Skip to content

Deepwalk fails when scale=30 #17

@kurapika0

Description

@kurapika0

I run into this problem when I try to deepwalk a scale=30 graph

[masdar:163668] *** Process received signal ***
[masdar:163668] Signal: Segmentation fault (11)
[masdar:163668] Signal code: Address not mapped (1)
[masdar:163668] Failing at address: 0x763685000040
[masdar:163668] [ 0] /lib/x86_64-linux-gnu/libc.so.6(+0x45330)[0x7653ab245330]
[masdar:163668] [ 1] /lib/x86_64-linux-gnu/libc.so.6(+0x1a17e1)[0x7653ab3a17e1]
[masdar:163668] [ 2] /data/home/sike/miniconda3/envs/g500/lib/libopen-pal.so.80(+0xdee9d)[0x7653ab9aae9d]
[masdar:163668] [ 3] /data/home/sike/miniconda3/envs/g500/lib/libmpi.so.40(mca_pml_ob1_recv_request_get_frag+0x6a)[0x7653abc5e58a]
[masdar:163668] [ 4] /data/home/sike/miniconda3/envs/g500/lib/libmpi.so.40(mca_pml_ob1_recv_request_progress_rget+0x322)[0x7653abc5ebe2]
[masdar:163668] [ 5] /data/home/sike/miniconda3/envs/g500/lib/libmpi.so.40(mca_pml_ob1_recv_req_start+0x4ee)[0x7653abc612ae]
[masdar:163668] [ 6] /data/home/sike/miniconda3/envs/g500/lib/libmpi.so.40(mca_pml_ob1_recv+0x12d)[0x7653abc5389d]
[masdar:163668] [ 7] /data/home/sike/miniconda3/envs/g500/lib/libmpi.so.40(MPI_Recv+0x167)[0x7653abac9897]
[masdar:163668] [ 8] /data/home/sike/graph/KnightKing/build/bin/deepwalk(+0x991e)[0x60da2145d91e]
[masdar:163668] [ 9] /data/home/sike/miniconda3/envs/g500/lib/libstdc++.so.6(+0xef5e4)[0x7653ab6ef5e4]
[masdar:163668] [10] /lib/x86_64-linux-gnu/libc.so.6(+0x9caa4)[0x7653ab29caa4]
[masdar:163668] [11] /lib/x86_64-linux-gnu/libc.so.6(+0x129c6c)[0x7653ab329c6c]
[masdar:163668] *** End of error message ***

The script is like

KK_BIN="/path/to/KnightKing/build/bin"

SCALE=30
VERTEX_COUNT=$((2 ** SCALE))
WALKERS=$((V / 1000000)); [[ "$W" -lt 1 ]] && W=1

# graph500.git
mpiexec -n "$NP" ./graph500_reference_bfs "$SCALE"
# ---------- bin → txt ----------
python3 "$PY_CONVERTER" "$BIN_FILE" "$TXT_FILE"
# ---------- txt → KnightKing .data ----------
"$KK_BIN/gconverter" -i "$TXT_FILE" -o "$DATA_FILE" -s unweighted
# ---------- DeepWalk ----------
"$KK_BIN/deepwalk" -g "$DATA_FILE" -v "$VERTEX_COUNT" -w "$WALKERS" -s unweighted -r 0.1 -l 10

The script run successfully for SCALE=16 but fails for SCALE=30.
Is this because KnightKing does not support such a large scale or there are some configuration problems on my side?

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