X-Git-Url: https://main.carlh.net/gitweb/?a=blobdiff_plain;f=libs%2Fardour%2Fsse_functions_xmm.cc;h=48212ea8e1b839363cc08e45e2ce2e86222cf72b;hb=751cc84dd2010a2c669a0155c0e2a0ce47d16592;hp=7b5ea143ecdd6e62bc48575d6e248c1731428897;hpb=75d2f51193f6fd25881a9c766db9078f3b68d80e;p=ardour.git

diff --git a/libs/ardour/sse_functions_xmm.cc b/libs/ardour/sse_functions_xmm.cc
index 7b5ea143ec..48212ea8e1 100644
--- a/libs/ardour/sse_functions_xmm.cc
+++ b/libs/ardour/sse_functions_xmm.cc
@@ -1,6 +1,6 @@
 /*
-    Copyright (C) 2007 Paul Davis
-    	Written by Sampo Savolainen
+    Copyright (C) 2007 Paul sDavis
+    Written by Sampo Savolainen
 
     This program is free software; you can redistribute it and/or modify
     it under the terms of the GNU General Public License as published by
@@ -19,10 +19,10 @@
 */
 
 #include <xmmintrin.h>
-#include <ardour/types.h>
+#include "ardour/types.h"
 
 void
-x86_sse_find_peaks(float *buf, nframes_t nframes, float *min, float *max)
+x86_sse_find_peaks(const ARDOUR::Sample* buf, ARDOUR::pframes_t nframes, float *min, float *max)
 {
 	__m128 current_max, current_min, work;
 
@@ -31,18 +31,41 @@ x86_sse_find_peaks(float *buf, nframes_t nframes, float *min, float *max)
 	current_max = _mm_set1_ps(*max);
 
 	// Work input until "buf" reaches 16 byte alignment
-	while ( ((unsigned long)buf) % 16 != 0 && nframes > 0) {
+	while ( ((intptr_t)buf) % 16 != 0 && nframes > 0) {
 
 		// Load the next float into the work buffer
 		work = _mm_set1_ps(*buf);
 
 		current_min = _mm_min_ps(current_min, work);
 		current_max = _mm_max_ps(current_max, work);
-		
+
 		buf++;
 		nframes--;
 	}
 
+        // use 64 byte prefetch for quadruple quads
+        while (nframes >= 16) {
+                __builtin_prefetch(buf+64,0,0);
+
+                work = _mm_load_ps(buf);
+                current_min = _mm_min_ps(current_min, work);
+                current_max = _mm_max_ps(current_max, work);
+                buf+=4;
+                work = _mm_load_ps(buf);
+                current_min = _mm_min_ps(current_min, work);
+                current_max = _mm_max_ps(current_max, work);
+                buf+=4;
+                work = _mm_load_ps(buf);
+                current_min = _mm_min_ps(current_min, work);
+                current_max = _mm_max_ps(current_max, work);
+                buf+=4;
+                work = _mm_load_ps(buf);
+                current_min = _mm_min_ps(current_min, work);
+                current_max = _mm_max_ps(current_max, work);
+                buf+=4;
+                nframes-=16;
+        }
+
 	// work through aligned buffers
 	while (nframes >= 4) {
 
@@ -54,7 +77,7 @@ x86_sse_find_peaks(float *buf, nframes_t nframes, float *min, float *max)
 		buf+=4;
 		nframes-=4;
 	}
-	
+
 	// work through the rest < 4 samples
 	while ( nframes > 0) {
 
@@ -63,7 +86,7 @@ x86_sse_find_peaks(float *buf, nframes_t nframes, float *min, float *max)
 
 		current_min = _mm_min_ps(current_min, work);
 		current_max = _mm_max_ps(current_max, work);
-		
+
 		buf++;
 		nframes--;
 	}