nir/opt_algebraic: optimize f2i32(fround_even(x)) to f2i32_rtne(x)

Add late optimization to fuse f2i32 and fround_even operations into a
single f2i32_rtne instruction when the intermediate fround_even result
is only used once. This eliminates redundant rounding since f2i32_rtne
performs round-to-nearest-even conversion directly.

Signed-off-by: Christian Gmeiner <cgmeiner@igalia.com>
Tested-by: Simon Perretta <simon.perretta@imgtec.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37426>

src/compiler/nir/nir_opt_algebraic.py

@@ -4024,6 +4024,10 @@ for s in [16, 32, 64]:
         (('bcsel@{}'.format(s), ('ine', 0, 'a@{}'.format(s)), 'b@{}'.format(s), 'c@{}'.format(s)), ('icsel_eqz', a, c, b), "options->has_icsel_eqz{} && !options->no_integers".format(s)),
     ])
 
+late_optimizations += [
+    (('f2i32', ('fround_even(is_used_once)', 'a@32')), ('f2i32_rtne', a), 'options->has_f2i32_rtne'),
+]
+
 distribute_src_mods = [
    # Try to remove some spurious negations rather than pushing them down.
    (('fmul', ('fneg', a), ('fneg', b)), ('fmul', a, b)),

src/compiler/nir/nir_shader_compiler_options.h

@@ -646,6 +646,9 @@ typedef struct nir_shader_compiler_options {
    /** Backend supports load_global_bounded intrinsics. */
    bool has_load_global_bounded;
 
+   /** Backend supports f2i32_rtne opcode. */
+   bool has_f2i32_rtne;
+
    /**
     * Is this the Intel vec4 backend?
     *