From 2b6ed975e28de79d7dc34f225238eae631199afc Mon Sep 17 00:00:00 2001
From: Henrik Rydgard <hrydgard@gmail.com>
Date: Sat, 6 Dec 2014 00:28:11 +0100
Subject: [PATCH] Improve fast sse matrix multiply

---
 math/fast/fast_matrix_sse.c | 10 +++++-----
 native.vcxproj              |  4 +++-
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/math/fast/fast_matrix_sse.c b/math/fast/fast_matrix_sse.c
index 166f71abce..9de79ce760 100644
--- a/math/fast/fast_matrix_sse.c
+++ b/math/fast/fast_matrix_sse.c
@@ -8,12 +8,12 @@
 
 void fast_matrix_mul_4x4_sse(float *dest, const float *a, const float *b) {
 	int i;
-	for (i = 0; i < 16; i += 4) {
-		__m128 a_col_1 = _mm_loadu_ps(a);
-		__m128 a_col_2 = _mm_loadu_ps(&a[4]);
-		__m128 a_col_3 = _mm_loadu_ps(&a[8]);
-		__m128 a_col_4 = _mm_loadu_ps(&a[12]);
+	__m128 a_col_1 = _mm_loadu_ps(a);
+	__m128 a_col_2 = _mm_loadu_ps(&a[4]);
+	__m128 a_col_3 = _mm_loadu_ps(&a[8]);
+	__m128 a_col_4 = _mm_loadu_ps(&a[12]);
 
+	for (i = 0; i < 16; i += 4) {
 		__m128 r_col = _mm_mul_ps(a_col_1, _mm_set1_ps(b[i]));
 		r_col = _mm_add_ps(r_col, _mm_mul_ps(a_col_2, _mm_set1_ps(b[i + 1])));
 		r_col = _mm_add_ps(r_col, _mm_mul_ps(a_col_3, _mm_set1_ps(b[i + 2])));
diff --git a/native.vcxproj b/native.vcxproj
index 59eaa83023..da563a52cd 100644
--- a/native.vcxproj
+++ b/native.vcxproj
@@ -754,7 +754,9 @@
     <ClCompile Include="math\math_util.cpp" />
     <ClCompile Include="math\fast\fast_math.c" />
     <ClCompile Include="math\fast\fast_matrix.c" />
-    <ClCompile Include="math\fast\fast_matrix_sse.c" />
+    <ClCompile Include="math\fast\fast_matrix_sse.c">
+      <AssemblerOutput Condition="'$(Configuration)|$(Platform)'=='Release|x64'">AssemblyAndSourceCode</AssemblerOutput>
+    </ClCompile>
     <ClCompile Include="midi\midi_input.cpp" />
     <ClCompile Include="net\http_client.cpp" />
     <ClCompile Include="net\resolve.cpp" />