From f1a8bcd876a6e5e714e512340e69916f6173610a Mon Sep 17 00:00:00 2001
From: Connor Olding <cloningdonor@gmail.com>
Date: Tue, 7 Apr 2015 13:55:12 -0700
Subject: [PATCH] use single-precision floats for ARM

NEON doesn't support doubles ;_;
anyway, this turns the two-times slowdown
caused by previous commits into a two-times speedup.
---
 Makefile       | 4 ++--
 include/util.h | 6 ++++++
 2 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/Makefile b/Makefile
index adfdb79..393afe1 100644
--- a/Makefile
+++ b/Makefile
@@ -70,8 +70,8 @@ linux: ALL_CXXFLAGS += $(OPT_FLAGS) -fpic
 linux: VST_FLAGS += -D__cdecl=
 linux: all
 
-linux-arm: ALL_CFLAGS += -Ofast -mfpu=neon -fpic
-linux-arm: ALL_CXXFLAGS += -Ofast -mfpu=neon -fpic
+linux-arm: ALL_CFLAGS += -Ofast -march=native -DFORCE_SINGLE -fpic
+linux-arm: ALL_CXXFLAGS += -Ofast -march=native -DFORCE_SINGLE -fpic
 linux-arm: VST_FLAGS += -D__cdecl=
 linux-arm: all
 
diff --git a/include/util.h b/include/util.h
index 665ff8e..3422c33 100644
--- a/include/util.h
+++ b/include/util.h
@@ -7,7 +7,13 @@
 #define INNER static inline
 #define PURE __attribute__((pure))
 #define CONST __attribute__((const))
+
+#ifndef FORCE_SINGLE
 typedef double v2df __attribute__((vector_size(16), aligned(16)));
+#else
+typedef float v2df __attribute__((vector_size(8), aligned(8)));
+#endif
+
 typedef float v4sf __attribute__((vector_size(16), aligned(16)));
 typedef unsigned long ulong; // __attribute((aligned(16)));