feat(dnn/arm): add fp32 nchw_nchw44 conv

GitOrigin-RevId: f19fe892d9f3e4c166d4835804bf5fc0ad31ccbc

dnn/src/arm_common/conv_bias/fp32/algos.h

@@ -6,7 +6,8 @@
  *
  * Unless required by applicable law or agreed to in writing,
  * software distributed under the License is distributed on an
- * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or
+ * implied.
  */
 
 #pragma once
@@ -156,7 +157,6 @@ private:
     uint32_t m_tile_size;
 };
 
-
 class ConvBiasImpl::AlgoF32Direct final : public AlgoBase {
     SmallVector<NCBKern> get_kimpls(const NCBKernSizeParam& param) const;
     bool m_large_group;
@@ -217,6 +217,24 @@ public:
             fallback::ConvBiasImpl* opr,
             const NCBKernSizeParam& param) const override;
 };
+
+class ConvBiasImpl::AlgoF32DirectStride2NCHWNCHW44 final : public AlgoBase {
+    SmallVector<NCBKern> get_kimpls(const NCBKernSizeParam& param) const;
+
+public:
+    AlgoF32DirectStride2NCHWNCHW44() {}
+    bool is_reproducible() const override { return true; }
+    const char* name() const override { return "F32_CONV_NCHW_NCHW44"; }
+    bool usable(fallback::ConvBiasImpl* opr, const NCBKernSizeParam& param,
+                AlgoSelectionStrategy algo_selection_strategy) const override;
+
+    size_t get_workspace(fallback::ConvBiasImpl* opr,
+                         const NCBKernSizeParam& param) const override;
+    virtual SmallVector<NCBKern> dispatch_kerns(
+            fallback::ConvBiasImpl* opr,
+            const NCBKernSizeParam& param) const override;
+};
+
 }  // namespace arm_common
 }  // namespace megdnn
 

dnn/src/arm_common/conv_bias/fp32/f32_direct_stride2_nchw_nchw44_algo.cpp

+/**
+ * \file
+ dnn/src/arm_common/conv_bias/fp32/f32_direct_stride2_nchw_nchw44_algo.cpp
+ * MegEngine is Licensed under the Apache License, Version 2.0 (the "License")
+ *
+ * Copyright (c) 2014-2020 Megvii Inc. All rights reserved.
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express
+ * or implied.
+ */
+
+#include "megdnn/oprs.h"
+#include "src/arm_common/conv_bias/fp32/algos.h"
+#include "src/arm_common/conv_bias/fp32/f32_direct_stride2_nchw_nchw44_kern.h"
+#include "src/arm_common/conv_bias/fp32/strategy.h"
+#include "src/arm_common/elemwise_op.h"
+#include "src/common/opr_delegate.h"
+
+#include "midout.h"
+
+using namespace megdnn;
+using namespace arm_common;
+using conv_fun = std::function<void(
+        WorkspaceBundle bundle, const ConvBiasImpl::NCBKernParam& kern_param,
+        const ConvBiasImpl::NCBKernIndex& ncb_index,
+        const CpuNDRange& workspace_ids, const CpuNDRange& ncb_range)>;
+MIDOUT_DECL(megdnn_arm_common_conv_bias_fp32_nchw_nchw44_stride2)
+namespace {
+static inline int block_helper(const int nthread, const int amount,
+                               const int per_unit_bytes) {
+    MEGDNN_MARK_USED_VAR(per_unit_bytes);
+    const int block_per_thread = div_ceil(amount, nthread);
+    const int best_block = 16;
+    const int max_block_num = div_ceil(block_per_thread, best_block);
+    const int min_block_num = std::max(max_block_num - 1, 1);
+    const int max_block = div_ceil(block_per_thread, max_block_num);
+    const int min_block = div_ceil(block_per_thread, min_block_num);
+    const int max_loss = std::abs(max_block_num * max_block - block_per_thread);
+    const int min_loss = std::abs(min_block_num * min_block - block_per_thread);
+    int block = max_loss > min_loss ? min_block : max_block;
+    return block;
+}
+static inline size_t get_perthread_cache_bytes(const int ic, const int ih2,
+                                               const int iw2) {
+    // border_size is used to avoid read illegal memory
+    int border_size = 64 * 2;
+    return ic * ih2 * iw2 * sizeof(float) + border_size;
+}
+static void get_rectified_size(
+        const megdnn::fallback::ConvBiasImpl::NCBKernSizeParam& param, int& ih2,
+        int& iw2, int& oh2, int& ow2) {
+    int iw = param.isz[1];
+    int oh = param.osz[0];
+    int ow = param.osz[1];
+
+    oh2 = oh;
+    ow2 = ow;
+    constexpr int cacheline = 64 / sizeof(float);
+    int block_oh = block_helper(param.nr_threads, oh, 0);
+    auto&& fm = param.filter_meta;
+    const int stride_h = static_cast<int>(fm.stride[0]);
+    const int filter_h = static_cast<int>(fm.spatial[0]);
+    ih2 = block_oh * stride_h + filter_h - stride_h;
+    iw2 = round_up(iw + 2 * static_cast<int>(fm.padding[1]), cacheline);
+}
+
+static WorkspaceBundle get_bundle(const ConvBiasImpl::NCBKernSizeParam& param) {
+    auto&& fm = param.filter_meta;
+    int group = fm.group;
+    int ic = fm.icpg;
+    int oc = fm.ocpg;
+    int fh = fm.spatial[0];
+    int fw = fm.spatial[1];
+    int ih2, iw2, oh2, ow2;
+    get_rectified_size(param, ih2, iw2, oh2, ow2);
+
+    int oh_block = block_helper(param.nr_threads, oh2, 0);
+    megdnn_assert(oh_block != 0, "oh_block!=0");
+    size_t src_size = get_perthread_cache_bytes(ic, ih2, iw2);
+    size_t weight_size = group * oc * ic * fh * fw * sizeof(float);
+    return {nullptr, {src_size * param.nr_threads, weight_size}};
+};
+
+static inline void copy_pad_src(float* sptr_base, const float* sptr_origin,
+                                int ph, int pw, int pad_right, int ih, int iw,
+                                int iw2, int pad_top, int pad_bottom, int ic,
+                                int ic_stride) {
+    MEGDNN_MARK_USED_VAR(ph);
+    rep(ic_idx, ic) {
+        const float* sptr = sptr_origin + ic_idx * ic_stride;
+        memset(sptr_base, 0, sizeof(float) * iw2 * pad_top);
+        sptr_base += iw2 * pad_top;
+        rep(ih_idx, ih) {
+            memset(sptr_base, 0, sizeof(float) * pw);
+            sptr_base += pw;
+            memcpy(sptr_base, sptr, sizeof(float) * iw);
+            sptr_base += iw;
+            sptr += iw;
+            memset(sptr_base, 0, sizeof(float) * pad_right);
+            sptr_base += pad_right;
+        }
+        memset(sptr_base, 0, sizeof(float) * iw2 * pad_bottom);
+        sptr_base += iw2 * pad_bottom;
+    }
+}
+static void pack_weight(WorkspaceBundle bundle,
+                        const ConvBiasImpl::NCBKernParam& kern_param,
+                        const ConvBiasImpl::NCBKernIndex& ncb_index) {
+    bundle.set(kern_param.workspace_ptr);
+    const int group_id = ncb_index.ndrange_id[0];
+    int fh = kern_param.filter_meta.spatial[0];
+    int fw = kern_param.filter_meta.spatial[1];
+    int oc = kern_param.filter_meta.ocpg;
+    int ic = kern_param.filter_meta.icpg;
+    int oc_block = oc;
+    int oc_idx = 0;
+    const float* fptr =
+            kern_param.filter<dt_float32>(group_id) + oc_idx * fh * fw * ic;
+    auto packed_weight = reinterpret_cast<float*>(bundle.get(1)) +
+                         group_id * oc * ic * fh * fw + oc_idx * ic * fh * fw;
+    conv_bias::pack_weight_fp32_nchw_nchw44(fptr, packed_weight, oc_block, fh,
+                                            fw, ic);
+}
+
+template <size_t filter, BiasMode bias_mode, typename Op>
+static void do_conv_kern(WorkspaceBundle bundle,
+                         const ConvBiasImpl::NCBKernParam& kern_param,
+                         const ConvBiasImpl::NCBKernIndex& ncb_index,
+                         const CpuNDRange&, const CpuNDRange&) {
+    const int oh = kern_param.osz[0];
+    const int ow = kern_param.osz[1];
+    const int fh = kern_param.filter_meta.spatial[0];
+    const int fw = kern_param.filter_meta.spatial[1];
+    const int ic = kern_param.filter_meta.icpg;
+    const int oc = kern_param.filter_meta.ocpg;
+    const int ih = kern_param.isz[0];
+    const int iw = kern_param.isz[1];
+    const int stride_h = kern_param.filter_meta.stride[0];
+    const int ph = kern_param.filter_meta.padding[0];
+    const int pw = kern_param.filter_meta.padding[1];
+    int ih2 = 0;
+    int iw2 = 0;
+    int oh2 = 0;
+    int ow2 = 0;
+    get_rectified_size(kern_param, ih2, iw2, oh2, ow2);
+    bundle.set(kern_param.workspace_ptr);
+
+    constexpr int pack_c = 4;
+    const int batch_id = ncb_index.ndrange_id[0];
+    const int group_id = ncb_index.ndrange_id[1];
+    int oc_idx = 0;
+    int oc_block = oc;
+    int oh_block = block_helper(kern_param.nr_threads, oh2, 0);
+    const int oh_idx = ncb_index.ndrange_id[2];
+    const int oh_block_real = std::min(oh - oh_idx * oh_block, oh_block);
+    const int ih_real = oh_block_real * stride_h + fh - stride_h;
+    const int src_top_pad = std::max(ph - oh_idx * oh_block * stride_h, 0);
+    const int src_bottom_pad = std::max(
+            (oh_idx * oh_block + oh_block_real - 1) * stride_h + fh - ih - ph,
+            0);
+    const int remain_right_pad = std::max(iw2 - iw - pw, 0);
+    const int src_offset = std::max(oh_idx * oh_block * stride_h - ph, 0) * iw;
+    const float* origin_sptr = static_cast<const float*>(kern_param.src<float>(
+                                       batch_id, group_id, 0, 1, 1)) +
+                               src_offset;
+    const size_t src_size = get_perthread_cache_bytes(ic, ih2, iw2);
+    float* sptr = reinterpret_cast<float*>((int8_t*)bundle.get(0) +
+                                           ncb_index.thread_id * src_size);
+
+    copy_pad_src(sptr, origin_sptr, ph, pw, remain_right_pad,
+                 ih_real - src_top_pad - src_bottom_pad, iw, iw2, src_top_pad,
+                 src_bottom_pad, ic, ih * iw);
+    // pack weight
+    auto packed_weight = reinterpret_cast<float*>(bundle.get(1)) +
+                         group_id * oc * ic * fh * fw + oc_idx * ic * fh * fw;
+    // get param
+    float_t* dst = kern_param.dst<float_t>(batch_id, group_id) +
+                   oh_idx * oh_block * ow * pack_c;
+    const float* bptr =
+            kern_param.bias<dt_float32>(batch_id, group_id) + oc_idx;
+    Op op;
+#define KERN1_NCHW44_CONV(filter)                                             \
+    conv_bias::conv_direct_stride2_##filter##x##filter##_fp32_nchw_nchw44<    \
+                                                                              \
+            bias_mode, Op>(sptr, packed_weight, bptr, nullptr, dst, oc_block, \
+                           ic, ih_real, iw2, oh, oh_block_real, ow, op, ph,   \
+                           pw)
+
+    DISPATCH_FILTER(filter, KERN1_NCHW44_CONV);
+#undef KERN1_NCHW44_CONV
+}
+
+}  // namespace
+
+/* ===================== stride2 algo ===================== */
+bool ConvBiasImpl::AlgoF32DirectStride2NCHWNCHW44::usable(
+        fallback::ConvBiasImpl*, const NCBKernSizeParam& param,
+        AlgoSelectionStrategy) const {
+    auto&& fm = param.filter_meta;
+    auto fh = fm.spatial[0];
+    int oc = fm.ocpg;
+    bool ok_type = ((param.src_type.enumv() == DTypeEnum::Float32 &&
+                     param.filter_type.enumv() == DTypeEnum::Float32 &&
+                     (param.dst_type.enumv() == DTypeEnum::Float32))) &&
+                   (fm.format == param::Convolution::Format::NCHW44);
+    bool ok_src_dst = fm.icpg < 4 && (oc % 4 == 0 && oc >= 4) && fm.group == 1;
+    bool ok_filter = fm.spatial_ndim == 2 && fh == fm.spatial[1] &&
+                     (fh == 3 || fh == 5 || fh == 7);
+    bool ok_slide = fm.dilation[0] == 1 && fm.dilation[1] == 1 &&
+                    fm.stride[0] == 2 && fm.stride[1] == 2;
+    bool ok_conv = !fm.should_flip && param.bias_mode != BiasMode::BIAS;
+    bool avaible = ok_type && ok_src_dst && ok_filter && ok_slide && ok_conv;
+    return avaible;
+}
+
+size_t ConvBiasImpl::AlgoF32DirectStride2NCHWNCHW44::get_workspace(
+        fallback::ConvBiasImpl*, const NCBKernSizeParam& param) const {
+    return get_bundle(param).total_size_in_bytes();
+}
+
+SmallVector<ConvBiasImpl::NCBKern>
+ConvBiasImpl::AlgoF32DirectStride2NCHWNCHW44::dispatch_kerns(
+        fallback::ConvBiasImpl*, const NCBKernSizeParam& param) const {
+    auto fm = param.filter_meta;
+    const int batch = param.n;
+    const int group = fm.group;
+    WorkspaceBundle wbundle = get_bundle(param);
+    conv_fun do_conv_fun = nullptr;
+    // NOTE: remain_w is not used to gen hash of midout for compatible with
+// shape runtime
+#define DO_CONV_KERN_FUN(filter, bias_mode, op)                        \
+    MIDOUT_BEGIN(megdnn_arm_common_conv_bias_fp32_nchw_nchw44_stride2, \
+                 midout_iv(#filter #bias_mode #op##_hash)) {           \
+        do_conv_fun = do_conv_kern<filter, bias_mode, op>;             \
+    }                                                                  \
+    MIDOUT_END();
+
+#define GET_OP_PARAM(filter, bias_mode)                               \
+    switch (param.nonlineMode) {                                      \
+        case param::ConvBias::NonlineMode::IDENTITY:                  \
+            DO_CONV_KERN_FUN(filter, bias_mode, NoneOp<dt_float32>)   \
+            break;                                                    \
+        case param::ConvBias::NonlineMode::RELU:                      \
+            DO_CONV_KERN_FUN(filter, bias_mode, ReluOp<dt_float32>)   \
+            break;                                                    \
+        case param::ConvBias::NonlineMode::H_SWISH:                   \
+            DO_CONV_KERN_FUN(filter, bias_mode, HSwishOp<dt_float32>) \
+            break;                                                    \
+        default:                                                      \
+            megdnn_assert(0);                                         \
+            break;                                                    \
+    }
+#define GET_BIAS_MODE_PARAM(filter)                                \
+    switch (param.bias_mode) {                                     \
+        case BiasMode::NO_BIAS:                                    \
+            GET_OP_PARAM(filter, BiasMode::NO_BIAS)                \
+            break;                                                 \
+        case BiasMode::BROADCAST_CHANNEL_BIAS:                     \
+            GET_OP_PARAM(filter, BiasMode::BROADCAST_CHANNEL_BIAS) \
+            break;                                                 \
+        default:                                                   \
+            megdnn_assert(0);                                      \
+            break;                                                 \
+    }
+
+#define DISPATCH_CONV_KERN()                \
+    switch (param.filter_meta.spatial[0]) { \
+        case 3:                             \
+            GET_BIAS_MODE_PARAM(3)          \
+            break;                          \
+        case 5:                             \
+            GET_BIAS_MODE_PARAM(5)          \
+            break;                          \
+        case 7:                             \
+            GET_BIAS_MODE_PARAM(7)          \
+            break;                          \
+        default:                            \
+            megdnn_assert(0);               \
+            break;                          \
+    }
+
+    DISPATCH_CONV_KERN();
+
+#undef DO_CONV_KERN_FUN
+#undef GET_REMAIN_W_PARAM
+#undef GET_OP_PARAM
+#undef GET_BIAS_MODE_PARAM
+#undef DISPATCH_CONV_KERN
+
+    megdnn_assert(do_conv_fun);
+
+    SmallVector<ConvBiasImpl::NCBKern> ret_kerns;
+    WorkspaceBundle bundle = wbundle;
+    int oh = param.osz[0];
+    int oh_block = block_helper(param.nr_threads, oh, 0);
+    auto do_pack_weight = [bundle](const NCBKernParam& kern_param,
+                                   const NCBKernIndex& ncb_index) {
+        pack_weight(bundle, kern_param, ncb_index);
+    };
+    ret_kerns.push_back({do_pack_weight, {static_cast<size_t>(group)}});
+    CpuNDRange ncb_range = {static_cast<size_t>(batch),
+                            static_cast<size_t>(group),
+                            static_cast<size_t>(div_ceil(oh, oh_block))};
+    auto do_conv = [bundle, do_conv_fun, ncb_range](
+                           const NCBKernParam& kern_param,
+                           const NCBKernIndex& ncb_index) {
+        do_conv_fun(bundle, kern_param, ncb_index, ncb_index.ndrange_id,
+                    ncb_range);
+    };
+    ret_kerns.push_back({do_conv, ncb_range});
+
+    return ret_kerns;
+}
+
+// vim: syntax=cpp.doxygen

dnn/src/arm_common/conv_bias/fp32/f32_direct_stride2_nchw_nchw44_kern.cpp

+/**
+ * \file
+ * dnn/src/arm_common/conv_bias/fp32/f32_direct_stride2_nchw_nchw44_kern.cpp
+ * MegEngine is Licensed under the Apache License, Version 2.0 (the "License")
+ *
+ * Copyright (c) 2014-2020 Megvii Inc. All rights reserved.
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or
+ * implied.
+ */
+
+#include "src/arm_common/conv_bias/fp32/f32_direct_stride2_nchw_nchw44_kern.h"
+#include "src/arm_common/conv_bias/intrinsic_helper.h"
+#include "src/arm_common/elemwise_op.h"
+#include "src/arm_common/simd_macro/marm_neon.h"
+#include "src/common/unroll_macro.h"
+#include "src/common/utils.h"
+#include "src/fallback/conv_bias/common.h"
+
+using namespace megdnn;
+using namespace arm_common;
+namespace {
+
+template <int src_idx, int weight_idx, int c_dim, typename Func, typename T,
+          typename T2, typename T3, typename T4>
+struct ShiftCalHelper {
+    static void impl(T& c, T2& src, T3& weight);
+};
+
+template <int src_idx, int weight_idx, typename Func, typename T, typename T2,
+          typename T3, typename T4>
+struct ShiftCalHelper<src_idx, weight_idx, 2, Func, T, T2, T3, T4> {
+    static void impl(T& c, T2& src, T3& weight) {
+        constexpr int stride = 2;
+#define cb(step)                                                     \
+    c[0][step] = Func::template impl<(step * stride + src_idx) % 4>( \
+            c[0][step], weight[0][weight_idx],                       \
+            src[(step * stride + src_idx) / 4]);                     \
+    c[1][step] = Func::template impl<(step * stride + src_idx) % 4>( \
+            c[1][step], weight[1][weight_idx],                       \
+            src[(step * stride + src_idx) / 4]);
+
+        UNROLL_CALL_RAW(8, cb);
+#undef cb
+    }
+};
+template <int src_idx, int weight_idx, typename Func, typename T, typename T2,
+          typename T3, typename T4>
+struct ShiftCalHelper<src_idx, weight_idx, 1, Func, T, T2, T3, T4> {
+    static void impl(T& c, T2& src, T3& weight) {
+        constexpr int stride = 2;
+#define cb(step)                                                     \
+    c[0][step] = Func::template impl<(step * stride + src_idx) % 4>( \
+            c[0][step], weight[0][weight_idx],                       \
+            src[(step * stride + src_idx) / 4]);
+
+        UNROLL_CALL_RAW(8, cb);
+#undef cb
+    }
+};
+
+template <int src_idx, int weight_idx, int c_dim, typename FUNC, typename T,
+          typename T2, typename T3>
+inline void cal_helper(T& c, T2& src, T3& weight) {
+    ShiftCalHelper<src_idx, weight_idx, c_dim, FUNC, T, T2, T3, int>::impl(
+            c, src, weight);
+};
+template <int oc>
+struct OCHelper {
+public:
+    static const int val = -1;
+};
+
+template <>
+struct OCHelper<4> {
+public:
+    static const int val = 1;
+};
+
+template <>
+struct OCHelper<8> {
+public:
+    static const int val = 2;
+};
+/**
+ *  oc8_ow8(m = 8, n = 8) and oc4_ow8(m = 4, n = 8) gemm like kernel
+ * */
+template <BiasMode bias_mode, typename Op, int remain_w, int filter_size,
+          int oc_block>
+struct KerNeonXXs2NchwNchw44FP32 {
+    static void impl(const float32_t* src_ptr, const float32_t* weight_ptr,
+                     const float32_t* bias_ptr, float32_t* dst_ptr, int ic,
+                     int ih, int iw, int ld_dst_oc, const Op& op);
+};
+template <BiasMode bias_mode, typename Op, int remain_w, int oc_block>
+struct KerNeonXXs2NchwNchw44FP32<bias_mode, Op, remain_w, 7, oc_block> {
+    static void impl(const float32_t* src_ptr, const float32_t* weight_ptr,
+                     const float32_t* bias_ptr, float32_t* dst_ptr, int ic,
+                     int ih, int iw, int ld_dst_oc, const Op& op) {
+        constexpr int loop_ic_step = 1;
+        constexpr int filter_size = 7;
+        constexpr int oc_step = 4;
+        constexpr int simd_len = 4;
+        constexpr int src_reg_size = 6;
+
+        constexpr int ld_weight_fw = oc_step * filter_size;
+        const int ld_weight_oc = oc_step * filter_size * filter_size * ic;
+        const int ld_weight_ic = oc_step * filter_size * filter_size;
+        const int ld_src_ic = ih * iw;
+        constexpr int c_dim = OCHelper<oc_block>::val;
+        float32x4_t c[c_dim][8];
+        init_ocx_ow8<c_dim, bias_mode>(c, bias_ptr, oc_step);
+
+        for (int ic_idx = 0; ic_idx < ic; ic_idx += loop_ic_step) {
+            float32x4_t src[src_reg_size];
+            float32x4_t weight[c_dim][filter_size];
+
+#define KERNEL_CB(step)                                              \
+    load_helper<src_reg_size, 0, simd_len, 0, Vld1q_f32>(            \
+            src, src_ptr + step * iw, 0);                            \
+    load_helper<filter_size, 0, oc_step, c_dim, Vld1q_f32>(          \
+            weight, weight_ptr + step * ld_weight_fw, ld_weight_oc); \
+    cal_helper<0, 0, c_dim, Vfmaq_laneq_f32>(c, src, weight);        \
+    cal_helper<1, 1, c_dim, Vfmaq_laneq_f32>(c, src, weight);        \
+    cal_helper<2, 2, c_dim, Vfmaq_laneq_f32>(c, src, weight);        \
+    cal_helper<3, 3, c_dim, Vfmaq_laneq_f32>(c, src, weight);        \
+    cal_helper<4, 4, c_dim, Vfmaq_laneq_f32>(c, src, weight);        \
+    cal_helper<5, 5, c_dim, Vfmaq_laneq_f32>(c, src, weight);        \
+    cal_helper<6, 6, c_dim, Vfmaq_laneq_f32>(c, src, weight);
+
+            UNROLL_CALL_RAW(7, KERNEL_CB)
+#undef KERNEL_CB
+
+            src_ptr += ld_src_ic;
+            weight_ptr += ld_weight_ic;
+        }
+        store_ocx_ow8_remain_static<c_dim, remain_w, Op>(c, op, dst_ptr,
+                                                         ld_dst_oc);
+    }
+};
+template <BiasMode bias_mode, typename Op, int remain_w, int oc_block>
+struct KerNeonXXs2NchwNchw44FP32<bias_mode, Op, remain_w, 5, oc_block> {
+    static void impl(const float32_t* src_ptr, const float32_t* weight_ptr,
+                     const float32_t* bias_ptr, float32_t* dst_ptr, int ic,
+                     int ih, int iw, int ld_dst_oc, const Op& op) {
+        constexpr int loop_ic_step = 1;
+        constexpr int filter_size = 5;
+        constexpr int oc_step = 4;
+        constexpr int simd_len = 4;
+        constexpr int src_reg_size = 5;
+
+        constexpr int ld_weight_fw = oc_step * filter_size;
+        const int ld_weight_oc = oc_step * filter_size * filter_size * ic;
+        const int ld_weight_ic = oc_step * filter_size * filter_size;
+        const int ld_src_ic = ih * iw;
+        constexpr int c_dim = OCHelper<oc_block>::val;
+        float32x4_t c[c_dim][8];
+        init_ocx_ow8<c_dim, bias_mode>(c, bias_ptr, oc_step);
+
+        for (int ic_idx = 0; ic_idx < ic; ic_idx += loop_ic_step) {
+            float32x4_t src[src_reg_size];
+            float32x4_t weight[c_dim][filter_size];
+
+#define KERNEL_CB(step)                                              \
+    load_helper<src_reg_size, 0, simd_len, 0, Vld1q_f32>(            \
+            src, src_ptr + step * iw, 0);                            \
+    load_helper<filter_size, 0, oc_step, c_dim, Vld1q_f32>(          \
+            weight, weight_ptr + step * ld_weight_fw, ld_weight_oc); \
+    cal_helper<0, 0, c_dim, Vfmaq_laneq_f32>(c, src, weight);        \
+    cal_helper<1, 1, c_dim, Vfmaq_laneq_f32>(c, src, weight);        \
+    cal_helper<2, 2, c_dim, Vfmaq_laneq_f32>(c, src, weight);        \
+    cal_helper<3, 3, c_dim, Vfmaq_laneq_f32>(c, src, weight);        \
+    cal_helper<4, 4, c_dim, Vfmaq_laneq_f32>(c, src, weight);
+            UNROLL_CALL_RAW(5, KERNEL_CB)
+#undef KERNEL_CB
+
+            src_ptr += ld_src_ic;
+            weight_ptr += ld_weight_ic;
+        }
+        store_ocx_ow8_remain_static<c_dim, remain_w, Op>(c, op, dst_ptr,
+                                                         ld_dst_oc);
+    }
+};
+
+template <BiasMode bias_mode, typename Op, int remain_w, int oc_block>
+struct KerNeonXXs2NchwNchw44FP32<bias_mode, Op, remain_w, 3, oc_block> {
+    static void impl(const float32_t* src_ptr, const float32_t* weight_ptr,
+                     const float32_t* bias_ptr, float32_t* dst_ptr, int ic,
+                     int ih, int iw, int ld_dst_oc, const Op& op) {
+        constexpr int loop_ic_step = 1;
+        constexpr int filter_size = 3;
+        constexpr int oc_step = 4;
+        constexpr int simd_len = 4;
+        constexpr int src_reg_size = 5;
+
+        constexpr int ld_weight_fw = oc_step * filter_size;
+        const int ld_weight_oc = oc_step * filter_size * filter_size * ic;
+        const int ld_weight_ic = oc_step * filter_size * filter_size;
+        const int ld_src_ic = ih * iw;
+        constexpr int c_dim = OCHelper<oc_block>::val;
+        float32x4_t c[c_dim][8];
+        init_ocx_ow8<c_dim, bias_mode>(c, bias_ptr, oc_step);
+
+        for (int ic_idx = 0; ic_idx < ic; ic_idx += loop_ic_step) {
+            float32x4_t src[src_reg_size];
+            float32x4_t weight[c_dim][filter_size];
+            // row 0
+            load_helper<5, 0, simd_len, 0, Vld1q_f32>(src, src_ptr, 0);
+            load_helper<3, 0, oc_step, c_dim, Vld1q_f32>(weight, weight_ptr,
+                                                         ld_weight_oc);
+            cal_helper<0, 0, c_dim, Vfmaq_laneq_f32>(c, src, weight);
+            cal_helper<1, 1, c_dim, Vfmaq_laneq_f32>(c, src, weight);
+            cal_helper<2, 2, c_dim, Vfmaq_laneq_f32>(c, src, weight);
+
+            // row 1
+            load_helper<5, 0, simd_len, 0, Vld1q_f32>(src, src_ptr + iw, 0);
+            load_helper<3, 0, oc_step, c_dim, Vld1q_f32>(
+                    weight, weight_ptr + 1 * ld_weight_fw, ld_weight_oc);
+            cal_helper<0, 0, c_dim, Vfmaq_laneq_f32>(c, src, weight);
+            cal_helper<1, 1, c_dim, Vfmaq_laneq_f32>(c, src, weight);
+            cal_helper<2, 2, c_dim, Vfmaq_laneq_f32>(c, src, weight);
+
+            // row 2
+            load_helper<5, 0, simd_len, 0, Vld1q_f32>(src, src_ptr + 2 * iw, 0);
+            load_helper<3, 0, oc_step, c_dim, Vld1q_f32>(
+                    weight, weight_ptr + 2 * ld_weight_fw, ld_weight_oc);
+            cal_helper<0, 0, c_dim, Vfmaq_laneq_f32>(c, src, weight);
+            cal_helper<1, 1, c_dim, Vfmaq_laneq_f32>(c, src, weight);
+            cal_helper<2, 2, c_dim, Vfmaq_laneq_f32>(c, src, weight);
+
+            src_ptr += ld_src_ic;
+            weight_ptr += ld_weight_ic;
+        }
+        store_ocx_ow8_remain_static<c_dim, remain_w, Op>(c, op, dst_ptr,
+                                                         ld_dst_oc);
+    }
+};
+
+}  // namespace
+
+void conv_bias::pack_weight_fp32_nchw_nchw44(const float32_t* in_ptr,
+                                             float32_t* dst_ptr, const int oc,
+                                             const int kh, const int kw,
+                                             const int ic) {
+    constexpr int oc_step = 4;
+    const int filter_oc_stride = kh * kw * ic;
+    const int filter_ic_stride = kh * kw * oc_step;
+    for (int oc_idx = 0; oc_idx < oc; oc_idx += oc_step) {
+        const float32_t* in_ptr_oc = in_ptr + oc_idx * filter_oc_stride;
+        float32_t* dst_ptr_oc = dst_ptr + oc_idx * filter_oc_stride;
+        for (int kh_idx = 0; kh_idx < kh; ++kh_idx) {
+            for (int kw_idx = 0; kw_idx < kw; ++kw_idx) {
+                for (int ic_idx = 0; ic_idx < ic; ++ic_idx) {
+                    float32x4_t vsrc = vld1q_f32(in_ptr_oc);
+                    vst1q_f32(dst_ptr_oc + ic_idx * filter_ic_stride, vsrc);
+                    in_ptr_oc += oc_step;
+                }
+                dst_ptr_oc += oc_step;
+            }
+        }
+    }
+}
+
+template <BiasMode bias_mode, typename Op, int filter_size>
+static void conv_direct_stride2_fp32_nchw_nchw44(
+        const float32_t* src, const float32_t* filter, const float32_t* bias,
+        float32_t*, float32_t* dst, const int oc, const int ic, const int ih,
+        const int iw, const int oh, const int oh_block, const int ow,
+        const Op& op, const int, const int) {
+    constexpr int fh = filter_size;
+    constexpr int fw = filter_size;
+    constexpr int ic_step = 1;
+    constexpr int big_oc_step = 8;
+    constexpr int oc_step = 4;
+    constexpr int ih_step = 1;
+    constexpr int oh_step = 1;
+    constexpr int ow_step = 8;
+    constexpr int stride_h = 2;
+    constexpr int stride_w = 2;
+    constexpr int pack_iw_len = 1;
+
+    const int img_stride = oh * ow;
+    const int ow_end = ow / ow_step * ow_step;
+    const int ow_remain = ow - ow_end;
+    const int oc_end = oc / big_oc_step * big_oc_step;
+    const int oc_remain = oc - oc_end;
+    const int ld_dst_oc = oc_step * img_stride;
+
+    using remain_fun = std::function<void(
+            const float32_t* src_ptr, const float32_t* weight_ptr,
+            const float32_t* bias_ptr, float32_t* dst_ptr, int ic, int ih,
+            int iw, int ld_dst_oc, const Op& op)>;
+    remain_fun kern_big_oc_remain = nullptr;
+    remain_fun kern_small_oc_remain = nullptr;
+
+    switch (ow_remain) {
+#define cb(step)                                                            \
+    case step:                                                              \
+        kern_big_oc_remain =                                                \
+                KerNeonXXs2NchwNchw44FP32<bias_mode, Op, step, filter_size, \
+                                          big_oc_step>::impl;               \
+        kern_small_oc_remain =                                              \
+                KerNeonXXs2NchwNchw44FP32<bias_mode, Op, step, filter_size, \
+                                          oc_step>::impl;                   \
+        break;
+
+        UNROLL_CALL_RAW(8, cb);
+        default:
+            megdnn_assert(0, "no remain %d for kern", ow_remain);
+    }
+    for (int oc_idx = 0; oc_idx < oc_end; oc_idx += big_oc_step) {
+        const int weight_offset = oc_idx * ic * fh * fw;
+        for (int oh_idx = 0; oh_idx < oh_block; oh_idx += oh_step) {
+            for (int ow_idx = 0; ow_idx < ow_end; ow_idx += ow_step) {
+                const int src_offset =
+                        (oh_idx * stride_h * iw + ow_idx * stride_w * ih_step) *
+                        ic_step * pack_iw_len;
+                const int dst_offset =
+                        oc_idx * img_stride + (oh_idx * ow + ow_idx) * oc_step;
+                KerNeonXXs2NchwNchw44FP32<
+                        bias_mode, Op, 0, filter_size,
+                        big_oc_step>::impl(src + src_offset,
+                                           filter + weight_offset,
+                                           bias + oc_idx, dst + dst_offset, ic,
+                                           ih, iw, ld_dst_oc, op);
+            }
+            if (ow_remain > 0) {
+                const int src_offset =
+                        (oh_idx * stride_h * iw + ow_end * stride_w * ih_step) *
+                        ic_step * pack_iw_len;
+                const int dst_offset =
+                        oc_idx * img_stride + (oh_idx * ow + ow_end) * oc_step;
+                kern_big_oc_remain(src + src_offset, filter + weight_offset,
+                                   bias + oc_idx, dst + dst_offset, ic, ih, iw,
+                                   ld_dst_oc, op);
+            }
+        }
+    }
+    if (oc_remain > 0) {
+        int oc_idx = oc_end;
+        const int weight_offset = oc_idx * ic * fh * fw;
+        for (int oh_idx = 0; oh_idx < oh_block; oh_idx += oh_step) {
+            for (int ow_idx = 0; ow_idx < ow_end; ow_idx += ow_step) {
+                const int src_offset =
+                        (oh_idx * stride_h * iw + ow_idx * stride_w * ih_step) *
+                        ic_step * pack_iw_len;
+                const int dst_offset =
+                        oc_idx * img_stride + (oh_idx * ow + ow_idx) * oc_step;
+                KerNeonXXs2NchwNchw44FP32<bias_mode, Op, 0, filter_size,
+                                          oc_step>::impl(src + src_offset,
+                                                         filter + weight_offset,
+                                                         bias + oc_idx,
+                                                         dst + dst_offset, ic,
+                                                         ih, iw, ld_dst_oc, op);
+            }
+            if (ow_remain > 0) {
+                const int src_offset =
+                        (oh_idx * stride_h * iw + ow_end * stride_w * ih_step) *
+                        ic_step * pack_iw_len;
+                const int dst_offset =
+                        oc_idx * img_stride + (oh_idx * ow + ow_end) * oc_step;
+                kern_small_oc_remain(src + src_offset, filter + weight_offset,
+                                     bias + oc_idx, dst + dst_offset, ic, ih,
+                                     iw, ld_dst_oc, op);
+            }
+        }
+    }
+}
+
+#define CONSTRUCT_FUNC(filter_size)                                               \
+    template <BiasMode bias_mode, typename Op>                                    \
+    void conv_bias::                                                              \
+            conv_direct_stride2_##filter_size##x##filter_size##_fp32_nchw_nchw44( \
+                    const float32_t* src, const float32_t* filter,                \
+                    const float32_t* bias, float32_t* temp, float32_t* dst,       \
+                    const int oc, const int ic, const int ih, const int iw,       \
+                    const int oh, const int oh_block, const int ow,               \
+                    const Op& op, const int ph, const int pw) {                   \
+        conv_direct_stride2_fp32_nchw_nchw44<bias_mode, Op, filter_size>(         \
+                src, filter, bias, temp, dst, oc, ic, ih, iw, oh, oh_block,       \
+                ow, op, ph, pw);                                                  \
+    }
+
+CONSTRUCT_FUNC(3);
+CONSTRUCT_FUNC(5);
+CONSTRUCT_FUNC(7);
+#undef CONSTRUCT_FUNC
+
+template <BiasMode bias_mode, typename Op>
+void conv_bias::conv_direct_stride2_2x2_fp32_nchw_nchw44(
+        const float32_t*, const float32_t*, const float32_t*, float32_t*,
+        float32_t*, const int, const int, const int, const int, const int,
+        const int, const int, const Op&, const int, const int) {
+    megdnn_assert(0, "not imple nchw_nchw44 2x2s2 conv");
+}
+
+#define INSTANTIATION(stride, i, bias, Op)                                   \
+    template void conv_bias::                                                \
+            conv_direct_##stride##_##i##x##i##_fp32_nchw_nchw44<bias, Op>(   \
+                    const float32_t*, const float32_t*, const float32_t*,    \
+                    float32_t*, float32_t*, const int, const int, const int, \
+                    const int, const int, const int, const int, const Op&,   \
+                    const int, const int);
+
+#define FOR_OP(stride, i, bias)                        \
+    INSTANTIATION(stride, i, bias, NoneOp<dt_float32>) \
+    INSTANTIATION(stride, i, bias, ReluOp<dt_float32>) \
+    INSTANTIATION(stride, i, bias, HSwishOp<dt_float32>)
+
+#define FOR_BIAS(stride, i)              \
+    FOR_OP(stride, i, BiasMode::NO_BIAS) \
+    FOR_OP(stride, i, BiasMode::BROADCAST_CHANNEL_BIAS)
+
+#define FOR_FILTER(stride) \
+    FOR_BIAS(stride, 2)    \
+    FOR_BIAS(stride, 3)    \
+    FOR_BIAS(stride, 5)    \
+    FOR_BIAS(stride, 7)
+
+FOR_FILTER(stride2)
+
+#undef FOR_STRIDE
+#undef FOR_FILTER
+#undef FOR_IC
+#undef FOR_BIAS
+#undef FOR_NONLINEAR
+#undef FOR_REMAIN
+#undef INSTANTIATION

dnn/src/arm_common/conv_bias/fp32/f32_direct_stride2_nchw_nchw44_kern.h

+/**
+ * \file dnn/src/arm_common/conv_bias/fp32/f32_direct_stride2_nchw_nchw44_kern.h
+ * MegEngine is Licensed under the Apache License, Version 2.0 (the "License")
+ *
+ * Copyright (c) 2014-2020 Megvii Inc. All rights reserved.
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or
+ * implied.
+ */
+
+#include "src/arm_common/conv_bias/opr_impl.h"
+#include "src/fallback/conv_bias/common.h"
+
+namespace megdnn {
+namespace arm_common {
+namespace conv_bias {
+#define KERN(stride, i, layout)                                                \
+    template <BiasMode bias_mode, typename Op>                                 \
+    void conv_direct_##stride##_##i##x##i##_fp32_nchw_##layout(                \
+            const float* src, const float* filter, const float* bias,          \
+            float* temp, float* dst, const int oc, const int ic, const int ih, \
+            const int iw, const int oh, const int oh_block, const int ow,      \
+            const Op& op, const int ph, const int pw);
+
+KERN(stride2, 2, nchw44)
+KERN(stride2, 3, nchw44)
+KERN(stride2, 5, nchw44)
+KERN(stride2, 7, nchw44)
+#undef KERN
+void pack_weight_fp32_nchw_nchw44(const float_t* in_ptr, float_t* dst_ptr,
+                                  const int oc, const int kh, const int kw,
+                                  const int ic);
+
+}  // namespace conv_bias
+}  // namespace arm_common
+}  // namespace megdnn
\ No newline at end of file

dnn/src/arm_common/conv_bias/intrinsic_helper.h

@@ -174,7 +174,167 @@ inline void store_ocx_ow4_remain_static(T& c, const Op& op, int8_t* dst_ptr,
                                         int ld_dst_oc) {
     StoreOcxOw4Remain<c_dim, ow_remain, Op, T>::impl(c, op, dst_ptr, ld_dst_oc);
 }
+////////////////////Store_OCX_OW8_Remain/////////////////////////
+template <int c_dim, int ow_remain, typename Op, typename T>
+struct StoreOcxOw8Remain {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int ld_dst_oc);
+};
+
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<2, 0, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int ld_dst_oc) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op({{c[0][2], c[0][3]}}, dst_ptr + 8);
+        op({{c[0][4], c[0][5]}}, dst_ptr + 16);
+        op({{c[0][6], c[0][7]}}, dst_ptr + 24);
+
+        op({{c[1][0], c[1][1]}}, dst_ptr + ld_dst_oc);
+        op({{c[1][2], c[1][3]}}, dst_ptr + ld_dst_oc + 8);
+        op({{c[1][4], c[1][5]}}, dst_ptr + ld_dst_oc + 16);
+        op({{c[1][6], c[1][7]}}, dst_ptr + ld_dst_oc + 24);
+    }
+};
+
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<2, 7, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int ld_dst_oc) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op({{c[0][2], c[0][3]}}, dst_ptr + 8);
+        op({{c[0][4], c[0][5]}}, dst_ptr + 16);
+        op(c[0][6], dst_ptr + 24);
+
+        op({{c[1][0], c[1][1]}}, dst_ptr + ld_dst_oc);
+        op({{c[1][2], c[1][3]}}, dst_ptr + ld_dst_oc + 8);
+        op({{c[1][4], c[1][5]}}, dst_ptr + ld_dst_oc + 16);
+        op(c[1][6], dst_ptr + ld_dst_oc + 24);
+    }
+};
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<2, 6, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int ld_dst_oc) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op({{c[0][2], c[0][3]}}, dst_ptr + 8);
+        op({{c[0][4], c[0][5]}}, dst_ptr + 16);
+
+        op({{c[1][0], c[1][1]}}, dst_ptr + ld_dst_oc);
+        op({{c[1][2], c[1][3]}}, dst_ptr + ld_dst_oc + 8);
+        op({{c[1][4], c[1][5]}}, dst_ptr + ld_dst_oc + 16);
+    }
+};
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<2, 5, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int ld_dst_oc) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op({{c[0][2], c[0][3]}}, dst_ptr + 8);
+        op(c[0][4], dst_ptr + 16);
+
+        op({{c[1][0], c[1][1]}}, dst_ptr + ld_dst_oc);
+        op({{c[1][2], c[1][3]}}, dst_ptr + ld_dst_oc + 8);
+        op(c[1][4], dst_ptr + ld_dst_oc + 16);
+    }
+};
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<2, 4, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int ld_dst_oc) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op({{c[0][2], c[0][3]}}, dst_ptr + 8);
+
+        op({{c[1][0], c[1][1]}}, dst_ptr + ld_dst_oc);
+        op({{c[1][2], c[1][3]}}, dst_ptr + ld_dst_oc + 8);
+    }
+};
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<2, 3, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int ld_dst_oc) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op(c[0][2], dst_ptr + 8);
+
+        op({{c[1][0], c[1][1]}}, dst_ptr + ld_dst_oc);
+        op(c[1][2], dst_ptr + ld_dst_oc + 8);
+    }
+};
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<2, 2, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int ld_dst_oc) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op({{c[1][0], c[1][1]}}, dst_ptr + ld_dst_oc);
+    }
+};
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<2, 1, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int ld_dst_oc) {
+        op(c[0][0], dst_ptr);
+        op(c[1][0], dst_ptr + ld_dst_oc);
+    }
+};
+
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<1, 0, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op({{c[0][2], c[0][3]}}, dst_ptr + 8);
+        op({{c[0][4], c[0][5]}}, dst_ptr + 16);
+        op({{c[0][6], c[0][7]}}, dst_ptr + 24);
+    }
+};
+
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<1, 7, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op({{c[0][2], c[0][3]}}, dst_ptr + 8);
+        op({{c[0][4], c[0][5]}}, dst_ptr + 16);
+        op(c[0][6], dst_ptr + 24);
+    }
+};
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<1, 6, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op({{c[0][2], c[0][3]}}, dst_ptr + 8);
+        op({{c[0][4], c[0][5]}}, dst_ptr + 16);
+    }
+};
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<1, 5, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op({{c[0][2], c[0][3]}}, dst_ptr + 8);
+        op(c[0][4], dst_ptr + 16);
+    }
+};
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<1, 4, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op({{c[0][2], c[0][3]}}, dst_ptr + 8);
+    }
+};
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<1, 3, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+        op(c[0][2], dst_ptr + 8);
+    }
+};
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<1, 2, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int) {
+        op({{c[0][0], c[0][1]}}, dst_ptr);
+    }
+};
+template <typename Op, typename T>
+struct StoreOcxOw8Remain<1, 1, Op, T> {
+    static void impl(T& c, const Op& op, float32_t* dst_ptr, int) {
+        op(c[0][0], dst_ptr);
+    }
+};
 
+template <int c_dim, int ow_remain, typename Op, typename T>
+inline void store_ocx_ow8_remain_static(T& c, const Op& op, float32_t* dst_ptr,
+                                        int ld_dst_oc) {
+    StoreOcxOw8Remain<c_dim, ow_remain, Op, T>::impl(c, op, dst_ptr, ld_dst_oc);
+}
 ////////////////////Store_OC8_OW8_Remain/////////////////////////
 
 template <int ow_remain, typename Op>
@@ -299,14 +459,15 @@ struct Store_OC8_OW8_Remain<1, Op> {
     }
 };
 
-template <int ow_remain, typename Op>
-inline void store_oc8_ow8_remain_static(int32x4_t c[2][8], const Op& op,
-                                        int8_t* dst_ptr, int ld_dst_oc) {
+///////////
+
+template <int ow_remain, typename Op, typename T, typename T2>
+inline void store_oc8_ow8_remain_static(T& c, const Op& op, T2 dst_ptr,
+                                        int ld_dst_oc) {
     Store_OC8_OW8_Remain<ow_remain, Op>::impl(c, op, dst_ptr, ld_dst_oc);
 }
 
-///////////////////////////////////////////////////////
-
+//////////////////////////////////////
 template <BiasMode bias_mode>
 inline void init_oc4_ow8(int32x4_t c[8], const int32_t* bias_ptr) {
     if (bias_mode == BiasMode::BROADCAST_CHANNEL_BIAS) {
@@ -337,6 +498,49 @@ inline void init_oc8_ow8(int32x4_t c[2][8], const int32_t* bias_ptr,
 #undef BAIS_INIT
     }
 }
+/////////////////////////init_ocx_ow8////////////////////
+template <int c_dim, BiasMode bias_mode, typename T, typename T2>
+struct InitOcxOw8 {
+    static void impl(T& c, T2 bias_ptr, int oc_step);
+};
+template <BiasMode bias_mode, typename T, typename T2>
+struct InitOcxOw8<2, bias_mode, T, T2> {
+    static void impl(T& c, const float32_t* bias_ptr, int oc_step) {
+        if (bias_mode == BiasMode::BROADCAST_CHANNEL_BIAS) {
+#define BAIS_INIT(step)               \
+    c[0][step] = vld1q_f32(bias_ptr); \
+    c[1][step] = vld1q_f32(bias_ptr + oc_step);
+            UNROLL_CALL_RAW(8, BAIS_INIT);
+#undef BAIS_INIT
+        } else {
+#define BAIS_INIT(step)          \
+    c[0][step] = vdupq_n_f32(0); \
+    c[1][step] = vdupq_n_f32(0);
+            UNROLL_CALL_RAW(8, BAIS_INIT);
+#undef BAIS_INIT
+        }
+    }
+};
+template <BiasMode bias_mode, typename T, typename T2>
+struct InitOcxOw8<1, bias_mode, T, T2> {
+    static void impl(T& c, const float32_t* bias_ptr, int) {
+        if (bias_mode == BiasMode::BROADCAST_CHANNEL_BIAS) {
+#define BAIS_INIT(step) c[0][step] = vld1q_f32(bias_ptr);
+            UNROLL_CALL_RAW(8, BAIS_INIT);
+#undef BAIS_INIT
+        } else {
+#define BAIS_INIT(step) c[0][step] = vdupq_n_f32(0);
+            UNROLL_CALL_RAW(8, BAIS_INIT);
+#undef BAIS_INIT
+        }
+    }
+};
+
+template <int c_dim, BiasMode bias_mode, typename T, typename T2>
+inline void init_ocx_ow8(T& c, T2 bias_ptr, int oc_step) {
+    InitOcxOw8<c_dim, bias_mode, T, T2>::impl(c, bias_ptr, oc_step);
+}
+/////////////////////init_ocx_ow4/////////////////////
 template <int c_dim, BiasMode bias_mode, typename T>
 struct InitOcxOw4 {
     static void impl(T& c, const int32_t* bias_ptr, int oc_step);
@@ -383,57 +587,54 @@ inline void init_ocx_ow4(T& c, const int32_t* bias_ptr, int oc_step) {
 }
 ///////////////////////////////////////
 template <int weight_number, int base_offset, int ptr_step, int oc_block,
-          typename Func, typename T, typename... XT>
+          typename Func, typename T, typename T2, typename... XT>
 struct LoadHelper {
-    static void impl(T& weight, const int8_t* ptr, int oc_offset, XT... args);
+    static void impl(T& weight, T2 ptr, int oc_offset, XT... args);
 };
 
 #define WEIGHT_CB(step) \
     src[step] = Func::impl(ptr + base_offset + step * ptr_step, args...);
 
-template <int base_offset, int ptr_step, typename Func, typename T,
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2,
           typename... XT>
-struct LoadHelper<1, base_offset, ptr_step, 0, Func, T, XT...> {
-    static void impl(T& src, const int8_t* ptr, int oc_offset, XT... args) {
+struct LoadHelper<1, base_offset, ptr_step, 0, Func, T, T2, XT...> {
+    static void impl(T& src, T2 ptr, int, XT... args) {
         UNROLL_CALL_RAW(1, WEIGHT_CB);
     }
 };
-template <int base_offset, int ptr_step, typename Func, typename T,
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2,
           typename... XT>
-struct LoadHelper<2, base_offset, ptr_step, 0, Func, T, XT...> {
-    static void impl(T& src, const int8_t* ptr, int oc_offset, XT... args) {
+struct LoadHelper<2, base_offset, ptr_step, 0, Func, T, T2, XT...> {
+    static void impl(T& src, T2 ptr, int, XT... args) {
         UNROLL_CALL_RAW(2, WEIGHT_CB);
     }
 };
 
-template <int base_offset, int ptr_step, typename Func, typename T,
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2,
           typename... XT>
-struct LoadHelper<3, base_offset, ptr_step, 0, Func, T, XT...> {
-    static void impl(T& src, const int8_t* ptr, int oc_offset, XT... args) {
+struct LoadHelper<3, base_offset, ptr_step, 0, Func, T, T2, XT...> {
+    static void impl(T& src, T2 ptr, int, XT... args) {
         UNROLL_CALL_RAW(3, WEIGHT_CB);
     }
 };
-template <int base_offset, int ptr_step, typename Func, typename T,
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2,
           typename... XT>
-struct LoadHelper<4, base_offset, ptr_step, 0, Func, T, XT...> {
-    static void impl(T& src, const int8_t* ptr, int oc_offset, XT... args) {
-        MEGDNN_MARK_USED_VAR(oc_offset);
+struct LoadHelper<4, base_offset, ptr_step, 0, Func, T, T2, XT...> {
+    static void impl(T& src, T2 ptr, int, XT... args) {
         UNROLL_CALL_RAW(4, WEIGHT_CB);
     }
 };
-template <int base_offset, int ptr_step, typename Func, typename T,
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2,
           typename... XT>
-struct LoadHelper<5, base_offset, ptr_step, 0, Func, T, XT...> {
-    static void impl(T& src, const int8_t* ptr, int oc_offset, XT... args) {
-        MEGDNN_MARK_USED_VAR(oc_offset);
+struct LoadHelper<5, base_offset, ptr_step, 0, Func, T, T2, XT...> {
+    static void impl(T& src, T2 ptr, int, XT... args) {
         UNROLL_CALL_RAW(5, WEIGHT_CB);
     }
 };
-template <int base_offset, int ptr_step, typename Func, typename T,
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2,
           typename... XT>
-struct LoadHelper<6, base_offset, ptr_step, 0, Func, T, XT...> {
-    static void impl(T& src, const int8_t* ptr, int oc_offset, XT... args) {
-        MEGDNN_MARK_USED_VAR(oc_offset);
+struct LoadHelper<6, base_offset, ptr_step, 0, Func, T, T2, XT...> {
+    static void impl(T& src, T2 ptr, int, XT... args) {
         UNROLL_CALL_RAW(6, WEIGHT_CB);
     }
 };
@@ -441,27 +642,36 @@ struct LoadHelper<6, base_offset, ptr_step, 0, Func, T, XT...> {
 
 #define WEIGHT_CB(step) \
     src[0][step] = Func::impl(ptr + base_offset + step * ptr_step);
-template <int base_offset, int ptr_step, typename Func, typename T>
-struct LoadHelper<1, base_offset, ptr_step, 1, Func, T> {
-    static void impl(T& src, const int8_t* ptr, int oc_offset) {
-        MEGDNN_MARK_USED_VAR(oc_offset);
-        UNROLL_CALL_RAW(1, WEIGHT_CB);
-    }
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<1, base_offset, ptr_step, 1, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int) { UNROLL_CALL_RAW(1, WEIGHT_CB); }
 };
-template <int base_offset, int ptr_step, typename Func, typename T>
-struct LoadHelper<2, base_offset, ptr_step, 1, Func, T> {
-    static void impl(T& src, const int8_t* ptr, int oc_offset) {
-        MEGDNN_MARK_USED_VAR(oc_offset);
-        UNROLL_CALL_RAW(2, WEIGHT_CB);
-    }
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<2, base_offset, ptr_step, 1, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int) { UNROLL_CALL_RAW(2, WEIGHT_CB); }
 };
 
-template <int base_offset, int ptr_step, typename Func, typename T>
-struct LoadHelper<3, base_offset, ptr_step, 1, Func, T> {
-    static void impl(T& src, const int8_t* ptr, int oc_offset) {
-        MEGDNN_MARK_USED_VAR(oc_offset);
-        UNROLL_CALL_RAW(3, WEIGHT_CB);
-    }
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<3, base_offset, ptr_step, 1, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int) { UNROLL_CALL_RAW(3, WEIGHT_CB); }
+};
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<4, base_offset, ptr_step, 1, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int) { UNROLL_CALL_RAW(4, WEIGHT_CB); }
+};
+
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<5, base_offset, ptr_step, 1, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int) { UNROLL_CALL_RAW(5, WEIGHT_CB); }
+};
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<6, base_offset, ptr_step, 1, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int) { UNROLL_CALL_RAW(6, WEIGHT_CB); }
+};
+
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<7, base_offset, ptr_step, 1, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int) { UNROLL_CALL_RAW(7, WEIGHT_CB); }
 };
 
 #undef WEIGHT_CB
@@ -470,40 +680,63 @@ struct LoadHelper<3, base_offset, ptr_step, 1, Func, T> {
     src[0][step] = Func::impl(ptr + base_offset + step * ptr_step); \
     src[1][step] = Func::impl(ptr + base_offset + step * ptr_step + oc_offset);
 
-template <int base_offset, int ptr_step, typename Func, typename T>
-struct LoadHelper<1, base_offset, ptr_step, 2, Func, T> {
-    static void impl(T& src, const int8_t* ptr, int oc_offset) {
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<1, base_offset, ptr_step, 2, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int oc_offset) {
         UNROLL_CALL_RAW(1, WEIGHT_CB);
     }
 };
-template <int base_offset, int ptr_step, typename Func, typename T>
-struct LoadHelper<2, base_offset, ptr_step, 2, Func, T> {
-    static void impl(T& src, const int8_t* ptr, int oc_offset) {
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<2, base_offset, ptr_step, 2, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int oc_offset) {
         UNROLL_CALL_RAW(2, WEIGHT_CB);
     }
 };
 
-template <int base_offset, int ptr_step, typename Func, typename T>
-struct LoadHelper<3, base_offset, ptr_step, 2, Func, T> {
-    static void impl(T& src, const int8_t* ptr, int oc_offset) {
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<3, base_offset, ptr_step, 2, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int oc_offset) {
         UNROLL_CALL_RAW(3, WEIGHT_CB);
     }
 };
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<4, base_offset, ptr_step, 2, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int oc_offset) {
+        UNROLL_CALL_RAW(4, WEIGHT_CB);
+    }
+};
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<5, base_offset, ptr_step, 2, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int oc_offset) {
+        UNROLL_CALL_RAW(5, WEIGHT_CB);
+    }
+};
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<6, base_offset, ptr_step, 2, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int oc_offset) {
+        UNROLL_CALL_RAW(6, WEIGHT_CB);
+    }
+};
+template <int base_offset, int ptr_step, typename Func, typename T, typename T2>
+struct LoadHelper<7, base_offset, ptr_step, 2, Func, T, T2> {
+    static void impl(T& src, T2 ptr, int oc_offset) {
+        UNROLL_CALL_RAW(7, WEIGHT_CB);
+    }
+};
 
 #undef WEIGHT_CB
 
 template <int weight_number, int base_offset, int ptr_step, int c_dim,
-          typename Func, typename T>
-inline void load_helper(T& weight, const int8_t* ptr, int oc_offset) {
-    LoadHelper<weight_number, base_offset, ptr_step, c_dim, Func, T>::impl(
+          typename Func, typename T, typename T2>
+inline void load_helper(T& weight, T2 ptr, int oc_offset) {
+    LoadHelper<weight_number, base_offset, ptr_step, c_dim, Func, T, T2>::impl(
             weight, ptr, oc_offset);
 }
 
 template <int weight_number, int base_offset, int ptr_step, int c_dim,
-          typename Func, typename T, typename... XT>
-inline void load_helper_x(T& weight, const int8_t* ptr, int oc_offset,
-                          XT... args) {
-    LoadHelper<weight_number, base_offset, ptr_step, c_dim, Func, T,
+          typename Func, typename T, typename T2, typename... XT>
+inline void load_helper_x(T& weight, T2 ptr, int oc_offset, XT... args) {
+    LoadHelper<weight_number, base_offset, ptr_step, c_dim, Func, T, T2,
                XT...>::impl(weight, ptr, oc_offset, args...);
 }
 

dnn/src/arm_common/conv_bias/neon_struct.h

@@ -34,6 +34,9 @@ struct Vmlal_s16 {
 struct Vld1q_s8 {
     static int8x16_t impl(const int8_t* ptr) { return vld1q_s8(ptr); }
 };
+struct Vld1q_f32 {
+    static float32x4_t impl(const float32_t* ptr) { return vld1q_f32(ptr); }
+};
 struct Vld1_s8 {
     static int8x8_t impl(const int8_t* ptr) { return vld1_s8(ptr); }
 };
@@ -50,5 +53,13 @@ struct Vldq_tbl_low_s8 {
 struct Vld1_dup_s8_s16 {
     static int16x8_t impl(const int8_t* ptr) { return vld1_dup_s8_s16(ptr); }
 };
+
+struct Vfmaq_laneq_f32 {
+    template <const int lane>
+    static float32x4_t impl(float32x4_t a, float32x4_t b, float32x4_t v) {
+        return vfmaq_laneq_f32(a, b, v, lane);
+    }
+};
+
 }  // namespace
 }  // namespace megdnn
\ No newline at end of file

dnn/src/arm_common/conv_bias/opr_impl.cpp

@@ -71,6 +71,7 @@ class ConvBiasImpl::AlgoPack : NonCopyableObj {
     AlgoF32DirectStride2 f32_direct_stride2_small_group{false};
     AlgoF32DirectStride1 f32_direct_stride1_large_group{true};
     AlgoF32DirectStride1 f32_direct_stride1_small_group{false};
+    AlgoF32DirectStride2NCHWNCHW44 f32_direct_stride2_nchw_nchw44;
     AlgoI8x8x16Direct i8x8x16_direct_large_group{true};
     AlgoI8x8x16Direct i8x8x16_direct_small_group{false};
     AlgoI8x8x16Stride2 i8x8x16_stride2_large_group{true};
@@ -123,6 +124,7 @@ public:
         direct_algos.emplace_back(&i8x8x16_stride2_filter2);
         direct_algos.emplace_back(&i8x8x16_stride2_large_group);
         direct_algos.emplace_back(&i8x8x16_stride2_small_group);
+        direct_algos.emplace_back(&f32_direct_stride2_nchw_nchw44);
         direct_algos.emplace_back(&f32_direct_stride1_large_group);
         direct_algos.emplace_back(&f32_direct_stride1_small_group);
         direct_algos.emplace_back(&f32_direct_stride2_large_group);

dnn/src/arm_common/conv_bias/opr_impl.h

@@ -67,6 +67,7 @@ private:
     class AlgoF32Direct;
     class AlgoF32DirectStride1;
     class AlgoF32DirectStride2;
+    class AlgoF32DirectStride2NCHWNCHW44;
     class AlgoI8x8x16Direct;
     class AlgoI8x8x16Stride2;
     class AlgoI8x8x16Stride2Filter2;

dnn/src/arm_common/elemwise_helper/kimpl/hswish.h

@@ -45,13 +45,17 @@ struct HSwishOp;
             vst1q_##_func_suffix(dst, vitem.val[0]);                           \
             vst1q_##_func_suffix(dst + SIMD_WIDTH, vitem.val[1]);              \
         }                                                                      \
+        void operator()(const _neon_type& src, _ctype* dst) const {            \
+            auto vitem = operator()(src);                                      \
+            vst1q_##_func_suffix(dst, vitem);                                  \
+        }                                                                      \
         _neon_type2 operator()(const _neon_type2& src) const {                 \
             auto val1 = src.val[0];                                            \
             auto val2 = src.val[1];                                            \
             H_SWISH_KERN(_func_suffix, val1, val2);                            \
             return {{val1, val2}};                                             \
         }                                                                      \
-        _neon_type operator()(const _neon_type& src) {                         \
+        _neon_type operator()(const _neon_type& src) const {                   \
             auto val_zero = vdupq_n_##_func_suffix(0.f);                       \
             auto val_six = vdupq_n_##_func_suffix(6.f);                        \
             auto val_three = vdupq_n_##_func_suffix(3.f);                      \
@@ -64,6 +68,7 @@ struct HSwishOp;
                                         val_rec_six);                          \
         }                                                                      \
     };
+
 OP(dt_float32, float32x4_t, float32x4x2_t, f32, 4)
 #if __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
 OP(__fp16, float16x8_t, float16x8x2_t, f16, 8)

dnn/src/arm_common/elemwise_helper/kimpl/none.h

@@ -6,7 +6,8 @@
  *
  * Unless required by applicable law or agreed to in writing,
  * software distributed under the License is distributed on an
- * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or
+ * implied.
  */
 #pragma once
 
@@ -30,6 +31,13 @@ struct NoneOp;
         using NoneOpBase::operator();                                        \
         constexpr static size_t SIMD_WIDTH = _simd_width;                    \
         _neon_type2 operator()(const _neon_type2& src) const { return src; } \
+        void operator()(const _neon_type2& src, _ctype* dst) const {         \
+            vst1q_##_func_suffix(dst, src.val[0]);                           \
+            vst1q_##_func_suffix(dst + SIMD_WIDTH, src.val[1]);              \
+        }                                                                    \
+        void operator()(const _neon_type& src, _ctype* dst) const {          \
+            vst1q_##_func_suffix(dst, src);                                  \
+        }                                                                    \
         _neon_type operator()(const _neon_type& src) const { return src; }   \
     };
 

dnn/src/arm_common/elemwise_helper/kimpl/relu.h

@@ -47,11 +47,16 @@ struct ReluOp;
             auto vitem1 = vmaxq_##_func_suffix(src.val[1], vzero);     \
             return {{vitem0, vitem1}};                                 \
         }                                                              \
+        void operator()(const _neon_type& src, _ctype* dst) const {    \
+            auto vitem = operator()(src);                              \
+            vst1q_##_func_suffix(dst, vitem);                          \
+        }                                                              \
         _neon_type operator()(const _neon_type& src) const {           \
             auto vzero = vdupq_n_##_func_suffix(0);                    \
             return vmaxq_##_func_suffix(src, vzero);                   \
         }                                                              \
     };
+
 OP(dt_float32, float32x4_t, float32x4x2_t, f32, 4)
 #if __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
 OP(__fp16, float16x8_t, float16x8x2_t, f16, 8)

dnn/src/arm_common/simd_macro/marm_neon.h

@@ -479,6 +479,39 @@ UNROLL_CALL_RAW(4, cb);
 #undef cb
 }  // namespace
 #define vdup_laneq_s16(vec, lane) Vdup_laneq_s16_armv7<lane>::impl(vec)
+namespace {
+template <int lane>
+struct Vfmap_laneq_f32_armv7 {
+    static float32x4_t impl(float32x4_t a, float32x4_t b, float32x4_t v);
+};
+
+template <>
+struct Vfmap_laneq_f32_armv7<0> {
+    static float32x4_t impl(float32x4_t a, float32x4_t b, float32x4_t v) {
+        return vmlaq_lane_f32(a, b, vget_low_f32(v), 0);
+    }
+};
+template <>
+struct Vfmap_laneq_f32_armv7<1> {
+    static float32x4_t impl(float32x4_t a, float32x4_t b, float32x4_t v) {
+        return vmlaq_lane_f32(a, b, vget_low_f32(v), 1);
+    }
+};
+template <>
+struct Vfmap_laneq_f32_armv7<2> {
+    static float32x4_t impl(float32x4_t a, float32x4_t b, float32x4_t v) {
+        return vmlaq_lane_f32(a, b, vget_high_f32(v), 0);
+    }
+};
+template <>
+struct Vfmap_laneq_f32_armv7<3> {
+    static float32x4_t impl(float32x4_t a, float32x4_t b, float32x4_t v) {
+        return vmlaq_lane_f32(a, b, vget_high_f32(v), 1);
+    }
+};
+}  // namespace
+#define vfmaq_laneq_f32(a, b, v, lane) \
+    Vfmap_laneq_f32_armv7<lane>::impl(a, b, v)
 
 #endif
 

dnn/test/arm_common/conv_bias.cpp

@@ -85,7 +85,7 @@ TEST_F(ARM_COMMON, CONV_BIAS_MATMUL_QU8) {
 
 #if MEGDNN_WITH_BENCHMARK
 
-static void benchmark_convbias(Handle* handle) {
+static void benchmark_convbias(Handle* handle, bool is_fp32 = false) {
     constexpr size_t RUNS = 30;
 
     Benchmarker<ConvBias> benchmarker_int(handle);
@@ -102,15 +102,25 @@ static void benchmark_convbias(Handle* handle) {
     Benchmarker<ConvBias> benchmarker_float(handle);
     benchmarker_float.set_display(false).set_times(RUNS);
     benchmarker_float.set_before_exec_callback(
-            conv_bias::ConvBiasAlgoChecker<ConvBias>(".+"));
+            conv_bias::ConvBiasAlgoChecker<ConvBias>(
+                    "IM2COLMATMUL:AARCH64_F32K8X12X1:192"));
 
     Benchmarker<ConvBias> benchmarker_int_nchw44(handle);
-    benchmarker_int_nchw44.set_times(RUNS)
-            .set_dtype(0, dtype::QuantizedS8(2.5))
-            .set_dtype(1, dtype::QuantizedS8(2.5))
-            .set_dtype(2, dtype::QuantizedS32(6.25))
-            .set_dtype(4, dtype::QuantizedS8(60.25))
-            .set_display(false);
+    if (is_fp32) {
+        benchmarker_int_nchw44.set_times(RUNS)
+                .set_dtype(0, dtype::Float32())
+                .set_dtype(1, dtype::Float32())
+                .set_dtype(2, dtype::Float32())
+                .set_dtype(4, dtype::Float32())
+                .set_display(false);
+    } else {
+        benchmarker_int_nchw44.set_times(RUNS)
+                .set_dtype(0, dtype::QuantizedS8(2.5))
+                .set_dtype(1, dtype::QuantizedS8(2.5))
+                .set_dtype(2, dtype::QuantizedS32(6.25))
+                .set_dtype(4, dtype::QuantizedS8(60.25))
+                .set_display(false);
+    }
     benchmarker_int_nchw44.set_before_exec_callback(
             conv_bias::ConvBiasAlgoChecker<ConvBias>(".+"));
 
@@ -151,7 +161,6 @@ static void benchmark_convbias(Handle* handle) {
         auto int_nchw44_used = benchmarker_int_nchw44.set_param(param).exec(
                                        {src, filter, bias, {}, dst}) /
                                RUNS;
-
         float computations = IC * (FS * FS) * dst.total_nr_elems() * 2 * 1e-6;
         printf("run: %s %s %s->%s \n", src.to_string().c_str(),
                filter.to_string().c_str(), bias.to_string().c_str(),
@@ -160,32 +169,42 @@ static void benchmark_convbias(Handle* handle) {
                computations / float_used);
         printf("int_nchw: %f ms %f Gflops, ", int_used,
                computations / int_used);
-        printf("int_nchw44: %f ms %f Gflops %f speedup, ", int_nchw44_used,
-               computations / int_nchw44_used, int_used / int_nchw44_used);
+        auto speed_up = int_used / int_nchw44_used;
+        if (is_fp32) {
+            speed_up = float_used / int_nchw44_used;
+            printf("fp32_nchw44: %f ms %f Gflops %f speedup, ", int_nchw44_used,
+                   computations / int_nchw44_used, speed_up);
+        } else {
+            printf("int_nchw44: %f ms %f Gflops %f speedup, ", int_nchw44_used,
+                   computations / int_nchw44_used, speed_up);
+        }
         printf("\n");
     };
-    run(1, 3, 32, 224, 224, 3, 2, true);
-    run(1, 3, 64, 224, 224, 5, 2, true);
-    run(1, 3, 64, 224, 224, 7, 2, true);
-    run(1, 3, 32, 224, 224, 7, 2, true);
-    for (size_t stride : {1, 2}) {
-        printf("stride %zu\n", stride);
-        for (size_t filter_size : {2, 3, 5, 7}) {
-            for (size_t img_size : {32}) {
-                for (size_t channel : {8, 16, 32, 64, 128, 256}) {
-                    run(1, channel, channel, img_size, img_size, filter_size,
-                        stride, false);
+
+    if (is_fp32) {
+        run(1, 3, 32, 224, 224, 3, 2, true);
+        run(1, 3, 64, 224, 224, 7, 2, true);
+    } else {
+        for (size_t stride : {1, 2}) {
+            printf("stride %zu\n", stride);
+            for (size_t filter_size : {2, 3, 5, 7}) {
+                for (size_t img_size : {32}) {
+                    for (size_t channel : {8, 16, 32, 64, 128, 256}) {
+                        run(1, channel, channel, img_size, img_size,
+                            filter_size, stride, false);
+                    }
                 }
             }
         }
     }
 }
 TEST_F(ARM_COMMON, BENCHMARK_CONVBIAS_NCHW44) {
-    benchmark_convbias(handle());
+    benchmark_convbias(handle(), true);
 }
 TEST_F(ARM_COMMON_MULTI_THREADS, BENCHMARK_CONVBIAS_NCHW44) {
-    benchmark_convbias(handle());
+    benchmark_convbias(handle(), true);
 }
+
 #endif
 TEST_F(ARM_COMMON, CONV_BIAS_MATMUL_QS8) {
     using namespace conv_bias;
@@ -1464,7 +1483,8 @@ TEST_F(ARM_COMMON, BENCHMARK_CONV_BIAS_QUINT8_STRIDE2_WITHDOTPROD) {
 #if MEGDNN_WITH_BENCHMARK
 
 namespace {
-std::vector<conv_bias::TestArg> get_conv_bias_1x1_benchmark_args(size_t pack_size = 1) {
+std::vector<conv_bias::TestArg> get_conv_bias_1x1_benchmark_args(
+        size_t pack_size = 1) {
     using namespace conv_bias;
     std::vector<TestArg> args;
     param::ConvBias param;
@@ -1474,15 +1494,17 @@ std::vector<conv_bias::TestArg> get_conv_bias_1x1_benchmark_args(size_t pack_siz
     param.pad_w = 0;
     param.nonlineMode = param::ConvBias::NonlineMode::IDENTITY;
     auto bench_case = [&](size_t OC, size_t IC, size_t H, size_t W) {
-        if(pack_size == 1)
+        if (pack_size == 1)
             args.emplace_back(param, TensorShape{1, IC, H, W},
-                          TensorShape{OC, IC, 1, 1}, TensorShape{});
+                              TensorShape{OC, IC, 1, 1}, TensorShape{});
         else {
-            if(pack_size == 4)
+            if (pack_size == 4)
                 param.format = param::ConvBias::Format::NCHW44;
-            args.emplace_back(param, TensorShape{1, IC / pack_size, H, W, pack_size},
-                          TensorShape{OC / pack_size, IC / pack_size, 1, 1, pack_size, pack_size}, 
-                          TensorShape{});
+            args.emplace_back(param,
+                              TensorShape{1, IC / pack_size, H, W, pack_size},
+                              TensorShape{OC / pack_size, IC / pack_size, 1, 1,
+                                          pack_size, pack_size},
+                              TensorShape{});
         }
     };
 

dnn/test/arm_common/conv_bias_multi_thread.cpp

@@ -78,9 +78,10 @@ std::vector<conv_bias::TestArg> get_nchw44_conv_bias_args(
     std::vector<TestArg> args;
 
     auto pack = [&](size_t n, size_t oc, size_t ic, size_t h, size_t w,
-                    size_t kernel, size_t stride, size_t group, NLMode nlmode) {
+                    size_t kernel, size_t stride, size_t group, NLMode nlmode,
+                    int any_pad = -1) {
         constexpr int pack_c = 4;
-        const size_t pad = no_pad ? 0 : kernel / 2;
+        const size_t pad = any_pad >= 0 ? any_pad : kernel / 2;
         auto bias_mode = no_bias ? megdnn::BiasMode::NO_BIAS
                                  : megdnn::BiasMode::BROADCAST_CHANNEL_BIAS;
         auto oc_per_group = oc / group;
@@ -90,7 +91,8 @@ std::vector<conv_bias::TestArg> get_nchw44_conv_bias_args(
                         ic_per_group > 0;
         bool nchw_disable = group > 1 || ic_per_group >= 4;
         bool nchw44_disable = ic_per_group % pack_c != 0;
-        if (!(ok_group)) {
+        bool invalid_pad = (w + 2 * pad < kernel) || (h + 2 * pad < kernel);
+        if (!(ok_group) || invalid_pad) {
             return;
         }
         if ((is_input_nchw && nchw_disable) ||
@@ -107,6 +109,7 @@ std::vector<conv_bias::TestArg> get_nchw44_conv_bias_args(
         param.pad_h = pad;
         param.pad_w = pad;
         param.nonlineMode = nlmode;
+
         auto src_tensor_shape = TensorShape{n, ic / pack_c, h, w, pack_c};
         auto weight_tensor_shape = TensorShape{
                 oc / pack_c, ic / pack_c, kernel_h, kernel_w, pack_c, pack_c};
@@ -338,6 +341,11 @@ TEST_F(ARM_COMMON_MULTI_THREADS, CONVBIAS_DIRECT_FP32_STR2_SMALL_GROUP) {
     check_conv_bias(get_conv_bias_args({2, 3, 5, 7}, 2, false, false, false),
                     handle(), "F32STRD2_SMALL_GROUP");
 }
+TEST_F(ARM_COMMON_MULTI_THREADS, CONVBIAS_NCHW_NCHW44_F32) {
+    check_conv_bias(
+            get_nchw44_conv_bias_args({3, 5, 7}, 2, false, false, false, true),
+            handle(), "F32_CONV_NCHW_NCHW44");
+}
 /**********************************F16 direct************************/
 #if __ARM_FEATURE_FP16_VECTOR_ARITHMETIC
 TEST_F(ARM_COMMON_MULTI_THREADS, CONVBIAS_DIRECT_FP16_LARGE_GROUP) {