aring@bmc.uu.se (05/31/89)
A subroutine in microsoft C, compiled in MS Quick-C does not use the 8087 coprocessor on my XT-clone, when linked into a Quick-Basic 4.5 Quick-Library, and also not when linked to Quick-Basic 4.5 pograms from DOS. The Quick Basic programs, however, have no problem recognising the 8087, whether linked to the Quick Library or not. Further, this problem seems to be specific to Quick-Basic 4.5 (4.0 - no problem) and to the XT-clone. On an AT-clone (with an 8087 strapped at 4.77 MHz however) there no problem. I should also point out that the subroutine DOES use the 8087 when run from the Quick-C environment, and when linked to C programs from the object file. Has anyone an idea of what the problem might be ? I've tried all kind of linking and compilation switches and setting environment variables (i.e. NO87=) but nothing helps. In fact I've timed the calculations in Quick-Basic 4.0 with NO87=TRUE and get exactly the same time as when running in Quick-Basic 4.5 without setting NO87. So it may be assumed that it IS the recognition of the 8087 which is the problem and not some other source of delay or interrupt. Seccond question:I need an fft which is faster than a streight Cooley-Tukey implementation (in C). Gratefull for a reference or source. Avi Ring. Dpmt Physiology and Biophysics, BMC, Box 572, S-751 23, SWEDEN