From 5b2fb138bc5ff68af27d8435cbc01f83a09ee078 Mon Sep 17 00:00:00 2001
From: Eric Anholt <eric@anholt.net>
Date: Tue, 18 Aug 2015 22:38:34 -0700
Subject: [PATCH] nir: Add opcodes for saturated vector math.

This corresponds to instructions used on vc4 for its blending inside of
shaders.  I've seen these opcodes on other architectures before, but I
think it's the first time these are needed in Mesa.

v2: Rename to 'u' instead of 'i', since they're all 'u'norm (from review
    by jekstrand)
---
 src/glsl/nir/nir_opcodes.py       | 45 +++++++++++++++++++++++++++++++
 src/glsl/nir/nir_opt_algebraic.py |  6 +++++
 2 files changed, 51 insertions(+)

diff --git a/src/glsl/nir/nir_opcodes.py b/src/glsl/nir/nir_opcodes.py
index f2d584fe484..3c0f1da94af 100644
--- a/src/glsl/nir/nir_opcodes.py
+++ b/src/glsl/nir/nir_opcodes.py
@@ -468,6 +468,51 @@ binop("fmax", tfloat, "", "fmaxf(src0, src1)")
 binop("imax", tint, commutative + associative, "src1 > src0 ? src1 : src0")
 binop("umax", tunsigned, commutative + associative, "src1 > src0 ? src1 : src0")
 
+# Saturated vector add for 4 8bit ints.
+binop("usadd_4x8", tint, commutative + associative, """
+dst = 0;
+for (int i = 0; i < 32; i += 8) {
+   dst |= MIN2(((src0 >> i) & 0xff) + ((src1 >> i) & 0xff), 0xff) << i;
+}
+""")
+
+# Saturated vector subtract for 4 8bit ints.
+binop("ussub_4x8", tint, "", """
+dst = 0;
+for (int i = 0; i < 32; i += 8) {
+   int src0_chan = (src0 >> i) & 0xff;
+   int src1_chan = (src1 >> i) & 0xff;
+   if (src0_chan > src1_chan)
+      dst |= (src0_chan - src1_chan) << i;
+}
+""")
+
+# vector min for 4 8bit ints.
+binop("umin_4x8", tint, commutative + associative, """
+dst = 0;
+for (int i = 0; i < 32; i += 8) {
+   dst |= MIN2((src0 >> i) & 0xff, (src1 >> i) & 0xff) << i;
+}
+""")
+
+# vector max for 4 8bit ints.
+binop("umax_4x8", tint, commutative + associative, """
+dst = 0;
+for (int i = 0; i < 32; i += 8) {
+   dst |= MAX2((src0 >> i) & 0xff, (src1 >> i) & 0xff) << i;
+}
+""")
+
+# unorm multiply: (a * b) / 255.
+binop("umul_unorm_4x8", tint, commutative + associative, """
+dst = 0;
+for (int i = 0; i < 32; i += 8) {
+   int src0_chan = (src0 >> i) & 0xff;
+   int src1_chan = (src1 >> i) & 0xff;
+   dst |= ((src0_chan * src1_chan) / 255) << i;
+}
+""")
+
 binop("fpow", tfloat, "", "powf(src0, src1)")
 
 binop_horiz("pack_half_2x16_split", 1, tunsigned, 1, tfloat, 1, tfloat,
diff --git a/src/glsl/nir/nir_opt_algebraic.py b/src/glsl/nir/nir_opt_algebraic.py
index 585e5e0ae98..6aa8b1f6cae 100644
--- a/src/glsl/nir/nir_opt_algebraic.py
+++ b/src/glsl/nir/nir_opt_algebraic.py
@@ -56,12 +56,16 @@ optimizations = [
    (('iabs', ('ineg', a)), ('iabs', a)),
    (('fadd', a, 0.0), a),
    (('iadd', a, 0), a),
+   (('usadd_4x8', a, 0), a),
+   (('usadd_4x8', a, ~0), ~0),
    (('fadd', ('fmul', a, b), ('fmul', a, c)), ('fmul', a, ('fadd', b, c))),
    (('iadd', ('imul', a, b), ('imul', a, c)), ('imul', a, ('iadd', b, c))),
    (('fadd', ('fneg', a), a), 0.0),
    (('iadd', ('ineg', a), a), 0),
    (('fmul', a, 0.0), 0.0),
    (('imul', a, 0), 0),
+   (('umul_unorm_4x8', a, 0), 0),
+   (('umul_unorm_4x8', a, ~0), a),
    (('fmul', a, 1.0), a),
    (('imul', a, 1), a),
    (('fmul', a, -1.0), ('fneg', a)),
@@ -201,6 +205,8 @@ optimizations = [
    # Subtracts
    (('fsub', a, ('fsub', 0.0, b)), ('fadd', a, b)),
    (('isub', a, ('isub', 0, b)), ('iadd', a, b)),
+   (('ussub_4x8', a, 0), a),
+   (('ussub_4x8', a, ~0), 0),
    (('fsub', a, b), ('fadd', a, ('fneg', b)), 'options->lower_sub'),
    (('isub', a, b), ('iadd', a, ('ineg', b)), 'options->lower_sub'),
    (('fneg', a), ('fsub', 0.0, a), 'options->lower_negate'),
-- 
2.30.2