On Fri, Jan 07, 2005 at 10:34:48AM +0100, Uros Bizjak wrote:
Some progress was achieved by separating SSE and x87 floating point
operators by TARGET_SSE_MATH, and it was again shown that a small
(ehm...) oversight in i386.md produced another big performance problem [3].
Regarding these reports, I would suggest this solution:
- with TARGET_SSE_MATH, _all_ SFmode values should be processed in SSE
registers.
- with TARGET_SSE_MATH && TARGET_SSE2, _all_ SFmode and DFmode values
should be processed in SSE registers.
Well, if you have __asm with "f", "t" or "u" constraints, you should still
put SFmode resp. DFmode values into the stacked registers.