From 7634ad2e2c1725c4c9571fa385b4ab822d8cd117 Mon Sep 17 00:00:00 2001
From: Holger Wirtz <dcoredump@googlemail.com>
Date: Tue, 29 Nov 2016 01:59:50 +0000
Subject: [PATCH] Changed some optimizations for Raspi.

---
 src/Makefile  | 33 ++++++++++++++++++++++++++++++---
 src/dexed.cpp |  6 +++++-
 2 files changed, 35 insertions(+), 4 deletions(-)

diff --git a/src/Makefile b/src/Makefile
index 32a8acb..e0d0f0f 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -8,15 +8,42 @@ LDFLAGS=-L/usr/local/lib -llvtk_plugin2
 ifeq ($(ARCH),)
    ARCH := $(shell uname -m)
 endif
+machine = $(shell sh -c 'uname -m 2>/dev/null || echo unknown')
 
-ifneq ($(findstring armv7,$(ARCH)),)
-   CXXFLAGS += -mfpu=neon -march=armv7-a -marm
+# Raspberry Pi B+, Zero, etc
+ifneq (,$(findstring armv6l,$(machine)))
+  CPU = -mcpu=arm1176jzf-s
+  FPU = -mfpu=vfp
+endif
+
+# Raspberry Pi 2 and 3
+ifneq (,$(findstring armv7l,$(machine)))
+  model = $(shell sh -c 'cat /sys/firmware/devicetree/base/model 2>/dev/null || echo unknown')
+ifneq (,$(findstring 3,$(model)))
+  CPU = -mcpu=cortex-a53
+  FPU = -mfpu=neon-fp-armv8
+else
+  CPU = -mcpu=cortex-a7 -mthumb
+  FPU = -mfpu=neon-vfpv4
+endif
+  FPU += -mneon-for-64bits
+endif
+
+# ARM A64
+ifneq (,$(findstring aarch64,$(machine)))
+  CPU = -mcpu=cortex-a53
+  PLATFORM += -mabi=lp64 -mcmodel=tiny
+endif
+
+ifneq ($(findstring arm,$(ARCH)),)
+   CXXFLAGS += -DRASPI
 endif
 
 ifeq ($(DEBUG), 1)
    CXXFLAGS += -DDEBUG $(CFLAGS)
 else
-   CXXFLAGS += -Ofast $(CFLAGS)
+   #CXXFLAGS += -Ofast $(CFLAGS) -DRASPI
+   CXXFLAGS += -Ofast $(CFLAGS) $(CPU) $(FPU)
 endif
 
 all: $(BUNDLE) Makefile
diff --git a/src/dexed.cpp b/src/dexed.cpp
index 19bac00..0cc74e9 100644
--- a/src/dexed.cpp
+++ b/src/dexed.cpp
@@ -372,8 +372,12 @@ void Dexed::GetSamples(uint32_t n_samples, float* buffer)
           voices[note].dx7_note->compute(audiobuf.get(), lfovalue, lfodelay, &controllers);
           for (uint32_t j=0; j < N; ++j) {
             int32_t val = audiobuf.get()[j];
-                        
+#ifndef RASPI
             val = val >> 4;
+#else
+            //val*=2;
+            val=val<<2; // hope this is an arithmetic shift!
+#endif
             int32_t clip_val = val < -(1 << 24) ? 0x8000 : val >= (1 << 24) ? 0x7fff : val >> 9;
             float f = float(clip_val) / float(0x8000);
             if(f>1.0)