diff options
| author | Karol Herbst <[email protected]> | 2018-07-12 15:02:27 +0200 |
|---|---|---|
| committer | Karol Herbst <[email protected]> | 2019-03-05 22:28:29 +0100 |
| commit | 272e927d0e9fed6e791d706ff5d895b6c2036fc0 (patch) | |
| tree | 2bbe3b7fbef9275137bd4669cb86b72181094379 /src/compiler/nir/nir_opt_algebraic.py | |
| parent | d0b47ec4df0eafe4f4afddc2a0594b392c27f426 (diff) | |
nir/spirv: initial handling of OpenCL.std extension opcodes
Not complete, mostly just adding things as I encounter them in CTS. But
not getting far enough yet to hit most of the OpenCL.std instructions.
Anyway, this is better than nothing and covers the most common builtins.
v2: add hadd proof from Jason
move some of the lowering into opt_algebraic and create new nir opcodes
simplify nextafter lowering
fix normalize lowering for inf
rework upsample to use nir_pack_bits
add missing files to build systems
v3: split lines of iadd/sub_sat expressions
Signed-off-by: Karol Herbst <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>
Diffstat (limited to 'src/compiler/nir/nir_opt_algebraic.py')
| -rw-r--r-- | src/compiler/nir/nir_opt_algebraic.py | 21 |
1 files changed, 21 insertions, 0 deletions
diff --git a/src/compiler/nir/nir_opt_algebraic.py b/src/compiler/nir/nir_opt_algebraic.py index e8191b72271..5b2e7ee2405 100644 --- a/src/compiler/nir/nir_opt_algebraic.py +++ b/src/compiler/nir/nir_opt_algebraic.py @@ -678,6 +678,12 @@ optimizations = [ ('bcsel', ('ilt', 31, 'bits'), 'insert', ('bfi', ('bfm', 'bits', 'offset'), 'insert', 'base')), 'options->lower_bitfield_insert'), + (('ihadd', a, b), ('iadd', ('iand', a, b), ('ishr', ('ixor', a, b), 1)), 'options->lower_hadd'), + (('uhadd', a, b), ('iadd', ('iand', a, b), ('ushr', ('ixor', a, b), 1)), 'options->lower_hadd'), + (('irhadd', a, b), ('isub', ('ior', a, b), ('ishr', ('ixor', a, b), 1)), 'options->lower_hadd'), + (('urhadd', a, b), ('isub', ('ior', a, b), ('ushr', ('ixor', a, b), 1)), 'options->lower_hadd'), + (('uadd_sat', a, b), ('bcsel', ('ult', ('iadd', a, b), a), -1, ('iadd', a, b)), 'options->lower_add_sat'), + (('usub_sat', a, b), ('bcsel', ('ult', a, b), 0, ('isub', a, b)), 'options->lower_add_sat'), # Alternative lowering that doesn't rely on bfi. (('bitfield_insert', 'base', 'insert', 'offset', 'bits'), @@ -794,6 +800,21 @@ optimizations = [ (('isign', a), ('imin', ('imax', a, -1), 1), 'options->lower_isign'), ] +# bit_size dependent lowerings +for bit_size in [8, 16, 32, 64]: + # convenience constants + intmax = (1 << (bit_size - 1)) - 1 + intmin = 1 << (bit_size - 1) + + optimizations += [ + (('iadd_sat@' + str(bit_size), a, b), + ('bcsel', ('ige', b, 1), ('bcsel', ('ilt', ('iadd', a, b), a), intmax, ('iadd', a, b)), + ('bcsel', ('ilt', a, ('iadd', a, b)), intmin, ('iadd', a, b))), 'options->lower_add_sat'), + (('isub_sat@' + str(bit_size), a, b), + ('bcsel', ('ilt', b, 0), ('bcsel', ('ilt', ('isub', a, b), a), intmax, ('isub', a, b)), + ('bcsel', ('ilt', a, ('isub', a, b)), intmin, ('isub', a, b))), 'options->lower_add_sat'), + ] + invert = OrderedDict([('feq', 'fne'), ('fne', 'feq'), ('fge', 'flt'), ('flt', 'fge')]) for left, right in itertools.combinations_with_replacement(invert.keys(), 2): |