Last active
December 22, 2015 23:59
-
-
Save zenoalbisser/6550188 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
commit 6058ce044452c0c5fbedbad565773c3a1b35d462 | |
Author: Zeno Albisser <zeno.albisser@digia.com> | |
Date: Thu Sep 12 22:36:42 2013 +0200 | |
debug that one | |
diff --git a/common.c b/common.c | |
index dec4589..56f9a53 100644 | |
--- a/common.c | |
+++ b/common.c | |
@@ -6,6 +6,7 @@ | |
#include <stdio.h> | |
#include <stdlib.h> | |
#include <string.h> | |
+#include <emmintrin.h> | |
#include "c63.h" | |
#include "tables.h" | |
@@ -14,29 +15,36 @@ void dequantize_idct_row(int16_t *in_data, uint8_t *prediction, int w, int h, | |
int y, uint8_t *out_data, uint8_t *quantization) | |
{ | |
int x; | |
- | |
int16_t block[8*8]; | |
+ __m128i in; | |
+ __m128i pred; | |
+ __m128i mask_min = _mm_set1_epi16(255); | |
+ __m128i mask_max = _mm_setzero_si128(); | |
/* Perform the dequantization and iDCT */ | |
for(x = 0; x < w; x += 8) | |
{ | |
- int i, j; | |
- | |
dequant_idct_block_8x8(in_data+(x*8), block, quantization); | |
- for (i = 0; i < 8; ++i) | |
+ int i; | |
+ for (i = 0; i < 8; i++) | |
{ | |
- for (j = 0; j < 8; ++j) | |
- { | |
- /* Add prediction block. Note: DCT is not precise - | |
- Clamp to legal values */ | |
- int16_t tmp = block[i*8+j] + (int16_t)prediction[i*w+j+x]; | |
+ // Fetch | |
+ in = _mm_loadu_si128(((__m128i*)&block[i*8])); // Convert the type by changing pointer | |
- if (tmp < 0) { tmp = 0; } | |
- else if (tmp > 255) { tmp = 255; } | |
+ pred = _mm_loadl_epi64(((__m128i*)&prediction[i*w+x])); // Convert the type by changing pointer | |
+ pred = _mm_unpacklo_epi8(pred, _mm_setzero_si128()); // and then get values | |
- out_data[i*w+j+x] = tmp; | |
- } | |
+ // Compute | |
+ in = _mm_adds_epi16(in, pred); | |
+ | |
+ // Pack and clamp | |
+ in = _mm_packus_epi16 (in, _mm_setzero_si128()); | |
+ | |
+ // Save | |
+ _mm_storel_epi64((__m128i*)(out_data+(i*w+x)), in); | |
} | |
} | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment