1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
|
diff --unified --recursive --text cuda-p1-0.22/Makefile cuda-pm1/Makefile
--- cuda-p1-0.22/Makefile 2018-11-18 15:54:23.000000000 -0500
+++ cuda-pm1/Makefile 2020-12-12 12:45:54.812604717 -0500
@@ -18,14 +18,14 @@
# Configure this line to specify which cuda architectures (compute capability) to target
# or leave all supported architectures for a bulky binary that supports all
-CUDA_ARCHES = 30 32 35 50 52 53 60 61 62 70 75
+CUDA_ARCHES = 35 50 52 53 60 61 62 70 75 80 86
# Expand CUDA_ARCHES
NVCC_ARCHES += $(foreach CARCH, $(CUDA_ARCHES), -gencode arch=compute_$(CARCH),code=sm_$(CARCH))
# *Always* include PTX for the highest level supported by this version of NVCC, to
# future-proof the binary for new architectures
-NVCC_ARCHES += -gencode arch=compute_75,code=compute_75
+NVCC_ARCHES += -gencode arch=compute_86,code=compute_86
# Use --ptxas-options -v to see register usage
# Use --maxrregcount to specify register usage
@@ -40,7 +40,7 @@
# follow their defaults, the necessary files will be installed in your PATH and
# LDPATH. Otherwise, you'll need to manually insert their paths here.
-LIBS = -lcufft -lcudart -lm src/mpir/.libs/libmpir.a src/mpir/.libs/libgmp.a
+LIBS = -L/opt/cuda/lib64/ -lcufft -lcudart -lm src/mpir/.libs/libmpir.a src/mpir/.libs/libgmp.a
LDFLAGS = $(COMMON_LDFLAGS) -fPIC -Wl,-O1 -Wl,--as-needed -Wl,--sort-common -Wl,--relax
DEBUG_LDFLAGS = $(COMMON_LDFLAGS) -fPIC
|