nir: Add optimization to use ROR/ROL instructions

v2: 1) Add more optimization rules for ROL/ROR (Matt Turner)
    2) Add lowering rules for ROL/ROR (Matt Turner)

Signed-off-by: Sagar Ghuge <sagar.ghuge@intel.com>
Reviewed-by: Matt Turner <mattst88@gmail.com>

diff --git a/src/compiler/nir/nir.h b/src/compiler/nir/nir.h
index df6dc5ba904b69d289926f1696c0f4a68b4b8bec..56b5c996af1efdef4453422d4d07237b53a474ba 100644 (file)
--- a/src/compiler/nir/nir.h
+++ b/src/compiler/nir/nir.h
@@ -2423,6 +2423,9 @@ typedef struct nir_shader_compiler_options {
    /* Lowers when 32x32->64 bit multiplication is not supported */
    bool lower_mul_2x32_64;
 
+   /* Lowers when rotate instruction is not supported */
+   bool lower_rotate;
+
    unsigned max_unroll_iterations;
 
    nir_lower_int64_options lower_int64_options;

diff --git a/src/compiler/nir/nir_opt_algebraic.py b/src/compiler/nir/nir_opt_algebraic.py
index a8f0a83741b5cc3720408d7f1566ce232598c62a..771a824743016ab8f821396b65090adb9afa490b 100644 (file)
--- a/src/compiler/nir/nir_opt_algebraic.py
+++ b/src/compiler/nir/nir_opt_algebraic.py
@@ -585,6 +585,18 @@ optimizations = [
    (('ushr', a, 0), a),
    (('iand', 0xff, ('ushr@32', a, 24)), ('ushr', a, 24)),
    (('iand', 0xffff, ('ushr@32', a, 16)), ('ushr', a, 16)),
+   (('ior', ('ishl@16', a, b), ('ushr@16', a, ('iadd', 16, ('ineg', b)))), ('urol', a, b), '!options->lower_rotate'),
+   (('ior', ('ishl@16', a, b), ('ushr@16', a, ('isub', 16, b))), ('urol', a, b), '!options->lower_rotate'),
+   (('ior', ('ishl@32', a, b), ('ushr@32', a, ('iadd', 32, ('ineg', b)))), ('urol', a, b), '!options->lower_rotate'),
+   (('ior', ('ishl@32', a, b), ('ushr@32', a, ('isub', 32, b))), ('urol', a, b), '!options->lower_rotate'),
+   (('ior', ('ushr@16', a, b), ('ishl@16', a, ('iadd', 16, ('ineg', b)))), ('uror', a, b), '!options->lower_rotate'),
+   (('ior', ('ushr@16', a, b), ('ishl@16', a, ('isub', 16, b))), ('uror', a, b), '!options->lower_rotate'),
+   (('ior', ('ushr@32', a, b), ('ishl@32', a, ('iadd', 32, ('ineg', b)))), ('uror', a, b), '!options->lower_rotate'),
+   (('ior', ('ushr@32', a, b), ('ishl@32', a, ('isub', 32, b))), ('uror', a, b), '!options->lower_rotate'),
+   (('urol@16', a, b), ('ior', ('ishl', a, b), ('ushr', a, ('isub', 16, b))), 'options->lower_rotate'),
+   (('urol@32', a, b), ('ior', ('ishl', a, b), ('ushr', a, ('isub', 32, b))), 'options->lower_rotate'),
+   (('uror@16', a, b), ('ior', ('ushr', a, b), ('ishl', a, ('isub', 16, b))), 'options->lower_rotate'),
+   (('uror@32', a, b), ('ior', ('ushr', a, b), ('ishl', a, ('isub', 32, b))), 'options->lower_rotate'),
    # Exponential/logarithmic identities
    (('~fexp2', ('flog2', a)), a), # 2^lg2(a) = a
    (('~flog2', ('fexp2', a)), a), # lg2(2^a) = a