[Mesa-dev] [PATCH 1/2] nir: add option to lower slt/sge/seq/sne

Rob Clark robdclark at gmail.com
Thu Apr 2 11:35:18 PDT 2015


From: Rob Clark <robclark at freedesktop.org>

In freedreno these get implemented as the matching f* instruction plus a
u2f to convert the result to float 1.0/0.0.  But less lines of code to
just let nir_opt_algebraic handle this for us, plus opens up some small
window for other opt passes to improve (ie. if some shader ended up with
both a flt and slt with same src args, for example).

v2: use b2f rather than u2f

Signed-off-by: Rob Clark <robclark at freedesktop.org>
Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>
Reviewed-by: Matt Turner <mattst88 at gmail.com>
---
 src/glsl/nir/nir.h                | 3 +++
 src/glsl/nir/nir_opt_algebraic.py | 4 ++++
 2 files changed, 7 insertions(+)

diff --git a/src/glsl/nir/nir.h b/src/glsl/nir/nir.h
index 6e2aa97..3d532b7 100644
--- a/src/glsl/nir/nir.h
+++ b/src/glsl/nir/nir.h
@@ -1372,6 +1372,9 @@ typedef struct nir_shader_compiler_options {
    /** lowers fneg and ineg to fsub and isub. */
    bool lower_negate;
 
+   /* lower {slt,sge,seq,sne} to {flt,fge,feq,fne} + b2f: */
+   bool lower_scmp;
+
    /**
     * Does the driver support real 32-bit integers?  (Otherwise, integers
     * are simulated by floats.)
diff --git a/src/glsl/nir/nir_opt_algebraic.py b/src/glsl/nir/nir_opt_algebraic.py
index 60d1160..a8c1745 100644
--- a/src/glsl/nir/nir_opt_algebraic.py
+++ b/src/glsl/nir/nir_opt_algebraic.py
@@ -96,6 +96,10 @@ optimizations = [
    (('fmin', ('fmax', ('fmin', ('fmax', a, 0.0), 1.0), 0.0), 1.0), ('fmin', ('fmax', a, 0.0), 1.0)),
    (('ior', ('flt', a, b), ('flt', a, c)), ('flt', a, ('fmax', b, c))),
    (('ior', ('fge', a, b), ('fge', a, c)), ('fge', a, ('fmin', b, c))),
+   (('slt', a, b), ('b2f', ('flt', a, b)), 'options->lower_scmp'),
+   (('sge', a, b), ('b2f', ('fge', a, b)), 'options->lower_scmp'),
+   (('seq', a, b), ('b2f', ('feq', a, b)), 'options->lower_scmp'),
+   (('sne', a, b), ('b2f', ('fne', a, b)), 'options->lower_scmp'),
    # Emulating booleans
    (('fmul', ('b2f', a), ('b2f', b)), ('b2f', ('iand', a, b))),
    (('fsat', ('fadd', ('b2f', a), ('b2f', b))), ('b2f', ('ior', a, b))),
-- 
2.1.0



More information about the mesa-dev mailing list