diff options
| author | Jason Ekstrand <[email protected]> | 2014-12-16 12:22:01 -0800 |
|---|---|---|
| committer | Jason Ekstrand <[email protected]> | 2015-01-15 07:20:22 -0800 |
| commit | 46f3e1ab504f016ab900c045f165a8376cf3fc0c (patch) | |
| tree | cb1afd74eeed6f431d061e9edf629df8d727b89c /src/glsl/nir/nir_opcodes.c | |
| parent | 2c7da78805175f36879111306ac37c12d33bf65b (diff) | |
nir/opcodes: Add algebraic properties metadata
This commit adds some algebraic properties to the metadata of each opcode
in NIR. In particular, you now know, just from the metadata, if a given
opcode is commutative or associative. This will be useful for algebraic
transformation passes that want to be able to match a + b as well as b + a
in one go.
v2: Make algebraic properties all caps. This was more consistent with the
intrinsics flags and seems better for flags in general.
Also, the enums are now declared with (1 << n) rather then hex values.
v3: fmin and fmax technically aren't commutative or associative. Things
get funny when one of the arguments is a NaN.
Reviewed-by: Connor Abbott <[email protected]>
Diffstat (limited to 'src/glsl/nir/nir_opcodes.c')
| -rw-r--r-- | src/glsl/nir/nir_opcodes.c | 3 |
1 files changed, 2 insertions, 1 deletions
diff --git a/src/glsl/nir/nir_opcodes.c b/src/glsl/nir/nir_opcodes.c index a62f989eed7..3a86641bc8a 100644 --- a/src/glsl/nir/nir_opcodes.c +++ b/src/glsl/nir/nir_opcodes.c @@ -28,7 +28,7 @@ #include "nir.h" #define OPCODE(_name, _num_inputs, _per_component, _output_size, _output_type, \ - _input_sizes, _input_types) \ + _input_sizes, _input_types, _algebraic_props) \ { \ .name = #_name, \ .num_inputs = _num_inputs, \ @@ -37,6 +37,7 @@ .output_type = _output_type, \ .input_sizes = _input_sizes, \ .input_types = _input_types, \ + .algebraic_properties = _algebraic_props, \ }, #define LAST_OPCODE(name) |