[Mesa-dev] [PATCH 1/3] nir/algebraic: Separate ffma lowering from fusing

[Kenneth Graunke]

On Thursday, May 5, 2016 5:51:18 PM PDT Jason Ekstrand wrote:
> The i965 driver has its own pass for fusing mul+add combinations that's
> much smarter than what nir_opt_algebraic can do so we don't want to get the
> nir_opt_algebraic one just because we didn't set lower_ffma.
> ---
>  src/compiler/nir/nir.h                      | 1 +
>  src/compiler/nir/nir_opt_algebraic.py       | 2 +-
>  src/gallium/drivers/freedreno/ir3/ir3_nir.c | 1 +
>  3 files changed, 3 insertions(+), 1 deletion(-)

Series is:
Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>

I always thought it was pretty sketchy that we were splitting up
actual fma() GLSL built-in calls.  The whole point of using fma()
is that you want higher precision.  Otherwise, there's a * b + c.

It may have been technically legal (maybe), but this seems like the
right thing to do.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: This is a digitally signed message part.
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20160510/9b8bc67c/attachment.sig>