Johannes, Olaf, Guenter....
thanks for these explanations. according to the very informative docs that Johannes sent, Intel doesn't consider this denormal bug to be their problem, and expect that software manufacturers will code around it.
also, looking at Olaf's patches, they make sense, but i lack the programming know-how to put them in the "critical spots".
so my question: could we "officially" incorperate these patches to fix the denormal problem into the PD release? such patches would have no detrimental effect on other processors/operating systems, right? perhaps they would even improve performance of other processors, since denormal numbers must eat CPU time of any machine to calculate, not just P4s.
best, derek
Johannes Taelman wrote:
Here's an overview of denormal handling: http://www.musicdsp.org/files/denormal.pdf