From 7f62a36f9b004ea35de52ecf99a695f90efb495c Mon Sep 17 00:00:00 2001
From: yumetodo <yume-wikijp@live.jp>
Date: Sun, 30 Oct 2016 15:37:32 +0900
Subject: [PATCH 1/2] specify const to promise no write accsess

---
 SigColorFastAviUtl/sigcolorfastaviutl.cpp | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/SigColorFastAviUtl/sigcolorfastaviutl.cpp b/SigColorFastAviUtl/sigcolorfastaviutl.cpp
index 33cb797..23240ab 100644
--- a/SigColorFastAviUtl/sigcolorfastaviutl.cpp
+++ b/SigColorFastAviUtl/sigcolorfastaviutl.cpp
@@ -196,7 +196,7 @@ BOOL func_proc_con(FILTER *fp, FILTER_PROC_INFO *fpip) // This is the main image
 	if (prevIsYC_Con)
 	{
 		/* Scan Y channel data */
-		int fh = fpip->h;
+		const int fh = fpip->h;
 
 #pragma loop( hint_parallel(0) )
 #pragma loop( ivdep )
@@ -212,8 +212,8 @@ BOOL func_proc_con(FILTER *fp, FILTER_PROC_INFO *fpip) // This is the main image
 	}
 	else //RGB mode
 	{
-		int fh = fpip->h;
-		int fw = fpip->w;
+		const int fh = fpip->h;
+		const int fw = fpip->w;
 #pragma loop( hint_parallel(0) )
 #pragma loop( ivdep )
 		for (int r = 0; r < fh; r++)
@@ -449,7 +449,7 @@ BOOL func_proc_sd(FILTER *fp, FILTER_PROC_INFO *fpip) // This is the main image
 	if (prevIsYC_SD)
 	{
 		/* Scan Y channel data */
-		int fh = fpip->h;
+		const int fh = fpip->h;
 
 #pragma loop( hint_parallel(0) )
 #pragma loop( ivdep )
@@ -465,8 +465,8 @@ BOOL func_proc_sd(FILTER *fp, FILTER_PROC_INFO *fpip) // This is the main image
 	}
 	else //RGB mode
 	{
-		int fh = fpip->h;
-		int fw = fpip->w;
+		const int fh = fpip->h;
+		const int fw = fpip->w;
 #pragma loop( hint_parallel(0) )
 #pragma loop( ivdep )
 		for (int r = 0; r < fh; r++)

From d74a3b874997d852c8933fe72a00ce97414b7d7e Mon Sep 17 00:00:00 2001
From: yumetodo <yume-wikijp@live.jp>
Date: Sun, 30 Oct 2016 15:41:54 +0900
Subject: [PATCH 2/2] allocate buffer pixel out side loop

auto-parallelizer's auto storage is very heavy. so, allocate buffer on heap as same as before but allocate only one time per thread
---
 SigColorFastAviUtl/sigcolorfastaviutl.cpp | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/SigColorFastAviUtl/sigcolorfastaviutl.cpp b/SigColorFastAviUtl/sigcolorfastaviutl.cpp
index 23240ab..908cad2 100644
--- a/SigColorFastAviUtl/sigcolorfastaviutl.cpp
+++ b/SigColorFastAviUtl/sigcolorfastaviutl.cpp
@@ -218,11 +218,11 @@ BOOL func_proc_con(FILTER *fp, FILTER_PROC_INFO *fpip) // This is the main image
 #pragma loop( ivdep )
 		for (int r = 0; r < fh; r++)
 		{
+			PIXEL* rgb = new PIXEL;
 #pragma loop( no_vector )
 			for (int c = 0; c < fw; c++)
 			{
 				PIXEL_YC* px = fpip->ycp_edit + r* fpip->max_w + c;
-				PIXEL* rgb= new PIXEL;
 				fp->exfunc->yc2rgb(rgb, px, 1);
 				// transform each channel is needed
 				//PIXEL t_rgb{ 0 };
@@ -243,8 +243,8 @@ BOOL func_proc_con(FILTER *fp, FILTER_PROC_INFO *fpip) // This is the main image
 				}
 				// convert back
 				fp->exfunc->rgb2yc(px, rgb, 1);
-				delete rgb;
 			}
+			delete rgb;
 		}
 	}
 #ifdef USECLOCK
@@ -471,11 +471,11 @@ BOOL func_proc_sd(FILTER *fp, FILTER_PROC_INFO *fpip) // This is the main image
 #pragma loop( ivdep )
 		for (int r = 0; r < fh; r++)
 		{
+			PIXEL* rgb = new PIXEL;
 #pragma loop( no_vector )
 			for (int c = 0; c < fw; c++)
 			{
 				PIXEL_YC* px = fpip->ycp_edit + r* fpip->max_w + c;
-				PIXEL* rgb = new PIXEL;
 				fp->exfunc->yc2rgb(rgb, px, 1);
 				// transform each channel is needed
 				//PIXEL t_rgb{ 0 };
@@ -496,8 +496,8 @@ BOOL func_proc_sd(FILTER *fp, FILTER_PROC_INFO *fpip) // This is the main image
 				}
 				// convert back
 				fp->exfunc->rgb2yc(px, rgb, 1);
-				delete rgb;
 			}
+			delete rgb;
 		}
 	}
 #ifdef USECLOCK