eliaskohout/ctensor
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Renamed the index_offsets array to stride. Performance improvements when creating the stride array.

Browse Source
This commit is contained in:
Elias Kohout
2023-09-07 11:54:55 +02:00
parent 685119e839
commit 526deb402e
2 changed files with 16 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
tensor.c
View File

@@ -18,7 +18,7 @@ void tensor_destroy(tensor t)
if (!tensor_is_empty(t)) { if (!tensor_is_empty(t)) {
free(t->size); free(t->size);
free(t->elements); free(t->elements);
free(t->index_offsets); free(t->stride);
} }
free(t); free(t);
} }
@@ -87,42 +87,40 @@ bool _tensor_set_size(tensor t, const uint32_t *size, uint8_t rank)
*/ */
uint32_t *temp_size; uint32_t *temp_size;
uint32_t *temp_index_offset; uint32_t *temp_stride;
dtype *temp_elements; dtype *temp_elements;
uint8_t i, j; uint8_t i;
uint32_t num_elem = 1; uint32_t num_elem = 1;
if(!_tensor_check_size(size, rank)) return false; if(!_tensor_check_size(size, rank)) return false;
/* Try allocating memory for the size/ index_offset array of the tensor */ /* Try allocating memory for the size/ stride array of the tensor */
for(i = 0; i < rank; i++) { for(i = 0; i < rank; i++) {
num_elem *= size[i]; num_elem *= size[i];
} }
temp_size = malloc(rank * sizeof(uint32_t)); temp_size = malloc(rank * sizeof(uint32_t));
temp_index_offset = malloc(rank * sizeof(uint32_t)); temp_stride = malloc(rank * sizeof(uint32_t));
temp_elements = malloc(num_elem * sizeof(dtype)); temp_elements = malloc(num_elem * sizeof(dtype));
if((temp_size == NULL && rank != 0) || (temp_index_offset == NULL && rank != 0) || temp_elements == NULL) { if((temp_size == NULL && rank != 0) || (temp_stride == NULL && rank != 0) || temp_elements == NULL) {
free(temp_size); free(temp_size);
free(temp_index_offset); free(temp_stride);
return false; return false;
} }
/* Freeing old memory. */ /* Freeing old memory. */
free(t->size); free(t->size);
free(t->index_offsets); free(t->stride);
free(t->elements); free(t->elements);
/* Setting the size array */ /* Setting the size array */
t->size = temp_size; t->size = temp_size;
if(rank != 0) memcpy(t->size, size, rank * sizeof(uint32_t)); if(rank != 0) memcpy(t->size, size, rank * sizeof(uint32_t));
t->rank = rank; t->rank = rank;
/* Setting the index_offset array */ /* Setting the stride array */
t->index_offsets = temp_index_offset; t->stride = temp_stride;
for(i = 0; i < t->rank; i++) { t->stride[0] = 1;
t->index_offsets[i] = 1; for(i = 1; i < t->rank; i++) {
for(j = i + 1; j < t->rank; j++) { t->stride[i] = t->stride[i - 1] * t->size[i - 1];
t->index_offsets[i] *= t->size[j];
}
} }
/* Setting the elements pointer and memory usage */ /* Setting the elements pointer and memory usage */
t->elements = temp_elements; t->elements = temp_elements;
@@ -154,7 +152,7 @@ bool tensor_set(tensor t, const uint32_t *index, dtype val)
for(i = 0; i < t->rank; i++) { for(i = 0; i < t->rank; i++) {
if(t->size[i] <= index[i]) return false; if(t->size[i] <= index[i]) return false;
offset += t->index_offsets[i] * index[i]; offset += t->stride[i] * index[i];
} }
t->elements[offset] = val; t->elements[offset] = val;
@@ -185,7 +183,7 @@ dtype tensor_get(const tensor t, const uint32_t *index, bool *success)
if(success != NULL) *success = false; if(success != NULL) *success = false;
return DTYPE_ZERO; return DTYPE_ZERO;
} }
offset += t->index_offsets[i] * index[i]; offset += t->stride[i] * index[i];
} }
if(success != NULL) *success = true; if(success != NULL) *success = true;

2
tensor.h
View File

@@ -16,7 +16,7 @@ typedef struct _tensor {
dtype *elements; dtype *elements;
uint8_t rank; uint8_t rank;
uint32_t *size; uint32_t *size;
uint32_t *index_offsets; uint32_t *stride;
uint32_t num_elem; uint32_t num_elem;
} *tensor; } *tensor;