neoscrypt: reduce spill load and increase pascal def intensity

1 MH/s reached on the 1070 ...
raziel1000 · Jul 9, 2016 · 4ca7b5a · 4ca7b5a
1 parent 26a862c
commit 4ca7b5a
Show file tree

Hide file tree

Showing 5 changed files with 184 additions and 194 deletions.
diff --git a/ccminer.vcxproj b/ccminer.vcxproj
@@ -258,7 +258,9 @@
     <CudaCompile Include="Algo256\cuda_bmw.cu">
       <MaxRegCount>76</MaxRegCount>
     </CudaCompile>
-    <CudaCompile Include="neoscrypt\cuda_neoscrypt.cu" />
+    <CudaCompile Include="neoscrypt\cuda_neoscrypt.cu">
+      <MaxRegCount>160</MaxRegCount>
+    </CudaCompile>
     <ClCompile Include="scrypt-jane.cpp" />
     <ClCompile Include="scrypt.cpp" />
     <ClCompile Include="skein2.cpp" />
@@ -540,4 +542,4 @@
   <Target Name="AfterClean">
     <Delete Files="@(FilesToCopy->'$(OutDir)%(Filename)%(Extension)')" TreatErrorsAsWarnings="true" />
   </Target>
-</Project>
+</Project>
diff --git a/configure.sh b/configure.sh
@@ -3,5 +3,5 @@
 extracflags="-march=native -D_REENTRANT -falign-functions=16 -falign-jumps=16 -falign-labels=16"
 
 CUDA_CFLAGS="-O3 -lineno -Xcompiler -Wall  -D_FORCE_INLINES" \
-	./configure CXXFLAGS="-O3 $extracflags" --with-cuda=/usr/local/cuda --with-nvml=libnvidia-ml.so
+	./configure CXXFLAGS="-O3 $extracflags" --with-cuda=/usr/local/cuda-8.0 --with-nvml=libnvidia-ml.so