[XNNPACK] disable ARM_BF16 and ARM_FP16_VECTOR (#94020)

Summary: This is not used and will cause build failure

Test Plan: CI

Differential Revision: D42982023

Pull Request resolved: https://github.com/pytorch/pytorch/pull/94020
Approved by: https://github.com/Skylion007, https://github.com/tiandiao123, https://github.com/digantdesai
This commit is contained in:
Hansong Zhang 2023-02-03 05:00:58 +00:00 committed by PyTorch MergeBot
parent dd7d47c4ac
commit d996acfbc2

View File

@ -632,6 +632,11 @@ if(USE_XNNPACK AND NOT USE_SYSTEM_XNNPACK)
set(XNNPACK_BUILD_BENCHMARKS OFF CACHE BOOL "")
set(XNNPACK_BUILD_TESTS OFF CACHE BOOL "")
# Disable ARM BF16 and FP16 vector for now; unused and causes build failures because
# these new ISA features may not be supported on older compilers
set(XNNPACK_ENABLE_ARM_BF16 OFF CACHE BOOL "")
set(XNNPACK_ENABLE_ARM_FP16_VECTOR OFF CACHE BOOL "")
# Setting this global PIC flag for all XNNPACK targets.
# This is needed for Object libraries within XNNPACK which must
# be PIC to successfully link this static libXNNPACK with pytorch