck::index_t N = ck::accumulate_n<ck::index_t>(
+        e_gs_ms_ns_lengths.begin() + NumDimM, NumDimN, 1, std::multiplies<>{});
 
-    ck::index_t K = std::accumulate(a_gs_ms_ks_lengths.begin() + NumDimM,
-                                    a_gs_ms_ks_lengths.begin() + NumDimM + NumDimK,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    ck::index_t K = ck::accumulate_n<ck::index_t>(
+        a_gs_ms_ks_lengths.begin() + NumDimM, NumDimK, 1, std::multiplies<>{});
 
     std::size_t flop      = std::size_t(2) * M * N * K;
     std::size_t num_btype = sizeof(ADataType) * M * K + sizeof(BDataType) * K * N +

example/26_contraction/contraction_bilinear_xdl_fp32.cpp

@@ -15,6 +15,7 @@
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 #include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/numeric.hpp"
 
 template <ck::index_t... Is>
 using S = ck::Sequence<Is...>;
@@ -358,20 +359,14 @@ int main(int argc, char* argv[])
 
     float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
 
-    ck::index_t M = std::accumulate(e_ms_ns_lengths.begin(),
-                                    e_ms_ns_lengths.begin() + NumDimM,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    ck::index_t M =
+        ck::accumulate_n<ck::index_t>(e_ms_ns_lengths.begin(), NumDimM, 1, std::multiplies<>{});
 
-    ck::index_t N = std::accumulate(e_ms_ns_lengths.begin() + NumDimM,
-                                    e_ms_ns_lengths.begin() + NumDimM + NumDimN,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    ck::index_t N = ck::accumulate_n<ck::index_t>(
+        e_ms_ns_lengths.begin() + NumDimM, NumDimN, 1, std::multiplies<>{});
 
-    ck::index_t K = std::accumulate(a_ms_ks_lengths.begin() + NumDimM,
-                                    a_ms_ks_lengths.begin() + NumDimM + NumDimK,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    ck::index_t K = ck::accumulate_n<ck::index_t>(
+        a_ms_ks_lengths.begin() + NumDimM, NumDimK, 1, std::multiplies<>{});
 
     std::size_t flop      = std::size_t(2) * M * N * K;
     std::size_t num_btype = sizeof(ADataType) * M * K + sizeof(BDataType) * K * N +

example/26_contraction/contraction_scale_xdl_fp32.cpp

@@ -15,6 +15,7 @@
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 #include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/numeric.hpp"
 
 template <ck::index_t... Is>
 using S = ck::Sequence<Is...>;
@@ -341,20 +342,14 @@ int main(int argc, char* argv[])
 
     float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
 
-    ck::index_t M = std::accumulate(e_ms_ns_lengths.begin(),
-                                    e_ms_ns_lengths.begin() + NumDimM,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    ck::index_t M =
+        ck::accumulate_n<ck::index_t>(e_ms_ns_lengths.begin(), NumDimM, 1, std::multiplies<>{});
 
-    ck::index_t N = std::accumulate(e_ms_ns_lengths.begin() + NumDimM,
-                                    e_ms_ns_lengths.begin() + NumDimM + NumDimN,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    ck::index_t N = ck::accumulate_n<ck::index_t>(
+        e_ms_ns_lengths.begin() + NumDimM, NumDimN, 1, std::multiplies<>{});
 
-    ck::index_t K = std::accumulate(a_ms_ks_lengths.begin() + NumDimM,
-                                    a_ms_ks_lengths.begin() + NumDimM + NumDimK,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    ck::index_t K = ck::accumulate_n<ck::index_t>(
+        a_ms_ks_lengths.begin() + NumDimM, NumDimK, 1, std::multiplies<>{});
 
     std::size_t flop = std::size_t(2) * M * N * K;
     std::size_t num_btype =

example/28_grouped_gemm_bias_e_permute/grouped_gemm_bias_e_permute_xdl_fp16.cpp

@@ -16,6 +16,7 @@
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 #include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/numeric.hpp"
 
 template <ck::index_t... Is>
 using S = ck::Sequence<Is...>;
@@ -302,20 +303,14 @@ int main(int argc, char* argv[])
         Tensor<DDataType> d_ms_ns(d_ms_ns_lengths, d_ms_ns_strides);
         Tensor<EDataType> e_ms_ns_device_result(e_ms_ns_lengths, e_ms_ns_strides);
 
-        ck::index_t M_ = std::accumulate(e_ms_ns_lengths.begin(),
-                                         e_ms_ns_lengths.begin() + NumDimM,
-                                         ck::index_t{1},
-                                         std::multiplies<ck::index_t>{});
+        ck::index_t M_ =
+            ck::accumulate_n<ck::index_t>(e_ms_ns_lengths.begin(), NumDimM, 1, std::multiplies<>{});
 
-        ck::index_t N_ = std::accumulate(e_ms_ns_lengths.begin() + NumDimM,
-                                         e_ms_ns_lengths.begin() + NumDimM + NumDimN,
-                                         ck::index_t{1},
-                                         std::multiplies<ck::index_t>{});
+        ck::index_t N_ = ck::accumulate_n<ck::index_t>(
+            e_ms_ns_lengths.begin() + NumDimM, NumDimN, 1, std::multiplies<>{});
 
-        ck::index_t K_ = std::accumulate(a_ms_ks_lengths.begin() + NumDimM,
-                                         a_ms_ks_lengths.begin() + NumDimM + NumDimK,
-                                         ck::index_t{1},
-                                         std::multiplies<ck::index_t>{});
+        ck::index_t K_ = ck::accumulate_n<ck::index_t>(
+            a_ms_ks_lengths.begin() + NumDimM, NumDimK, 1, std::multiplies<>{});
 
         a_tensors.push_back(a_ms_ks);
         b_tensors.push_back(b_ns_ks);

example/29_batched_gemm_bias_e_permute/batched_gemm_bias_e_permute_xdl_fp16.cpp

@@ -15,6 +15,7 @@
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 #include "ck/library/utility/host_tensor_generator.hpp"
+#include "ck/library/utility/numeric.hpp"
 
 template <ck::index_t... Is>
 using S = ck::Sequence<Is...>;
@@ -317,25 +318,17 @@ int main(int argc, char* argv[])
 
     float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
 
-    ck::index_t G = std::accumulate(e_gs_ms_ns_lengths.begin(),
-                                    e_gs_ms_ns_lengths.begin() + NumDimG,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
-
-    ck::index_t M = std::accumulate(e_gs_ms_ns_lengths.begin() + NumDimG,
-                                    e_gs_ms_ns_lengths.begin() + NumDimG + NumDimM,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
-
-    ck::index_t N = std::accumulate(e_gs_ms_ns_lengths.begin() + NumDimG + NumDimM,
-                                    e_gs_ms_ns_lengths.begin() + NumDimG + NumDimM + NumDimN,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
-
-    ck::index_t K = std::accumulate(a_gs_ms_ks_lengths.begin() + NumDimG + NumDimM,
-                                    a_gs_ms_ks_lengths.begin() + NumDimG + NumDimM + NumDimK,
-                                    ck::index_t{1},
-                                    std::multiplies<ck::index_t>{});
+    ck::index_t G =
+        ck::accumulate_n<ck::index_t>(e_gs_ms_ns_lengths.begin(), NumDimG, 1, std::multiplies<>{});
+
+    ck::index_t M = ck::accumulate_n<ck::index_t>(
+        e_gs_ms_ns_lengths.begin() + NumDimG, NumDimM, 1, std::multiplies<>{});
+
+    ck::index_t N = ck::accumulate_n<ck::index_t>(
+        e_gs_ms_ns_lengths.begin() + NumDimG + NumDimM, NumDimN, 1, std::multiplies<>{});
+
+    ck::index_t K = ck::accumulate_n<ck::index_t>(
+        a_gs_ms_ks_lengths.begin() + NumDimG + NumDimM, NumDimK, 1, std::multiplies<>{});
 
     std::size_t flop      = std::size_t(2) * G * M * N * K;
     std::size_t num_btype = sizeof(ADataType) * G * M * K + sizeof(BDataType) * G * K * N +

example/30_grouped_conv_fwd_multiple_d/common.hpp

@@ -16,6 +16,7 @@
 #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
+#include "ck/library/utility/algorithm.hpp"
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"

example/30_grouped_conv_fwd_multiple_d/run_grouped_conv_fwd_example.inc

@@ -116,7 +116,7 @@ bool run_grouped_conv_fwd(const ExecutionConfig& config,
     std::array<ck::index_t, NDimSpatial> input_left_pads{};
     std::array<ck::index_t, NDimSpatial> input_right_pads{};
 
-    auto copy = [](auto& x, auto& y) { std::copy(x.begin(), x.end(), y.begin()); };
+    auto copy = [](auto& x, auto& y) { ck::ranges::copy(x, y.begin()); };
 
     copy(in_g_n_c_wis_desc.GetLengths(), a_g_n_c_wis_lengths);
     copy(in_g_n_c_wis_desc.GetStrides(), a_g_n_c_wis_strides);

example/38_grouped_conv_bwd_data_multiple_d/common.hpp

@@ -15,6 +15,7 @@
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 #include "ck/library/reference_tensor_operation/cpu/reference_conv_bwd_data.hpp"
+#include "ck/library/utility/algorithm.hpp"
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/convolution_host_tensor_descriptor_helper.hpp"
 #include "ck/library/utility/convolution_parameter.hpp"

example/38_grouped_conv_bwd_data_multiple_d/run_grouped_conv_bwd_data_example.inc

@@ -52,7 +52,7 @@ bool run_conv_bwd_data(const ExecutionConfig& config,
     std::array<ck::index_t, NDimSpatial> input_left_pads{};
     std::array<ck::index_t, NDimSpatial> input_right_pads{};
 
-    auto copy = [](auto& x, auto& y) { std::copy(x.begin(), x.end(), y.begin()); };
+    auto copy = [](auto& x, auto& y) { ck::ranges::copy(x, y.begin()); };
 
     copy(out_g_n_k_wos_desc.GetLengths(), a_g_n_k_wos_lengths);
     copy(out_g_n_k_wos_desc.GetStrides(), a_g_n_k_wos_strides);

example/41_grouped_conv_conv_fwd/run_grouped_conv_conv_fwd_example.inc

@@ -120,18 +120,14 @@ bool run_grouped_conv_conv_fwd(bool do_verification,
     const ck::index_t gemm_batch = a0_g_n_c_wis_lengths[0];
 
     const ck::index_t gemm0_m_length =
-        e1_g_n_k_wos_lengths[1] * std::accumulate(e1_g_n_k_wos_lengths.begin() + 3,
-                                                  e1_g_n_k_wos_lengths.begin() + 3 + NDimSpatial,
-                                                  ck::index_t{1},
-                                                  std::multiplies<ck::index_t>{});
+        e1_g_n_k_wos_lengths[1] *
+        ck::accumulate_n<ck::index_t>(
+            e1_g_n_k_wos_lengths.begin() + 3, NDimSpatial, 1, std::multiplies<>{});
 
     const ck::index_t gemm0_n_length = b0_g_k_c_xs_lengths[1];
 
-    const ck::index_t gemm0_k_length =
-        std::accumulate(b0_g_k_c_xs_lengths.begin() + 2,
-                        b0_g_k_c_xs_lengths.begin() + 2 + NDimSpatial + 1,
-                        ck::index_t{1},
-                        std::multiplies<ck::index_t>{});
+    const ck::index_t gemm0_k_length = ck::accumulate_n<ck::index_t>(
+        b0_g_k_c_xs_lengths.begin() + 2, NDimSpatial + 1, 1, std::multiplies<>{});
 
     const ck::index_t gemm1_n_length = b1_g_k_c_xs_lengths[1];
 

example/44_conv2d_fwd_quant/conv2d_fwd_xdl_bias_relu_perlayer_quantization_int8.cpp

@@ -6,6 +6,7 @@
 #include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_d_xdl_cshuffle.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
+#include "ck/library/utility/algorithm.hpp"
 #include "ck/library/utility/check_err.hpp"
 #include "ck/library/utility/device_memory.hpp"
 #include "ck/library/utility/host_tensor.hpp"
@@ -144,7 +145,7 @@ bool run_grouped_conv_fwd(bool do_verification,
     std::array<ck::index_t, NDimSpatial> input_left_pads{};
     std::array<ck::index_t, NDimSpatial> input_right_pads{};
 
-    auto copy = [](auto& x, auto& y) { std::copy(x.begin(), x.end(), y.begin()); };
+    auto copy = [](auto& x, auto& y) { ck::ranges::copy(x, y.begin()); };
 
     copy(in_g_n_c_wis_desc.GetLengths(), a_g_n_c_wis_lengths);
     copy(in_g_n_c_wis_desc.GetStrides(), a_g_n_c_wis_strides);