Skip to content

Makefile options for NVIDIA Jetson Nano #350

Description

@sigkill
[2022-02-02.22:15:38] [Info] Initializing NVIDIA Tegra X1
[2022-02-02.22:15:38] [Info] Generating 327,680 starting points (12.5MB)
[2022-02-02.22:15:39] [Info] 10.0%
[2022-02-02.22:15:40] [Info] 20.0%
[2022-02-02.22:15:40] [Info] 30.0%
[2022-02-02.22:15:40] [Info] 40.0%
[2022-02-02.22:15:40] [Info] 50.0%
[2022-02-02.22:15:41] [Info] 60.0%
[2022-02-02.22:15:41] [Info] 70.0%
[2022-02-02.22:15:41] [Info] 80.0%
[2022-02-02.22:15:41] [Info] 90.0%
[2022-02-02.22:15:41] [Info] 100.0%
[2022-02-02.22:15:41] [Info] Done
[2022-02-02.22:15:41] [Info] Loading addresses from '/root/addresses'
[2022-02-02.22:15:41] [Info] 88 addresses loaded (0.0MB)
[2022-02-02.22:15:41] [Info] Allocating bloom filter (0.0MB)
NVIDIA Tegra X1  1318 / 3964MB | 88 targets 5.82 MKey/s (324,403,200 total) [00:00:53][2022-02-02.22:16:38] [Info] Checkpoint
NVIDIA Tegra X1  1319 / 3964MB | 88 targets 5.86 MKey/s (680,263,680 total) [00:01:53][2022-02-02.22:17:39] [Info] Checkpoint
NVIDIA Tegra X1  1319 / 3964MB | 88 targets 5.84 MKey/s (734,003,200 total) [00:02:02]

Makefile contents:

diff --git a/Makefile b/Makefile
index 6375830..e385f51 100644
--- a/Makefile
+++ b/Makefile
@@ -10,12 +10,12 @@ LIBS+=-L$(LIBDIR)

 # C++ options
 CXX=g++
-CXXFLAGS=-O2 -std=c++11
+CXXFLAGS=-O3 -std=c++11 -march=native

 # CUDA variables
-COMPUTE_CAP=30
+COMPUTE_CAP=53
 NVCC=nvcc
-NVCCFLAGS=-std=c++11 -gencode=arch=compute_${COMPUTE_CAP},code=\"sm_${COMPUTE_CAP}\" -Xptxas="-v" -Xcompiler "${CXXFLAGS}"
+NVCCFLAGS=-std=c++11 -gencode arch=compute_53,code=sm_53 -Xptxas="-v" -Xcompiler "${CXXFLAGS}"
 CUDA_HOME=/usr/local/cuda
 CUDA_LIB=${CUDA_HOME}/lib64
 CUDA_INCLUDE=${CUDA_HOME}/include

Activity

  1. sigkill commented on Feb 3, 2022

    @sigkill
    Author

    Works with cuda 10.2 ... haven't tested 11.4 yet.

  2. sigkill commented on Feb 3, 2022

    @sigkill
    Author

    When using the PR from #226
    I get about 2.0MKey/s more.

    NVIDIA Tegra X1  1398 / 3964MB | 88 targets 7.11 MKey/s (168,157,184,000 total) [06:33:07][2022-02-03.05:23:52] [Info] Checkpoint
    NVIDIA Tegra X1  1398 / 3964MB | 88 targets 7.08 MKey/s (168,583,168,000 total) [06:34:07][2022-02-03.05:24:52] [Info] Checkpoint
    NVIDIA Tegra X1  1398 / 3964MB | 88 targets 7.09 MKey/s (169,017,671,680 total) [06:35:09][2022-02-03.05:25:54] [Info] Checkpoint
    NVIDIA Tegra X1  1398 / 3964MB | 88 targets 7.04 MKey/s (169,443,655,680 total) [06:36:09][2022-02-03.05:26:54] [Info] Checkpoint
    NVIDIA Tegra X1  1398 / 3964MB | 88 targets 7.14 MKey/s (169,873,899,520 total) [06:37:09][2022-02-03.05:27:55] [Info] Checkpoint
    NVIDIA Tegra X1  1398 / 3964MB | 88 targets 7.16 MKey/s (170,316,922,880 total) [06:38:11][2022-02-03.05:28:57] [Info] Checkpoint
    NVIDIA Tegra X1  1398 / 3964MB | 88 targets 7.16 MKey/s (170,436,198,400 total) [06:38:27]```
    
  3. ALexus003 commented on Jun 2, 2022

    @ALexus003

    с CUDA 10.1 не работает ?

  4. sigkill commented on Jun 5, 2022

    @sigkill
    Author

    С 10.1 не пробовал, но подозреваю, что должно работать. I was able to get 7.45 MKey/s.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions