Move leftover Makefiles to new SM scheme

FZJ-JSC · Nov 14, 2024 · 58d7ddd · 58d7ddd
1 parent 043528f
commit 58d7ddd
Show file tree

Hide file tree

Showing 16 changed files with 29 additions and 13 deletions.
diff --git a/03-H_Multi_GPU_Parallelization/.master/Makefile b/03-H_Multi_GPU_Parallelization/.master/Makefile
@@ -6,6 +6,7 @@ CUDA_HOME ?= /usr/local/cuda
 ifndef MPI_HOME
 $(error MPI_HOME is not set)
 endif
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -15,7 +16,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/03-H_Multi_GPU_Parallelization/solutions/Makefile b/03-H_Multi_GPU_Parallelization/solutions/Makefile
@@ -6,6 +6,7 @@ CUDA_HOME ?= /usr/local/cuda
 ifndef MPI_HOME
 $(error MPI_HOME is not set)
 endif
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -15,7 +16,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/03-H_Multi_GPU_Parallelization/solutions/advanced/Makefile b/03-H_Multi_GPU_Parallelization/solutions/advanced/Makefile
@@ -6,6 +6,7 @@ CUDA_HOME ?= /usr/local/cuda
 ifndef MPI_HOME
 $(error MPI_HOME is not set)
 endif
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -15,7 +16,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/03-H_Multi_GPU_Parallelization/tasks/Makefile b/03-H_Multi_GPU_Parallelization/tasks/Makefile
@@ -6,6 +6,7 @@ CUDA_HOME ?= /usr/local/cuda
 ifndef MPI_HOME
 $(error MPI_HOME is not set)
 endif
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -15,7 +16,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/06-H_Overlap_Communication_and_Computation_MPI/.master/Makefile b/06-H_Overlap_Communication_and_Computation_MPI/.master/Makefile
@@ -4,6 +4,7 @@ NVCC=nvcc
 MPICXX=mpicxx
 JSC_SUBMIT_CMD ?= srun --gres=gpu:4 --ntasks-per-node 4
 CUDA_HOME ?= /usr/local/cuda
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -13,7 +14,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/06-H_Overlap_Communication_and_Computation_MPI/solutions/Makefile b/06-H_Overlap_Communication_and_Computation_MPI/solutions/Makefile
@@ -4,6 +4,7 @@ NVCC=nvcc
 MPICXX=mpicxx
 JSC_SUBMIT_CMD ?= srun --gres=gpu:4 --ntasks-per-node 4
 CUDA_HOME ?= /usr/local/cuda
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -13,7 +14,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/06-H_Overlap_Communication_and_Computation_MPI/tasks/Makefile b/06-H_Overlap_Communication_and_Computation_MPI/tasks/Makefile
@@ -4,6 +4,7 @@ NVCC=nvcc
 MPICXX=mpicxx
 JSC_SUBMIT_CMD ?= srun --gres=gpu:4 --ntasks-per-node 4
 CUDA_HOME ?= /usr/local/cuda
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -13,7 +14,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/08-H_NCCL_NVSHMEM/.master/NCCL/Makefile b/08-H_NCCL_NVSHMEM/.master/NCCL/Makefile
@@ -5,6 +5,7 @@ JSC_SUBMIT_CMD ?= srun --cpu-bind=socket --gres=gpu:4 --ntasks-per-node 4
 MPICXX=mpicxx
 CUDA_HOME ?= /usr/local/cuda
 NCCL_HOME ?= /usr
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -14,7 +15,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/08-H_NCCL_NVSHMEM/.master/NVSHMEM/Makefile b/08-H_NCCL_NVSHMEM/.master/NVSHMEM/Makefile
@@ -11,6 +11,7 @@ endif
 ifndef MPI_HOME
 $(error MPI_HOME is not set)
 endif
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -20,7 +21,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/08-H_NCCL_NVSHMEM/solutions/NCCL/Makefile b/08-H_NCCL_NVSHMEM/solutions/NCCL/Makefile
@@ -5,6 +5,7 @@ JSC_SUBMIT_CMD ?= srun --cpu-bind=socket --gres=gpu:4 --ntasks-per-node 4
 MPICXX=mpicxx
 CUDA_HOME ?= /usr/local/cuda
 NCCL_HOME ?= /usr
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -14,7 +15,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/08-H_NCCL_NVSHMEM/solutions/NVSHMEM/Makefile b/08-H_NCCL_NVSHMEM/solutions/NVSHMEM/Makefile
@@ -11,6 +11,7 @@ endif
 ifndef MPI_HOME
 $(error MPI_HOME is not set)
 endif
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -20,7 +21,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/08-H_NCCL_NVSHMEM/tasks/NCCL/Makefile b/08-H_NCCL_NVSHMEM/tasks/NCCL/Makefile
@@ -5,6 +5,7 @@ JSC_SUBMIT_CMD ?= srun --cpu-bind=socket --gres=gpu:4 --ntasks-per-node 4
 MPICXX=mpicxx
 CUDA_HOME ?= /usr/local/cuda
 NCCL_HOME ?= /usr
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -14,7 +15,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/08-H_NCCL_NVSHMEM/tasks/NVSHMEM/Makefile b/08-H_NCCL_NVSHMEM/tasks/NVSHMEM/Makefile
@@ -11,6 +11,7 @@ endif
 ifndef MPI_HOME
 $(error MPI_HOME is not set)
 endif
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37
@@ -20,7 +21,7 @@ GENCODE_SM60    := -gencode arch=compute_60,code=sm_60
 GENCODE_SM70    := -gencode arch=compute_70,code=sm_70
 GENCODE_SM80    := -gencode arch=compute_80,code=sm_80 -gencode arch=compute_80,code=compute_80
 GENCODE_SM90    := -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90,code=compute_90
-GENCODE_FLAGS	:= $(GENCODE_SM80) $(GENCODE_SM90)
+GENCODE_FLAGS	:= $(GENCODE_SM$(_JSCCOURSE_GPU_ARCH))
 ifdef DISABLE_CUB
         NVCC_FLAGS = -Xptxas --optimize-float-atomics
 else

diff --git a/...Graphs_and_Device-initiated_Communication_with_NVSHMEM/.master/Using_CUDA_Graphs/Makefile b/...Graphs_and_Device-initiated_Communication_with_NVSHMEM/.master/Using_CUDA_Graphs/Makefile
@@ -5,6 +5,7 @@ JSC_SUBMIT_CMD ?= srun --cpu-bind=socket --gres=gpu:4 --ntasks-per-node 4
 MPICXX=mpicxx
 CUDA_HOME ?= /usr/local/cuda
 NCCL_HOME ?= /usr
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37

diff --git a/...aphs_and_Device-initiated_Communication_with_NVSHMEM/solutions/Using_CUDA_Graphs/Makefile b/...aphs_and_Device-initiated_Communication_with_NVSHMEM/solutions/Using_CUDA_Graphs/Makefile
@@ -5,6 +5,7 @@ JSC_SUBMIT_CMD ?= srun --cpu-bind=socket --gres=gpu:4 --ntasks-per-node 4
 MPICXX=mpicxx
 CUDA_HOME ?= /usr/local/cuda
 NCCL_HOME ?= /usr
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37

diff --git a/...A_Graphs_and_Device-initiated_Communication_with_NVSHMEM/tasks/Using_CUDA_Graphs/Makefile b/...A_Graphs_and_Device-initiated_Communication_with_NVSHMEM/tasks/Using_CUDA_Graphs/Makefile
@@ -5,6 +5,7 @@ JSC_SUBMIT_CMD ?= srun --cpu-bind=socket --gres=gpu:4 --ntasks-per-node 4
 MPICXX=mpicxx
 CUDA_HOME ?= /usr/local/cuda
 NCCL_HOME ?= /usr
+_JSCCOURSE_GPU_ARCH?=80
 GENCODE_SM30	:= -gencode arch=compute_30,code=sm_30
 GENCODE_SM35	:= -gencode arch=compute_35,code=sm_35
 GENCODE_SM37	:= -gencode arch=compute_37,code=sm_37