Re: [PATCH] aarch64: prefer using csinv, csneg in zero extend contexts

Alex Coplan writes: >> -Original Message- >> From: Richard Sandiford >> Sent: 06 May 2020 11:28 >> To: Alex Coplan >> Cc: gcc-patches@gcc.gnu.org; Richard Earnshaw ; >> Marcus Shawcroft ; Kyrylo Tkachov >> ; nd >> Subject: Re: [PATCH] a

RE: [PATCH] aarch64: prefer using csinv, csneg in zero extend contexts

> -Original Message- > From: Richard Sandiford > Sent: 06 May 2020 11:28 > To: Alex Coplan > Cc: gcc-patches@gcc.gnu.org; Richard Earnshaw ; > Marcus Shawcroft ; Kyrylo Tkachov > ; nd > Subject: Re: [PATCH] aarch64: prefer using csinv, csneg in zero extend &g

Re: [PATCH] aarch64: prefer using csinv, csneg in zero extend contexts

Alex Coplan writes: >> -Original Message- >> From: Richard Sandiford >> Sent: 30 April 2020 15:13 >> To: Alex Coplan >> Cc: gcc-patches@gcc.gnu.org; Richard Earnshaw ; >> Marcus Shawcroft ; Kyrylo Tkachov >> ; nd >> Subject: Re: [PATCH] a

RE: [PATCH] aarch64: prefer using csinv, csneg in zero extend contexts

> -Original Message- > From: Richard Sandiford > Sent: 30 April 2020 15:13 > To: Alex Coplan > Cc: gcc-patches@gcc.gnu.org; Richard Earnshaw ; > Marcus Shawcroft ; Kyrylo Tkachov > ; nd > Subject: Re: [PATCH] aarch64: prefer using csinv, csneg in zero extend >

Re: [PATCH] aarch64: prefer using csinv, csneg in zero extend contexts

Alex Coplan writes: >> At the risk of feature creep :-) a useful third pattern could be >> to combine a zero-extended operator result with an existing DImode value. >> In that case, the existing DImode value really can be "rZ" and should >> always be in the "else" arm of the if_then_else. E.g.: >

RE: [PATCH] aarch64: prefer using csinv, csneg in zero extend contexts

gcc.gnu.org; Richard Earnshaw ; > Marcus Shawcroft ; Kyrylo Tkachov > ; nd > Subject: Re: [PATCH] aarch64: prefer using csinv, csneg in zero extend > contexts > > Yeah. The thing that surprised me was that the non-extending form > has the operator in the "then" arm o

Re: [PATCH] aarch64: prefer using csinv, csneg in zero extend contexts

Thanks for doing this. Alex Coplan writes: > Hello, > > The attached patch adds an optimization to the AArch64 backend to catch > additional cases where we can use csinv and csneg. > > Given the C code: > > unsigned long long inv(unsigned a, unsigned b, unsigned c) > { > return a ? b : ~c; > }

[PATCH] aarch64: prefer using csinv, csneg in zero extend contexts

Hello, The attached patch adds an optimization to the AArch64 backend to catch additional cases where we can use csinv and csneg. Given the C code: unsigned long long inv(unsigned a, unsigned b, unsigned c) { return a ? b : ~c; } Prior to this patch, AArch64 GCC at -O2 generates: inv: