+#include <math.h>
+#include <string.h>
+#include <stdlib.h>
+
+#include "idct_reference.h"
+
+void* idct_reference_alloc(const uint32_t* quant_table)
+{
+ uint32_t* qt_copy = (uint32_t*)malloc(DCTSIZE2 * sizeof(uint32_t));
+ // FIXME: check for NULL return
+
+ memcpy(qt_copy, quant_table, DCTSIZE2 * sizeof(uint32_t));
+
+ return qt_copy;
+}
+
+void idct_reference_free(void* userdata)
+{
+ free(userdata);
+}
+
+void idct_reference(const int16_t* input, const void* userdata, uint8_t* output)
+{
+ const uint32_t* quant_table = (const uint32_t*)userdata;
+ double temp[DCTSIZE2];
+
+ for (unsigned y = 0; y < 8; ++y) {
+ for (unsigned x = 0; x < 8; ++x) {
+ double acc = 0.0;
+ for (unsigned u = 0; u < 8; ++u) {
+ for (unsigned v = 0; v < 8; ++v) {
+ double c_u = (u == 0) ? 1/sqrt(2.0) : 1.0;
+ double c_v = (v == 0) ? 1/sqrt(2.0) : 1.0;
+ acc += c_u * c_v
+ * input[u * DCTSIZE + v] * quant_table[u * DCTSIZE + v]
+ * cos((2 * x + 1) * v * M_PI / 16.0)
+ * cos((2 * y + 1) * u * M_PI / 16.0);
+ }
+ }
+ temp[y * DCTSIZE + x] = 0.25 * acc;
+ }
+ }
+
+ for (unsigned y = 0; y < 8; ++y) {
+ for (unsigned x = 0; x < 8; ++x) {
+ double val = temp[y * DCTSIZE + x];
+ if (val < 0.0) {
+ output[y * DCTSIZE + x] = 0;
+ } else if (val >= 255.0) {
+ output[y * DCTSIZE + x] = 255;
+ } else {
+ output[y * DCTSIZE + x] = (uint8_t)(val + 0.5);
+ }
+ }
+ }
+}