[Mesa-dev] [PATCH 3/3] nir/algebraic: Add lowering for ldexp

[Jason Ekstrand]

The algorithm used is different from both the naieve suggestion from the
GLSL spec and the one used in GLSL IR today.  Unfortunately, the GLSL IR
implementation doesn't handle some of the corner cases correctly and
neither does a naieve f * 2.0^exp implementation.  Assuming that hardware
does the sane thing when multiplying by an exact power of two, this
implementation should actually be correct.  It does pass all of the Vulkan
CTS tests (which a simple port of the GLSL IR implementation does not).

Cc: Matt Turner <mattst88 at gmail.com>
---
 src/compiler/nir/nir_opt_algebraic.py | 31 +++++++++++++++++++++++++++++++
 1 file changed, 31 insertions(+)

diff --git a/src/compiler/nir/nir_opt_algebraic.py b/src/compiler/nir/nir_opt_algebraic.py
index 2749b06..7d1937e 100644
--- a/src/compiler/nir/nir_opt_algebraic.py
+++ b/src/compiler/nir/nir_opt_algebraic.py
@@ -371,6 +371,37 @@ optimizations = [
      'options->lower_unpack_snorm_4x8'),
 ]
 
+def fexp2i(exp):
+   # We assume that exp is already in the range [-126, 127].
+   return ('ishl', ('iadd', exp, 127), 23)
+
+def ldexp32(f, exp):
+   # First, we clamp exp to a reasonable range.  The maximum possible range
+   # for a normal exponent is [-126, 127] and, throwing in denormals, you get
+   # a maximum range of [-149, 127].  This means that we can potentially have
+   # a swing of +-276.  If you start with FLT_MAX, you actually have to do
+   # ldexp(FLT_MAX, -278) to get it to flush all the way to zero.  The GLSL
+   # spec, on the other hand, only requires that we handle an exponent value
+   # in the range [-127, 128].  This implementation is *mostly* correct; it
+   # handles a range on exp of [-252, 254] which allows you to create any
+   # value (including denorms if the hardware supports it) and to adjust the
+   # exponent of any normal value to anything you want.
+   exp = ('imin', ('imax', exp, -252), 254)
+
+   # Now we compute two powers of 2, one for exp/2 and one for exp-exp/2.
+   # (We use ishr which isn't the same for -1, but the -1 case still works
+   # since we use exp-exp/2 as the second exponent.)  While the spec
+   # technically defines ldexp as f * 2.0^exp, simply multiplying once doesn't
+   # work with denormals and doesn't allow for the full swing in exponents
+   # that you can get with normalized values.  Instead, we create two powers
+   # of two and multiply by them each in turn.  That way the effective range
+   # of our exponent is doubled.
+   pow2_1 = fexp2i(('ishr', exp, 1))
+   pow2_2 = fexp2i(('isub', exp, ('ishr', exp, 1)))
+   return ('fmul', ('fmul', f, pow2_1), pow2_2)
+
+optimizations += [(('ldexp', 'x', 'exp'), ldexp32('x', 'exp'))]
+
 # Unreal Engine 4 demo applications open-codes bitfieldReverse()
 def bitfield_reverse(u):
     step1 = ('ior', ('ishl', u, 16), ('ushr', u, 16))
-- 
2.5.0.400.gff86faf