Date: Mon, 05 Jan 2026 23:31:44 +0000 From: Matthias Andree <mandree@FreeBSD.org> To: ports-committers@FreeBSD.org, dev-commits-ports-all@FreeBSD.org, dev-commits-ports-main@FreeBSD.org Subject: git: 90d4fc1621bb - main - graphics/openexr: fix build on Sandy Bridge CPUs Message-ID: <695c49e0.34d99.6c8eb2f9@gitrepo.freebsd.org>
index | next in thread | raw e-mail
The branch main has been updated by mandree: URL: https://cgit.FreeBSD.org/ports/commit/?id=90d4fc1621bb463d19be5a6459f90d644258188d commit 90d4fc1621bb463d19be5a6459f90d644258188d Author: Matthias Andree <mandree@FreeBSD.org> AuthorDate: 2026-01-05 23:24:13 +0000 Commit: Matthias Andree <mandree@FreeBSD.org> CommitDate: 2026-01-05 23:31:22 +0000 graphics/openexr: fix build on Sandy Bridge CPUs OpenEXR attempts to use F16C intrinsics when AVX available and at least one of these is true: F16C available, GCC used, or clang used. Intel Sandy Bridge processors do have AVX extensions, but no F16C instruction set extension, so compilers (clang 19, 20, gcc 14) complain: work/openexr-3.4.4/src/lib/OpenEXRCore/unpack.c:37:18: error: always_inline function '_mm256_cvtph_ps' requires target feature 'f16c', but would be inlined into function 'half_to_float_buffer' that is compiled without support for 'f16c' 37 | out, _mm256_cvtph_ps (_mm_loadu_si128 ((const __m128i*) in))); | ^ Not bumping PORTREVISION because this is not a supported configuration. https://github.com/AcademySoftwareFoundation/openexr/issues/2231 Reported by: "UnitedMarsupials" MFH: 2025Q4 --- graphics/openexr/files/patch-src_lib_OpenEXRCore_unpack.c | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/graphics/openexr/files/patch-src_lib_OpenEXRCore_unpack.c b/graphics/openexr/files/patch-src_lib_OpenEXRCore_unpack.c new file mode 100644 index 000000000000..62afea351beb --- /dev/null +++ b/graphics/openexr/files/patch-src_lib_OpenEXRCore_unpack.c @@ -0,0 +1,11 @@ +--- src/lib/OpenEXRCore/unpack.c.orig 2025-11-17 01:18:11 UTC ++++ src/lib/OpenEXRCore/unpack.c +@@ -15,7 +15,7 @@ + + /* TODO: learn arm neon intrinsics for this */ + #if (defined(__x86_64__) || defined(_M_X64)) +-# if defined(__AVX__) && (defined(__F16C__) || defined(__GNUC__) || defined(__clang__)) ++# if defined(__AVX__) && defined(__F16C__) && (defined(__GNUC__) || defined(__clang__)) + # define USE_F16C_INTRINSICS + # elif (defined(__GNUC__) || defined(__clang__)) + # define ENABLE_F16C_TESThome | help
Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?695c49e0.34d99.6c8eb2f9>