From e6ffb126a539e80207ff48b3499bc51843023902 Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Wed, 12 Feb 2025 21:10:22 -0500
Subject: [PATCH 01/24] dev environment
---
compile_flags.txt | 2 ++
dev.sh | 35 +++++++++++++++++++++++++++++++++++
2 files changed, 37 insertions(+)
create mode 100644 compile_flags.txt
create mode 100755 dev.sh
diff --git a/compile_flags.txt b/compile_flags.txt
new file mode 100644
index 0000000..449b102
--- /dev/null
+++ b/compile_flags.txt
@@ -0,0 +1,2 @@
+-I./include
+-I./build/include
diff --git a/dev.sh b/dev.sh
new file mode 100755
index 0000000..a86d79b
--- /dev/null
+++ b/dev.sh
@@ -0,0 +1,35 @@
+#!/usr/bin/env sh
+
+CONTAINER='binsparse-reference-c'
+IMAGE='docker.io/junikimm717/nvim2025:finch'
+DIR="$(realpath "$(dirname "$0")")"
+
+case "$1" in
+ pull|p)
+ podman pull "$IMAGE"
+ (podman container ls | grep "$CONTAINER" > /dev/null 2>&1) && podman container rm -fv "$CONTAINER"
+ ;;
+ clear|c)
+ (podman container ls -a | grep "$CONTAINER" > /dev/null 2>&1)\
+ && {
+ podman container kill "$CONTAINER" > /dev/null 2>&1;
+ podman container rm "$CONTAINER"
+ }
+ ;;
+ *)
+ set +x
+ if ! (podman container ls -a | grep "$CONTAINER" > /dev/null 2>&1); then
+ podman run\
+ -dt\
+ --name "$CONTAINER"\
+ --group-add keep-groups\
+ -v "$DIR:/workspace"\
+ --privileged\
+ --rm\
+ "$IMAGE"
+ fi || exit 1
+ podman exec\
+ -e ENV=/root/.profile\
+ -it "$CONTAINER" /bin/bash
+ ;;
+esac
From 388acbdd896989364f25fe8ffd3d76072c803140 Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Sat, 15 Feb 2025 18:20:25 -0500
Subject: [PATCH 02/24] read tensor implemented
---
compile_flags.txt | 2 +
include/binsparse/read_tensor.h | 17 +++
include/binsparse/tensor.h | 84 ++++++++++++++
src/CMakeLists.txt | 1 +
src/read_matrix.c | 1 +
src/read_tensor.c | 194 ++++++++++++++++++++++++++++++++
6 files changed, 299 insertions(+)
create mode 100644 include/binsparse/read_tensor.h
create mode 100644 include/binsparse/tensor.h
create mode 100644 src/read_tensor.c
diff --git a/compile_flags.txt b/compile_flags.txt
index 449b102..ba24856 100644
--- a/compile_flags.txt
+++ b/compile_flags.txt
@@ -1,2 +1,4 @@
-I./include
-I./build/include
+-DBSP_USE_HDF5
+-I/usr/include/hdf5/serial
diff --git a/include/binsparse/read_tensor.h b/include/binsparse/read_tensor.h
new file mode 100644
index 0000000..0e9a2f9
--- /dev/null
+++ b/include/binsparse/read_tensor.h
@@ -0,0 +1,17 @@
+#pragma once
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifdef BSP_USE_HDF5
+#include <hdf5.h>
+
+bsp_tensor_t bsp_read_tensor_from_group(hid_t f);
+#endif
+
+bsp_tensor_t bsp_read_tensor(const char* file_name, const char* group);
+
+#ifdef __cplusplus
+}
+#endif
diff --git a/include/binsparse/tensor.h b/include/binsparse/tensor.h
new file mode 100644
index 0000000..e70ef04
--- /dev/null
+++ b/include/binsparse/tensor.h
@@ -0,0 +1,84 @@
+#pragma once
+
+#include <binsparse/array.h>
+#include <binsparse/structure.h>
+
+typedef enum {
+ BSP_TENSOR_SPARSE = 0,
+ BSP_TENSOR_DENSE = 1,
+ BSP_TENSOR_ELEMENT = 2,
+} bsp_level_kind_t;
+
+typedef struct {
+ bsp_level_kind_t kind;
+ // data here should be bsp_element_t*, bsp_sparse_t*, or bsp_dense_t*
+ void* data;
+} bsp_level_t;
+
+// corresponds to BSP_TENSOR_ELEMENT
+typedef struct {
+ bsp_array_t values;
+} bsp_element_t;
+
+// corresponds to BSP_TENSOR_DENSE
+typedef struct {
+ int rank;
+ bsp_array_t pointers_to;
+ bsp_array_t* indices;
+ bsp_level_t* child;
+} bsp_sparse_t;
+
+typedef struct {
+ int rank;
+ bsp_level_t* child;
+} bsp_dense_t;
+
+typedef struct {
+ int rank;
+ size_t* dims;
+ size_t nnz;
+ bool is_iso;
+
+ bsp_level_t* level;
+ // don't think too much about this at the moment.
+ bsp_structure_t structure;
+} bsp_tensor_t;
+
+static inline bsp_tensor_t bsp_construct_default_tensor_t() {
+ bsp_tensor_t tensor;
+ tensor.structure = BSP_GENERAL;
+ tensor.is_iso = false;
+ tensor.nnz = tensor.rank = 0;
+ tensor.dims = NULL;
+
+ tensor.level = NULL;
+ return tensor;
+}
+
+static void bsp_destroy_level_t(bsp_level_t* level) {
+ switch (level->kind) {
+ case BSP_TENSOR_ELEMENT:;
+ bsp_element_t* element = level->data;
+ bsp_destroy_array_t(element->values);
+ free(element);
+ break;
+ case BSP_TENSOR_DENSE:;
+ bsp_dense_t* dense = level->data;
+ bsp_destroy_level_t(dense->child);
+ free(dense);
+ break;
+ case BSP_TENSOR_SPARSE:;
+ bsp_sparse_t* sparse = level->data;
+ bsp_destroy_array_t(*sparse->indices);
+ bsp_destroy_array_t(*sparse->pointers_to);
+ bsp_destroy_level_t(sparse->child);
+ free(sparse);
+ break;
+ default:;
+ }
+}
+
+static inline void bsp_destroy_tensor_t(bsp_tensor_t tensor) {
+ bsp_destroy_level_t(tensor.level);
+ free(tensor.dims);
+}
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 50ffb26..22a19ff 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -4,5 +4,6 @@
target_sources(binsparse-rc PRIVATE
src/read_matrix.c
+ src/read_tensor.c
src/write_matrix.c
)
diff --git a/src/read_matrix.c b/src/read_matrix.c
index 464210c..bf2b96d 100644
--- a/src/read_matrix.c
+++ b/src/read_matrix.c
@@ -143,6 +143,7 @@ bsp_matrix_t bsp_read_matrix_from_group(hid_t f) {
bsp_matrix_format_t format = bsp_get_matrix_format(format_string);
+ // isn't this never true?
assert(format != 0);
matrix.format = format;
diff --git a/src/read_tensor.c b/src/read_tensor.c
new file mode 100644
index 0000000..5ee7dea
--- /dev/null
+++ b/src/read_tensor.c
@@ -0,0 +1,194 @@
+/*
+ * SPDX-FileCopyrightText: 2024 Binsparse Developers
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#include <assert.h>
+#include <binsparse/hdf5_wrapper.h>
+#include <binsparse/matrix.h>
+#include <binsparse/matrix_market/matrix_market_read.h>
+#include <binsparse/tensor.h>
+#include <cJSON/cJSON.h>
+#include <math.h>
+#include <string.h>
+
+static char* key_with_index(const char* key, size_t index) {
+ int keylen = strlen(key);
+ int strsize = keylen * sizeof(char) +
+ (int) ((ceil(log10(index + 1)) + 1) * sizeof(char));
+ char* res = (char*) malloc(strsize);
+ for (int i = 0; i < keylen; i++) {
+ res[i] = key[i];
+ }
+ sprintf(res + keylen, "%ld", index);
+ res[strsize] = '\0';
+ return res;
+}
+
+bsp_tensor_t bsp_read_tensor_from_group(hid_t f) {
+ bsp_tensor_t tensor = bsp_construct_default_tensor_t();
+
+ char* json_string = bsp_read_attribute(f, (char*) "binsparse");
+
+ cJSON* j = cJSON_Parse(json_string);
+
+ assert(j != NULL);
+ assert(cJSON_IsObject(j));
+
+ cJSON* binsparse = cJSON_GetObjectItemCaseSensitive(j, "binsparse");
+ assert(cJSON_IsObject(binsparse));
+
+ cJSON* version_ = cJSON_GetObjectItemCaseSensitive(binsparse, "version");
+
+ assert(version_ != NULL);
+
+ assert(cJSON_IsString(version_));
+
+ // TODO: check version.
+
+ cJSON* format_ = cJSON_GetObjectItemCaseSensitive(binsparse, "format");
+ assert(format_ != NULL);
+ char* format_string = cJSON_GetStringValue(format_);
+
+ // nnz computation
+ cJSON* nnz_ =
+ cJSON_GetObjectItemCaseSensitive(binsparse, "number_of_stored_values");
+ assert(nnz_ != NULL);
+ size_t nnz = cJSON_GetNumberValue(nnz_);
+ tensor.nnz = nnz;
+
+ // check tensor shape.
+ cJSON* shape_ = cJSON_GetObjectItemCaseSensitive(binsparse, "shape");
+ assert(shape_ != NULL);
+ tensor.rank = cJSON_GetArraySize(shape_);
+ size_t* dims = (size_t*) malloc(tensor.rank * sizeof(size_t));
+ for (int idx = 0; idx < tensor.rank; idx++) {
+ dims[idx] = cJSON_GetNumberValue(cJSON_GetArrayItem(shape_, idx));
+ }
+ tensor.dims = dims;
+ assert(tensor.rank > 0);
+
+ cJSON* data_types_ =
+ cJSON_GetObjectItemCaseSensitive(binsparse, "data_types");
+ assert(data_types_ != NULL);
+
+ cJSON* binsparse_tensor =
+ cJSON_GetObjectItemCaseSensitive(binsparse, "tensor");
+ assert(binsparse_tensor != NULL);
+ cJSON* json_level =
+ cJSON_GetObjectItemCaseSensitive(binsparse_tensor, "level");
+ assert(json_level != NULL);
+
+ bsp_level_t* cur_level = malloc(sizeof(bsp_level_t));
+ tensor.level = cur_level;
+
+ // this is effectively a pointer on dims.
+ size_t depth = 0;
+
+ while (depth < tensor.rank + 1) {
+ cJSON* type_object =
+ cJSON_GetObjectItemCaseSensitive(json_level, "level_kind");
+ char* type = type_object ? cJSON_GetStringValue(type_object) : NULL;
+ assert(type != NULL);
+
+ // base case: working with an element.
+ if (strcmp(type, "element") == 0) {
+ bsp_array_t values = bsp_read_array(f, (char*) "values");
+ cur_level->kind = BSP_TENSOR_ELEMENT;
+ bsp_element_t* data = malloc(sizeof(bsp_element_t));
+ data->values = values;
+ cur_level->data = data;
+ depth++;
+ break;
+ }
+
+ // compute what the rank of our current level is, and update our pointer
+ // accordingly.
+ cJSON* rank_obj = cJSON_GetObjectItemCaseSensitive(json_level, "rank");
+ int rank = cJSON_GetNumberValue(rank_obj);
+
+ if (strcmp(type, "dense") == 0) {
+ cur_level->kind = BSP_TENSOR_DENSE;
+
+ bsp_dense_t* data = malloc(sizeof(bsp_dense_t));
+ data->rank = rank;
+ data->child = malloc(sizeof(bsp_level_t));
+
+ cur_level->data = data;
+ cur_level = data->child;
+ } else if (strcmp(type, "sparse") == 0) {
+ cur_level->kind = BSP_TENSOR_SPARSE;
+
+ bsp_sparse_t* data = malloc(sizeof(bsp_sparse_t));
+
+ // initialize pointers_to.
+ {
+ char* pointers_key = key_with_index("pointers_to_", depth);
+ data->pointers_to = bsp_read_array(f, pointers_key);
+ free(pointers_key);
+ }
+
+ // initialize indices
+ data->indices = malloc(rank * sizeof(bsp_array_t));
+ for (int idx = 0; idx < rank; idx++) {
+ char* indices_key = key_with_index("indices_", depth + rank);
+ data->indices[idx] = bsp_read_array(f, indices_key);
+ free(indices_key);
+ }
+
+ data->rank = rank;
+ data->child = malloc(sizeof(bsp_level_t));
+ cur_level->data = data;
+ cur_level = data->child;
+
+ } else {
+ assert(false);
+ }
+ // update the depth here.
+ depth += rank;
+ }
+ if (cJSON_HasObjectItem(binsparse, "structure")) {
+ cJSON* structure_ =
+ cJSON_GetObjectItemCaseSensitive(binsparse, "structure");
+ char* structure = cJSON_GetStringValue(structure_);
+ tensor.structure = bsp_get_structure(structure);
+ }
+
+ cJSON_Delete(j);
+ free(json_string);
+
+ return tensor;
+}
+
+static inline size_t bsp_final_dot(const char* str) {
+ size_t dot_idx = 0;
+ for (size_t i = 0; str[i] != '\0'; i++) {
+ if (str[i] == '.') {
+ dot_idx = i;
+ }
+ }
+ return dot_idx;
+}
+
+bsp_tensor_t bsp_read_tensor(const char* file_name, const char* group) {
+ if (group == NULL) {
+ size_t idx = bsp_final_dot(file_name);
+ if (strcmp(file_name + idx, ".hdf5") == 0 ||
+ strcmp(file_name + idx, ".h5") == 0) {
+ hid_t f = H5Fopen(file_name, H5F_ACC_RDONLY, H5P_DEFAULT);
+ bsp_tensor_t tensor = bsp_read_tensor_from_group(f);
+ H5Fclose(f);
+ return tensor;
+ } else {
+ assert(false);
+ }
+ } else {
+ hid_t f = H5Fopen(file_name, H5F_ACC_RDONLY, H5P_DEFAULT);
+ hid_t g = H5Gopen1(f, group);
+ bsp_tensor_t matrix = bsp_read_tensor_from_group(g);
+ H5Gclose(g);
+ H5Fclose(f);
+ return matrix;
+ }
+}
From 29a58e72511759908139513435ddf8c1c3e9d2e7 Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Wed, 19 Feb 2025 17:58:33 -0500
Subject: [PATCH 03/24] segfault debugging
---
examples/CMakeLists.txt | 1 +
examples/tensor_test.c | 19 +++++++++++++++++++
include/binsparse/read_tensor.h | 2 ++
include/binsparse/tensor.h | 14 +++++++++++---
src/read_tensor.c | 27 +++++++++++++++------------
5 files changed, 48 insertions(+), 15 deletions(-)
create mode 100644 examples/tensor_test.c
diff --git a/examples/CMakeLists.txt b/examples/CMakeLists.txt
index 7e5bbd1..1df709c 100644
--- a/examples/CMakeLists.txt
+++ b/examples/CMakeLists.txt
@@ -19,5 +19,6 @@ add_example(bsp-ls)
add_example(benchmark_read)
add_example(benchmark_read_parallel)
add_example(benchmark_write)
+add_example(tensor_test)
add_subdirectory(cpp)
diff --git a/examples/tensor_test.c b/examples/tensor_test.c
new file mode 100644
index 0000000..091367f
--- /dev/null
+++ b/examples/tensor_test.c
@@ -0,0 +1,19 @@
+#include<binsparse/tensor.h>
+#include<binsparse/read_tensor.h>
+
+int main(int argc, char **argv) {
+ if (argc < 2) {
+ fprintf(stderr, "usage: ./tensor_test [file_name.h5]\n");
+ return 1;
+ }
+ char* file_name = argv[1];
+ bsp_tensor_t tensor = bsp_read_tensor(file_name, NULL);
+ printf("rank: %d\n", tensor.rank);
+ printf("dims:");
+ for (int i = 0; i < tensor.rank; i++) {
+ printf("%ld, ", tensor.dims[i]);
+ }
+ printf("\n");
+ bsp_destroy_tensor_t(tensor);
+ return 0;
+}
diff --git a/include/binsparse/read_tensor.h b/include/binsparse/read_tensor.h
index 0e9a2f9..9d419ec 100644
--- a/include/binsparse/read_tensor.h
+++ b/include/binsparse/read_tensor.h
@@ -4,6 +4,8 @@
extern "C" {
#endif
+char* key_with_index(const char* key, size_t index);
+
#ifdef BSP_USE_HDF5
#include <hdf5.h>
diff --git a/include/binsparse/tensor.h b/include/binsparse/tensor.h
index e70ef04..9fa577e 100644
--- a/include/binsparse/tensor.h
+++ b/include/binsparse/tensor.h
@@ -23,7 +23,7 @@ typedef struct {
// corresponds to BSP_TENSOR_DENSE
typedef struct {
int rank;
- bsp_array_t pointers_to;
+ bsp_array_t* pointers_to;
bsp_array_t* indices;
bsp_level_t* child;
} bsp_sparse_t;
@@ -56,6 +56,8 @@ static inline bsp_tensor_t bsp_construct_default_tensor_t() {
}
static void bsp_destroy_level_t(bsp_level_t* level) {
+ if (level == NULL)
+ return;
switch (level->kind) {
case BSP_TENSOR_ELEMENT:;
bsp_element_t* element = level->data;
@@ -69,8 +71,14 @@ static void bsp_destroy_level_t(bsp_level_t* level) {
break;
case BSP_TENSOR_SPARSE:;
bsp_sparse_t* sparse = level->data;
- bsp_destroy_array_t(*sparse->indices);
- bsp_destroy_array_t(*sparse->pointers_to);
+
+ if (sparse->pointers_to != NULL)
+ bsp_destroy_array_t(*sparse->pointers_to);
+ if (sparse->indices != NULL) {
+ for (int i = 0; i < sparse->rank; i++) {
+ bsp_destroy_array_t(sparse->indices[i]);
+ }
+ }
bsp_destroy_level_t(sparse->child);
free(sparse);
break;
diff --git a/src/read_tensor.c b/src/read_tensor.c
index 5ee7dea..48996d5 100644
--- a/src/read_tensor.c
+++ b/src/read_tensor.c
@@ -13,10 +13,13 @@
#include <math.h>
#include <string.h>
-static char* key_with_index(const char* key, size_t index) {
+/*
+Returns "{key}-{index}"
+*/
+char* key_with_index(const char* key, size_t index) {
int keylen = strlen(key);
int strsize = keylen * sizeof(char) +
- (int) ((ceil(log10(index + 1)) + 1) * sizeof(char));
+ (int) ((ceil(log10(index + 1)) + 1) * sizeof(char)) + 1;
char* res = (char*) malloc(strsize);
for (int i = 0; i < keylen; i++) {
res[i] = key[i];
@@ -45,12 +48,6 @@ bsp_tensor_t bsp_read_tensor_from_group(hid_t f) {
assert(cJSON_IsString(version_));
- // TODO: check version.
-
- cJSON* format_ = cJSON_GetObjectItemCaseSensitive(binsparse, "format");
- assert(format_ != NULL);
- char* format_string = cJSON_GetStringValue(format_);
-
// nnz computation
cJSON* nnz_ =
cJSON_GetObjectItemCaseSensitive(binsparse, "number_of_stored_values");
@@ -117,22 +114,26 @@ bsp_tensor_t bsp_read_tensor_from_group(hid_t f) {
cur_level->data = data;
cur_level = data->child;
+
} else if (strcmp(type, "sparse") == 0) {
cur_level->kind = BSP_TENSOR_SPARSE;
bsp_sparse_t* data = malloc(sizeof(bsp_sparse_t));
- // initialize pointers_to.
- {
+ // initialize pointers_to, but only if the depth is not zero.
+ if (depth != 0) {
char* pointers_key = key_with_index("pointers_to_", depth);
- data->pointers_to = bsp_read_array(f, pointers_key);
+ data->pointers_to = malloc(sizeof(bsp_array_t));
+ *data->pointers_to = bsp_read_array(f, pointers_key);
free(pointers_key);
+ } else {
+ data->pointers_to = NULL;
}
// initialize indices
data->indices = malloc(rank * sizeof(bsp_array_t));
for (int idx = 0; idx < rank; idx++) {
- char* indices_key = key_with_index("indices_", depth + rank);
+ char* indices_key = key_with_index("indices_", depth + idx);
data->indices[idx] = bsp_read_array(f, indices_key);
free(indices_key);
}
@@ -146,6 +147,8 @@ bsp_tensor_t bsp_read_tensor_from_group(hid_t f) {
assert(false);
}
// update the depth here.
+ json_level = cJSON_GetObjectItemCaseSensitive(json_level, "level");
+ assert(json_level != NULL);
depth += rank;
}
if (cJSON_HasObjectItem(binsparse, "structure")) {
From 436863e6e527a1aa9f615593acd18c061c231d3f Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Wed, 19 Feb 2025 21:51:08 -0500
Subject: [PATCH 04/24] tensor writer implemented
TODO: you need to apparently set the format key.
---
examples/tensor_test.c | 14 +--
include/binsparse/tensor.h | 23 +++++
include/binsparse/write_tensor.h | 30 +++++++
src/CMakeLists.txt | 1 +
src/write_tensor.c | 143 +++++++++++++++++++++++++++++++
5 files changed, 205 insertions(+), 6 deletions(-)
create mode 100644 include/binsparse/write_tensor.h
create mode 100644 src/write_tensor.c
diff --git a/examples/tensor_test.c b/examples/tensor_test.c
index 091367f..bfa0bfc 100644
--- a/examples/tensor_test.c
+++ b/examples/tensor_test.c
@@ -1,19 +1,21 @@
-#include<binsparse/tensor.h>
-#include<binsparse/read_tensor.h>
+#include <binsparse/tensor.h>
+#include <binsparse/read_tensor.h>
+#include <binsparse/write_tensor.h>
-int main(int argc, char **argv) {
- if (argc < 2) {
- fprintf(stderr, "usage: ./tensor_test [file_name.h5]\n");
+int main(int argc, char** argv) {
+ if (argc < 3) {
+ fprintf(stderr, "usage: ./tensor_test [file_name.h5] [output_file_name.h5]\n");
return 1;
}
char* file_name = argv[1];
- bsp_tensor_t tensor = bsp_read_tensor(file_name, NULL);
+ bsp_tensor_t tensor = bsp_read_tensor(argv[1], NULL);
printf("rank: %d\n", tensor.rank);
printf("dims:");
for (int i = 0; i < tensor.rank; i++) {
printf("%ld, ", tensor.dims[i]);
}
printf("\n");
+ bsp_write_tensor(argv[2], tensor, NULL, NULL, 9);
bsp_destroy_tensor_t(tensor);
return 0;
}
diff --git a/include/binsparse/tensor.h b/include/binsparse/tensor.h
index 9fa577e..32e16a3 100644
--- a/include/binsparse/tensor.h
+++ b/include/binsparse/tensor.h
@@ -86,6 +86,29 @@ static void bsp_destroy_level_t(bsp_level_t* level) {
}
}
+static bsp_array_t bsp_get_tensor_values(bsp_tensor_t tensor) {
+ bsp_level_t* level = tensor.level;
+ while (level != NULL) {
+ switch (level->kind) {
+ case BSP_TENSOR_ELEMENT:;
+ bsp_element_t* element = level->data;
+ return element->values;
+ break;
+ case BSP_TENSOR_SPARSE:;
+ bsp_sparse_t* sparse = level->data;
+ level = sparse->child;
+ break;
+ case BSP_TENSOR_DENSE:;
+ bsp_dense_t* dense = level->data;
+ level = dense->child;
+ break;
+ default:;
+ }
+ }
+ // this should never happen!
+ assert(false);
+}
+
static inline void bsp_destroy_tensor_t(bsp_tensor_t tensor) {
bsp_destroy_level_t(tensor.level);
free(tensor.dims);
diff --git a/include/binsparse/write_tensor.h b/include/binsparse/write_tensor.h
new file mode 100644
index 0000000..64497c2
--- /dev/null
+++ b/include/binsparse/write_tensor.h
@@ -0,0 +1,30 @@
+/*
+ * SPDX-FileCopyrightText: 2024 Binsparse Developers
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#pragma once
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// TODO: make cJSON optional.
+
+#include<binsparse/tensor.h>
+#include <cJSON/cJSON.h>
+
+#ifdef BSP_USE_HDF5
+#include <hdf5.h>
+
+int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
+ int compression_level);
+#endif
+
+int bsp_write_tensor(const char* fname, bsp_tensor_t tensor, const char* group,
+ cJSON* user_json, int compression_level);
+
+#ifdef __cplusplus
+}
+#endif
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 22a19ff..4f52770 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -6,4 +6,5 @@ target_sources(binsparse-rc PRIVATE
src/read_matrix.c
src/read_tensor.c
src/write_matrix.c
+ src/write_tensor.c
)
diff --git a/src/write_tensor.c b/src/write_tensor.c
new file mode 100644
index 0000000..2c1a983
--- /dev/null
+++ b/src/write_tensor.c
@@ -0,0 +1,143 @@
+#include <assert.h>
+#include <binsparse/tensor.h>
+#include <unistd.h>
+
+#include <binsparse/binsparse.h>
+#include <binsparse/read_tensor.h>
+#include <cJSON/cJSON.h>
+
+static cJSON* init_tensor_json(bsp_tensor_t tensor, cJSON* user_json) {
+ cJSON* j = cJSON_CreateObject();
+ assert(j != NULL);
+
+ cJSON* binsparse = cJSON_CreateObject();
+
+ assert(binsparse != NULL);
+
+ cJSON_AddItemToObject(j, "binsparse", binsparse);
+
+ cJSON* userJsonItem;
+
+ cJSON_ArrayForEach(userJsonItem, user_json) {
+ cJSON_AddItemToObject(j, userJsonItem->string, userJsonItem);
+ }
+
+ cJSON_AddStringToObject(binsparse, "version", BINSPARSE_VERSION);
+
+ cJSON* shape = cJSON_AddArrayToObject(binsparse, "shape");
+ for (int i = 0; i < tensor.rank; i++) {
+ cJSON_AddItemToArray(shape, cJSON_CreateNumber(tensor.dims[i]));
+ }
+
+ cJSON_AddNumberToObject(binsparse, "number_of_stored_values", tensor.nnz);
+
+ if (tensor.structure != BSP_GENERAL) {
+ cJSON_AddStringToObject(binsparse, "structure",
+ bsp_get_structure_string(tensor.structure));
+ }
+
+ return j;
+}
+
+int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
+ int compression_level) {
+ // bsp_matrix_t matrix;
+ cJSON* j = init_tensor_json(tensor, user_json);
+ cJSON* binsparse = cJSON_GetObjectItemCaseSensitive(j, "binsparse");
+ assert(binsparse != NULL);
+ cJSON* data_types = cJSON_AddObjectToObject(binsparse, "data_types");
+ bsp_array_t values = bsp_get_tensor_values(tensor);
+
+ if (!tensor.is_iso) {
+ cJSON_AddStringToObject(data_types, "values",
+ bsp_get_type_string(values.type));
+ } else {
+ char* base_type_string = bsp_get_type_string(values.type);
+ size_t len = strlen(base_type_string) + 6;
+ char* type_string = (char*) malloc(sizeof(char) * len);
+
+ strncpy(type_string, "iso[", len);
+ strncpy(type_string + 4, base_type_string, len - 4);
+ strncpy(type_string + len - 2, "]", 2);
+
+ cJSON_AddStringToObject(data_types, "values", type_string);
+
+ free(type_string);
+ }
+
+ // attempt to write an array.
+ int result = bsp_write_array(f, (char*) "values", values, compression_level);
+ if (result != 0) {
+ cJSON_Delete(j);
+ return result;
+ }
+
+ int rank = 0;
+ bsp_level_t* level = tensor.level;
+ while (level->kind != BSP_TENSOR_ELEMENT) {
+ switch (level->kind) {
+ case BSP_TENSOR_SPARSE:;
+ bsp_sparse_t* sparse = level->data;
+ size_t layer_rank = sparse->rank;
+ if (sparse->pointers_to != NULL) {
+ cJSON_AddStringToObject(data_types,
+ key_with_index("pointers_to_", rank),
+ bsp_get_type_string(sparse->pointers_to->type));
+ result = bsp_write_array(f, key_with_index("pointers_to_", rank),
+ *sparse->pointers_to, compression_level);
+ if (result != 0) {
+ cJSON_Delete(j);
+ return result;
+ }
+ }
+
+ for (int i = 0; i < layer_rank; i++) {
+ cJSON_AddStringToObject(data_types,
+ key_with_index("indices_", rank + i),
+ bsp_get_type_string(sparse->indices[i].type));
+ result = bsp_write_array(f, key_with_index("indices_", rank + i),
+ sparse->indices[i], compression_level);
+ if (result != 0) {
+ cJSON_Delete(j);
+ return result;
+ }
+ }
+
+ rank += layer_rank;
+ level = sparse->child;
+ break;
+ case BSP_TENSOR_DENSE:;
+ rank += ((bsp_dense_t*) level->data)->rank;
+ level = ((bsp_dense_t*) level->data)->child;
+ break;
+ default:;
+ }
+ }
+
+ char* json_string = cJSON_Print(j);
+ bsp_write_attribute(f, (char*) "binsparse", json_string);
+ free(json_string);
+
+ return 0;
+}
+
+int bsp_write_tensor(const char* fname, bsp_tensor_t tensor, const char* group,
+ cJSON* user_json, int compression_level) {
+ if (group == NULL) {
+ hid_t f = H5Fcreate(fname, H5F_ACC_TRUNC, H5P_DEFAULT, H5P_DEFAULT);
+ bsp_write_tensor_to_group(f, tensor, user_json, compression_level);
+ H5Fclose(f);
+ } else {
+ hid_t f;
+ if (access(fname, F_OK) == 0) {
+ f = H5Fopen(fname, H5F_ACC_RDWR, H5P_DEFAULT);
+ } else {
+ f = H5Fcreate(fname, H5F_ACC_TRUNC, H5P_DEFAULT, H5P_DEFAULT);
+ }
+ hid_t g = H5Gcreate1(f, group, H5P_DEFAULT);
+ bsp_write_tensor_to_group(g, tensor, user_json, compression_level);
+ H5Gclose(g);
+ H5Fclose(f);
+ }
+ return 0;
+}
From 7150a6ef23b8913070ee341271952100dbbefa85 Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Wed, 19 Feb 2025 23:40:52 -0500
Subject: [PATCH 05/24] tensor reader/writer successfully bootstrapped
---
src/write_tensor.c | 28 ++++++++++++++++++++++++++--
1 file changed, 26 insertions(+), 2 deletions(-)
diff --git a/src/write_tensor.c b/src/write_tensor.c
index 2c1a983..7a2ccc7 100644
--- a/src/write_tensor.c
+++ b/src/write_tensor.c
@@ -11,9 +11,12 @@ static cJSON* init_tensor_json(bsp_tensor_t tensor, cJSON* user_json) {
assert(j != NULL);
cJSON* binsparse = cJSON_CreateObject();
-
assert(binsparse != NULL);
+ cJSON* binsparse_tensor = cJSON_CreateObject();
+ assert(binsparse_tensor != NULL);
+
+ cJSON_AddItemToObject(binsparse, "tensor", binsparse_tensor);
cJSON_AddItemToObject(j, "binsparse", binsparse);
cJSON* userJsonItem;
@@ -43,8 +46,13 @@ int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
int compression_level) {
// bsp_matrix_t matrix;
cJSON* j = init_tensor_json(tensor, user_json);
+ // tensor:
cJSON* binsparse = cJSON_GetObjectItemCaseSensitive(j, "binsparse");
assert(binsparse != NULL);
+ cJSON* binsparse_tensor =
+ cJSON_GetObjectItemCaseSensitive(binsparse, "tensor");
+ assert(binsparse_tensor != NULL);
+
cJSON* data_types = cJSON_AddObjectToObject(binsparse, "data_types");
bsp_array_t values = bsp_get_tensor_values(tensor);
@@ -74,11 +82,16 @@ int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
int rank = 0;
bsp_level_t* level = tensor.level;
- while (level->kind != BSP_TENSOR_ELEMENT) {
+ cJSON* json_level = cJSON_AddObjectToObject(binsparse_tensor, "level");
+ while (true) {
+ int reached_end = 0;
switch (level->kind) {
case BSP_TENSOR_SPARSE:;
bsp_sparse_t* sparse = level->data;
size_t layer_rank = sparse->rank;
+ cJSON_AddStringToObject(json_level, "level_kind", "sparse");
+ cJSON_AddNumberToObject(json_level, "rank", layer_rank);
+
if (sparse->pointers_to != NULL) {
cJSON_AddStringToObject(data_types,
key_with_index("pointers_to_", rank),
@@ -107,11 +120,22 @@ int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
level = sparse->child;
break;
case BSP_TENSOR_DENSE:;
+ cJSON_AddStringToObject(json_level, "level_kind", "dense");
+ cJSON_AddNumberToObject(json_level, "rank",
+ ((bsp_dense_t*) level->data)->rank);
rank += ((bsp_dense_t*) level->data)->rank;
level = ((bsp_dense_t*) level->data)->child;
break;
+ case BSP_TENSOR_ELEMENT:;
+ cJSON_AddStringToObject(json_level, "level_kind", "element");
+ reached_end = 1;
+ break;
default:;
}
+
+ if (reached_end)
+ break;
+ json_level = cJSON_AddObjectToObject(json_level, "level");
}
char* json_string = cJSON_Print(j);
From b1e6d12357719046d916ad11fe9f078396d96c51 Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Thu, 20 Feb 2025 17:40:10 -0500
Subject: [PATCH 06/24] testing framework for tensor read/write implemented
---
.gitignore | 1 +
examples/cpp/CMakeLists.txt | 1 +
examples/cpp/tensor_test.cpp | 1 +
examples/simple_read.c | 2 +-
include/binsparse/tensor.h | 30 ++++++++++-------
src/write_tensor.c | 9 ++++--
test/CMakeLists.txt | 1 +
test/julia/CMakeLists.txt | 13 ++++++++
test/julia/tensor_test.jl | 62 ++++++++++++++++++++++++++++++++++++
9 files changed, 104 insertions(+), 16 deletions(-)
create mode 100644 examples/cpp/tensor_test.cpp
create mode 100644 test/julia/CMakeLists.txt
create mode 100644 test/julia/tensor_test.jl
diff --git a/.gitignore b/.gitignore
index a1e6811..bfb11fb 100644
--- a/.gitignore
+++ b/.gitignore
@@ -6,3 +6,4 @@ scripts
venv
build
._*
+tensor_test_files
diff --git a/examples/cpp/CMakeLists.txt b/examples/cpp/CMakeLists.txt
index 2b8dcdf..d6ad9e1 100644
--- a/examples/cpp/CMakeLists.txt
+++ b/examples/cpp/CMakeLists.txt
@@ -15,5 +15,6 @@ add_example(mtx2bsp)
add_example(bsp2mtx)
add_example(check_equivalence)
add_example(bsp-ls)
+add_example(tensor_test)
add_example(benchmark_read)
add_example(benchmark_write)
diff --git a/examples/cpp/tensor_test.cpp b/examples/cpp/tensor_test.cpp
new file mode 100644
index 0000000..677d71e
--- /dev/null
+++ b/examples/cpp/tensor_test.cpp
@@ -0,0 +1 @@
+#include "../tensor_test.c"
diff --git a/examples/simple_read.c b/examples/simple_read.c
index f2a1184..a4ba9bb 100644
--- a/examples/simple_read.c
+++ b/examples/simple_read.c
@@ -7,7 +7,7 @@
#include <binsparse/binsparse.h>
int main(int argc, char** argv) {
- char* file_name = "test.hdf5";
+ char* file_name = (char*) "test.hdf5";
hid_t f = H5Fopen(file_name, H5F_ACC_RDWR, H5P_DEFAULT);
diff --git a/include/binsparse/tensor.h b/include/binsparse/tensor.h
index 32e16a3..87c1b54 100644
--- a/include/binsparse/tensor.h
+++ b/include/binsparse/tensor.h
@@ -59,18 +59,20 @@ static void bsp_destroy_level_t(bsp_level_t* level) {
if (level == NULL)
return;
switch (level->kind) {
- case BSP_TENSOR_ELEMENT:;
- bsp_element_t* element = level->data;
+ case BSP_TENSOR_ELEMENT: {
+ bsp_element_t* element = (bsp_element_t*) level->data;
bsp_destroy_array_t(element->values);
free(element);
break;
- case BSP_TENSOR_DENSE:;
- bsp_dense_t* dense = level->data;
+ }
+ case BSP_TENSOR_DENSE: {
+ bsp_dense_t* dense = (bsp_dense_t*) level->data;
bsp_destroy_level_t(dense->child);
free(dense);
break;
- case BSP_TENSOR_SPARSE:;
- bsp_sparse_t* sparse = level->data;
+ }
+ case BSP_TENSOR_SPARSE: {
+ bsp_sparse_t* sparse = (bsp_sparse_t*) level->data;
if (sparse->pointers_to != NULL)
bsp_destroy_array_t(*sparse->pointers_to);
@@ -82,6 +84,7 @@ static void bsp_destroy_level_t(bsp_level_t* level) {
bsp_destroy_level_t(sparse->child);
free(sparse);
break;
+ }
default:;
}
}
@@ -90,18 +93,21 @@ static bsp_array_t bsp_get_tensor_values(bsp_tensor_t tensor) {
bsp_level_t* level = tensor.level;
while (level != NULL) {
switch (level->kind) {
- case BSP_TENSOR_ELEMENT:;
- bsp_element_t* element = level->data;
+ case BSP_TENSOR_ELEMENT: {
+ bsp_element_t* element = (bsp_element_t*) level->data;
return element->values;
break;
- case BSP_TENSOR_SPARSE:;
- bsp_sparse_t* sparse = level->data;
+ }
+ case BSP_TENSOR_SPARSE: {
+ bsp_sparse_t* sparse = (bsp_sparse_t*) level->data;
level = sparse->child;
break;
- case BSP_TENSOR_DENSE:;
- bsp_dense_t* dense = level->data;
+ }
+ case BSP_TENSOR_DENSE: {
+ bsp_dense_t* dense = (bsp_dense_t*) level->data;
level = dense->child;
break;
+ }
default:;
}
}
diff --git a/src/write_tensor.c b/src/write_tensor.c
index 7a2ccc7..1d468a7 100644
--- a/src/write_tensor.c
+++ b/src/write_tensor.c
@@ -86,7 +86,7 @@ int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
while (true) {
int reached_end = 0;
switch (level->kind) {
- case BSP_TENSOR_SPARSE:;
+ case BSP_TENSOR_SPARSE: {
bsp_sparse_t* sparse = level->data;
size_t layer_rank = sparse->rank;
cJSON_AddStringToObject(json_level, "level_kind", "sparse");
@@ -119,17 +119,20 @@ int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
rank += layer_rank;
level = sparse->child;
break;
- case BSP_TENSOR_DENSE:;
+ }
+ case BSP_TENSOR_DENSE: {
cJSON_AddStringToObject(json_level, "level_kind", "dense");
cJSON_AddNumberToObject(json_level, "rank",
((bsp_dense_t*) level->data)->rank);
rank += ((bsp_dense_t*) level->data)->rank;
level = ((bsp_dense_t*) level->data)->child;
break;
- case BSP_TENSOR_ELEMENT:;
+ }
+ case BSP_TENSOR_ELEMENT: {
cJSON_AddStringToObject(json_level, "level_kind", "element");
reached_end = 1;
break;
+ }
default:;
}
diff --git a/test/CMakeLists.txt b/test/CMakeLists.txt
index 0dd7ef0..45ff378 100644
--- a/test/CMakeLists.txt
+++ b/test/CMakeLists.txt
@@ -3,3 +3,4 @@
# SPDX-License-Identifier: BSD-3-Clause
add_subdirectory(bash)
+add_subdirectory(julia)
diff --git a/test/julia/CMakeLists.txt b/test/julia/CMakeLists.txt
new file mode 100644
index 0000000..4c2c9d8
--- /dev/null
+++ b/test/julia/CMakeLists.txt
@@ -0,0 +1,13 @@
+# SPDX-FileCopyrightText: 2024 Binsparse Developers
+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+find_program(JULIA_PROGRAM julia)
+
+enable_testing()
+
+if(JULIA_PROGRAM)
+ add_test(NAME tensors.tensor_test COMMAND ${JULIA_PROGRAM} ${CMAKE_CURRENT_SOURCE_DIR}/tensor_test.jl)
+
+ add_test(NAME tensors.cpp.tensor_test COMMAND ${JULIA_PROGRAM} ${CMAKE_CURRENT_SOURCE_DIR}/tensor_test.jl cpp)
+endif()
diff --git a/test/julia/tensor_test.jl b/test/julia/tensor_test.jl
new file mode 100644
index 0000000..04cb1d2
--- /dev/null
+++ b/test/julia/tensor_test.jl
@@ -0,0 +1,62 @@
+using Finch;
+using HDF5;
+
+# dir = @__DIR__();
+dir = pwd();
+
+print("Args:", ARGS, "\n");
+
+if length(ARGS) == 0
+ tensor_test = joinpath(dir, "../../examples/tensor_test")
+ test_files = joinpath(dir, "../../tensor_test_files")
+else
+ tensor_test = joinpath(dir, "../../examples/cpp/tensor_test-cpp")
+ test_files = joinpath(dir, "../../tensor_test_files/cpp")
+end
+
+mkpath(test_files)
+
+function tensortest(tensor::Tensor, input::AbstractString, output::AbstractString)
+ print("tensor_test ", input, " -> ", output, "\n")
+ fwrite(input, tensor)
+ run(`$tensor_test $input $output`)
+ new_tensor = fread(output)
+ # @assert new_tensor == tensor
+end
+
+tensortest(
+ Tensor(
+ Dense(SparseList(SparseList(Element(0.0)))),
+ fsprand(10, 10, 10, 0.1)
+ ),
+ joinpath(test_files, "input1.bsp.h5"),
+ joinpath(test_files, "output1.bsp.h5")
+)
+
+
+tensortest(
+ Tensor(
+ Dense(SparseCOO{2}(Element(0.0))),
+ fsprand(10, 10, 10, 0.1)
+ ),
+ joinpath(test_files, "input2.bsp.h5"),
+ joinpath(test_files, "output2.bsp.h5")
+)
+
+tensortest(
+ Tensor(
+ Dense(Dense(Dense(Element(0.0)))),
+ fsprand(10, 10, 10, 0.1)
+ ),
+ joinpath(test_files, "input3.bsp.h5"),
+ joinpath(test_files, "output3.bsp.h5")
+)
+
+tensortest(
+ Tensor(
+ SparseCOO{2}(Element(0.0)),
+ fsprand(10, 10, 0.1)
+ ),
+ joinpath(test_files, "input4.bsp.h5"),
+ joinpath(test_files, "output4.bsp.h5")
+)
From 138647f3ba585f87532e64f5dc059b4fa19aae56 Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Fri, 21 Feb 2025 08:27:43 -0500
Subject: [PATCH 07/24] tests fixed
---
test/julia/tensor_test.jl | 8 +++-----
1 file changed, 3 insertions(+), 5 deletions(-)
diff --git a/test/julia/tensor_test.jl b/test/julia/tensor_test.jl
index 04cb1d2..216a0ec 100644
--- a/test/julia/tensor_test.jl
+++ b/test/julia/tensor_test.jl
@@ -1,11 +1,8 @@
using Finch;
using HDF5;
-# dir = @__DIR__();
dir = pwd();
-print("Args:", ARGS, "\n");
-
if length(ARGS) == 0
tensor_test = joinpath(dir, "../../examples/tensor_test")
test_files = joinpath(dir, "../../tensor_test_files")
@@ -20,8 +17,9 @@ function tensortest(tensor::Tensor, input::AbstractString, output::AbstractStrin
print("tensor_test ", input, " -> ", output, "\n")
fwrite(input, tensor)
run(`$tensor_test $input $output`)
- new_tensor = fread(output)
- # @assert new_tensor == tensor
+ # for whatever reason, fread returns a swizzlearray
+ new_tensor = fread(output).body
+ @assert new_tensor == tensor
end
tensortest(
From 3309097e4f7228e09cc6931704fd33058c2edeca Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Sat, 22 Feb 2025 11:50:49 -0500
Subject: [PATCH 08/24] parse transposes too
---
include/binsparse/tensor.h | 5 ++++-
src/read_tensor.c | 18 +++++++++++++++---
src/write_tensor.c | 17 +++++++++++------
test/julia/tensor_test.jl | 5 ++---
4 files changed, 32 insertions(+), 13 deletions(-)
diff --git a/include/binsparse/tensor.h b/include/binsparse/tensor.h
index 87c1b54..60dad01 100644
--- a/include/binsparse/tensor.h
+++ b/include/binsparse/tensor.h
@@ -36,6 +36,7 @@ typedef struct {
typedef struct {
int rank;
size_t* dims;
+ size_t* transpose;
size_t nnz;
bool is_iso;
@@ -50,6 +51,7 @@ static inline bsp_tensor_t bsp_construct_default_tensor_t() {
tensor.is_iso = false;
tensor.nnz = tensor.rank = 0;
tensor.dims = NULL;
+ tensor.transpose = NULL;
tensor.level = NULL;
return tensor;
@@ -117,5 +119,6 @@ static bsp_array_t bsp_get_tensor_values(bsp_tensor_t tensor) {
static inline void bsp_destroy_tensor_t(bsp_tensor_t tensor) {
bsp_destroy_level_t(tensor.level);
- free(tensor.dims);
+ if (tensor.dims != NULL) free(tensor.dims);
+ if (tensor.transpose != NULL) free(tensor.transpose);
}
diff --git a/src/read_tensor.c b/src/read_tensor.c
index 48996d5..ec552a2 100644
--- a/src/read_tensor.c
+++ b/src/read_tensor.c
@@ -70,11 +70,23 @@ bsp_tensor_t bsp_read_tensor_from_group(hid_t f) {
cJSON_GetObjectItemCaseSensitive(binsparse, "data_types");
assert(data_types_ != NULL);
- cJSON* binsparse_tensor =
+ cJSON* binsparse_custom =
cJSON_GetObjectItemCaseSensitive(binsparse, "tensor");
- assert(binsparse_tensor != NULL);
+ assert(binsparse_custom != NULL);
+
+ cJSON* transpose_ =
+ cJSON_GetObjectItemCaseSensitive(binsparse_custom, "transpose");
+ if (transpose_ != NULL) {
+ size_t* transpose = (size_t*) malloc(tensor.rank * sizeof(size_t));
+ for (int idx = 0; idx < tensor.rank; idx++) {
+ transpose[idx] =
+ cJSON_GetNumberValue((cJSON_GetArrayItem(transpose_, idx)));
+ }
+ tensor.transpose = transpose;
+ }
+
cJSON* json_level =
- cJSON_GetObjectItemCaseSensitive(binsparse_tensor, "level");
+ cJSON_GetObjectItemCaseSensitive(binsparse_custom, "level");
assert(json_level != NULL);
bsp_level_t* cur_level = malloc(sizeof(bsp_level_t));
diff --git a/src/write_tensor.c b/src/write_tensor.c
index 1d468a7..d1f272e 100644
--- a/src/write_tensor.c
+++ b/src/write_tensor.c
@@ -13,10 +13,10 @@ static cJSON* init_tensor_json(bsp_tensor_t tensor, cJSON* user_json) {
cJSON* binsparse = cJSON_CreateObject();
assert(binsparse != NULL);
- cJSON* binsparse_tensor = cJSON_CreateObject();
- assert(binsparse_tensor != NULL);
+ cJSON* binsparse_custom = cJSON_CreateObject();
+ assert(binsparse_custom != NULL);
- cJSON_AddItemToObject(binsparse, "tensor", binsparse_tensor);
+ cJSON_AddItemToObject(binsparse, "tensor", binsparse_custom);
cJSON_AddItemToObject(j, "binsparse", binsparse);
cJSON* userJsonItem;
@@ -32,6 +32,11 @@ static cJSON* init_tensor_json(bsp_tensor_t tensor, cJSON* user_json) {
cJSON_AddItemToArray(shape, cJSON_CreateNumber(tensor.dims[i]));
}
+ cJSON* transpose = cJSON_AddArrayToObject(binsparse_custom, "transpose");
+ for (int i = 0; i < tensor.rank; i++) {
+ cJSON_AddItemToArray(transpose, cJSON_CreateNumber(tensor.transpose[i]));
+ }
+
cJSON_AddNumberToObject(binsparse, "number_of_stored_values", tensor.nnz);
if (tensor.structure != BSP_GENERAL) {
@@ -49,9 +54,9 @@ int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
// tensor:
cJSON* binsparse = cJSON_GetObjectItemCaseSensitive(j, "binsparse");
assert(binsparse != NULL);
- cJSON* binsparse_tensor =
+ cJSON* binsparse_custom =
cJSON_GetObjectItemCaseSensitive(binsparse, "tensor");
- assert(binsparse_tensor != NULL);
+ assert(binsparse_custom != NULL);
cJSON* data_types = cJSON_AddObjectToObject(binsparse, "data_types");
bsp_array_t values = bsp_get_tensor_values(tensor);
@@ -82,7 +87,7 @@ int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
int rank = 0;
bsp_level_t* level = tensor.level;
- cJSON* json_level = cJSON_AddObjectToObject(binsparse_tensor, "level");
+ cJSON* json_level = cJSON_AddObjectToObject(binsparse_custom, "level");
while (true) {
int reached_end = 0;
switch (level->kind) {
diff --git a/test/julia/tensor_test.jl b/test/julia/tensor_test.jl
index 216a0ec..844c758 100644
--- a/test/julia/tensor_test.jl
+++ b/test/julia/tensor_test.jl
@@ -17,9 +17,8 @@ function tensortest(tensor::Tensor, input::AbstractString, output::AbstractStrin
print("tensor_test ", input, " -> ", output, "\n")
fwrite(input, tensor)
run(`$tensor_test $input $output`)
- # for whatever reason, fread returns a swizzlearray
- new_tensor = fread(output).body
- @assert new_tensor == tensor
+ output_tensor = fread(output)
+ @assert tensor == output_tensor
end
tensortest(
From 4e50d1486a88283cc8cff5eb607584046268408b Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Thu, 27 Feb 2025 21:15:41 -0500
Subject: [PATCH 09/24] json reading/writing adjusted for binsparse compliance
---
src/read_tensor.c | 4 ++--
src/write_tensor.c | 10 +++++-----
2 files changed, 7 insertions(+), 7 deletions(-)
diff --git a/src/read_tensor.c b/src/read_tensor.c
index ec552a2..01453ac 100644
--- a/src/read_tensor.c
+++ b/src/read_tensor.c
@@ -71,7 +71,7 @@ bsp_tensor_t bsp_read_tensor_from_group(hid_t f) {
assert(data_types_ != NULL);
cJSON* binsparse_custom =
- cJSON_GetObjectItemCaseSensitive(binsparse, "tensor");
+ cJSON_GetObjectItemCaseSensitive(binsparse, "custom");
assert(binsparse_custom != NULL);
cJSON* transpose_ =
@@ -97,7 +97,7 @@ bsp_tensor_t bsp_read_tensor_from_group(hid_t f) {
while (depth < tensor.rank + 1) {
cJSON* type_object =
- cJSON_GetObjectItemCaseSensitive(json_level, "level_kind");
+ cJSON_GetObjectItemCaseSensitive(json_level, "level_desc");
char* type = type_object ? cJSON_GetStringValue(type_object) : NULL;
assert(type != NULL);
diff --git a/src/write_tensor.c b/src/write_tensor.c
index d1f272e..52a4877 100644
--- a/src/write_tensor.c
+++ b/src/write_tensor.c
@@ -16,7 +16,7 @@ static cJSON* init_tensor_json(bsp_tensor_t tensor, cJSON* user_json) {
cJSON* binsparse_custom = cJSON_CreateObject();
assert(binsparse_custom != NULL);
- cJSON_AddItemToObject(binsparse, "tensor", binsparse_custom);
+ cJSON_AddItemToObject(binsparse, "custom", binsparse_custom);
cJSON_AddItemToObject(j, "binsparse", binsparse);
cJSON* userJsonItem;
@@ -55,7 +55,7 @@ int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
cJSON* binsparse = cJSON_GetObjectItemCaseSensitive(j, "binsparse");
assert(binsparse != NULL);
cJSON* binsparse_custom =
- cJSON_GetObjectItemCaseSensitive(binsparse, "tensor");
+ cJSON_GetObjectItemCaseSensitive(binsparse, "custom");
assert(binsparse_custom != NULL);
cJSON* data_types = cJSON_AddObjectToObject(binsparse, "data_types");
@@ -94,7 +94,7 @@ int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
case BSP_TENSOR_SPARSE: {
bsp_sparse_t* sparse = level->data;
size_t layer_rank = sparse->rank;
- cJSON_AddStringToObject(json_level, "level_kind", "sparse");
+ cJSON_AddStringToObject(json_level, "level_desc", "sparse");
cJSON_AddNumberToObject(json_level, "rank", layer_rank);
if (sparse->pointers_to != NULL) {
@@ -126,7 +126,7 @@ int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
break;
}
case BSP_TENSOR_DENSE: {
- cJSON_AddStringToObject(json_level, "level_kind", "dense");
+ cJSON_AddStringToObject(json_level, "level_desc", "dense");
cJSON_AddNumberToObject(json_level, "rank",
((bsp_dense_t*) level->data)->rank);
rank += ((bsp_dense_t*) level->data)->rank;
@@ -134,7 +134,7 @@ int bsp_write_tensor_to_group(hid_t f, bsp_tensor_t tensor, cJSON* user_json,
break;
}
case BSP_TENSOR_ELEMENT: {
- cJSON_AddStringToObject(json_level, "level_kind", "element");
+ cJSON_AddStringToObject(json_level, "level_desc", "element");
reached_end = 1;
break;
}
From 1683216a4b961ec61c8a8880b70927d79940950f Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Sun, 16 Mar 2025 14:21:24 -0400
Subject: [PATCH 10/24] change to shared library
---
CMakeLists.txt | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/CMakeLists.txt b/CMakeLists.txt
index a0c78b7..231146d 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -13,7 +13,7 @@ set(CMAKE_CXX_STANDARD 20)
set(CMAKE_C_FLAGS "-O3 -march=native")
-add_library(binsparse-rc STATIC)
+add_library(binsparse-rc SHARED)
add_subdirectory(include)
add_subdirectory(src)
From c6dcbb42da7cd506f77a741a6dc2c5b28ace9fe3 Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Mon, 31 Mar 2025 12:32:30 -0400
Subject: [PATCH 11/24] so it turns out that we need int32 indices
---
include/binsparse/tensor.h | 4 ++++
test/julia/tensor_test.jl | 32 ++++++++++++++++++++++++++++++++
2 files changed, 36 insertions(+)
diff --git a/include/binsparse/tensor.h b/include/binsparse/tensor.h
index 60dad01..3909318 100644
--- a/include/binsparse/tensor.h
+++ b/include/binsparse/tensor.h
@@ -23,7 +23,11 @@ typedef struct {
// corresponds to BSP_TENSOR_DENSE
typedef struct {
int rank;
+ // pointers_to, while it will only ever point to one bsp_array_t, must be kept
+ // as a pointer (rather than a struct) because there are cases where it MUST
+ // be null.
bsp_array_t* pointers_to;
+ // indices is supposed to be an array of bsp_array_t's.
bsp_array_t* indices;
bsp_level_t* child;
} bsp_sparse_t;
diff --git a/test/julia/tensor_test.jl b/test/julia/tensor_test.jl
index 844c758..3f9812f 100644
--- a/test/julia/tensor_test.jl
+++ b/test/julia/tensor_test.jl
@@ -57,3 +57,35 @@ tensortest(
joinpath(test_files, "input4.bsp.h5"),
joinpath(test_files, "output4.bsp.h5")
)
+
+tensortest(
+ Tensor(
+ Dense(SparseList{Int32}(Element{0.0,Float64,Int32}())),
+ [
+ 0 1 0 3;
+ 1 0 0 4;
+ ]
+ ),
+ joinpath(test_files, "inputcsr.bsp.h5"),
+ joinpath(test_files, "outputcsr.bsp.h5")
+)
+tensortest(
+ Tensor(
+ Dense(SparseList{Int32}(Element{0.0,Float64,Int32}())),
+ fsprand(10, 10, 0.1)
+ ),
+ joinpath(test_files, "inputcsr2.bsp.h5"),
+ joinpath(test_files, "outputcsr2.bsp.h5")
+)
+
+tensortest(
+ Tensor(
+ Dense(Dense(Element(0.0))),
+ [
+ 1 1 2 3;
+ 6 1 5 4;
+ ]
+ ),
+ joinpath(test_files, "inputdense.bsp.h5"),
+ joinpath(test_files, "outputdense.bsp.h5")
+)
From 467e819436bd8351ed66963b32e4cf185326eacf Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Mon, 31 Mar 2025 15:23:03 -0400
Subject: [PATCH 12/24] change everything to int32
---
test/julia/tensor_test.jl | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/test/julia/tensor_test.jl b/test/julia/tensor_test.jl
index 3f9812f..7671ffa 100644
--- a/test/julia/tensor_test.jl
+++ b/test/julia/tensor_test.jl
@@ -23,7 +23,7 @@ end
tensortest(
Tensor(
- Dense(SparseList(SparseList(Element(0.0)))),
+ Dense(SparseList{Int32}(SparseList{Int32}(Element{0.0,Float64,Int32}()))),
fsprand(10, 10, 10, 0.1)
),
joinpath(test_files, "input1.bsp.h5"),
@@ -42,7 +42,7 @@ tensortest(
tensortest(
Tensor(
- Dense(Dense(Dense(Element(0.0)))),
+ Dense(Dense(Dense(Element{0.0,Float64,Int32}()))),
fsprand(10, 10, 10, 0.1)
),
joinpath(test_files, "input3.bsp.h5"),
From 054253842acf88612ceb24516a4e0310da2556c4 Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Mon, 7 Apr 2025 11:59:09 -0400
Subject: [PATCH 13/24] make ready for pr
---
CMakeLists.txt | 44 --------------------------------------------
dev.sh | 35 -----------------------------------
src/read_matrix.c | 1 -
3 files changed, 80 deletions(-)
delete mode 100644 CMakeLists.txt
delete mode 100755 dev.sh
diff --git a/CMakeLists.txt b/CMakeLists.txt
deleted file mode 100644
index 231146d..0000000
--- a/CMakeLists.txt
+++ /dev/null
@@ -1,44 +0,0 @@
-# SPDX-FileCopyrightText: 2024 Binsparse Developers
-#
-# SPDX-License-Identifier: BSD-3-Clause
-
-cmake_minimum_required(VERSION 3.5)
-project(binsparse-rc)
-
-cmake_policy(SET CMP0079 NEW)
-
-set(CMAKE_C_STANDARD 11)
-
-set(CMAKE_CXX_STANDARD 20)
-
-set(CMAKE_C_FLAGS "-O3 -march=native")
-
-add_library(binsparse-rc SHARED)
-
-add_subdirectory(include)
-add_subdirectory(src)
-
-# NOTE: For now, both HDF5 and cJSON are `PUBLIC`, meaning that anything that
-# depends on `binsparse-rc` will also link/include HDF5 and cJSON. We can change
-# these to `PRIVATE` to use them only when building binsparse-rc.
-
-find_package(HDF5 REQUIRED COMPONENTS C)
-target_link_libraries(binsparse-rc PUBLIC ${HDF5_C_LIBRARIES})
-target_include_directories(binsparse-rc PUBLIC . ${HDF5_INCLUDE_DIRS})
-
-include(FetchContent)
-FetchContent_Declare(
- cJSON
- GIT_REPOSITORY https://github.com/DaveGamble/cJSON.git
- GIT_TAG v1.7.17
-)
-FetchContent_MakeAvailable(cJSON)
-
-configure_file(${cJSON_SOURCE_DIR}/cJSON.h ${CMAKE_BINARY_DIR}/include/cJSON/cJSON.h)
-target_include_directories(${PROJECT_NAME} PUBLIC ${CMAKE_BINARY_DIR}/include)
-target_link_libraries(${PROJECT_NAME} PUBLIC cjson)
-
-if(CMAKE_PROJECT_NAME STREQUAL PROJECT_NAME)
- add_subdirectory(examples)
- add_subdirectory(test)
-endif()
diff --git a/dev.sh b/dev.sh
deleted file mode 100755
index a86d79b..0000000
--- a/dev.sh
+++ /dev/null
@@ -1,35 +0,0 @@
-#!/usr/bin/env sh
-
-CONTAINER='binsparse-reference-c'
-IMAGE='docker.io/junikimm717/nvim2025:finch'
-DIR="$(realpath "$(dirname "$0")")"
-
-case "$1" in
- pull|p)
- podman pull "$IMAGE"
- (podman container ls | grep "$CONTAINER" > /dev/null 2>&1) && podman container rm -fv "$CONTAINER"
- ;;
- clear|c)
- (podman container ls -a | grep "$CONTAINER" > /dev/null 2>&1)\
- && {
- podman container kill "$CONTAINER" > /dev/null 2>&1;
- podman container rm "$CONTAINER"
- }
- ;;
- *)
- set +x
- if ! (podman container ls -a | grep "$CONTAINER" > /dev/null 2>&1); then
- podman run\
- -dt\
- --name "$CONTAINER"\
- --group-add keep-groups\
- -v "$DIR:/workspace"\
- --privileged\
- --rm\
- "$IMAGE"
- fi || exit 1
- podman exec\
- -e ENV=/root/.profile\
- -it "$CONTAINER" /bin/bash
- ;;
-esac
diff --git a/src/read_matrix.c b/src/read_matrix.c
index bf2b96d..464210c 100644
--- a/src/read_matrix.c
+++ b/src/read_matrix.c
@@ -143,7 +143,6 @@ bsp_matrix_t bsp_read_matrix_from_group(hid_t f) {
bsp_matrix_format_t format = bsp_get_matrix_format(format_string);
- // isn't this never true?
assert(format != 0);
matrix.format = format;
From 25126850e4d6da8b2d8f689c1fa8c66dff15c521 Mon Sep 17 00:00:00 2001
From: Benjamin Brock <benjamin.brock@intel.com>
Date: Mon, 7 Apr 2025 13:33:58 -0700
Subject: [PATCH 14/24] Fix formatting
---
examples/tensor_test.c | 5 +++--
include/binsparse/tensor.h | 6 ++++--
include/binsparse/write_tensor.h | 2 +-
3 files changed, 8 insertions(+), 5 deletions(-)
diff --git a/examples/tensor_test.c b/examples/tensor_test.c
index bfa0bfc..cd30060 100644
--- a/examples/tensor_test.c
+++ b/examples/tensor_test.c
@@ -1,10 +1,11 @@
-#include <binsparse/tensor.h>
#include <binsparse/read_tensor.h>
+#include <binsparse/tensor.h>
#include <binsparse/write_tensor.h>
int main(int argc, char** argv) {
if (argc < 3) {
- fprintf(stderr, "usage: ./tensor_test [file_name.h5] [output_file_name.h5]\n");
+ fprintf(stderr,
+ "usage: ./tensor_test [file_name.h5] [output_file_name.h5]\n");
return 1;
}
char* file_name = argv[1];
diff --git a/include/binsparse/tensor.h b/include/binsparse/tensor.h
index 3909318..8201e67 100644
--- a/include/binsparse/tensor.h
+++ b/include/binsparse/tensor.h
@@ -123,6 +123,8 @@ static bsp_array_t bsp_get_tensor_values(bsp_tensor_t tensor) {
static inline void bsp_destroy_tensor_t(bsp_tensor_t tensor) {
bsp_destroy_level_t(tensor.level);
- if (tensor.dims != NULL) free(tensor.dims);
- if (tensor.transpose != NULL) free(tensor.transpose);
+ if (tensor.dims != NULL)
+ free(tensor.dims);
+ if (tensor.transpose != NULL)
+ free(tensor.transpose);
}
diff --git a/include/binsparse/write_tensor.h b/include/binsparse/write_tensor.h
index 64497c2..27e1f29 100644
--- a/include/binsparse/write_tensor.h
+++ b/include/binsparse/write_tensor.h
@@ -12,7 +12,7 @@ extern "C" {
// TODO: make cJSON optional.
-#include<binsparse/tensor.h>
+#include <binsparse/tensor.h>
#include <cJSON/cJSON.h>
#ifdef BSP_USE_HDF5
From 5ac75e6da0696f341b71e8d2cb49459df8087deb Mon Sep 17 00:00:00 2001
From: Juni Kim <junikimm717@gmail.com>
Date: Sun, 20 Apr 2025 17:14:52 -0400
Subject: [PATCH 15/24] add cmakelists back in
---
CMakeLists.txt | 44 ++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 44 insertions(+)
create mode 100644 CMakeLists.txt
diff --git a/CMakeLists.txt b/CMakeLists.txt
new file mode 100644
index 0000000..231146d
--- /dev/null
+++ b/CMakeLists.txt
@@ -0,0 +1,44 @@
+# SPDX-FileCopyrightText: 2024 Binsparse Developers
+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+cmake_minimum_required(VERSION 3.5)
+project(binsparse-rc)
+
+cmake_policy(SET CMP0079 NEW)
+
+set(CMAKE_C_STANDARD 11)
+
+set(CMAKE_CXX_STANDARD 20)
+
+set(CMAKE_C_FLAGS "-O3 -march=native")
+
+add_library(binsparse-rc SHARED)
+
+add_subdirectory(include)
+add_subdirectory(src)
+
+# NOTE: For now, both HDF5 and cJSON are `PUBLIC`, meaning that anything that
+# depends on `binsparse-rc` will also link/include HDF5 and cJSON. We can change
+# these to `PRIVATE` to use them only when building binsparse-rc.
+
+find_package(HDF5 REQUIRED COMPONENTS C)
+target_link_libraries(binsparse-rc PUBLIC ${HDF5_C_LIBRARIES})
+target_include_directories(binsparse-rc PUBLIC . ${HDF5_INCLUDE_DIRS})
+
+include(FetchContent)
+FetchContent_Declare(
+ cJSON
+ GIT_REPOSITORY https://github.com/DaveGamble/cJSON.git
+ GIT_TAG v1.7.17
+)
+FetchContent_MakeAvailable(cJSON)
+
+configure_file(${cJSON_SOURCE_DIR}/cJSON.h ${CMAKE_BINARY_DIR}/include/cJSON/cJSON.h)
+target_include_directories(${PROJECT_NAME} PUBLIC ${CMAKE_BINARY_DIR}/include)
+target_link_libraries(${PROJECT_NAME} PUBLIC cjson)
+
+if(CMAKE_PROJECT_NAME STREQUAL PROJECT_NAME)
+ add_subdirectory(examples)
+ add_subdirectory(test)
+endif()
From 093add15ee37969643a0b4218f68a2692e1b16c9 Mon Sep 17 00:00:00 2001
From: Benjamin Brock <benjamin.brock@intel.com>
Date: Wed, 6 Aug 2025 15:29:08 -0700
Subject: [PATCH 16/24] Add SPDX headers, remove `compile_flags.txt`. (Use
`-DCMAKE_EXPORT_COMPILE_COMMANDS=ON` to generate automatically.)
---
.gitignore | 1 +
CMakeLists.txt | 10 +++++-----
compile_flags.txt | 4 ----
examples/cpp/tensor_test.cpp | 6 ++++++
examples/tensor_test.c | 6 ++++++
include/binsparse/read_tensor.h | 6 ++++++
include/binsparse/tensor.h | 6 ++++++
src/CMakeLists.txt | 4 ++--
src/write_tensor.c | 6 ++++++
test/julia/tensor_test.jl | 4 ++++
10 files changed, 42 insertions(+), 11 deletions(-)
delete mode 100644 compile_flags.txt
diff --git a/.gitignore b/.gitignore
index 8fc2fe2..8dddad7 100644
--- a/.gitignore
+++ b/.gitignore
@@ -5,5 +5,6 @@
scripts
venv
build*
+compile_flags.txt
._*
tensor_test_files
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 349bbe8..6ea7531 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -46,11 +46,6 @@ target_include_directories(${PROJECT_NAME}
$<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}>
${HDF5_INCLUDE_DIRS})
-if(CMAKE_PROJECT_NAME STREQUAL PROJECT_NAME)
- add_subdirectory(examples)
- add_subdirectory(test)
-endif()
-
# Installation rules - these are always needed when the library is built
install(TARGETS binsparse
EXPORT binsparse-targets
@@ -87,3 +82,8 @@ install(FILES
"${CMAKE_CURRENT_BINARY_DIR}/binsparse-config.cmake"
"${CMAKE_CURRENT_BINARY_DIR}/binsparse-config-version.cmake"
DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/binsparse)
+
+if(CMAKE_PROJECT_NAME STREQUAL PROJECT_NAME)
+ add_subdirectory(examples)
+ add_subdirectory(test)
+endif()
diff --git a/compile_flags.txt b/compile_flags.txt
deleted file mode 100644
index ba24856..0000000
--- a/compile_flags.txt
+++ /dev/null
@@ -1,4 +0,0 @@
--I./include
--I./build/include
--DBSP_USE_HDF5
--I/usr/include/hdf5/serial
diff --git a/examples/cpp/tensor_test.cpp b/examples/cpp/tensor_test.cpp
index 677d71e..893c334 100644
--- a/examples/cpp/tensor_test.cpp
+++ b/examples/cpp/tensor_test.cpp
@@ -1 +1,7 @@
+/*
+ * SPDX-FileCopyrightText: 2024 Binsparse Developers
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
#include "../tensor_test.c"
diff --git a/examples/tensor_test.c b/examples/tensor_test.c
index cd30060..5d3ba6e 100644
--- a/examples/tensor_test.c
+++ b/examples/tensor_test.c
@@ -1,3 +1,9 @@
+/*
+ * SPDX-FileCopyrightText: 2024 Binsparse Developers
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
#include <binsparse/read_tensor.h>
#include <binsparse/tensor.h>
#include <binsparse/write_tensor.h>
diff --git a/include/binsparse/read_tensor.h b/include/binsparse/read_tensor.h
index 9d419ec..579c5c7 100644
--- a/include/binsparse/read_tensor.h
+++ b/include/binsparse/read_tensor.h
@@ -1,3 +1,9 @@
+/*
+ * SPDX-FileCopyrightText: 2024 Binsparse Developers
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
#pragma once
#ifdef __cplusplus
diff --git a/include/binsparse/tensor.h b/include/binsparse/tensor.h
index 8201e67..74863b2 100644
--- a/include/binsparse/tensor.h
+++ b/include/binsparse/tensor.h
@@ -1,3 +1,9 @@
+/*
+ * SPDX-FileCopyrightText: 2024 Binsparse Developers
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
#pragma once
#include <binsparse/array.h>
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index ee41ab5..0b82bd2 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -4,7 +4,7 @@
target_sources(binsparse PRIVATE
read_matrix.c
- src/read_tensor.c
+ read_tensor.c
write_matrix.c
- src/write_tensor.c
+ write_tensor.c
)
diff --git a/src/write_tensor.c b/src/write_tensor.c
index 52a4877..11a84ef 100644
--- a/src/write_tensor.c
+++ b/src/write_tensor.c
@@ -1,3 +1,9 @@
+/*
+ * SPDX-FileCopyrightText: 2024 Binsparse Developers
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
#include <assert.h>
#include <binsparse/tensor.h>
#include <unistd.h>
diff --git a/test/julia/tensor_test.jl b/test/julia/tensor_test.jl
index 7671ffa..cbba45f 100644
--- a/test/julia/tensor_test.jl
+++ b/test/julia/tensor_test.jl
@@ -1,3 +1,7 @@
+# SPDX-FileCopyrightText: 2024 Binsparse Developers
+#
+# SPDX-License-Identifier: BSD-3-Clause
+
using Finch;
using HDF5;
From 0e75d8fb729c351ef28f8097a8c28175e94d1c44 Mon Sep 17 00:00:00 2001
From: Benjamin Brock <benjamin.brock@intel.com>
Date: Wed, 6 Aug 2025 15:35:10 -0700
Subject: [PATCH 17/24] Add missing includes.
---
include/binsparse/read_tensor.h | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/include/binsparse/read_tensor.h b/include/binsparse/read_tensor.h
index 579c5c7..bcf3145 100644
--- a/include/binsparse/read_tensor.h
+++ b/include/binsparse/read_tensor.h
@@ -6,6 +6,14 @@
#pragma once
+#include <binsparse/tensor.h>
+
+#ifndef __cplusplus
+#include <stddef.h>
+#else
+#include <cstddef>
+#endif
+
#ifdef __cplusplus
extern "C" {
#endif
From 35435891beaf99d725b3e90175cf78488acffbc9 Mon Sep 17 00:00:00 2001
From: Benjamin Brock <benjamin.brock@intel.com>
Date: Wed, 6 Aug 2025 16:00:36 -0700
Subject: [PATCH 18/24] Add Tensor tests.
---
.github/workflows/ci.yml | 16 ++++++++++++++--
1 file changed, 14 insertions(+), 2 deletions(-)
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 365a69d..0ae2e73 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -38,8 +38,14 @@ jobs:
cmake -B build
- name: Build
run: VERBOSE=true make -C build -j `nproc`
- - name: Test
+ - name: Test-Matrix
run: ctest --test-dir ./build/test/bash
+ - name: Install-Julia
+ run: curl -fsSL https://install.julialang.org | sh
+ - name: Install-Finch
+ run: julia -e 'using Pkg; Pkg.add("Finch")'
+ - name: Test-Tensor
+ run: ctest --test-dir ./build/test/julia
clang:
runs-on: 'ubuntu-latest'
@@ -55,5 +61,11 @@ jobs:
cmake -B build
- name: Build
run: VERBOSE=true make -C build -j `nproc`
- - name: Test
+ - name: Test-Matrix
run: ctest --test-dir ./build/test/bash
+ - name: Install-Julia
+ run: curl -fsSL https://install.julialang.org | sh
+ - name: Install-Finch
+ run: julia -e 'using Pkg; Pkg.add("Finch")'
+ - name: Test-Tensor
+ run: ctest --test-dir ./build/test/julia
From 3abac30bd130f1e9eeabde66f374262c989de958 Mon Sep 17 00:00:00 2001
From: Benjamin Brock <benjamin.brock@intel.com>
Date: Wed, 6 Aug 2025 16:15:22 -0700
Subject: [PATCH 19/24] Fix Julia install
---
.github/workflows/ci.yml | 41 +++++++++++++---------------------------
1 file changed, 13 insertions(+), 28 deletions(-)
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 0ae2e73..ef207c3 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -23,12 +23,20 @@ jobs:
- run: pip install -r requirements.txt
- name: Checks
uses: pre-commit/action@v3.0.0
-
- gcc:
+
+ build:
runs-on: 'ubuntu-latest'
+ strategy:
+ matrix:
+ include:
+ - cc: gcc-12
+ cxx: g++-12
+ - cc: clang
+ cxx: clang++
+ name: ${{ matrix.cc }}
env:
- CXX: g++-12
- CC: gcc-12
+ CC: ${{ matrix.cc }}
+ CXX: ${{ matrix.cxx }}
steps:
- uses: actions/checkout@v4
- name: CMake
@@ -41,30 +49,7 @@ jobs:
- name: Test-Matrix
run: ctest --test-dir ./build/test/bash
- name: Install-Julia
- run: curl -fsSL https://install.julialang.org | sh
- - name: Install-Finch
- run: julia -e 'using Pkg; Pkg.add("Finch")'
- - name: Test-Tensor
- run: ctest --test-dir ./build/test/julia
-
- clang:
- runs-on: 'ubuntu-latest'
- env:
- CXX: clang++
- CC: clang
- steps:
- - uses: actions/checkout@v4
- - name: CMake
- run: |
- sudo apt-get update
- sudo apt-get install libhdf5-dev clang
- cmake -B build
- - name: Build
- run: VERBOSE=true make -C build -j `nproc`
- - name: Test-Matrix
- run: ctest --test-dir ./build/test/bash
- - name: Install-Julia
- run: curl -fsSL https://install.julialang.org | sh
+ run: curl -fsSL https://install.julialang.org | sh -s -- -y
- name: Install-Finch
run: julia -e 'using Pkg; Pkg.add("Finch")'
- name: Test-Tensor
From 53f2a76573681ce0f31298d674a47e3c68dbad50 Mon Sep 17 00:00:00 2001
From: Benjamin Brock <benjamin.brock@intel.com>
Date: Wed, 6 Aug 2025 16:18:10 -0700
Subject: [PATCH 20/24] Fix whitespace
---
.github/workflows/ci.yml | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index ef207c3..86769f5 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -23,7 +23,7 @@ jobs:
- run: pip install -r requirements.txt
- name: Checks
uses: pre-commit/action@v3.0.0
-
+
build:
runs-on: 'ubuntu-latest'
strategy:
From 79b7f955a96c9a837440a2b18c996dfd703ab939 Mon Sep 17 00:00:00 2001
From: Benjamin Brock <benjamin.brock@intel.com>
Date: Wed, 6 Aug 2025 16:33:18 -0700
Subject: [PATCH 21/24] Turn off Julia pre-compilation, install HDF5.
---
.github/workflows/ci.yml | 14 ++++++++------
1 file changed, 8 insertions(+), 6 deletions(-)
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 86769f5..938bf93 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -42,15 +42,17 @@ jobs:
- name: CMake
run: |
sudo apt-get update
- sudo apt-get install libhdf5-dev $CXX $CC
+ sudo apt-get install libhdf5-dev gcc-12 g++-12 clang
cmake -B build
- name: Build
run: VERBOSE=true make -C build -j `nproc`
- - name: Test-Matrix
+ - name: Matrix Tests
run: ctest --test-dir ./build/test/bash
- - name: Install-Julia
+ - name: Install Julia
run: curl -fsSL https://install.julialang.org | sh -s -- -y
- - name: Install-Finch
- run: julia -e 'using Pkg; Pkg.add("Finch")'
- - name: Test-Tensor
+ - name: Disable Julia precompilation
+ run: julia -e 'using PrecompileTools, Preferences; set_preferences!(PrecompileTools, "precompile_workloads" => false; force=true)'
+ - name: Install Julia Packages
+ run: julia -e 'using Pkg; Pkg.add(["Finch", "HDF5"])'
+ - name: Tensor Tests
run: ctest --test-dir ./build/test/julia
From 8980791a1316684eb6b5fda850ee7efa6e5b0df6 Mon Sep 17 00:00:00 2001
From: Benjamin Brock <benjamin.brock@intel.com>
Date: Wed, 6 Aug 2025 16:39:49 -0700
Subject: [PATCH 22/24] Add missing package.
---
.github/workflows/ci.yml | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 938bf93..3d8680c 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -51,7 +51,7 @@ jobs:
- name: Install Julia
run: curl -fsSL https://install.julialang.org | sh -s -- -y
- name: Disable Julia precompilation
- run: julia -e 'using PrecompileTools, Preferences; set_preferences!(PrecompileTools, "precompile_workloads" => false; force=true)'
+ run: julia -e 'using Pkg; Pkg.add("PrecompileTools"); using PrecompileTools, Preferences; set_preferences!(PrecompileTools, "precompile_workloads" => false; force=true)'
- name: Install Julia Packages
run: julia -e 'using Pkg; Pkg.add(["Finch", "HDF5"])'
- name: Tensor Tests
From 0c3f3f83b8d1afbe14a4556014262d1f06e94ef4 Mon Sep 17 00:00:00 2001
From: Benjamin Brock <benjamin.brock@intel.com>
Date: Wed, 6 Aug 2025 16:41:42 -0700
Subject: [PATCH 23/24] Add missing Julia package.
---
.github/workflows/ci.yml | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 3d8680c..516614f 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -51,7 +51,7 @@ jobs:
- name: Install Julia
run: curl -fsSL https://install.julialang.org | sh -s -- -y
- name: Disable Julia precompilation
- run: julia -e 'using Pkg; Pkg.add("PrecompileTools"); using PrecompileTools, Preferences; set_preferences!(PrecompileTools, "precompile_workloads" => false; force=true)'
+ run: julia -e 'using Pkg; Pkg.add(["PrecompileTools", "Preferences"]); using PrecompileTools, Preferences; set_preferences!(PrecompileTools, "precompile_workloads" => false; force=true)'
- name: Install Julia Packages
run: julia -e 'using Pkg; Pkg.add(["Finch", "HDF5"])'
- name: Tensor Tests
From e2c21a0cd59a5c9ac46a2057418cd4835a54d71a Mon Sep 17 00:00:00 2001
From: Benjamin Brock <benjamin.brock@intel.com>
Date: Wed, 6 Aug 2025 17:44:51 -0700
Subject: [PATCH 24/24] Correct comment.
---
include/binsparse/tensor.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/include/binsparse/tensor.h b/include/binsparse/tensor.h
index 74863b2..4d63512 100644
--- a/include/binsparse/tensor.h
+++ b/include/binsparse/tensor.h
@@ -26,7 +26,7 @@ typedef struct {
bsp_array_t values;
} bsp_element_t;
-// corresponds to BSP_TENSOR_DENSE
+// corresponds to BSP_TENSOR_SPARSE
typedef struct {
int rank;
// pointers_to, while it will only ever point to one bsp_array_t, must be kept
@@ -38,6 +38,7 @@ typedef struct {
bsp_level_t* child;
} bsp_sparse_t;
+// corresponds to BSP_TENSOR_DENSE
typedef struct {
int rank;
bsp_level_t* child;