From 009c3ee3f74d6babfaae56f80bbac3f95a2d4cb1 Mon Sep 17 00:00:00 2001
From: "Unknown W. Brackets" <checkins@unknownbrackets.org>
Date: Fri, 21 Mar 2014 02:23:09 -0700
Subject: [PATCH] vertexjit: Save S16-S31 registers in ARM.

---
 Common/ArmEmitter.h           | 6 ++++++
 GPU/GLES/VertexDecoderArm.cpp | 6 ++++++
 2 files changed, 12 insertions(+)

diff --git a/Common/ArmEmitter.h b/Common/ArmEmitter.h
index 4b8f73169e..c2910aee73 100644
--- a/Common/ArmEmitter.h
+++ b/Common/ArmEmitter.h
@@ -591,6 +591,12 @@ public:
 	void VSTMIA(ARMReg dest, bool WriteBack, ARMReg firstreg, int numregs);
 	void VLDMDB(ARMReg dest, bool WriteBack, ARMReg firstreg, int numregs);
 	void VSTMDB(ARMReg dest, bool WriteBack, ARMReg firstreg, int numregs);
+	void VPUSH(ARMReg firstvreg, int numvregs) {
+		VSTMDB(R_SP, true, firstvreg, numvregs);
+	}
+	void VPOP(ARMReg firstvreg, int numvregs) {
+		VLDMIA(R_SP, true, firstvreg, numvregs);
+	}
 	void VLDR(ARMReg Dest, ARMReg Base, s16 offset);
 	void VSTR(ARMReg Src,  ARMReg Base, s16 offset);
 	void VCMP(ARMReg Vd, ARMReg Vm);
diff --git a/GPU/GLES/VertexDecoderArm.cpp b/GPU/GLES/VertexDecoderArm.cpp
index e0c4b79f17..633cdfd436 100644
--- a/GPU/GLES/VertexDecoderArm.cpp
+++ b/GPU/GLES/VertexDecoderArm.cpp
@@ -157,6 +157,9 @@ JittedVertexDecoder VertexDecoderJitCache::Compile(const VertexDecoder &dec) {
 	SetCC(CC_AL);
 
 	PUSH(6, R4, R5, R6, R7, R8, R_LR);
+	if (cpu_info.bNEON) {
+		VPUSH(D8, 8);
+	}
 
 	// Keep the scale/offset in a few fp registers if we need it.
 	// This step can be NEON-ized but the savings would be miniscule.
@@ -244,6 +247,9 @@ JittedVertexDecoder VertexDecoderJitCache::Compile(const VertexDecoder &dec) {
 	SUBS(counterReg, counterReg, 1);
 	B_CC(CC_NEQ, loopStart);
 
+	if (cpu_info.bNEON) {
+		VPOP(D8, 8);
+	}
 	POP(6, R4, R5, R6, R7, R8, R_PC);
 
 	FlushLitPool();