Apparently the presence of a hardware FPU is no longer implied by
-march=armv7-a (or armv7 I assume), and so adding -mfpu=neon is
necessary when using hardware floating point in gcc/g++.
Change-Id: I59c5b58933fae2e4e5a747b2af128b801acc812e
Reviewed-on: https://gem5-review.googlesource.com/c/public/gem5/+/27747
Maintainer: Gabe Black <gabeblack@google.com>
Tested-by: kokoro <noreply+kokoro@google.com>
Reviewed-by: Jason Lowe-Power <power.jg@gmail.com>
env['ABI'] = 'arm'
get_abi_opt('CROSS_COMPILE', 'arm-linux-gnueabihf-')
-env.Append(CXXFLAGS='-march=armv7-a')
+env.Append(CXXFLAGS=[ '-march=armv7-a', '-mfpu=neon' ])
env['CALL_TYPE']['inst'].impl('m5op.S', 'verify_inst.cc', default=True)
env['ABI'] = 'thumb'
get_abi_opt('CROSS_COMPILE', 'arm-linux-gnueabihf-')
-env.Append(CXXFLAGS=[ '-mthumb', '-march=armv7' ])
+env.Append(CXXFLAGS=[ '-mthumb', '-march=armv7', '-mfpu=neon' ])
env['CALL_TYPE']['inst'].impl('m5op.S', 'verify_inst.cc', default=True)