Re: CFI directives and dynamic stack alignment

2015-08-10 Thread Alan Modra
On Mon, Aug 03, 2015 at 02:48:09PM -0700, Steve Ellcey wrote: > When I generate code to dynamically align the stack my code looks like > this: > > fn2: > .frame $fp,32,$31 # vars= 0, regs= 2/0, args= 16, gp= 8 > .mask 0xc000,-4 > .fmask 0x,0 > .

Re: [gimple] PHI-opt bug, advice needed

2015-08-10 Thread Richard Biener
On August 10, 2015 11:49:15 AM GMT+02:00, Nathan Sidwell wrote: >Richard, >I got the min/max VRP optimization I mentioned working, but it >uncovered a bug >in phiopts where that pass is constructing MIN and MAX exprs. (Funnily >enough, >I only found that failure due to a broken test of decimal

[gimple] PHI-opt bug, advice needed

2015-08-10 Thread Nathan Sidwell
Richard, I got the min/max VRP optimization I mentioned working, but it uncovered a bug in phiopts where that pass is constructing MIN and MAX exprs. (Funnily enough, I only found that failure due to a broken test of decimal fp). We start with the following bit of code: // X has unknown rang