Re: bit tweaks [was: Re: [nfsd4] potentially hardware breaking regression in 4.14-rc and 4.13.11]

On Tue, Nov 14, 2017 at 3:53 PM, Rasmus Villemoes wrote: > > Odd. 7.2 and 7.1 (both of which I've just compiled from source, no > special configure flags or anything) generate exactly the same (good) > code for fs/namespace.o after patching. I also tried with >

Re: bit tweaks [was: Re: [nfsd4] potentially hardware breaking regression in 4.14-rc and 4.13.11]

On Tue, Nov 14, 2017 at 3:53 PM, Rasmus Villemoes wrote: > > Odd. 7.2 and 7.1 (both of which I've just compiled from source, no > special configure flags or anything) generate exactly the same (good) > code for fs/namespace.o after patching. I also tried with > CONFIG_CC_OPTIMIZE_FOR_SIZE=y just

Re: bit tweaks [was: Re: [nfsd4] potentially hardware breaking regression in 4.14-rc and 4.13.11]

On 14 November 2017 at 23:43, Linus Torvalds wrote: > On Tue, Nov 14, 2017 at 2:24 PM, Rasmus Villemoes > wrote: >> >> Can you be more specific? That's not what I see with gcc 7.1. > > I have gcc-7.2.1, and it made a horrible mess of the

Re: bit tweaks [was: Re: [nfsd4] potentially hardware breaking regression in 4.14-rc and 4.13.11]

On 14 November 2017 at 23:43, Linus Torvalds wrote: > On Tue, Nov 14, 2017 at 2:24 PM, Rasmus Villemoes > wrote: >> >> Can you be more specific? That's not what I see with gcc 7.1. > > I have gcc-7.2.1, and it made a horrible mess of the do_mount() code. Odd. 7.2 and 7.1 (both of which I've

Re: bit tweaks [was: Re: [nfsd4] potentially hardware breaking regression in 4.14-rc and 4.13.11]

On Tue, Nov 14, 2017 at 2:24 PM, Rasmus Villemoes wrote: > > Can you be more specific? That's not what I see with gcc 7.1. I have gcc-7.2.1, and it made a horrible mess of the do_mount() code. Look for the comment "/* Separate the per-mountpoint flags */" and do the

Re: bit tweaks [was: Re: [nfsd4] potentially hardware breaking regression in 4.14-rc and 4.13.11]

On Tue, Nov 14, 2017 at 2:24 PM, Rasmus Villemoes wrote: > > Can you be more specific? That's not what I see with gcc 7.1. I have gcc-7.2.1, and it made a horrible mess of the do_mount() code. Look for the comment "/* Separate the per-mountpoint flags */" and do the obvious conversion of the

Re: bit tweaks [was: Re: [nfsd4] potentially hardware breaking regression in 4.14-rc and 4.13.11]

On 14 November 2017 at 00:54, Linus Torvalds wrote: > On Mon, Nov 13, 2017 at 3:30 PM, Linus Torvalds > wrote: >> >> So let's just rewrite that mnt_flags conversion that way, justr to get >> gcc to generate the obvious code. > > Oh

Re: bit tweaks [was: Re: [nfsd4] potentially hardware breaking regression in 4.14-rc and 4.13.11]

On 14 November 2017 at 00:54, Linus Torvalds wrote: > On Mon, Nov 13, 2017 at 3:30 PM, Linus Torvalds > wrote: >> >> So let's just rewrite that mnt_flags conversion that way, justr to get >> gcc to generate the obvious code. > > Oh wow. I tried to do the same thing in fs/namespace.c where it

Re: bit tweaks [was: Re: [nfsd4] potentially hardware breaking regression in 4.14-rc and 4.13.11]

On Mon, Nov 13, 2017 at 3:30 PM, Linus Torvalds wrote: > > So let's just rewrite that mnt_flags conversion that way, justr to get > gcc to generate the obvious code. Oh wow. I tried to do the same thing in fs/namespace.c where it does the reverse bit translation,

Re: bit tweaks [was: Re: [nfsd4] potentially hardware breaking regression in 4.14-rc and 4.13.11]

On Mon, Nov 13, 2017 at 3:30 PM, Linus Torvalds wrote: > > So let's just rewrite that mnt_flags conversion that way, justr to get > gcc to generate the obvious code. Oh wow. I tried to do the same thing in fs/namespace.c where it does the reverse bit translation, and gcc makes a _horrible_ mess

Re: bit tweaks [was: Re: [nfsd4] potentially hardware breaking regression in 4.14-rc and 4.13.11]

On Mon, Nov 13, 2017 at 2:59 PM, Rasmus Villemoes wrote: >> Sadly, gcc makes a mess of it and actually generates code that looks >> like the original C.[...] > > Actually, new enough gcc (7.1, I think) does contain a pattern that does > this, but unfortunately only if

Re: bit tweaks [was: Re: [nfsd4] potentially hardware breaking regression in 4.14-rc and 4.13.11]

On Mon, Nov 13, 2017 at 2:59 PM, Rasmus Villemoes wrote: >> Sadly, gcc makes a mess of it and actually generates code that looks >> like the original C.[...] > > Actually, new enough gcc (7.1, I think) does contain a pattern that does > this, but unfortunately only if one spells it > > y |= (x