Opened 3 years ago

Closed 2 years ago

#29575 closed defect (invalid)

debian-jessie: fflas_ffpack again

Reported by: Matthias Köppe Owned by:
Priority: major Milestone: sage-duplicate/invalid/wontfix
Component: packages: standard Keywords:
Cc: Dima Pasechnik, Antonio Rojas, Timo Kaufmann, François Bissey, Volker Braun, ​tmonteil Merged in:
Authors: Reviewers: Jonathan Kliem, Matthias Koeppe
Report Upstream: N/A Work issues:
Branch: Commit:
Dependencies: Stopgaps:

Status badges

Description

Again (even after #29102, unrelated to #29161):

debian-jessie-minimal (https://github.com/mkoeppe/sage/runs/617198894):

../../../fflas-ffpack/fflas/fflas_simd/simd512_float.inl: In static member function 'static Simd512_impl<true, false, true, 4>::vect_t Simd512_impl<true, false, true, 4>::blendv(Simd512_impl<true, false, true, 4>::vect_t, Simd512_impl<true, false, true, 4>::vect_t, Simd512_impl<true, false, true, 4>::vect_t)':
../../../fflas-ffpack/fflas/fflas_simd/simd512_float.inl:189:48: error: '_mm512_castps512_ps256' was not declared in this scope
         __m256 lowa  = _mm512_castps512_ps256(a);
                                                ^
../../../fflas-ffpack/fflas/fflas_simd/simd512_float.inl:190:82: error: '_mm512_castps_pd' was not declared in this scope
         __m256 higha = _mm256_castpd_ps(_mm512_extractf64x4_pd(_mm512_castps_pd(a),1));
                                                                                  ^
../../../fflas-ffpack/fflas/fflas_simd/simd512_float.inl:201:51: error: '_mm512_castps256_ps512' was not declared in this scope
         __m512 res = _mm512_castps256_ps512(reslow);

Change History (8)

comment:1 Changed 3 years ago by Matthias Köppe

Priority: majorblocker

comment:2 Changed 3 years ago by gh-kliem

This looks like this should be handled in #27122.

_mm512_castps512_ps256 is an AVX512F instruction (see https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm512_castps512_ps256&expand=625) and it looks like gcc 4.9.2 just cannot handle it.

comment:3 Changed 3 years ago by gh-kliem

And I think this only happened on the test branch because #27122 was tested and therefore -march=native in CFLAGS.

comment:4 Changed 3 years ago by Matthias Köppe

Yes, I agree

comment:5 Changed 3 years ago by Matthias Köppe

Milestone: sage-9.1sage-duplicate/invalid/wontfix
Priority: blockermajor
Status: newneeds_review

comment:6 Changed 3 years ago by gh-kliem

Status: needs_reviewpositive_review

comment:7 Changed 3 years ago by Samuel Lelièvre

Reviewers: Jonathan Kliem, Matthias Koeppe

comment:8 Changed 2 years ago by Frédéric Chapoton

Resolution: invalid
Status: positive_reviewclosed
Note: See TracTickets for help on using tickets.