Repository Issue Activity (beta feature)

#

#

#1120 Autotune over epilogue subtiling

#2044 Kernel perf regression using latest main branch with configs autotuned from v1.0.0

#

#

#2013 [AMD 350x] Helion Layernorm kernel performance is worse compared to inductor for some shapes in backward pass

#

#2041 Support register-backed tensor caching across `hl.static_range` iterations (equivalent to Triton's `VAR_ARGS_ARRAY`)

#1661 [AMD 350X] Helion DCPP kernel has numerical accuracy discrepancy compared to inductor's

#

#

#

#

#

#

#1425 [AMD 350X] Helion Layernorm kernel performance is worse compared to inductor

#

#

#

#

#

#

#

#