On 10/04/2023 15:09, Jerry Evans wrote:
Having been badly burned by rubbish Intel math hardware back in the day*, I was fascinated to more latterly come across this set of related resources:https://urldefense.proofpoint.com/v2/url?u=https-3A__web.stanford.edu_class_ee380_Abstracts_170201.html&d=DwICaQ&c=009klHSCxuh5AI1vNQzSO0KGjl4nbi2Q0M1QLJX9BeE&r=TRvFbpof3kTa2q5hdjI2hccynPix7hNL2n0I6DmlDy0&m=-Kn71eXIpKloARhvNgZWAbdJVCmCOr158TDTrutp3yu-qm1lwIVrN6fg-cLut2jB&s=7k4APvkpMu1OX2rgmJOovN6FO1HdmtQsLdgF0zbZ9gI&e= https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_libcg_bfp&d=DwICaQ&c=009klHSCxuh5AI1vNQzSO0KGjl4nbi2Q0M1QLJX9BeE&r=TRvFbpof3kTa2q5hdjI2hccynPix7hNL2n0I6DmlDy0&m=-Kn71eXIpKloARhvNgZWAbdJVCmCOr158TDTrutp3yu-qm1lwIVrN6fg-cLut2jB&s=2SBZjna2q0YhnJ2M5bAC-9lRpDTqmlB9q02QGIQo1Tc&e= .On the subject of plugins setting FPU flags and modes. All I can say is yak. A terrible idea as, IIRC, even changing modes was slow. And a guarantee of absolutely zero. FPU modality (and denormal avoidance/mitigation) is precisely the kind of thing that should be specified by the host to downstream processors, just like bit depth, sample rate and buffer sizes. * Pentium 4 anyone? Cry me a river.
boul blip blop?
