volk icon indicating copy to clipboard operation
volk copied to clipboard

volk_32fc_s32fc_rotator2puppet_32fc fails in {a,u}_sse4_1

Open marcusmueller opened this issue 5 months ago • 10 comments

found this while working on packaging: ctest fails in a container, and pretty drastically so:

offset 3 in1: 0.85403 + 0.188718j in2: -0.00175621 + 0.0954956j tolerance was: 0.001

|in2| is too small by a factor of roughly 10.

Full output

105/218 Test #105: qa_volk_32fc_s32fc_rotator2puppet_32fc ......................................................................................***Failed    0.01 sec
RUN_VOLK_TESTS: volk_32fc_s32fc_rotator2puppet_32fc(131071,1)
generic completed in 0.233065 ms
a_sse4_1 completed in 0.110075 ms
u_sse4_1 completed in 0.109965 ms
a_avx completed in 0.056566 ms
u_avx completed in 0.056075 ms
a_avx_fma completed in 0.061925 ms
u_avx_fma completed in 0.062336 ms
offset 2 in1: -0.84885 + 0.194543j  in2: -0.805366 + 0.269719j tolerance was: 0.001
offset 3 in1: 0.85403 + 0.188718j  in2: -0.00175621 + 0.0954956j tolerance was: 0.001
offset 4 in1: -0.845072 + -0.717403j  in2: -0.921176 + -0.513012j tolerance was: 0.001
offset 5 in1: 0.395931 + 0.139802j  in2: -0.00499856 + 0.000294006j tolerance was: 0.001
offset 6 in1: 0.58138 + 0.301571j  in2: 0.597208 + 0.111701j tolerance was: 0.001
offset 7 in1: 0.373094 + 0.818401j  in2: 0.000608006 + -0.00100112j tolerance was: 0.001
offset 8 in1: 0.41406 + 0.391089j  in2: 0.481277 + 0.184129j tolerance was: 0.001
offset 9 in1: 0.850985 + -0.569144j  in2: 2.67737e-05 + -0.000143105j tolerance was: 0.001
offset 10 in1: -0.537605 + -0.63551j  in2: -0.682274 + -0.271972j tolerance was: 0.001
offset 11 in1: -0.392492 + 0.269725j  in2: -7.39303e-06 + 1.97787e-07j tolerance was: 0.001
volk_32fc_s32fc_rotator2puppet_32fc: fail on arch a_sse4_1
offset 2 in1: -0.84885 + 0.194543j  in2: -0.805366 + 0.269719j tolerance was: 0.001
offset 3 in1: 0.85403 + 0.188718j  in2: -0.00175621 + 0.0954956j tolerance was: 0.001
offset 4 in1: -0.845072 + -0.717403j  in2: -0.921176 + -0.513012j tolerance was: 0.001
offset 5 in1: 0.395931 + 0.139802j  in2: -0.00499856 + 0.000294006j tolerance was: 0.001
offset 6 in1: 0.58138 + 0.301571j  in2: 0.597208 + 0.111701j tolerance was: 0.001
offset 7 in1: 0.373094 + 0.818401j  in2: 0.000608006 + -0.00100112j tolerance was: 0.001
offset 8 in1: 0.41406 + 0.391089j  in2: 0.481277 + 0.184129j tolerance was: 0.001
offset 9 in1: 0.850985 + -0.569144j  in2: 2.67737e-05 + -0.000143105j tolerance was: 0.001
offset 10 in1: -0.537605 + -0.63551j  in2: -0.682274 + -0.271972j tolerance was: 0.001
offset 11 in1: -0.392492 + 0.269725j  in2: -7.39303e-06 + 1.97787e-07j tolerance was: 0.001
volk_32fc_s32fc_rotator2puppet_32fc: fail on arch u_sse4_1
Best aligned arch: u_avx
Best unaligned arch: u_avx

marcusmueller avatar Aug 09 '25 12:08 marcusmueller