search for: mutatingfma

I wonder if we can do this in a separate analysis MachineFunction SSA pass. 1) SelectionDAG will generate a pseudo instruction MutatingFMA. When it's generated it's allowed to have d = a * b + c form, where d doesn't have to be in {a, b, c}. 2) Later, the proposed pass uses an algorithm to decide for instruction MI: `%vreg0 = MutatingFMA %vreg1, %vreg2, %vreg3`, it should tie %vreg0 with, say %vreg2. Then it tells the sche...

...On Fri, Mar 4, 2016 at 5:14 PM Tim Shen <timshen at google.com> wrote: > In 5), by saying "In the same block" I mean we should replace uses of > vreg0 with vreg2 afterwards, but only limited in the same basicblock. > > In 6), we need another tiny piece to actually lower MutatingFMA to real > forms, which should be trivial. > > On Fri, Mar 4, 2016 at 5:09 PM Tim Shen <timshen at google.com> wrote: > >> I wonder if we can do this in a separate analysis MachineFunction SSA >> pass. >> 1) SelectionDAG will generate a pseudo instruction Mutating...

...at 5:14 PM Tim Shen <timshen at google.com> wrote: > >> In 5), by saying "In the same block" I mean we should replace uses of >> vreg0 with vreg2 afterwards, but only limited in the same basicblock. >> >> In 6), we need another tiny piece to actually lower MutatingFMA to real >> forms, which should be trivial. >> >> On Fri, Mar 4, 2016 at 5:09 PM Tim Shen <timshen at google.com> wrote: >> >>> I wonder if we can do this in a separate analysis MachineFunction SSA >>> pass. >>> 1) SelectionDAG will generate a...

Ping? On Mon, Feb 22, 2016 at 1:06 PM Tim Shen <timshen at google.com> wrote: > On Fri, Feb 19, 2016 at 5:10 PM Tim Shen <timshen at google.com> wrote: > >> I wonder if we can fix this by making the transformation simpler, that >> is, instead of doing: >> > > I wrote a prototype (see attach) for this idea, it actually improves some > of the test cases