LangRef: Clarify behavior of minnum/maxnum

[arsenm]

Summary

Make it clearer minnum(+0, +0) cannot return -0. Also remove
a note about the result always being quiet which is directly
contradicted by the following paragraph.

[foad]

llvm/docs/LangRef.rst

14905

returns either one...

[nikic]

Also remove a note about the result always being quiet which is directly contradicted by the following paragraph.

Where is the contradiction?

[arsenm]

In D148792#4283344, @nikic wrote:

Also remove a note about the result always being quiet which is directly contradicted by the following paragraph.

Where is the contradiction?

The entire following paragraph. "Unlike the IEEE-754 2008 behavior, this does not distinguish between
signaling and quiet NaN inputs. If a target's implementation follows
the standard and returns a quiet NaN if either input is a signaling
NaN, the intrinsic lowering is responsible for quieting the inputs to
correctly return the non-NaN input (e.g. by using the equivalent of
`llvm.canonicalize`)."

This is not returning an "always quiet nan" and matches the libm behavior

[nikic]

In D148792#4283345, @arsenm wrote:

In D148792#4283344, @nikic wrote:

Also remove a note about the result always being quiet which is directly contradicted by the following paragraph.

Where is the contradiction?

The entire following paragraph. "Unlike the IEEE-754 2008 behavior, this does not distinguish between
signaling and quiet NaN inputs. If a target's implementation follows
the standard and returns a quiet NaN if either input is a signaling
NaN, the intrinsic lowering is responsible for quieting the inputs to
correctly return the non-NaN input (e.g. by using the equivalent of
`llvm.canonicalize`)."

This is not returning an "always quiet nan" and matches the libm behavior

Not sure I follow. If a NaN is returned, the NaN will always be quiet. The paragraph only clarifies that even if one input is SNaN a NaN will only be returned if the other input is NaN as well.

[arsenm]

In D148792#4283350, @nikic wrote:

Not sure I follow. If a NaN is returned, the NaN will always be quiet.

This is distinguishing quiet and signaling nans. Not only is this inconsistent with the IR policy for unconstrained FP operations, it's not the behavior of libm fmin/fmax.

These intrinsics have the fmin/fmax behavior and were incorrectly named. The first sentence describes the IEEE754 behavior, and the second paragraph elaborates that they do not.

[nikic]

In D148792#4283393, @arsenm wrote:

In D148792#4283350, @nikic wrote:

Not sure I follow. If a NaN is returned, the NaN will always be quiet.

This is distinguishing quiet and signaling nans. Not only is this inconsistent with the IR policy for unconstrained FP operations, it's not the behavior of libm fmin/fmax.

These intrinsics have the fmin/fmax behavior and were incorrectly named. The first sentence describes the IEEE754 behavior, and the second paragraph elaborates that they do not.

I see, thanks for the explanation!

[arsenm]

ping

[foad]

llvm/docs/LangRef.rst

14905

ping

[arsenm]

llvm/docs/LangRef.rst

14905

Apparently I fixed this for max only

[arsenm]

fa5e6697c4094c5d40578e7cc07dfebd3f448013

llvm/docs/LangRef.rst

 
 Semantics:
 """"""""""
 
 Follows the IEEE-754 semantics for minNum, except for handling of
 signaling NaNs. This match's the behavior of libm's fmin.
 
 If either operand is a NaN, returns the other non-NaN operand. Returns
-NaN only if both operands are NaN. The returned NaN is always
-quiet. If the operands compare equal, returns a value that compares
-equal to both operands. This means that fmin(+/-0.0, +/-0.0) could
-return either -0.0 or 0.0.
+NaN only if both operands are NaN. If the operands compare equal,
+either one of the operands. For example, this means that fmin(+0.0,
+-0.0) could return either operand.
 
 Unlike the IEEE-754 2008 behavior, this does not distinguish between
 signaling and quiet NaN inputs. If a target's implementation follows
 the standard and returns a quiet NaN if either input is a signaling
 NaN, the intrinsic lowering is responsible for quieting the inputs to
 correctly return the non-NaN input (e.g. by using the equivalent of
 ``llvm.canonicalize``).
 

 type.
 
 Semantics:
 """"""""""
 Follows the IEEE-754 semantics for maxNum except for the handling of
 signaling NaNs. This matches the behavior of libm's fmax.
 
 If either operand is a NaN, returns the other non-NaN operand. Returns
-NaN only if both operands are NaN. The returned NaN is always
-quiet. If the operands compare equal, returns a value that compares
-equal to both operands. This means that fmax(+/-0.0, +/-0.0) could
-return either -0.0 or 0.0.
+NaN only if both operands are NaN. If the operands compare equal,
+returns either one of the operands. For example, this means that
+fmax(+0.0, -0.0) could return either -0.0 or 0.0.
 
 Unlike the IEEE-754 2008 behavior, this does not distinguish between
 signaling and quiet NaN inputs. If a target's implementation follows
 the standard and returns a quiet NaN if either input is a signaling
 NaN, the intrinsic lowering is responsible for quieting the inputs to
 correctly return the non-NaN input (e.g. by using the equivalent of
 ``llvm.canonicalize``).