separate non-Lua part of matrix code to a dedicated dir

29 files changed, 1805 insertions, 537 deletions

diff --git a/nerv/Makefile b/nerv/Makefile
index 4008453..06a439a 100644
--- a/nerv/Makefile
+++ b/nerv/Makefile
@@ -1,58 +1,65 @@
 .PHONY: build install clean
 SHELL := /bin/bash
 BUILD_DIR := $(CURDIR)/build
-OBJS := nerv.o lib/luaT/luaT.o lib/io/chunk_file.o common.o \
-		matrix/mmatrix.o matrix/cumatrix.o matrix/init.o matrix/cukernel.o \
+LIB_PATH := $(LUA_BINDIR)/../lib
+OBJ_DIR := $(BUILD_DIR)/objs
+CORE_OBJS := common.o lib/io/chunk_file.o lib/matrix/mmatrix.o lib/matrix/cumatrix.o lib/matrix/cukernel.o
+NERV_OBJS := nerv.o \
+		matrix/mmatrix.o matrix/cumatrix.o matrix/init.o \
 		io/init.o io/chunk_file.o \
 		examples/oop_example.o
-LIBS := libnerv.so
+LUAT_OBJS := lib/luaT/luaT.o
+CORE_OBJS := $(addprefix $(OBJ_DIR)/,$(CORE_OBJS))
+NERV_OBJS := $(addprefix $(OBJ_DIR)/,$(NERV_OBJS))
+LUAT_OBJS := $(addprefix $(OBJ_DIR)/,$(LUAT_OBJS))
+OBJS := $(CORE_OBJS) $(NERV_OBJS) $(LUAT_OBJS)
+LIBS := $(INST_LIBDIR)/libnerv.so $(LIB_PATH)/libnervcore.so $(LIB_PATH)/libluaT.so
 LUA_LIBS := matrix/init.lua io/init.lua init.lua \
 			layer/init.lua layer/affine.lua layer/sigmoid.lua layer/softmax_ce.lua \
 			layer/window.lua layer/bias.lua layer/combiner.lua layer/mse.lua \
 			nn/init.lua nn/layer_repo.lua nn/param_repo.lua nn/layer_dag.lua \
 			io/sgd_buffer.lua
 INCLUDE := -I $(LUA_INCDIR) -DLUA_USE_APICHECK
-CUDA_BASE := /usr/local/cuda-6.5
-#CUDA_BASE := /usr/local/cuda-5.0
+#CUDA_BASE := /usr/local/cuda-6.5
+CUDA_BASE := /usr/local/cuda-5.0
 CUDA_INCLUDE := -I $(CUDA_BASE)/include/
 INCLUDE += $(CUDA_INCLUDE)
 LDFLAGS := -L$(CUDA_BASE)/lib64/  -Wl,-rpath=$(CUDA_BASE)/lib64/ -lcudart -lcublas
 CFLAGS := -Wall -Wextra -O2
-OBJ_DIR := $(BUILD_DIR)/objs
-SUBDIR := matrix io layer examples nn lib/io lib/luaT
+SUBDIR := matrix io layer examples nn lib/io lib/luaT lib/matrix
 NVCC := $(CUDA_BASE)/bin/nvcc
 NVCC_FLAGS := -Xcompiler -fPIC,-Wall,-Wextra
 
 LUA_DIR = $(INST_LUADIR)/nerv
-OBJS := $(addprefix $(OBJ_DIR)/,$(OBJS))
 OBJ_SUBDIR := $(addprefix $(OBJ_DIR)/,$(SUBDIR))
 LUA_SUBDIR := $(addprefix $(LUA_DIR)/,$(SUBDIR))
-LIBS := $(addprefix $(INST_LIBDIR)/,$(LIBS))
 LUA_LIBS := $(addprefix $(LUA_DIR)/,$(LUA_LIBS))
 
 build: $(OBJ_DIR) $(OBJ_SUBDIR) $(OBJS)
+
 $(OBJ_DIR) $(LUA_DIR) $(OBJ_SUBDIR) $(LUA_SUBDIR):
 	-mkdir -p $@
 $(OBJ_DIR)/%.o: %.c $(patsubst /%.o,/%.c,$@)
 	gcc -c -o $@ $< $(INCLUDE) -fPIC $(CFLAGS)
-$(OBJ_DIR)/matrix/cukernel.o: matrix/cukernel.cu
+$(OBJ_DIR)/lib/matrix/cukernel.o: lib/matrix/cukernel.cu
 	$(NVCC) -c -o $@ $< $(INCLUDE) $(NVCC_FLAGS)
 $(LUA_DIR)/%.lua: %.lua
 	cp $< $@
 #$(OBJ_DIR)/luaT.o:
 #	gcc -c -o $@ luaT/luaT.c $(INCLUDE) -fPIC
-$(LIBS): $(OBJS)
-	gcc -shared -o $@ $(OBJS) $(LDFLAGS)
+$(LIB_PATH)/libnervcore.so: $(CORE_OBJS)
+	gcc -shared -o $@ $^ $(LDFLAGS)
+$(LIB_PATH)/libluaT.so: $(LUAT_OBJS)
+	gcc -shared -o $@ $^ $(LDFLAGS)
+$(INST_LIBDIR)/libnerv.so: $(NERV_OBJS) $(LIB_PATH)/libnervcore.so $(LIB_PATH)/libluaT.so
+	gcc -shared -o $@ $(NERV_OBJS) $(LDFLAGS) -Wl,-rpath=$(LIB_PATH) -L$(LIB_PATH) -lnervcore -lluaT
 
 $(OBJ_DIR)/matrix/cumatrix.o: matrix/generic/cumatrix.c matrix/generic/matrix.c matrix/generic/cukernel.cu
 $(OBJ_DIR)/matrix/mmatrix.o: matrix/generic/mmatrix.c matrix/generic/matrix.c
-$(OBJ_DIR)/matrix/cukernel.o: matrix/generic/cukernel.cu
-
-.PHONY: speech
 
-speech:
-	-mkdir -p build/objs/speech/tnet_io
-	$(MAKE) -C speech/ BUILD_DIR=$(BUILD_DIR) LIB_DIR=$(LIB_DIR) OBJ_DIR=$(CURDIR)/build/objs/speech/ LUA_DIR=$(LUA_DIR)
+$(OBJ_DIR)/lib/matrix/cumatrix.o: lib/matrix/generic/cumatrix.c lib/matrix/generic/matrix.c lib/matrix/generic/cukernel.cu
+$(OBJ_DIR)/lib/matrix/mmatrix.o: lib/matrix/generic/mmatrix.c lib/matrix/generic/matrix.c
+$(OBJ_DIR)/lib/matrix/cukernel.o: lib/matrix/generic/cukernel.cu
 
 clean:
 	-rm -rf $(OBJ_DIR)
diff --git a/nerv/common.c b/nerv/common.c
index b4e39e6..0584438 100644
--- a/nerv/common.c
+++ b/nerv/common.c
@@ -11,6 +11,31 @@ int nerv_error(lua_State *L, const char *err_mesg_fmt, ...) {
     return 0;
 }
 
+int nerv_error_status(lua_State *L, Status *status) {
+    const char *mmesg = NULL;
+    switch (status->err_code)
+    {
+        case MAT_GENERAL_ERR: mmesg = "general error"; break;
+        case MAT_INSUF_MEM: mmesg = "insufficient memory"; break;
+        case MAT_INVALID_FORMAT: mmesg = "invalid matrix format"; break;
+        case MAT_WRITE_ERROR: mmesg = "error while writing matrix"; break;
+        case MAT_INVALID_COPY_INTERVAL: mmesg = "invalid copy interval"; break;
+        case MAT_MISMATCH_DIM: mmesg = "mismatching matrix dimension"; break;
+        case MAT_WRONG_MULT_DIM: mmesg = "wrong multipier dimension"; break;
+        case MAT_ROW_VECTOR_EXP: mmesg = "row vector expected"; break;
+        case MAT_COL_VECTOR_EXP: mmesg = "column vector expected"; break;
+        case MAT_IDX_VECTOR_EXP: mmesg = "index vector expected"; break;
+        case MAT_INVALID_IDX: mmesg = "invalid index"; break;
+        case MAT_CUDA_ERR: mmesg = "cuda error"; break;
+        case MAT_CUBLAS_ERR: mmesg = "cublas error"; break;
+    }
+    if (status->msg)
+        nerv_error(L, "%s: %s @%s:%d", mmesg, status->msg,
+                                        status->file, status->lineno);
+    else
+        nerv_error(L, "%s @%s:%d", mmesg, status->file, status->lineno);
+}
+
 int nerv_error_method_not_implemented(lua_State *L) {
     return nerv_error(L, "method not implemented"); 
 }
diff --git a/nerv/common.h b/nerv/common.h
index 6657dc4..413ca51 100644
--- a/nerv/common.h
+++ b/nerv/common.h
@@ -7,6 +7,50 @@
 #include <stdio.h>
 #include <stdlib.h>
 
+enum {
+    MAT_NORMAL,
+    MAT_GENERAL_ERR,
+    MAT_INSUF_MEM,
+    MAT_INVALID_FORMAT,
+    MAT_WRITE_ERROR,
+    MAT_INVALID_COPY_INTERVAL,
+    MAT_MISMATCH_DIM,
+    MAT_WRONG_MULT_DIM,
+    MAT_ROW_VECTOR_EXP,
+    MAT_COL_VECTOR_EXP,
+    MAT_IDX_VECTOR_EXP,
+    MAT_INVALID_IDX,
+    MAT_CUDA_ERR,
+    MAT_CUBLAS_ERR
+};
+
+typedef struct Status {
+    int err_code;
+    const char *file;
+    int lineno;
+    const char *msg;
+} Status;
+
+#define NERV_SET_STATUS(status, code, m) \
+    do { \
+        (status)->err_code = code; \
+        (status)->msg = m; \
+        (status)->file = __FILE__; \
+        (status)->lineno = __LINE__; \
+    } while (0)
+
+#define NERV_EXIT_STATUS(status, code, msg) \
+    do { \
+        NERV_SET_STATUS(status, code, msg); \
+        return; \
+    } while (0)
+
+#define NERV_LUA_CHECK_STATUS(L, status) \
+    do { \
+        if (status.err_code != MAT_NORMAL) \
+            nerv_error_status(L, &status); \
+    } while (0)
+
 typedef struct HashNode {
     const char *key;
     void *val;
@@ -31,6 +75,7 @@ void hashmap_clear(HashMap *h);
 size_t bkdr_hash(const char *key);
 
 int nerv_error(lua_State *L, const char *err_mesg_fmt, ...); 
+int nerv_error_status(lua_State *L, Status *status); 
 int nerv_error_method_not_implemented(lua_State *L);
 void luaN_append_methods(lua_State *L, const luaL_Reg *mlist);
 #endif
diff --git a/nerv/lib/io/chunk_file.c b/nerv/lib/io/chunk_file.c
index a305962..e70ffc9 100644
--- a/nerv/lib/io/chunk_file.c
+++ b/nerv/lib/io/chunk_file.c
@@ -91,6 +91,7 @@ static ChunkFile *open_write(const char *fn, int *status) {
     }
     cfp = (ChunkFile *)malloc(sizeof(ChunkFile));
     cfp->fp = fp;
+    cfp->info = NULL;
     cfp->status = CF_WRITE;
     *status = CF_NORMAL;
     return cfp;
@@ -111,8 +112,6 @@ static ChunkFile *open_read(const char *fn, int *status) {
         return NULL;
     }
     cfp = (ChunkFile *)malloc(sizeof(ChunkFile));
-    cfp->fp = fp;
-    cfp->status = CF_READ;
     offset = ftello(fp);
     /* fprintf(stderr, "%d\n", (int)offset); */
     for (i = 0;; offset += chunk_len, i++)
@@ -144,7 +143,9 @@ static ChunkFile *open_read(const char *fn, int *status) {
         head = cip;
     }
     *status = CF_NORMAL;
+    cfp->fp = fp;
     cfp->info = head;
+    cfp->status = CF_READ;
     return cfp;
 }
 
@@ -208,13 +209,16 @@ void nerv_chunk_file_close(ChunkFile *cfp) {
 
 void nerv_chunk_file_destroy(ChunkFile *cfp) {
     ChunkInfo *i, *ni;
-    if (cfp->status != CF_CLOSED) fclose(cfp->fp);
-    for (i = cfp->info; i; i = ni)
+    if (cfp->info)
     {
-        ni = i->next;
-        free(i->metadata);
-        free(i);
+        for (i = cfp->info; i; i = ni)
+        {
+            ni = i->next;
+            free(i->metadata);
+            free(i);
+        }
     }
+    if (cfp->status != CF_CLOSED) fclose(cfp->fp);
     free(cfp);
 }
 
diff --git a/nerv/matrix/cuda_helper.h b/nerv/lib/matrix/cuda_helper.h
index fde6f18..8041efb 100644
--- a/nerv/matrix/cuda_helper.h
+++ b/nerv/lib/matrix/cuda_helper.h
@@ -4,33 +4,68 @@
 #include "cuda_runtime.h"
 #include "driver_types.h"
 #include "cublas_v2.h"
-#define CUBLAS_SAFE_SYNC_CALL(call) \
+
+#define CUBLAS_SAFE_SYNC_CALL_RET(call, status) \
+    do { \
+        cublasStatus_t  err = (call); \
+        if (err != CUBLAS_STATUS_SUCCESS) \
+        { \
+            NERV_SET_STATUS(status, MAT_CUBLAS_ERR, cublasGetErrorString(err)); \
+            return 0; \
+        } \
+        cudaDeviceSynchronize(); \
+    } while (0)
+
+#define CUBLAS_SAFE_SYNC_CALL(call, status) \
     do { \
         cublasStatus_t  err = (call); \
         if (err != CUBLAS_STATUS_SUCCESS) \
-            nerv_error(L, "cumatrix cublas error: %s at %s:%d", \
-                        cublasGetErrorString(err), __FILE__, __LINE__); \
+            NERV_EXIT_STATUS(status, MAT_CUBLAS_ERR, cublasGetErrorString(err)); \
         cudaDeviceSynchronize(); \
     } while (0)
 
-#define CUDA_SAFE_CALL(call) \
+#define CUDA_SAFE_CALL_RET(call, status) \
+    do { \
+        cudaError_t err = (call); \
+        if (err != cudaSuccess) \
+        { \
+            NERV_SET_STATUS(status, MAT_CUDA_ERR, cudaGetErrorString(err)); \
+            return 0; \
+        } \
+    } while (0)
+
+#define CUDA_SAFE_CALL(call, status) \
     do { \
         cudaError_t err = (call); \
         if (err != cudaSuccess) \
-            nerv_error(L, "cumatrix CUDA error: %s at %s:%d", \
-                            cudaGetErrorString(err), __FILE__, __LINE__); \
+            NERV_EXIT_STATUS(status, MAT_CUDA_ERR, cudaGetErrorString(err)); \
+    } while (0)
+
+#define CUDA_SAFE_SYNC_CALL(call, status) \
+    do { \
+        CUDA_SAFE_CALL(call, status); \
+        cudaDeviceSynchronize(); \
     } while (0)
 
-#define CUDA_SAFE_SYNC_CALL(call) \
+#define CUDA_SAFE_SYNC_CALL_RET(call, status) \
     do { \
-        CUDA_SAFE_CALL(call); \
+        CUDA_SAFE_CALL_RET(call, status); \
         cudaDeviceSynchronize(); \
     } while (0)
 
-#define CHECK_SAME_DIMENSION(a, b) \
+#define CHECK_SAME_DIMENSION(a, b, status) \
+    do { \
+        if (!(a->nrow == b->nrow && a->ncol == b->ncol)) \
+            NERV_EXIT_STATUS(status, MAT_MISMATCH_DIM, 0); \
+    } while (0)
+
+#define CHECK_SAME_DIMENSION_RET(a, b, status) \
     do { \
         if (!(a->nrow == b->nrow && a->ncol == b->ncol)) \
-            nerv_error(L, "matrices should be of the same dimension"); \
+        { \
+            NERV_SET_STATUS(status, MAT_MISMATCH_DIM, 0); \
+            return 0; \
+        } \
     } while (0)
 
 static const char *cublasGetErrorString(cublasStatus_t err) {
diff --git a/nerv/matrix/cukernel.cu b/nerv/lib/matrix/cukernel.cu
index a19030a..a19030a 100644
--- a/nerv/matrix/cukernel.cu
+++ b/nerv/lib/matrix/cukernel.cu
diff --git a/nerv/matrix/cukernel.h b/nerv/lib/matrix/cukernel.h
index 8a1494f..8a1494f 100644
--- a/nerv/matrix/cukernel.h
+++ b/nerv/lib/matrix/cukernel.h
diff --git a/nerv/lib/matrix/cumatrix.c b/nerv/lib/matrix/cumatrix.c
new file mode 100644
index 0000000..9641197
--- /dev/null
+++ b/nerv/lib/matrix/cumatrix.c
@@ -0,0 +1,69 @@
+#define NERV_GENERIC_CUMATRIX
+#include "../../common.h"
+#include "cuda_helper.h"
+#include <string.h>
+#define PROFILE_HASHMAP_SIZE 123457
+static cublasHandle_t cublas_handle;
+static cudaEvent_t profile_start, profile_stop;
+static HashMap *profile;
+
+void nerv_cumatrix_print_profile() {
+    size_t i;
+    fprintf(stderr, "*** [nerv cumatrix profile] **\n");
+    for (i = 0; i < profile->size; i++)
+    {
+        HashNode *ptr;
+        for (ptr = profile->bucket[i]; ptr; ptr = ptr->next)
+        {
+            fprintf(stderr, "%s:\t%.6f\n", ptr->key, *(float *)ptr->val);
+        }
+    }
+}
+
+void nerv_cumatrix_clear_profile() {
+    hashmap_clear(profile);
+}
+
+void accu_profile(const char *name, float delta) {
+    float *val = hashmap_getval(profile, name);
+    if (!val)
+    {
+        val = malloc(sizeof(float));
+        *val = 0;
+        hashmap_setval(profile, name, val);
+    }
+    *val += delta;
+}
+
+void nerv_cumatrix_init() {
+    cublasCreate(&cublas_handle);
+    cudaEventCreate(&profile_start);
+    cudaEventCreate(&profile_stop);
+    profile = hashmap_create(PROFILE_HASHMAP_SIZE, bkdr_hash, strcmp);
+}
+
+#define MATRIX_USE_FLOAT
+#define cuda_matrix_(NAME) cuda_matrix_float_##NAME
+#define nerv_matrix_(NAME) nerv_matrix_cuda_float_##NAME
+#define cudak_(NAME) cudak_float_ ## NAME
+#define NERV_CUBLAS_(NAME) cublasS##NAME
+#define MATRIX_CUMATRIX_HOST_TNAME nerv_matrix_host_float_tname
+#include "generic/cumatrix.c"
+#undef NERV_CUBLAS_
+#undef cudak_
+#undef nerv_matrix_
+#undef cuda_matrix_
+#undef MATRIX_USE_FLOAT
+#undef MATRIX_ELEM
+#undef MATRIX_ELEM_PTR
+#undef MATRIX_ELEM_FMT
+#undef MATRIX_ELEM_WRITE_FMT
+#undef MATRIX_CUMATRIX_HOST_TNAME
+
+#define MATRIX_USE_DOUBLE
+#define cuda_matrix_(NAME) cuda_matrix_double_##NAME
+#define nerv_matrix_(NAME) nerv_matrix_cuda_double_##NAME
+#define cudak_(NAME) cudak_double_ ## NAME
+#define NERV_CUBLAS_(NAME) cublasD##NAME
+#define MATRIX_CUMATRIX_HOST_TNAME nerv_matrix_host_double_tname
+#include "generic/cumatrix.c"
diff --git a/nerv/lib/matrix/cumatrix.h b/nerv/lib/matrix/cumatrix.h
new file mode 100644
index 0000000..9f71507
--- /dev/null
+++ b/nerv/lib/matrix/cumatrix.h
@@ -0,0 +1,6 @@
+#ifndef NERV_CUMATRIX_H
+#define NERV_CUMATRIX_H
+void nerv_cumatrix_print_profile();
+void nerv_cumatrix_clear_profile();
+void nerv_cumatrix_init();
+#endif
diff --git a/nerv/lib/matrix/generic/cukernel.cu b/nerv/lib/matrix/generic/cukernel.cu
new file mode 100644
index 0000000..6111193
--- /dev/null
+++ b/nerv/lib/matrix/generic/cukernel.cu
@@ -0,0 +1,571 @@
+#ifdef NERV_GENERIC_CUKERNEL
+#include <assert.h>
+#include <stdio.h>
+#include "../matrix.h"
+#include "cuda.h"
+#include "float.h"
+#define CUDA_THREADS_N 16
+#define CUDA_THREADS_NN ((CUDA_THREADS_N) * (CUDA_THREADS_N))
+#define CEIL_DIV(a, b) (((a) + (b) - 1) / (b))
+__global__ void cudak_(log_elem)(const MATRIX_ELEM *a, MATRIX_ELEM *b, 
+                                int nrow, int ncol, int stride) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    long idx;
+    MATRIX_ELEM tmp;
+    if (i >= nrow || j >= ncol) return;
+    idx = j + i * stride;
+    tmp = a[idx];
+    if(tmp < FLT_MIN) tmp = FLT_MIN;
+    b[idx] = log(tmp);
+}
+
+__global__ void cudak_(mul_elem)(const MATRIX_ELEM *a, const MATRIX_ELEM *b,
+                                MATRIX_ELEM *c, 
+                                int nrow, int ncol, int stride) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    long idx;
+    if (i >= nrow || j >= ncol) return;
+    idx = j + i * stride;
+    c[idx] = a[idx] * b[idx];
+}
+
+__global__ void cudak_(sigmoid)(const MATRIX_ELEM *a, MATRIX_ELEM *b,
+                        int nrow, int ncol, int stride) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    long idx;
+    if (i >= nrow || j >= ncol) return;
+    idx = j + i * stride;
+    b[idx] = 1.0 / (1.0 + exp(-a[idx]));
+}
+
+__global__ void cudak_(sigmoid_grad)(const MATRIX_ELEM *output,
+                                    const MATRIX_ELEM *err,
+                                    MATRIX_ELEM *nerr,
+                                    int nrow, int ncol, int stride) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    long idx;
+    if (i >= nrow || j >= ncol) return;
+    idx = j + i * stride;
+    nerr[idx] = output[idx] * (1.0 - output[idx]) * err[idx];
+}
+
+__global__ void cudak_(softmax_final)(const MATRIX_ELEM *a, MATRIX_ELEM *b,
+                        const MATRIX_ELEM *max, const MATRIX_ELEM *deno,
+                        int nrow, int ncol, int stride, int mstride) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    long idx;
+    if (i >= nrow || j >= ncol) return;
+    idx = j + i * stride;
+    b[idx] = exp(a[idx] - max[0 + i * mstride]) / deno[0 + i * mstride];
+}
+
+__global__ void cudak_(block_reduce_rowsum)(const MATRIX_ELEM *input,
+                                            MATRIX_ELEM *output,
+                                            const int istride, const int ostride,
+                                            const int n) {
+    extern __shared__ MATRIX_ELEM cudak_(arr)[];
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    cudak_(arr)[threadIdx.x] = j < n ? input[j + istride * blockIdx.y] : 0;
+    __syncthreads();
+    for (int offset = blockDim.x >> 1;  offset; offset >>= 1)
+    {
+        if (threadIdx.x < offset)
+            cudak_(arr)[threadIdx.x] += cudak_(arr)[threadIdx.x + offset];
+        __syncthreads();
+    }
+    if (threadIdx.x == 0)
+        output[blockIdx.x + ostride * blockIdx.y] = cudak_(arr)[0];
+}
+
+__global__ void cudak_(block_reduce_colsum)(const MATRIX_ELEM *input,
+                                MATRIX_ELEM *output,
+                                const int istride, const int ostride,
+                                const int n) {
+    extern __shared__ MATRIX_ELEM cudak_(arr)[];
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    cudak_(arr)[threadIdx.y] = i < n ? input[blockIdx.x + istride * i] : 0;
+    __syncthreads();
+    for (int offset = blockDim.y >> 1;  offset; offset >>= 1)
+    {
+        if (threadIdx.y < offset)
+            cudak_(arr)[threadIdx.y] += cudak_(arr)[threadIdx.y + offset];
+        __syncthreads();
+    }
+    if (threadIdx.y == 0)
+        output[blockIdx.x + ostride * blockIdx.y] = cudak_(arr)[0];
+}
+
+__global__ void cudak_(block_reduce_colsame)(const MATRIX_ELEM *input,
+                                            const MATRIX_ELEM *ref_input,
+                                            MATRIX_ELEM *output,
+                                            const int istride, const int ostride,
+                                            const int n) {
+    extern __shared__ MATRIX_ELEM cudak_(arr)[];
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    cudak_(arr)[threadIdx.y] = (i < n && input[blockIdx.x + istride * i] == \
+                                        ref_input[blockIdx.x + istride * i]) ? 1.0 : 0;
+    __syncthreads();
+    for (int offset = blockDim.y >> 1;  offset; offset >>= 1)
+    {
+        if (threadIdx.y < offset)
+            cudak_(arr)[threadIdx.y] += cudak_(arr)[threadIdx.y + offset];
+        __syncthreads();
+    }
+    if (threadIdx.y == 0)
+        output[blockIdx.x + ostride * blockIdx.y] = cudak_(arr)[0];
+}
+
+__global__ void cudak_(block_reduce_softmax_rowsum)(const MATRIX_ELEM *input,
+                                        MATRIX_ELEM *output,
+                                        const MATRIX_ELEM *max,
+                                        const int istride, const int ostride,
+                                        const int mstride, const int n) {
+    extern __shared__ MATRIX_ELEM cudak_(arr)[];
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    cudak_(arr)[threadIdx.x] = j < n ? exp(input[j + istride * blockIdx.y] - \
+                                    max[0 + mstride * blockIdx.y]) : 0;
+    __syncthreads();
+    for (int offset = blockDim.x >> 1;  offset; offset >>= 1)
+    {
+        if (threadIdx.x < offset)
+            cudak_(arr)[threadIdx.x] += cudak_(arr)[threadIdx.x + offset];
+        __syncthreads();
+    }
+    if (threadIdx.x == 0)
+        output[blockIdx.x + ostride * blockIdx.y] = cudak_(arr)[0];
+}
+
+__global__ void cudak_(block_reduce_rowmax)(const MATRIX_ELEM *input,
+                                            MATRIX_ELEM *output,
+                                            const int istride, const int ostride,
+                                            const int n) {
+    extern __shared__ MATRIX_ELEM cudak_(arr)[];
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    cudak_(arr)[threadIdx.x] = j < n ? input[j + istride * blockIdx.y] : -FLT_MAX;
+    __syncthreads();
+    for (int offset = blockDim.x >> 1;  offset; offset >>= 1)
+    {
+        if (threadIdx.x < offset)
+        {
+            MATRIX_ELEM l = cudak_(arr)[threadIdx.x],
+                        r = cudak_(arr)[threadIdx.x + offset];
+            if (r > l)
+                cudak_(arr)[threadIdx.x] = r;
+        }
+        __syncthreads();
+    }
+    if (threadIdx.x == 0)
+        output[blockIdx.x + ostride * blockIdx.y] = cudak_(arr)[0];
+}
+
+__global__ void cudak_(block_reduce_rowmax_idx)(const MATRIX_ELEM *input,
+                                                const MATRIX_ELEM *idx_input,
+                                                MATRIX_ELEM *output,
+                                                MATRIX_ELEM *idx_output,
+                                                const int istride, const int ostride,
+                                                const int n) {
+    extern __shared__ MATRIX_ELEM cudak_(arr)[];
+    MATRIX_ELEM *arr_val = cudak_(arr);
+    MATRIX_ELEM *arr_idx = arr_val + blockDim.x;
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    arr_val[threadIdx.x] = j < n ? input[j + istride * blockIdx.y] : -FLT_MAX;
+    arr_idx[threadIdx.x] = j < n ? idx_input[j + istride * blockIdx.y] : 0;
+    __syncthreads();
+    for (int offset = blockDim.x >> 1;  offset; offset >>= 1)
+    {
+        if (threadIdx.x < offset)
+        {
+            MATRIX_ELEM l = arr_val[threadIdx.x],
+                        r = arr_val[threadIdx.x + offset];
+            if (r > l)
+            {
+                arr_val[threadIdx.x] = r;
+                arr_idx[threadIdx.x] = arr_idx[threadIdx.x + offset];
+            }
+        }
+        __syncthreads();
+    }
+    if (threadIdx.x == 0)
+    {
+        output[blockIdx.x + ostride * blockIdx.y] = arr_val[0];
+        idx_output[blockIdx.x + ostride * blockIdx.y] = arr_idx[0];
+    }
+}
+
+__global__ void cudak_(add_row)(const MATRIX_ELEM *a, MATRIX_ELEM *b,
+                                int nrow, int ncol, int stride, double beta) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    if (i >= nrow || j >= ncol) return;
+    b[j + i * stride] += beta * a[j];
+}
+
+__global__ void cudak_(fill)(MATRIX_ELEM *a,
+                            int nrow, int ncol, int stride, double val) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    if (i >= nrow || j >= ncol) return;
+    a[j + i * stride] = val;
+}
+
+__global__ void cudak_(expand_frm)(const MATRIX_ELEM *a, MATRIX_ELEM *b,
+                                    int nrow, int ncol,
+                                    int enrow, int encol,
+                                    int stride, int estride,
+                                    int context) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    int ridx;
+    if (i >= enrow || j >= encol) return;
+    ridx = i + j / ncol - context;
+    if (ridx < 0) ridx = 0;
+    else if (ridx >= nrow) ridx = nrow - 1;
+    b[j + i * estride] = a[j % ncol + ridx * stride];
+}
+
+__global__ void cudak_(rearrange_frm)(const MATRIX_ELEM *a, MATRIX_ELEM *b,
+                                    int nrow, int ncol,
+                                    int stride, int step, int orig_dim) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    if (i >= nrow || j >= ncol) return;
+    b[j + i * stride] = a[j / step + (j % step) * orig_dim + i * stride];
+}
+
+__global__ void cudak_(scale_rows_by_col)(const MATRIX_ELEM *a, MATRIX_ELEM *b,
+                                        int nrow, int ncol,
+                                        int astride, int bstride) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    if (i >= nrow || j >= ncol) return;
+    b[j + i * bstride] *= a[i * astride];
+}
+
+__global__ void cudak_(scale_rows_by_row)(const MATRIX_ELEM *a, MATRIX_ELEM *b,
+                                        int nrow, int ncol,
+                                        int stride) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    if (i >= nrow || j >= ncol) return;
+    b[j + i * stride] *= a[j];
+}
+
+__global__ void cudak_(decompress)(const MATRIX_ELEM *a, MATRIX_ELEM *b,
+                                    int nrow, int ncol,
+                                    int stride_a, int stride_b) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    if (i >= nrow || j >= ncol) return;
+    b[lrintf(a[j + i * stride_a]) + i * stride_b] = 1.0;
+}
+
+__global__ void cudak_(gen_col_idx)(MATRIX_ELEM *b,
+                                    int nrow, int ncol, int stride) {
+    int j = blockIdx.x * blockDim.x + threadIdx.x;
+    int i = blockIdx.y * blockDim.y + threadIdx.y;
+    if (i >= nrow || j >= ncol) return;
+    b[j + i * stride] = j;
+}
+
+extern "C" {
+#include "../cukernel.h"
+    void cudak_(cuda_log_elem)(const Matrix *a, Matrix *b) {
+        dim3 threadsPerBlock(CUDA_THREADS_N, CUDA_THREADS_N);
+        dim3 numBlocks(CEIL_DIV(b->ncol, threadsPerBlock.x),
+                CEIL_DIV(b->nrow, threadsPerBlock.y));
+        cudak_(log_elem)<<<numBlocks, threadsPerBlock>>> \
+            (MATRIX_ELEM_PTR(a), MATRIX_ELEM_PTR(b),
+             b->nrow, b->ncol, b->stride / sizeof(MATRIX_ELEM));
+        cudaStreamSynchronize(0);
+    }
+
+    void cudak_(cuda_mul_elem)(const Matrix *a, const Matrix *b,
+                                Matrix *c) {
+        dim3 threadsPerBlock(CUDA_THREADS_N, CUDA_THREADS_N);
+        dim3 numBlocks(CEIL_DIV(b->ncol, threadsPerBlock.x),
+                CEIL_DIV(b->nrow, threadsPerBlock.y));
+        cudak_(mul_elem)<<<numBlocks, threadsPerBlock>>> \
+            (MATRIX_ELEM_PTR(a), MATRIX_ELEM_PTR(b),
+             MATRIX_ELEM_PTR(c),
+             b->nrow, b->ncol, b->stride / sizeof(MATRIX_ELEM));
+        cudaStreamSynchronize(0);
+    }
+
+    void cudak_(cuda_sigmoid)(const Matrix *a, Matrix *b) {
+        dim3 threadsPerBlock(CUDA_THREADS_N, CUDA_THREADS_N);
+        dim3 numBlocks(CEIL_DIV(b->ncol, threadsPerBlock.x),
+                CEIL_DIV(b->nrow, threadsPerBlock.y));
+        cudak_(sigmoid)<<<numBlocks, threadsPerBlock>>> \
+            (MATRIX_ELEM_PTR(a), MATRIX_ELEM_PTR(b), b->nrow, b->ncol,
+            b->stride / sizeof(MATRIX_ELEM));
+        cudaStreamSynchronize(0);
+    }
+
+    void cudak_(cuda_sigmoid_grad)(const Matrix *output,
+                                    const Matrix *err, Matrix *nerr) {
+        dim3 threadsPerBlock(CUDA_THREADS_N, CUDA_THREADS_N);
+        dim3 numBlocks(CEIL_DIV(nerr->ncol, threadsPerBlock.x),
+                CEIL_DIV(nerr->nrow, threadsPerBlock.y));
+        cudak_(sigmoid_grad)<<<numBlocks, threadsPerBlock>>> \
+            (MATRIX_ELEM_