diff options
author | Ian Romanick <[email protected]> | 2019-06-03 15:22:15 -0700 |
---|---|---|
committer | Ian Romanick <[email protected]> | 2019-07-11 10:20:03 -0700 |
commit | 1259f6d80215152e647b5c32db4fad32e420d495 (patch) | |
tree | 353a8c5eae4be4e57c748073e9a327ba9e579b9d /src/compiler/nir | |
parent | 3a1fdca5ad206f578b0d54a490a8bf6f199c8851 (diff) |
nir: intel/vec4: Add flag to disable some algebraic optimizations
A couple patches later in this series use the flag to avoid a few
thousand shader-db regresions on all vec4 platforms.
I'm not particularly enamored with the name of this flag. However, I
suspect the Intel vec4 backend is the only backend that will benefit
from it. Specifically, the cases where this helps are all cases where
we want to prevent nir_opt_algebraic from rearranging instructions to
create 3-source instructions, such as ffma and flrp, with additional
immediate value or uniform sources.
The earlier commit "intel/vec4: Try to emit a single load for multiple
3-src instruction operands" solves most of the problems caused by
additional immediate values, but the restrictions on register strides
that cause problems for uniforms and shader inputs persist.
Reviewed-by: Matt Turner <[email protected]>
Diffstat (limited to 'src/compiler/nir')
-rw-r--r-- | src/compiler/nir/nir.h | 10 |
1 files changed, 10 insertions, 0 deletions
diff --git a/src/compiler/nir/nir.h b/src/compiler/nir/nir.h index 4fa50940e17..bc7ae1cc03e 100644 --- a/src/compiler/nir/nir.h +++ b/src/compiler/nir/nir.h @@ -2630,6 +2630,16 @@ typedef struct nir_shader_compiler_options { /* Lowers when rotate instruction is not supported */ bool lower_rotate; + /** + * Is this the Intel vec4 backend? + * + * Used to inhibit algebraic optimizations that are known to be harmful on + * the Intel vec4 backend. This is generally applicable to any + * optimization that might cause more immediate values to be used in + * 3-source (e.g., ffma and flrp) instructions. + */ + bool intel_vec4; + unsigned max_unroll_iterations; nir_lower_int64_options lower_int64_options; |