sleuth_prep hangs on "summarizing bootstraps" when num_cores > 1 and gene_mode = TRUE

[ajw2329]

Hello,

First of all, thanks for building/maintaining this great tool!

I have recently encountered an issue with sleuth_prep when attempting to using gene_mode = TRUE with multiple cores. The function prints the "summarizing bootstraps" message and as many dots as I have samples, after which CPU usage drops to 0 and the function just hangs.

The same call to sleuth_prep using num_cores = 1 finishes in quite a short amount of time (certainly under 10 minutes) while the multicore call would hang for at least hours (I've not seen it complete). Setting num_cores > 1 works just fine if gene_mode is unset. I have tested this both in a jupyter notebook and also directly in an R terminal.

Here is the call to sleuth_prep:

human_so <- 
    sleuth_prep(
        human_sample_info, 
        target_mapping = human_ttg,
        aggregation_column = 'gene', 
        gene_mode = TRUE,
        extra_bootstrap_summary = TRUE,
        num_cores = 12)

and here are the messages:

'gene_mode' is TRUE. Sleuth will do counts aggregation at the gene level for downstream normalization, transformation, and modeling steps, as well as for plotting and results.
reading in kallisto results
dropping unused factor levels
.............
normalizing est_counts
63056 targets passed the filter
normalizing tpm
merging in metadata
aggregating by column: gene
30474 genes passed the filter
summarizing bootstraps
.............

At which point the function hangs.

Thanks very much for any help and please let me know if you need sessionInfo() or any other details. The sleuth version is sleuth_0.30.0 and R is 3.5.1. I am running things on a server with 1TB RAM and 64 cores and I have no restrictions on thread, RAM usage.