On the NVIDIA 3080 GPU, running WindowPost multiple times causes the program to crash due to VRAM not being released.
2026-05-20T11:37:24.428 DEBUG filcrypto::util::types > get_gpu_devices: start
2026-05-20T11:37:24.778 DEBUG rust_gpu_tools::cuda::utils > Loaded CUDA devices: [Device { vendor: Nvidia, name: "NVIDIA GeForce RTX 3080", memory: 10394533888, compute_units: 68, compute_capability: (8, 6), pci_id: PciId(16640), uuid: Some(92e81032-b40b-2091-5b6f-fd9ff7a15e74), context: UnownedContext { inner: 0x7f8968001840 } }]
2026-05-20T11:37:24.779 DEBUG rust_gpu_tools::device > loaded devices: [Device { vendor: Nvidia, name: "NVIDIA GeForce RTX 3080", memory: 10394533888, compute_units: 68, compute_capability: Some((8, 6)), pci_id: PciId(16640), uuid: Some(92e81032-b40b-2091-5b6f-fd9ff7a15e74), cuda: Some(Device { vendor: Nvidia, name: "NVIDIA GeForce RTX 3080", memory: 10394533888, compute_units: 68, compute_capability: (8, 6), pci_id: PciId(16640), uuid: Some(92e81032-b40b-2091-5b6f-fd9ff7a15e74), context: UnownedContext { inner: 0x7f8968001840 } }) }]
2026-05-20T11:37:24.779 DEBUG filcrypto::util::types > get_gpu_devices: end
2026-05-20T11:37:24.779Z INFO advmgr sealer/worker_local.go:862 Detected GPU devices. {"count": 1}
2026-05-20T11:37:24.783Z INFO worker worker/worker.go:790 Worker registered successfully, waiting for tasks
2026-05-20T11:42:14.414 DEBUG filcrypto::util::types > generate_single_window_post_with_vanilla: start
2026-05-20T11:42:14.593 INFO filecoin_proofs::api::window_post > generate_single_window_post_with_vanilla:start
2026-05-20T11:42:14.593 INFO filecoin_proofs::caches > trying parameters memory cache for: Window_POST[34359738368]
2026-05-20T11:42:14.593 INFO filecoin_proofs::caches > no params in memory cache for Window_POST[34359738368]
2026-05-20T11:42:14.658 INFO storage_proofs_core::parameter_cache > parameter set identifier for cache: FallbackPoSt::PublicParams{sector_size: 34359738368, challenge_count: 10, sector_count: 2349}
2026-05-20T11:42:14.659 INFO storage_proofs_core::parameter_cache > ensuring that all ancestor directories for: "/mnt/data/proofs/v28-proof-of-spacetime-fallback-merkletree-poseidon_hasher-8-8-0-0377ded656c6f524f1618760bffe4e0a1c51d5a70c4509eedae8a27555733edc.params" exist
2026-05-20T11:42:14.659 INFO storage_proofs_core::parameter_cache > checking cache_path: "/mnt/data/proofs/v28-proof-of-spacetime-fallback-merkletree-poseidon_hasher-8-8-0-0377ded656c6f524f1618760bffe4e0a1c51d5a70c4509eedae8a27555733edc.params" for parameters
2026-05-20T11:42:14.659 INFO storage_proofs_core::parameter_cache > Verify production parameters is false
2026-05-20T11:42:57.126 INFO storage_proofs_core::parameter_cache > read parameters into SuprasSeal from cache "/mnt/data/proofs/v28-proof-of-spacetime-fallback-merkletree-poseidon_hasher-8-8-0-0377ded656c6f524f1618760bffe4e0a1c51d5a70c4509eedae8a27555733edc.params"
2026-05-20T11:42:57.161 INFO filecoin_proofs::api::post_util > single_partition_vanilla_proofs:start
2026-05-20T11:42:57.161 DEBUG filecoin_proofs::api::post_util > processing partition: 0
2026-05-20T11:42:57.185 INFO filecoin_proofs::api::post_util > single_partition_vanilla_proofs:finish
2026-05-20T11:42:58.080 INFO storage_proofs_core::compound_proof > snark_proof:start
2026-05-20T11:42:58.236 INFO bellperson::groth16::prover::supraseal > Bellperson 0.27.0 with SupraSeal is being used!
2026-05-20T11:44:04.904 INFO bellperson::groth16::prover::supraseal > synthesis time: 66.668150612s
2026-05-20T11:44:04.905 INFO bellperson::groth16::prover::supraseal > starting proof timer
2026-05-20T11:47:28.542 INFO bellperson::groth16::prover::supraseal > prover time: 203.636956421s
2026-05-20T11:47:31.429 INFO storage_proofs_core::compound_proof > snark_proof:finish
2026-05-20T11:47:31.430 INFO filecoin_proofs::api::window_post > generate_single_window_post_with_vanilla:finish
2026-05-20T11:47:31.473 DEBUG filcrypto::util::types > generate_single_window_post_with_vanilla: end
2026-05-20T11:50:11.754 DEBUG filcrypto::util::types > generate_single_window_post_with_vanilla: start
2026-05-20T11:50:12.181 INFO filecoin_proofs::api::window_post > generate_single_window_post_with_vanilla:start
2026-05-20T11:50:12.182 INFO filecoin_proofs::caches > trying parameters memory cache for: Window_POST[34359738368]
2026-05-20T11:50:12.182 INFO filecoin_proofs::caches > found params in memory cache for Window_POST[34359738368]
2026-05-20T11:50:12.182 INFO filecoin_proofs::api::post_util > single_partition_vanilla_proofs:start
2026-05-20T11:50:12.182 DEBUG filecoin_proofs::api::post_util > processing partition: 0
2026-05-20T11:50:12.295 INFO filecoin_proofs::api::post_util > single_partition_vanilla_proofs:finish
2026-05-20T11:50:13.208 INFO storage_proofs_core::compound_proof > snark_proof:start
2026-05-20T11:50:13.463 INFO bellperson::groth16::prover::supraseal > Bellperson 0.27.0 with SupraSeal is being used!
2026-05-20T11:50:55.904 INFO bellperson::groth16::prover::supraseal > synthesis time: 42.440174104s
2026-05-20T11:50:55.904 INFO bellperson::groth16::prover::supraseal > starting proof timer
thread '<unnamed>' (3001907) panicked at /root/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/supraseal-c2-0.1.2/src/lib.rs:220:9:
cudaMalloc(&d_ptr, n * sizeof(T))@sppark-0.1.14/sppark/util/gpu_t.cuh:331 failed: "out of memory"
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
2026-05-20T11:50:59.530Z ERROR advmgr sealer/worker_local.go:778 generating window PoSt failed {"error": "Rust panic: no unwind information"}
2026-05-20T11:50:59.535Z WARN rpc go-jsonrpc@v0.10.1/handler.go:425 error in RPC call to 'Filecoin.GenerateWindowPoSt': generate window PoSt with vanilla proofs:
github.com/filecoin-project/lotus/storage/sealer.(*LocalWorker).GenerateWindowPoStAdv
/root/go/src/github.com/filecoin-project/lotus/storage/sealer/worker_local.go:779
- Rust panic: no unwind information
On the NVIDIA 3080 GPU, running WindowPost multiple times causes the program to crash due to VRAM not being released.