RE: [PATCH v3 11/11] Hexagon (target/hexagon) Use direct block chaining for tight loops

[Taylor Simpson]

> -----Original Message-----
> From: Richard Henderson <richard.henderson@linaro.org>
> Sent: Friday, November 4, 2022 8:44 PM
> To: Taylor Simpson <tsimpson@quicinc.com>; qemu-devel@nongnu.org
> Cc: philmd@linaro.org; ale@rev.ng; anjo@rev.ng; Brian Cain
> <bcain@quicinc.com>; Matheus Bernardino (QUIC)
> <quic_mathbern@quicinc.com>
> Subject: Re: [PATCH v3 11/11] Hexagon (target/hexagon) Use direct block
> chaining for tight loops
> 
> On 11/5/22 06:26, Taylor Simpson wrote:
> > Direct block chaining is documented here
> > https://qemu.readthedocs.io/en/latest/devel/tcg.html#direct-block-chai
> > ning
> >
> > Hexagon inner loops end with the endloop0 instruction To go back to
> > the beginning of the loop, this instructions writes to PC from
> > register SA0 (start address 0).  To use direct block chaining, we have
> > to assign PC with a constant value.  So, we specialize the code
> > generation when the start of the translation block is equal to SA0.
> >
> > When this is the case, we defer the compare/branch from endloop0 to
> > gen_end_tb.  When this is done, we can assign the start address of the
> > TB to PC.
> >
> > Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>
> > ---
> >   target/hexagon/cpu.h       | 17 ++++++++----
> >   target/hexagon/gen_tcg.h   |  3 ++
> >   target/hexagon/translate.h |  1 +
> >   target/hexagon/genptr.c    | 57
> ++++++++++++++++++++++++++++++++++++++
> >   target/hexagon/translate.c | 34 +++++++++++++++++++++++
> >   5 files changed, 107 insertions(+), 5 deletions(-)
> >
> > diff --git a/target/hexagon/cpu.h b/target/hexagon/cpu.h index
> > ff8c26272d..5260e0f127 100644
> > --- a/target/hexagon/cpu.h
> > +++ b/target/hexagon/cpu.h
> > @@ -152,16 +152,23 @@ struct ArchCPU {
> >
> >   #include "cpu_bits.h"
> >
> > +typedef union {
> > +    uint32_t i;
> > +    struct {
> > +        bool is_tight_loop:1;
> > +    };
> > +} HexStateFlags;
> 
> I don't see this as an improvement on manual flags handling, as it makes the
> flags value be dependent on host bit-field ordering.  This makes it more
> difficult to compare traces across hosts.

I coded this originally with manual handling but decided this would be easier 
to read/understand/maintain - especially as we add more flags and some have 
more than 1 bit.

I haven't noticed the flags in any of the logs.  Where are they printed?


> 
> Otherwise,
> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
> 
> 
> r~