Re: [PATCH][AArch64] Change FP reassociation width

2017-06-14 Thread James Greenhalgh
On Tue, Jun 13, 2017 at 10:43:05AM +0100, Wilco Dijkstra wrote: > Richard Earnshaw (lists) wrote: > > > > Why 1 and not 2?  Many processors have 2 fp pipes and forcing this down > > to a sequential stream is not obviously the right thing. > > 1 was faster than 2. Like I said, the reassociation is

Re: [PATCH][AArch64] Change FP reassociation width

2017-06-13 Thread Wilco Dijkstra
Richard Earnshaw (lists) wrote: > > Why 1 and not 2?  Many processors have 2 fp pipes and forcing this down > to a sequential stream is not obviously the right thing. 1 was faster than 2. Like I said, the reassociation is too aggressive and even splits multiply-add rather than keeping them. Until

Re: [PATCH][AArch64] Change FP reassociation width

2017-06-13 Thread Richard Earnshaw (lists)
On 12/06/17 11:50, Wilco Dijkstra wrote: > Currently the FP reassociation width is set to 4 on AArch64. On recent > GCCs this has become more aggressive in splitting expressions. This means > many FMAs are split into FMUL and FADD. The reassociation increases register > pressure, in some benchma

[PATCH][AArch64] Change FP reassociation width

2017-06-12 Thread Wilco Dijkstra
Currently the FP reassociation width is set to 4 on AArch64. On recent GCCs this has become more aggressive in splitting expressions. This means many FMAs are split into FMUL and FADD. The reassociation increases register pressure, in some benchmarks so much that inner loops start to spill. This