Re: [PATCH v2] target/ppc: Fix for optimized vsl/vsr instructions

2019-10-05 Thread David Gibson
On Fri, Oct 04, 2019 at 01:52:41PM -0500, Paul Clarke wrote: > On 10/4/19 8:43 AM, Stefan Brankovic wrote: > > In previous implementation, invocation of TCG shift function could request > > shift of TCG variable by 64 bits when variable 'sh' is 0, which is not > > supported in TCG (values can be

Re: [PATCH v2] target/ppc: Fix for optimized vsl/vsr instructions

2019-10-04 Thread Paul Clarke
On 10/4/19 8:43 AM, Stefan Brankovic wrote: > In previous implementation, invocation of TCG shift function could request > shift of TCG variable by 64 bits when variable 'sh' is 0, which is not > supported in TCG (values can be shifted by 0 to 63 bits). This patch fixes > this by using two

Re: [PATCH v2] target/ppc: Fix for optimized vsl/vsr instructions

2019-10-04 Thread Aleksandar Markovic
> Reported-by: Paul Clark Stefan, Paul's full name is Paul A. Clarke. Thanks for the fix! Aleksandar

[PATCH v2] target/ppc: Fix for optimized vsl/vsr instructions

2019-10-04 Thread Stefan Brankovic
This patch fixes bug in optimized vsl/vsr instructions reported by Mark Cave-Ayland and Paul Clarke. Sorry for not responding earlier, I was absent last couple of days. I also integrated some suggestions made by Aleksandar Markovic. New soultion is tested and still has noticable performance

[PATCH v2] target/ppc: Fix for optimized vsl/vsr instructions

2019-10-04 Thread Stefan Brankovic
In previous implementation, invocation of TCG shift function could request shift of TCG variable by 64 bits when variable 'sh' is 0, which is not supported in TCG (values can be shifted by 0 to 63 bits). This patch fixes this by using two separate invocation of TCG shift functions, with maximum