[Mesa-dev] [PATCH] nir: bump loop unroll limit to 96.

[Timothy Arceri]

On 11/10/17 10:50, Dave Airlie wrote:
> From: Dave Airlie <airlied at redhat.com>
> 
> With the ssao demo from Vulkan demos:
> radv/rx480: 440->440fps
> anv/haswell: 24->34 fps
> 
> The demo does a 0->32 loop across a ubo with 32 members.
> 
> Signed-off-by: Dave Airlie <airlied at redhat.com>

Reviewed-by: Timothy Arceri <tarceri at itsqueeze.com>

> ---
>   src/compiler/nir/nir_opt_loop_unroll.c | 4 +++-
>   1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/src/compiler/nir/nir_opt_loop_unroll.c b/src/compiler/nir/nir_opt_loop_unroll.c
> index 79d04f9..dae5bfc 100644
> --- a/src/compiler/nir/nir_opt_loop_unroll.c
> +++ b/src/compiler/nir/nir_opt_loop_unroll.c
> @@ -33,8 +33,10 @@
>    * to give about the same results. Around 5 instructions per node.  But some
>    * loops that would unroll with GLSL IR fail to unroll if we set this to 25 so
>    * we set it to 26.
> + * This was bumped to 96 because it unrolled more loops with a positive
> + * effect (vulkan ssao demo).
>    */
> -#define LOOP_UNROLL_LIMIT 26
> +#define LOOP_UNROLL_LIMIT 96
>   
>   /* Prepare this loop for unrolling by first converting to lcssa and then
>    * converting the phis from the loops first block and the block that follows
>