On 18/08/2024 21.24, Paul Gevers wrote:
Source: nvidia-nccl
ppc64el uploaded arm64 FTBFS with /usr/include/aarch64-linux-gnu/bits/math-vector.h(96): error: identifier "__Float32x4_t" is undefined typedef __Float32x4_t __f32x4_t; ^ /usr/include/aarch64-linux-gnu/bits/math-vector.h(97): error: identifier "__Float64x2_t" is undefined typedef __Float64x2_t __f64x2_t; ^ /usr/include/aarch64-linux-gnu/bits/math-vector.h(106): error: identifier "__SVFloat32_t" is undefined typedef __SVFloat32_t __sv_f32_t; ^ /usr/include/aarch64-linux-gnu/bits/math-vector.h(107): error: identifier "__SVFloat64_t" is undefined typedef __SVFloat64_t __sv_f64_t; ^ /usr/include/aarch64-linux-gnu/bits/math-vector.h(108): error: identifier "__SVBool_t" is undefined typedef __SVBool_t __sv_bool_t; ^ I've seen this before ... (tensorpipe-cuda) ... might be a libc6 regression Andreas