Browse Source

scrypt: assign specific compute models to the kernels

2upstream
Tanguy Pruvot 10 years ago
parent
commit
9208888c57
  1. 29
      Makefile.am
  2. 28
      ccminer.vcxproj

29
Makefile.am

@ -114,20 +114,37 @@ x17/cuda_x17_sha512.o: x17/cuda_x17_sha512.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $< $(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
quark/cuda_quark_blake512.o: quark/cuda_quark_blake512.cu quark/cuda_quark_blake512.o: quark/cuda_quark_blake512.cu
$(NVCC) $(nvcc_FLAGS) -I cudpp-2.1/include --maxrregcount=80 -o $@ -c $< $(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
quark/cuda_quark_keccak512.o: quark/cuda_quark_keccak512.cu quark/cuda_quark_keccak512.o: quark/cuda_quark_keccak512.cu
$(NVCC) $(nvcc_FLAGS) -I cudpp-2.1/include --maxrregcount=88 -o $@ -c $< $(NVCC) $(nvcc_FLAGS) --maxrregcount=88 -o $@ -c $<
quark/cuda_quark_compactionTest.o: quark/cuda_quark_compactionTest.cu quark/cuda_quark_compactionTest.o: quark/cuda_quark_compactionTest.cu
$(NVCC) $(nvcc_FLAGS) -I cudpp-2.1/include --maxrregcount=80 -o $@ -c $< $(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
JHA/cuda_jha_compactionTest.o: JHA/cuda_jha_compactionTest.cu JHA/cuda_jha_compactionTest.o: JHA/cuda_jha_compactionTest.cu
$(NVCC) $(nvcc_FLAGS) -I cudpp-2.1/include --maxrregcount=80 -o $@ -c $< $(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
# This kernel need also an older SM to be able to autotune kernels # This object does not use cuda device code but call the different kernels (autotune)
scrypt/salsa_kernel.o: scrypt/salsa_kernel.cu scrypt/salsa_kernel.o: scrypt/salsa_kernel.cu
$(NVCC) -I . -gencode=arch=compute_20,code=\"sm_21,compute_20\" -o $@ -c $< $(NVCC) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_20,code=\"sm_21,compute_20\" -o $@ -c $<
# These kernels are for older devices (SM)
scrypt/test_kernel.o: scrypt/test_kernel.cu
$(NVCC) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_20,code=\"sm_20,compute_20\" -o $@ -c $<
scrypt/fermi_kernel.o: scrypt/fermi_kernel.cu
$(NVCC) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_20,code=\"sm_21,compute_20\" -o $@ -c $<
scrypt/kepler_kernel.o: scrypt/kepler_kernel.cu
$(NVCC) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_30,code=\"sm_30,compute_30\" -o $@ -c $<
scrypt/nv_kernel.o: scrypt/nv_kernel.cu
$(NVCC) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_30,code=\"sm_30,compute_30\" -o $@ -c $<
scrypt/titan_kernel.o: scrypt/titan_kernel.cu
$(NVCC) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_35,code=\"sm_35,compute_35\" -o $@ -c $<
skein.o: skein.cu skein.o: skein.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=64 -o $@ -c $< $(NVCC) $(nvcc_FLAGS) --maxrregcount=64 -o $@ -c $<

28
ccminer.vcxproj

@ -352,21 +352,29 @@
<CudaCompile Include="cuda_nist5.cu"> <CudaCompile Include="cuda_nist5.cu">
</CudaCompile> </CudaCompile>
<CudaCompile Include="scrypt\blake.cu" /> <CudaCompile Include="scrypt\blake.cu" />
<CudaCompile Include="scrypt\keccak.cu" />
<CudaCompile Include="scrypt\sha256.cu" />
<CudaCompile Include="scrypt\salsa_kernel.cu">
<CodeGeneration>compute_20,sm_21</CodeGeneration>
</CudaCompile>
<CudaCompile Include="scrypt\fermi_kernel.cu"> <CudaCompile Include="scrypt\fermi_kernel.cu">
<CodeGeneration Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">compute_20,sm_21;compute_30,sm_30;compute_35,sm_35;compute_50,sm_50;compute_52,sm_52</CodeGeneration> <CodeGeneration>compute_20,sm_21</CodeGeneration>
</CudaCompile>
<CudaCompile Include="scrypt\kepler_kernel.cu">
<CodeGeneration>compute_30,sm_30</CodeGeneration>
</CudaCompile>
<CudaCompile Include="scrypt\nv_kernel.cu">
<CodeGeneration>compute_30,sm_30</CodeGeneration>
</CudaCompile> </CudaCompile>
<CudaCompile Include="scrypt\keccak.cu" />
<CudaCompile Include="scrypt\kepler_kernel.cu" />
<CudaCompile Include="scrypt\nv_kernel.cu" />
<CudaCompile Include="scrypt\nv_kernel2.cu"> <CudaCompile Include="scrypt\nv_kernel2.cu">
<CodeGeneration Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">compute_35,sm_35;compute_50,sm_50;compute_52,sm_52</CodeGeneration> <CodeGeneration>compute_35,sm_35;compute_50,sm_50;compute_52,sm_52</CodeGeneration>
</CudaCompile> </CudaCompile>
<CudaCompile Include="scrypt\salsa_kernel.cu"> <CudaCompile Include="scrypt\test_kernel.cu">
<CodeGeneration Condition="'$(Configuration)|$(Platform)'=='Release|Win32'">compute_20,sm_21</CodeGeneration> <CodeGeneration>compute_20,sm_21</CodeGeneration>
</CudaCompile>
<CudaCompile Include="scrypt\titan_kernel.cu">
<CodeGeneration>compute_35,sm_35</CodeGeneration>
</CudaCompile> </CudaCompile>
<CudaCompile Include="scrypt\sha256.cu" />
<CudaCompile Include="scrypt\test_kernel.cu" />
<CudaCompile Include="scrypt\titan_kernel.cu" />
<CudaCompile Include="zr5.cu" /> <CudaCompile Include="zr5.cu" />
<CudaCompile Include="heavy\cuda_blake512.cu"> <CudaCompile Include="heavy\cuda_blake512.cu">
</CudaCompile> </CudaCompile>

Loading…
Cancel
Save