On Thu, 2017-01-12 at 15:20 +1100, Timothy Arceri wrote:
> If one of the inputs to the multiplcation in ffma is the result of
> an fmul there is a chance that we can reuse the result of that
> fmul in other ffma calls if we do the multiplication in the right
> order.
>
> For example it is a fairly common pattern for shaders to do something
> similar to this:
>
> const float a = 0.5;
> in vec4 b;
> in float c;
>
> ...
>
> b.x = b.x * c;
> b.y = b.y * c;
>
> ...
>
> b.x = b.x * a + a;
> b.y = b.y * a + a;
>
> So by simply detecting that constant a is part of the multiplication
> in ffma and switching it with previous fmul that updates b we end up
> with:
>
> ...
>
> c = a * c;
>
> ...
>
> b.x = b.x * c + a;
> b.y = b.y * c + a;
>
> shader-db results BDW:
>
> total instructions in shared programs: 13056473 -> 13038614 (-0.14%)
> instructions in affected programs: 2433641 -> 2415782 (-0.73%)
> helped: 10114
> HURT: 330
>
> total cycles in shared programs: 256515402 -> 256331740 (-0.07%)
> cycles in affected programs: 137476650 -> 137292988 (-0.13%)
> helped: 10802
> HURT: 3871
>
> total spills in shared programs: 14923 -> 14675 (-1.66%)
> spills in affected programs: 7976 -> 7728 (-3.11%)
> helped: 280
> HURT: 27
>
> total fills in shared programs: 20141 -> 19691 (-2.23%)
> fills in affected programs: 11408 -> 10958 (-3.94%)
> helped: 282
> HURT: 27
>
> LOST: 6
> GAINED: 1
> ---
> src/compiler/nir/nir_opt_algebraic.py | 2 ++
> src/compiler/nir/nir_search_helpers.h | 11 +++++++++++
> 2 files changed, 13 insertions(+)
>
> diff --git a/src/compiler/nir/nir_opt_algebraic.py
> b/src/compiler/nir/nir_opt_algebraic.py
> index b5974a7..3cddda3 100644
> --- a/src/compiler/nir/nir_opt_algebraic.py
> +++ b/src/compiler/nir/nir_opt_algebraic.py
> @@ -331,6 +331,8 @@ optimizations = [
> (('~fadd', '#a', ('fadd', b, '#c')), ('fadd', ('fadd', a, c),
> b)),
> (('iadd', '#a', ('iadd', b, '#c')), ('iadd', ('iadd', a, c), b)),
>
> + (('fadd', ('fmul(is_used_once)', ('fmul(is_used_once)',
> 'a(is_not_const)', 'b(is_not_const)'), '#c'), d), ('fadd', ('fmul',
> ('fmul', a, '#c'), b), d)),
> +
> # Misc. lowering
> (('fmod@32', a, b), ('fsub', a, ('fmul', b, ('ffloor', ('fdiv',
> a, b)))), 'options->lower_fmod32'),
> (('fmod@64', a, b), ('fsub', a, ('fmul', b, ('ffloor', ('fdiv',
> a, b)))), 'options->lower_fmod64'),
> diff --git a/src/compiler/nir/nir_search_helpers.h
> b/src/compiler/nir/nir_search_helpers.h
> index ddaff52..f89e417 100644
> --- a/src/compiler/nir/nir_search_helpers.h
> +++ b/src/compiler/nir/nir_search_helpers.h
> @@ -115,6 +115,17 @@ is_zero_to_one(nir_alu_instr *instr, unsigned
> src, unsigned num_components,
> }
>
> static inline bool
> +is_not_const(nir_alu_instr *instr, unsigned src, unsigned
> num_components)
I'm surprised this compiled and ran. There should be an extra param
here:
const uint8_t *swizzle
Fixed locally.
> +{
> + nir_const_value *val = nir_src_as_const_value(instr-
> >src[src].src);
> +
> + if (val)
> + return false;
> +
> + return true;
> +}
> +
> +static inline bool
> is_used_more_than_once(nir_alu_instr *instr)
> {
> bool zero_if_use = list_empty(&instr->dest.dest.ssa.if_uses);
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
