Re: [03/nn] Allow vector CONSTs

2017-10-27 Thread Richard Sandiford
Jeff Law writes: > On 10/23/2017 05:18 AM, Richard Sandiford wrote: >> This patch allows (const ...) wrappers to be used for rtx vector >> constants, as an alternative to const_vector. This is useful >> for SVE, where the number of elements isn't known until runtime. > Right. It's constant, but

Re: [03/nn] Allow vector CONSTs

2017-10-25 Thread Jeff Law
On 10/23/2017 05:18 AM, Richard Sandiford wrote: > This patch allows (const ...) wrappers to be used for rtx vector > constants, as an alternative to const_vector. This is useful > for SVE, where the number of elements isn't known until runtime. Right. It's constant, but not knowable at compile t

[03/nn] Allow vector CONSTs

2017-10-23 Thread Richard Sandiford
This patch allows (const ...) wrappers to be used for rtx vector constants, as an alternative to const_vector. This is useful for SVE, where the number of elements isn't known until runtime. It could also be useful in future for fixed-length vectors, to reduce the amount of memory needed to repre

Allow vector CONSTs

2017-09-23 Thread Richard Sandiford
This patch allows (const ...) wrappers to be used for rtx vector constants, as an alternative to const_vector. This is useful for SVE, where the number of elements isn't known until runtime. It could also be useful in future for fixed-length vectors, to reduce the amount of memory needed to repres