Re: [PATCH 1/2] arm: Fix vcond_mask expander for MVE (PR target/100757)

On 16/07/2021 16:06, Richard Sandiford via Gcc-patches wrote: Hi, Sorry for the slow review. I'd initially held off from reviewing because it sounded like you were trying to treat predicates as MODE_VECTOR_BOOL instead. Is that right? If so, how did that go? Yes, that's part of PR 101325

Re: [PATCH 1/2] arm: Fix vcond_mask expander for MVE (PR target/100757)

Hi, Sorry for the slow review. I'd initially held off from reviewing because it sounded like you were trying to treat predicates as MODE_VECTOR_BOOL instead. Is that right? If so, how did that go? It does feel like the right long-term direction. Treating masks as integers for AVX seems to mak

Re: [PATCH 1/2] arm: Fix vcond_mask expander for MVE (PR target/100757)

Ping? Le ven. 2 juil. 2021 à 10:53, Christophe Lyon a écrit : > Hi, > > On Wed, 9 Jun 2021 at 17:04, Richard Sandiford > wrote: > > > > Christophe Lyon writes: > > > The problem in this PR is that we call VPSEL with a mask of vector > > > type instead of HImode. This happens because operand 3

Re: [PATCH 1/2] arm: Fix vcond_mask expander for MVE (PR target/100757)

On Fri, 2 Jul 2021 at 10:53, Christophe Lyon wrote: > > Hi, > > On Wed, 9 Jun 2021 at 17:04, Richard Sandiford > wrote: > > > > Christophe Lyon writes: > > > The problem in this PR is that we call VPSEL with a mask of vector > > > type instead of HImode. This happens because operand 3 in vcond_m

Re: [PATCH 1/2] arm: Fix vcond_mask expander for MVE (PR target/100757)

Hi, On Wed, 9 Jun 2021 at 17:04, Richard Sandiford wrote: > > Christophe Lyon writes: > > The problem in this PR is that we call VPSEL with a mask of vector > > type instead of HImode. This happens because operand 3 in vcond_mask > > is the pre-computed vector comparison and has vector type. The

Re: [PATCH 1/2] arm: Fix vcond_mask expander for MVE (PR target/100757)

Christophe Lyon writes: > The problem in this PR is that we call VPSEL with a mask of vector > type instead of HImode. This happens because operand 3 in vcond_mask > is the pre-computed vector comparison and has vector type. The fix is > to transfer this value to VPR.P0 by comparing operand 3 with

[PATCH 1/2] arm: Fix vcond_mask expander for MVE (PR target/100757)

The problem in this PR is that we call VPSEL with a mask of vector type instead of HImode. This happens because operand 3 in vcond_mask is the pre-computed vector comparison and has vector type. The fix is to transfer this value to VPR.P0 by comparing operand 3 with a vector of constant 1 of the sa