I am aligning a modest number of short reads (10.5 M) to a large genome (~4 Gb) with bwa mem. I expect a high error rate and expect many multiply mapping reads, so I relax some of the parameters as follows:

bwa mem -t 32 -k 15 -r 1.25 -c 100000000 -B 1 -O 0 -E 1 -L 500 -T 12

I keep getting errors related to memory allocation

[M::process] read 4529482 sequences (320000032 bp)...
[M::process] read 3965734 sequences (279216556 bp)...
[mem_chain2aln] Failed to allocate 22528 bytes at bwamem.c line 715: Cannot allocate memory
[mem_chain2aln] Failed to allocate 22528 bytes at bwamem.c line 715: Cannot allocate memory
[mem_chain2aln] Failed to allocate 22528 bytes at bwamem.c line 715: Cannot allocate memory
[mem_chain2aln] Failed to allocate 22528 bytes at bwamem.c line 715: Cannot allocate memory

This process is running on a 16-core node (32 with hyperthreading) with 128 GB memory and ~1 TB disk space.

How can it be running out of memory?