From 9f5744d4c053fd57719ff0954a4ead5591f5e158 Mon Sep 17 00:00:00 2001
From: Tanguy Pruvot <tanguy.pruvot@gmail.com>
Date: Wed, 17 Jun 2015 03:56:22 +0200
Subject: [PATCH] luffa/cube: fine tuning of maxregcount for the 750Ti

This allow to get 69 regs used (tested on linux) 69 or 72 make
the compiler to use 64 regs which is not enough on the 750Ti
for optimal performance...
---
 Makefile.am                       | 2 +-
 ccminer.vcxproj                   | 2 +-
 x11/cuda_x11_luffa512_Cubehash.cu | 2 --
 3 files changed, 2 insertions(+), 4 deletions(-)

diff --git a/Makefile.am b/Makefile.am
index 2cf49ba..ea8c9b7 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -105,7 +105,7 @@ x11/cuda_x11_luffa512.o: x11/cuda_x11_luffa512.cu
 	$(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
 
 x11/cuda_x11_luffa512_Cubehash.o: x11/cuda_x11_luffa512_Cubehash.cu
-	$(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
+	$(NVCC) $(nvcc_FLAGS) --maxrregcount=76 -o $@ -c $<
 
 x13/cuda_x13_hamsi512.o: x13/cuda_x13_hamsi512.cu
 	$(NVCC) $(nvcc_FLAGS) --maxrregcount=72 -o $@ -c $<
diff --git a/ccminer.vcxproj b/ccminer.vcxproj
index d42e263..235e36c 100644
--- a/ccminer.vcxproj
+++ b/ccminer.vcxproj
@@ -455,7 +455,7 @@
     </CudaCompile>
     <CudaCompile Include="x11\cuda_x11_luffa512.cu" />
     <CudaCompile Include="x11\cuda_x11_luffa512_Cubehash.cu">
-      <MaxRegCount>80</MaxRegCount>
+      <MaxRegCount>76</MaxRegCount>
     </CudaCompile>
     <CudaCompile Include="x11\cuda_x11_shavite512.cu">
       <MaxRegCount>128</MaxRegCount>
diff --git a/x11/cuda_x11_luffa512_Cubehash.cu b/x11/cuda_x11_luffa512_Cubehash.cu
index 1e7b914..cab0062 100644
--- a/x11/cuda_x11_luffa512_Cubehash.cu
+++ b/x11/cuda_x11_luffa512_Cubehash.cu
@@ -731,8 +731,6 @@ static void finalization512(uint32_t *statebuffer, uint32_t *statechainv, uint32
 __global__
 #if __CUDA_ARCH__ > 500
 __launch_bounds__(256, 4)
-#else
-__launch_bounds__(256, 3)
 #endif
 void x11_luffaCubehash512_gpu_hash_64(uint32_t threads, uint32_t *g_hash)
 {