GATK AVX accelerated pairHMM Segfault issue with Long reads - Consistently around the same region

[gokalpcelik]

Hi. We have a user post related to HaplotypeCaller and long reads causing segfaults consistently around the same region (T2T genome chr7:135400000-135500000) in 5 samples.

https://gatk.broadinstitute.org/hc/en-us/community/posts/18200128189723-HaplotyopeCaller-SIGSEGV-error-in-small-region-across-multiple-samples

Segfault messages are summarized below

#
# A fatal error has been detected by the Java Runtime Environment:
#
#  SIGSEGV (0xb) at pc=0x00007e376e5f70fb, pid=141529, tid=141589
#
# JRE version: OpenJDK Runtime Environment 21.9 (18.0.1+10) (build 18.0.1+10)
# Java VM: OpenJDK 64-Bit Server VM 21.9 (18.0.1+10, mixed mode, sharing, tiered, compressed class ptrs, g1 gc, linux-amd64)
# Problematic frame:
# C  [libgkl_pairhmm_omp13802390137529139622.so+0x1c0fb]  double compute_full_prob_avx512d<double>(testcase*)+0x37b

Disabling AVX accelerated pairHMM solved the problem and user was able to call variants from the region.

What else can be done for it ?