X-Git-Url: https://main.carlh.net/gitweb/?a=blobdiff_plain;f=libopenjpeg%2Fmct.c;h=870993b06d250b01f4a8163b960dd00934190224;hb=6e0162a8a67932e59489b6f6a3bec01d64d2db8b;hp=f41c9ab05baed69375bd5bcebd2f3e360f4ed0a8;hpb=95bc884365deb41c357583874c23d82eac7cad2d;p=openjpeg.git

diff --git a/libopenjpeg/mct.c b/libopenjpeg/mct.c
index f41c9ab0..870993b0 100644
--- a/libopenjpeg/mct.c
+++ b/libopenjpeg/mct.c
@@ -1,9 +1,10 @@
 /*
+ * Copyright (c) 2002-2007, Communications and Remote Sensing Laboratory, Universite catholique de Louvain (UCL), Belgium
+ * Copyright (c) 2002-2007, Professor Benoit Macq
  * Copyright (c) 2001-2003, David Janssens
  * Copyright (c) 2002-2003, Yannick Verschueren
- * Copyright (c) 2003-2005, Francois Devaux and Antonin Descampe
- * Copyright (c) 2005, Hervé Drolon, FreeImage Team
- * Copyright (c) 2002-2005, Communications and remote sensing Laboratory, Universite catholique de Louvain, Belgium
+ * Copyright (c) 2003-2007, Francois-Olivier Devaux and Antonin Descampe
+ * Copyright (c) 2005, Herve Drolon, FreeImage Team
  * All rights reserved.
  *
  * Redistribution and use in source and binary forms, with or without
@@ -28,6 +29,10 @@
  * POSSIBILITY OF SUCH DAMAGE.
  */
 
+#ifdef __SSE__
+#include <xmmintrin.h>
+#endif
+
 #include "opj_includes.h"
 
 /* <summary> */
@@ -43,16 +48,20 @@ static const double mct_norms_real[3] = { 1.732, 1.805, 1.573 };
 /* <summary> */
 /* Foward reversible MCT. */
 /* </summary> */
-void mct_encode(int *c0, int *c1, int *c2, int n) {
+void mct_encode(
+		int* restrict c0,
+		int* restrict c1,
+		int* restrict c2,
+		int n)
+{
 	int i;
-	for (i = 0; i < n; i++) {
-		int r, g, b, y, u, v;
-		r = c0[i];
-		g = c1[i];
-		b = c2[i];
-		y = (r + (g << 1) + b) >> 2;
-		u = b - g;
-		v = r - g;
+	for(i = 0; i < n; ++i) {
+		int r = c0[i];
+		int g = c1[i];
+		int b = c2[i];
+		int y = (r + (g * 2) + b) >> 2;
+		int u = b - g;
+		int v = r - g;
 		c0[i] = y;
 		c1[i] = u;
 		c2[i] = v;
@@ -62,16 +71,20 @@ void mct_encode(int *c0, int *c1, int *c2, int n) {
 /* <summary> */
 /* Inverse reversible MCT. */
 /* </summary> */
-void mct_decode(int *c0, int *c1, int *c2, int n) {
+void mct_decode(
+		int* restrict c0,
+		int* restrict c1, 
+		int* restrict c2, 
+		int n)
+{
 	int i;
-	for (i = 0; i < n; i++) {
-		int y, u, v, r, g, b;
-		y = c0[i];
-		u = c1[i];
-		v = c2[i];
-		g = y - ((u + v) >> 2);
-		r = v + g;
-		b = u + g;
+	for (i = 0; i < n; ++i) {
+		int y = c0[i];
+		int u = c1[i];
+		int v = c2[i];
+		int g = y - ((u + v) >> 2);
+		int r = v + g;
+		int b = u + g;
 		c0[i] = r;
 		c1[i] = g;
 		c2[i] = b;
@@ -88,16 +101,20 @@ double mct_getnorm(int compno) {
 /* <summary> */
 /* Foward irreversible MCT. */
 /* </summary> */
-void mct_encode_real(int *c0, int *c1, int *c2, int n) {
+void mct_encode_real(
+		int* restrict c0,
+		int* restrict c1,
+		int* restrict c2,
+		int n)
+{
 	int i;
-	for (i = 0; i < n; i++) {
-		int r, g, b, y, u, v;
-		r = c0[i];
-		g = c1[i];
-		b = c2[i];
-		y = fix_mul(r, 2449) + fix_mul(g, 4809) + fix_mul(b, 934);
-		u = -fix_mul(r, 1382) - fix_mul(g, 2714) + fix_mul(b, 4096);
-		v = fix_mul(r, 4096) - fix_mul(g, 3430) - fix_mul(b, 666);
+	for(i = 0; i < n; ++i) {
+		int r = c0[i];
+		int g = c1[i];
+		int b = c2[i];
+		int y =  fix_mul(r, 2449) + fix_mul(g, 4809) + fix_mul(b, 934);
+		int u = -fix_mul(r, 1382) - fix_mul(g, 2714) + fix_mul(b, 4096);
+		int v =  fix_mul(r, 4096) - fix_mul(g, 3430) - fix_mul(b, 666);
 		c0[i] = y;
 		c1[i] = u;
 		c2[i] = v;
@@ -107,16 +124,58 @@ void mct_encode_real(int *c0, int *c1, int *c2, int n) {
 /* <summary> */
 /* Inverse irreversible MCT. */
 /* </summary> */
-void mct_decode_real(int *c0, int *c1, int *c2, int n) {
+void mct_decode_real(
+		float* restrict c0,
+		float* restrict c1,
+		float* restrict c2,
+		int n)
+{
 	int i;
-	for (i = 0; i < n; i++) {
-		int y, u, v, r, g, b;
-		y = c0[i];
-		u = c1[i];
-		v = c2[i];
-		r = y + fix_mul(v, 11485);
-		g = y - fix_mul(u, 2819) - fix_mul(v, 5850);
-		b = y + fix_mul(u, 14516);
+#ifdef __SSE__
+	__m128 vrv, vgu, vgv, vbu;
+	vrv = _mm_set1_ps(1.402f);
+	vgu = _mm_set1_ps(0.34413f);
+	vgv = _mm_set1_ps(0.71414f);
+	vbu = _mm_set1_ps(1.772f);
+	for (i = 0; i < (n >> 3); ++i) {
+		__m128 vy, vu, vv;
+		__m128 vr, vg, vb;
+
+		vy = _mm_load_ps(c0);
+		vu = _mm_load_ps(c1);
+		vv = _mm_load_ps(c2);
+		vr = _mm_add_ps(vy, _mm_mul_ps(vv, vrv));
+		vg = _mm_sub_ps(_mm_sub_ps(vy, _mm_mul_ps(vu, vgu)), _mm_mul_ps(vv, vgv));
+		vb = _mm_add_ps(vy, _mm_mul_ps(vu, vbu));
+		_mm_store_ps(c0, vr);
+		_mm_store_ps(c1, vg);
+		_mm_store_ps(c2, vb);
+		c0 += 4;
+		c1 += 4;
+		c2 += 4;
+
+		vy = _mm_load_ps(c0);
+		vu = _mm_load_ps(c1);
+		vv = _mm_load_ps(c2);
+		vr = _mm_add_ps(vy, _mm_mul_ps(vv, vrv));
+		vg = _mm_sub_ps(_mm_sub_ps(vy, _mm_mul_ps(vu, vgu)), _mm_mul_ps(vv, vgv));
+		vb = _mm_add_ps(vy, _mm_mul_ps(vu, vbu));
+		_mm_store_ps(c0, vr);
+		_mm_store_ps(c1, vg);
+		_mm_store_ps(c2, vb);
+		c0 += 4;
+		c1 += 4;
+		c2 += 4;
+	}
+	n &= 7;
+#endif
+	for(i = 0; i < n; ++i) {
+		float y = c0[i];
+		float u = c1[i];
+		float v = c2[i];
+		float r = y + (v * 1.402f);
+		float g = y - (u * 0.34413f) - (v * (0.71414f));
+		float b = y + (u * 1.772f);
 		c0[i] = r;
 		c1[i] = g;
 		c2[i] = b;