[llvm] LangRef: Clarify llvm.minnum and llvm.maxnum about sNaN and signed zero (PR #112852)
YunQiang Su via llvm-commits
llvm-commits at lists.llvm.org
Tue Dec 3 18:34:07 PST 2024
https://github.com/wzssyqa updated https://github.com/llvm/llvm-project/pull/112852
>From 991056836991efdfe4d0ccf7bdd9d77321c11832 Mon Sep 17 00:00:00 2001
From: YunQiang Su <yunqiang at isrc.iscas.ac.cn>
Date: Fri, 18 Oct 2024 16:16:43 +0800
Subject: [PATCH 01/11] LangRef: Clarify llvm.minnum and llvm.maxnum about sNaN
The documents claims that it ignores sNaN, while in the code it may
be different.
- as the finally callback, it use libc call fmin(3)/fmax(3).
while C23 clarify that fmin(3)/fmax(3) should return NaN for sNaN vs NUM.
- on some architectures, such as aarch64, it converts to `fmaxnm`,
which returns qNaN for sNaN vs NUM.
- on RISC-V (SPEC 2019+), it converts to `fmax`, which returns NUM
for sNaN vs NUM.
Since we have introduced llvm.minimumnum and llvm.maximumnum, which
follow IEEE 754-2019's minimumNumber/maximumNumber.
So, it's time for use to clarify llvm.minnum and llvm.maxnum.
Let's define it to the libc's defination.
---
llvm/docs/LangRef.rst | 75 ++++++++++++++++++-------------------------
1 file changed, 31 insertions(+), 44 deletions(-)
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index abfd2fdfb9de71..57df13e464a32e 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -16574,21 +16574,13 @@ type.
Semantics:
""""""""""
+Follows the IEEE754 2008 semantics for minNum, except for handling of
++0.0 vs -0.0. This matches the behavior of libm's fmin.
-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. If the operands compare equal,
-returns either one of the operands. For example, this means that
-fmin(+0.0, -0.0) returns 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``).
+If either operand is a qNaN, returns the other non-NaN operand. Returns
+NaN only if both operands are NaN or either operand is sNaN.
+If the operands compare equal, returns either one of the operands.
+For example, this means that fmin(+0.0, -0.0) returns either operand.
.. _i_maxnum:
@@ -16625,20 +16617,13 @@ type.
Semantics:
""""""""""
-Follows the IEEE-754 semantics for maxNum except for the handling of
-signaling NaNs. This matches the behavior of libm's fmax.
+Follows the IEEE754 2008 semantics for maxNum, except for handling of
++0.0 vs -0.0. 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. If the operands compare equal,
-returns either one of the operands. For example, this means that
-fmax(+0.0, -0.0) returns 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``).
+NaN only if both operands are NaN or either operand is sNaN.
+If the operands compare equal, returns either one of the operands.
+For example, this means that fmin(+0.0, -0.0) returns either operand.
.. _i_minimum:
@@ -19517,12 +19502,12 @@ The '``llvm.vector.reduce.fmax.*``' intrinsics do a floating-point
matches the element-type of the vector input.
This instruction has the same comparison semantics as the '``llvm.maxnum.*``'
-intrinsic. That is, the result will always be a number unless all elements of
-the vector are NaN. For a vector with maximum element magnitude 0.0 and
-containing both +0.0 and -0.0 elements, the sign of the result is unspecified.
+intrinsic. If the intrinsic call has the ``nnan`` fast-math flag, then the
+operation can assume that NaNs are not present in the input vector.
-If the intrinsic call has the ``nnan`` fast-math flag, then the operation can
-assume that NaNs are not present in the input vector.
+It is deprecated, since the different order of inputs may produce different
+outputs, and it is hard to optimize with Vector or SIMD extensions.
+Use '``llvm.vector.reduce.fmaximum``' or '``llvm.vector.reduce.fmaximumnum``' instead.
Arguments:
""""""""""
@@ -19550,12 +19535,12 @@ The '``llvm.vector.reduce.fmin.*``' intrinsics do a floating-point
matches the element-type of the vector input.
This instruction has the same comparison semantics as the '``llvm.minnum.*``'
-intrinsic. That is, the result will always be a number unless all elements of
-the vector are NaN. For a vector with minimum element magnitude 0.0 and
-containing both +0.0 and -0.0 elements, the sign of the result is unspecified.
+intrinsic. If the intrinsic call has the ``nnan`` fast-math flag, then the
+operation can assume that NaNs are not present in the input vector.
-If the intrinsic call has the ``nnan`` fast-math flag, then the operation can
-assume that NaNs are not present in the input vector.
+It is deprecated, since the different order of inputs may produce different
+outputs, and it is hard to optimize with Vector or SIMD extensions.
+Use '``llvm.vector.reduce.fminimum``' or '``llvm.vector.reduce.fminimumnum``' instead.
Arguments:
""""""""""
@@ -23169,13 +23154,14 @@ result type. If only ``nnan`` is set then the neutral value is ``-Infinity``.
This instruction has the same comparison semantics as the
:ref:`llvm.vector.reduce.fmax <int_vector_reduce_fmax>` intrinsic (and thus the
-'``llvm.maxnum.*``' intrinsic). That is, the result will always be a number
-unless all elements of the vector and the starting value are ``NaN``. For a
-vector with maximum element magnitude ``0.0`` and containing both ``+0.0`` and
-``-0.0`` elements, the sign of the result is unspecified.
+'``llvm.maxnum.*``' intrinsic).
To ignore the start value, the neutral value can be used.
+It is deprecated, since the different order of inputs may produce different
+outputs, and it is hard to optimize with Vector or SIMD extensions.
+Use '``llvm.vp.vector.reduce.fmaximum``' or '``llvm.vp.vector.reduce.fmaximumnum``' instead.
+
Examples:
"""""""""
@@ -23239,13 +23225,14 @@ result type. If only ``nnan`` is set then the neutral value is ``+Infinity``.
This instruction has the same comparison semantics as the
:ref:`llvm.vector.reduce.fmin <int_vector_reduce_fmin>` intrinsic (and thus the
-'``llvm.minnum.*``' intrinsic). That is, the result will always be a number
-unless all elements of the vector and the starting value are ``NaN``. For a
-vector with maximum element magnitude ``0.0`` and containing both ``+0.0`` and
-``-0.0`` elements, the sign of the result is unspecified.
+'``llvm.minnum.*``' intrinsic).
To ignore the start value, the neutral value can be used.
+It is deprecated, since the different order of inputs may produce different
+outputs, and it is hard to optimize with Vector or SIMD extensions.
+Use '``llvm.vp.vector.reduce.fminimum``' or '``llvm.vp.vector.reduce.fminimumnum``' instead.
+
Examples:
"""""""""
>From 7f8d4e7b3f01f33d315d3e5cc9fa1cdd5bad6ba9 Mon Sep 17 00:00:00 2001
From: YunQiang Su <wzssyqa at gmail.com>
Date: Sat, 19 Oct 2024 01:33:13 +0800
Subject: [PATCH 02/11] minNum doesn't care about +0 vs -0
---
llvm/docs/LangRef.rst | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index 57df13e464a32e..51d96a797bfa13 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -16574,8 +16574,8 @@ type.
Semantics:
""""""""""
-Follows the IEEE754 2008 semantics for minNum, except for handling of
-+0.0 vs -0.0. This matches the behavior of libm's fmin.
+Follows the IEEE754 2008 semantics for minNum.
+This also matches the behavior of libm's fmin.
If either operand is a qNaN, returns the other non-NaN operand. Returns
NaN only if both operands are NaN or either operand is sNaN.
@@ -16617,8 +16617,8 @@ type.
Semantics:
""""""""""
-Follows the IEEE754 2008 semantics for maxNum, except for handling of
-+0.0 vs -0.0. This matches the behavior of libm's fmax.
+Follows the IEEE754 2008 semantics for maxNum.
+This also 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 or either operand is sNaN.
>From 656f88868883c0ef984d633a6afa78953bc597fe Mon Sep 17 00:00:00 2001
From: YunQiang Su <syq at debian.org>
Date: Sun, 20 Oct 2024 12:13:21 +0800
Subject: [PATCH 03/11] add history about libm
---
llvm/docs/LangRef.rst | 12 ++++++++++--
1 file changed, 10 insertions(+), 2 deletions(-)
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index 51d96a797bfa13..ad3a78d6d9fe12 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -16575,7 +16575,11 @@ type.
Semantics:
""""""""""
Follows the IEEE754 2008 semantics for minNum.
-This also matches the behavior of libm's fmin.
+This also matches the current (C23) behavior of libm's fmin.
+
+Historically, libc returns NUM for NUM vs (sNaN or qNaN), and may return
+sNaN for qNaN vs sNaN. Withe recent libc versions, libc follows IEEE754-2008:
+NUM vs sNaN -> qNaN; NUM vs qNaN -> NUM; qNaN vs sNaN -> qNaN; sNaN vs sNaN -> qNaN.
If either operand is a qNaN, returns the other non-NaN operand. Returns
NaN only if both operands are NaN or either operand is sNaN.
@@ -16618,7 +16622,11 @@ type.
Semantics:
""""""""""
Follows the IEEE754 2008 semantics for maxNum.
-This also matches the behavior of libm's fmax.
+This also matches the current (C23) behavior of libm's fmax.
+
+Historically, libc returns NUM for NUM vs (sNaN or qNaN), and may return
+sNaN for qNaN vs sNaN. Withe recent libc versions, libc follows IEEE754-2008:
+NUM vs sNaN -> qNaN; NUM vs qNaN -> NUM; qNaN vs sNaN -> qNaN; sNaN vs sNaN -> qNaN.
If either operand is a NaN, returns the other non-NaN operand. Returns
NaN only if both operands are NaN or either operand is sNaN.
>From a77539fcec5024a99b9fe297ecd6ea76d6b5a860 Mon Sep 17 00:00:00 2001
From: YunQiang Su <yunqiang at isrc.iscas.ac.cn>
Date: Mon, 21 Oct 2024 08:56:22 +0800
Subject: [PATCH 04/11] fmin requires +0>-0
---
llvm/docs/LangRef.rst | 24 ++++++++++++++++--------
1 file changed, 16 insertions(+), 8 deletions(-)
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index ad3a78d6d9fe12..2a0923c37da412 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -16530,7 +16530,7 @@ versions of the intrinsics respect the exception behavior.
- qNaN, invalid exception
* - ``+0.0 vs -0.0``
- - either one
+ - +0.0(max)/-0.0(min)
- +0.0(max)/-0.0(min)
- +0.0(max)/-0.0(min)
@@ -16574,8 +16574,14 @@ type.
Semantics:
""""""""""
-Follows the IEEE754 2008 semantics for minNum.
-This also matches the current (C23) behavior of libm's fmin.
+Follows the IEEE754 2008 semantics for minNum with +0.0>-0.0.
+This is more strict than current (C23) behavior of libm's fmin.
+Some applications like Clang, can call '``llvm.minnum.*``' with '``nsz``' attribute
+to archive the same behaivor of libm's fmin.
+
+For some architecturs, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, they have the
+strict same instructions; thus it is quite simple for these architectures.
+For other architectures, the custom or expand methods may provide '``nsz``' flavor.
Historically, libc returns NUM for NUM vs (sNaN or qNaN), and may return
sNaN for qNaN vs sNaN. Withe recent libc versions, libc follows IEEE754-2008:
@@ -16621,12 +16627,14 @@ type.
Semantics:
""""""""""
-Follows the IEEE754 2008 semantics for maxNum.
-This also matches the current (C23) behavior of libm's fmax.
+Follows the IEEE754 2008 semantics for maxNum with +0.0>-0.0.
+This is more strict than current (C23) behavior of libm's fmax.
+Some applications like Clang, can call '``llvm.maxnum.*``' with '``nsz``' attribute
+to archive the same behaivor of libm's fmax.
-Historically, libc returns NUM for NUM vs (sNaN or qNaN), and may return
-sNaN for qNaN vs sNaN. Withe recent libc versions, libc follows IEEE754-2008:
-NUM vs sNaN -> qNaN; NUM vs qNaN -> NUM; qNaN vs sNaN -> qNaN; sNaN vs sNaN -> qNaN.
+For some architecturs, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, they have the
+strict same instructions; thus it is quite simple for these architectures.
+For other architectures, the custom or expand methods may provide '``nsz``' flavor.
If either operand is a NaN, returns the other non-NaN operand. Returns
NaN only if both operands are NaN or either operand is sNaN.
>From 673efdcbe9aa552e7805926fd046a32a08d4839d Mon Sep 17 00:00:00 2001
From: YunQiang Su <yunqiang at isrc.iscas.ac.cn>
Date: Mon, 21 Oct 2024 09:11:35 +0800
Subject: [PATCH 05/11] some fix
---
llvm/docs/LangRef.rst | 12 +++++++-----
1 file changed, 7 insertions(+), 5 deletions(-)
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index 2a0923c37da412..d2f4f4f88fab22 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -16580,17 +16580,16 @@ Some applications like Clang, can call '``llvm.minnum.*``' with '``nsz``' attrib
to archive the same behaivor of libm's fmin.
For some architecturs, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, they have the
-strict same instructions; thus it is quite simple for these architectures.
+strictly same instructions; thus it is quite simple for these architectures.
For other architectures, the custom or expand methods may provide '``nsz``' flavor.
Historically, libc returns NUM for NUM vs (sNaN or qNaN), and may return
-sNaN for qNaN vs sNaN. Withe recent libc versions, libc follows IEEE754-2008:
+sNaN for qNaN vs sNaN. With the recent libc versions, libc follows IEEE754-2008:
NUM vs sNaN -> qNaN; NUM vs qNaN -> NUM; qNaN vs sNaN -> qNaN; sNaN vs sNaN -> qNaN.
If either operand is a qNaN, returns the other non-NaN operand. Returns
NaN only if both operands are NaN or either operand is sNaN.
If the operands compare equal, returns either one of the operands.
-For example, this means that fmin(+0.0, -0.0) returns either operand.
.. _i_maxnum:
@@ -16633,13 +16632,16 @@ Some applications like Clang, can call '``llvm.maxnum.*``' with '``nsz``' attrib
to archive the same behaivor of libm's fmax.
For some architecturs, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, they have the
-strict same instructions; thus it is quite simple for these architectures.
+strictly same instructions; thus it is quite simple for these architectures.
For other architectures, the custom or expand methods may provide '``nsz``' flavor.
+Historically, libc returns NUM for NUM vs (sNaN or qNaN), and may return
+sNaN for qNaN vs sNaN. With the recent libc versions, libc follows IEEE754-2008:
+NUM vs sNaN -> qNaN; NUM vs qNaN -> NUM; qNaN vs sNaN -> qNaN; sNaN vs sNaN -> qNaN.
+
If either operand is a NaN, returns the other non-NaN operand. Returns
NaN only if both operands are NaN or either operand is sNaN.
If the operands compare equal, returns either one of the operands.
-For example, this means that fmin(+0.0, -0.0) returns either operand.
.. _i_minimum:
>From 8a4d9a46e00a1d354ba357eb3b0798fe7b49ead1 Mon Sep 17 00:00:00 2001
From: YunQiang Su <yunqiang at isrc.iscas.ac.cn>
Date: Wed, 23 Oct 2024 09:25:07 +0800
Subject: [PATCH 06/11] add inconsistent note
---
llvm/docs/LangRef.rst | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index d2f4f4f88fab22..f2a46de27651d2 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -16587,6 +16587,11 @@ Historically, libc returns NUM for NUM vs (sNaN or qNaN), and may return
sNaN for qNaN vs sNaN. With the recent libc versions, libc follows IEEE754-2008:
NUM vs sNaN -> qNaN; NUM vs qNaN -> NUM; qNaN vs sNaN -> qNaN; sNaN vs sNaN -> qNaN.
+Note that that arithmetic on an sNaN doesn't consistently produce a qNaN,
+so arithmetic feeding into a minnum can produce inconsistent results.
+Such as `fmin(sNaN+0.0, 1.0)` can produce qNaN or 1.0 depending on whether `+0.0`
+is optimized out.
+
If either operand is a qNaN, returns the other non-NaN operand. Returns
NaN only if both operands are NaN or either operand is sNaN.
If the operands compare equal, returns either one of the operands.
@@ -16639,6 +16644,11 @@ Historically, libc returns NUM for NUM vs (sNaN or qNaN), and may return
sNaN for qNaN vs sNaN. With the recent libc versions, libc follows IEEE754-2008:
NUM vs sNaN -> qNaN; NUM vs qNaN -> NUM; qNaN vs sNaN -> qNaN; sNaN vs sNaN -> qNaN.
+Note that that arithmetic on an sNaN doesn't consistently produce a qNaN,
+so arithmetic feeding into a maxnum can produce inconsistent results.
+Such as `fmax(sNaN+0.0, 1.0)` can produce qNaN or 1.0 depending on whether `+0.0`
+is optimized out.
+
If either operand is a NaN, returns the other non-NaN operand. Returns
NaN only if both operands are NaN or either operand is sNaN.
If the operands compare equal, returns either one of the operands.
>From 676c95839e42c5d317c048295466dd2e9852ccdc Mon Sep 17 00:00:00 2001
From: YunQiang Su <yunqiang at isrc.iscas.ac.cn>
Date: Wed, 30 Oct 2024 19:23:17 +0800
Subject: [PATCH 07/11] do more clarify
---
llvm/docs/LangRef.rst | 46 ++++++++++++++++----------
llvm/include/llvm/CodeGen/ISDOpcodes.h | 17 +++++++---
2 files changed, 40 insertions(+), 23 deletions(-)
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index f2a46de27651d2..d9b4d2cfb61b3a 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -16574,12 +16574,21 @@ type.
Semantics:
""""""""""
-Follows the IEEE754 2008 semantics for minNum with +0.0>-0.0.
-This is more strict than current (C23) behavior of libm's fmin.
-Some applications like Clang, can call '``llvm.minnum.*``' with '``nsz``' attribute
-to archive the same behaivor of libm's fmin.
+Follows the IEEE-754 semantics for minNum, except that -0.0 < +0.0 for the purposes
+of this intrinsic. As for signaling NaNs, per the IEEE-754 semantics, if either operand
+is an sNaN, the result is always a qNaN. This matches the recommended behavior for the libm
+function fmin, although not all implementations have implemented these recommended behaviors.
-For some architecturs, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, they have the
+If either operand is a qNaN, returns the other non-NaN operand. Returns
+NaN only if both operands are NaN or either operand is sNaN.
+
+If the operands compare equal, returns either one of the operands.
+
+Returns -0.0 for +0.0 vs -0.0. libm doesn't require it, so that
+some applications like Clang, can call '``llvm.minnum.*``' with '``nsz``' attribute
+to archive the required behaivors of libm's fmin.
+
+Some architecturs, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, have the
strictly same instructions; thus it is quite simple for these architectures.
For other architectures, the custom or expand methods may provide '``nsz``' flavor.
@@ -16592,10 +16601,6 @@ so arithmetic feeding into a minnum can produce inconsistent results.
Such as `fmin(sNaN+0.0, 1.0)` can produce qNaN or 1.0 depending on whether `+0.0`
is optimized out.
-If either operand is a qNaN, returns the other non-NaN operand. Returns
-NaN only if both operands are NaN or either operand is sNaN.
-If the operands compare equal, returns either one of the operands.
-
.. _i_maxnum:
'``llvm.maxnum.*``' Intrinsic
@@ -16631,12 +16636,21 @@ type.
Semantics:
""""""""""
-Follows the IEEE754 2008 semantics for maxNum with +0.0>-0.0.
-This is more strict than current (C23) behavior of libm's fmax.
-Some applications like Clang, can call '``llvm.maxnum.*``' with '``nsz``' attribute
-to archive the same behaivor of libm's fmax.
+Follows the IEEE-754 semantics for minNum, except that -0.0 < +0.0 for the purposes
+of this intrinsic. As for signaling NaNs, per the IEEE-754 semantics, if either operand
+is an sNaN, the result is always a qNaN. This matches the recommended behavior for the libm
+function fmin, although not all implementations have implemented these recommended behaviors.
-For some architecturs, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, they have the
+If either operand is a qNaN, returns the other non-NaN operand. Returns
+NaN only if both operands are NaN or either operand is sNaN.
+
+If the operands compare equal, returns either one of the operands.
+
+Returns -0.0 for +0.0 vs -0.0. libm doesn't require it, so that
+some applications like Clang, can call '``llvm.minnum.*``' with '``nsz``' attribute
+to archive the required behaivors of libm's fmin.
+
+Some architecturs, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, have the
strictly same instructions; thus it is quite simple for these architectures.
For other architectures, the custom or expand methods may provide '``nsz``' flavor.
@@ -16649,10 +16663,6 @@ so arithmetic feeding into a maxnum can produce inconsistent results.
Such as `fmax(sNaN+0.0, 1.0)` can produce qNaN or 1.0 depending on whether `+0.0`
is optimized out.
-If either operand is a NaN, returns the other non-NaN operand. Returns
-NaN only if both operands are NaN or either operand is sNaN.
-If the operands compare equal, returns either one of the operands.
-
.. _i_minimum:
'``llvm.minimum.*``' Intrinsic
diff --git a/llvm/include/llvm/CodeGen/ISDOpcodes.h b/llvm/include/llvm/CodeGen/ISDOpcodes.h
index 0b6d155b6d161e..179cab137b5095 100644
--- a/llvm/include/llvm/CodeGen/ISDOpcodes.h
+++ b/llvm/include/llvm/CodeGen/ISDOpcodes.h
@@ -1021,13 +1021,20 @@ enum NodeType {
LRINT,
LLRINT,
- /// FMINNUM/FMAXNUM - Perform floating-point minimum or maximum on two
- /// values.
+ /// FMINNUM/FMAXNUM - Perform floating-point minimum maximum on two values,
+ /// following IEEE-754 definitions.
///
- /// In the case where a single input is a NaN (either signaling or quiet),
- /// the non-NaN input is returned.
+ /// If one input is a signaling NaN, returns a quiet NaN. This matches
+ /// IEEE-754 2008's minnum/maxnum behavior for signaling NaNs (which differs
+ /// from 2019).
+ ///
+ /// These treat -0 as ordered less than +0, matching the behavior of IEEE-754
+ /// 2019's minimumNumber/maximumNumber.
///
- /// The return value of (FMINNUM 0.0, -0.0) could be either 0.0 or -0.0.
+ /// Note that that arithmetic on an sNaN doesn't consistently produce a qNaN,
+ /// so arithmetic feeding into a minnum/maxnum can produce inconsistent
+ /// results. FMAXIMUN/FMINIMUM or FMAXIMUMNUM/FMINIMUMNUM may be better choice
+ /// for non-distinction of sNaN/qNaN handling.
FMINNUM,
FMAXNUM,
>From 4b03b8bc381c3e6720a36a05590a6889d5e72243 Mon Sep 17 00:00:00 2001
From: YunQiang Su <yunqiang at isrc.iscas.ac.cn>
Date: Fri, 1 Nov 2024 09:02:34 +0800
Subject: [PATCH 08/11] some improvement on wording
---
llvm/docs/LangRef.rst | 34 +++++++++++---------------
llvm/include/llvm/CodeGen/ISDOpcodes.h | 4 +--
2 files changed, 16 insertions(+), 22 deletions(-)
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index d9b4d2cfb61b3a..1d6b34fc4a12b0 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -16582,15 +16582,12 @@ function fmin, although not all implementations have implemented these recommend
If either operand is a qNaN, returns the other non-NaN operand. Returns
NaN only if both operands are NaN or either operand is sNaN.
-If the operands compare equal, returns either one of the operands.
+This behavior is more strict than the definition in C and IEEE 754, where either zero may be returned.
+To achieve the same permissiveness, the backend may implement the nsz attribute, and one may use the nsz
+attribute on the intrinsic call.
-Returns -0.0 for +0.0 vs -0.0. libm doesn't require it, so that
-some applications like Clang, can call '``llvm.minnum.*``' with '``nsz``' attribute
-to archive the required behaivors of libm's fmin.
-
-Some architecturs, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, have the
-strictly same instructions; thus it is quite simple for these architectures.
-For other architectures, the custom or expand methods may provide '``nsz``' flavor.
+Some architectures, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, have instructions that match
+these semantics exactly; thus it is quite simple for these architectures.
Historically, libc returns NUM for NUM vs (sNaN or qNaN), and may return
sNaN for qNaN vs sNaN. With the recent libc versions, libc follows IEEE754-2008:
@@ -16644,15 +16641,12 @@ function fmin, although not all implementations have implemented these recommend
If either operand is a qNaN, returns the other non-NaN operand. Returns
NaN only if both operands are NaN or either operand is sNaN.
-If the operands compare equal, returns either one of the operands.
-
-Returns -0.0 for +0.0 vs -0.0. libm doesn't require it, so that
-some applications like Clang, can call '``llvm.minnum.*``' with '``nsz``' attribute
-to archive the required behaivors of libm's fmin.
+This behavior is more strict than the definition in C and IEEE 754, where either zero may be returned.
+To achieve the same permissiveness, the backend may implement the nsz attribute, and one may use the nsz
+attribute on the intrinsic call.
-Some architecturs, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, have the
-strictly same instructions; thus it is quite simple for these architectures.
-For other architectures, the custom or expand methods may provide '``nsz``' flavor.
+Some architectures, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, have instructions that match
+these semantics exactly; thus it is quite simple for these architectures.
Historically, libc returns NUM for NUM vs (sNaN or qNaN), and may return
sNaN for qNaN vs sNaN. With the recent libc versions, libc follows IEEE754-2008:
@@ -19544,7 +19538,7 @@ intrinsic. If the intrinsic call has the ``nnan`` fast-math flag, then the
operation can assume that NaNs are not present in the input vector.
It is deprecated, since the different order of inputs may produce different
-outputs, and it is hard to optimize with Vector or SIMD extensions.
+outputs, and it is hard to optimize with vector or SIMD extensions.
Use '``llvm.vector.reduce.fmaximum``' or '``llvm.vector.reduce.fmaximumnum``' instead.
Arguments:
@@ -19577,7 +19571,7 @@ intrinsic. If the intrinsic call has the ``nnan`` fast-math flag, then the
operation can assume that NaNs are not present in the input vector.
It is deprecated, since the different order of inputs may produce different
-outputs, and it is hard to optimize with Vector or SIMD extensions.
+outputs, and it is hard to optimize with vector or SIMD extensions.
Use '``llvm.vector.reduce.fminimum``' or '``llvm.vector.reduce.fminimumnum``' instead.
Arguments:
@@ -23197,7 +23191,7 @@ This instruction has the same comparison semantics as the
To ignore the start value, the neutral value can be used.
It is deprecated, since the different order of inputs may produce different
-outputs, and it is hard to optimize with Vector or SIMD extensions.
+outputs, and it is hard to optimize with vector or SIMD extensions.
Use '``llvm.vp.vector.reduce.fmaximum``' or '``llvm.vp.vector.reduce.fmaximumnum``' instead.
Examples:
@@ -23268,7 +23262,7 @@ This instruction has the same comparison semantics as the
To ignore the start value, the neutral value can be used.
It is deprecated, since the different order of inputs may produce different
-outputs, and it is hard to optimize with Vector or SIMD extensions.
+outputs, and it is hard to optimize with vector or SIMD extensions.
Use '``llvm.vp.vector.reduce.fminimum``' or '``llvm.vp.vector.reduce.fminimumnum``' instead.
Examples:
diff --git a/llvm/include/llvm/CodeGen/ISDOpcodes.h b/llvm/include/llvm/CodeGen/ISDOpcodes.h
index 179cab137b5095..e70118f8ce1d2d 100644
--- a/llvm/include/llvm/CodeGen/ISDOpcodes.h
+++ b/llvm/include/llvm/CodeGen/ISDOpcodes.h
@@ -1022,10 +1022,10 @@ enum NodeType {
LLRINT,
/// FMINNUM/FMAXNUM - Perform floating-point minimum maximum on two values,
- /// following IEEE-754 definitions.
+ /// following IEEE-754 definitions except for signed zero behavior.
///
/// If one input is a signaling NaN, returns a quiet NaN. This matches
- /// IEEE-754 2008's minnum/maxnum behavior for signaling NaNs (which differs
+ /// IEEE-754 2008's minNum/maxNum behavior for signaling NaNs (which differs
/// from 2019).
///
/// These treat -0 as ordered less than +0, matching the behavior of IEEE-754
>From 50f1184fb754680ec30535c79c2c8d3db8e273b4 Mon Sep 17 00:00:00 2001
From: YunQiang Su <yunqiang at isrc.iscas.ac.cn>
Date: Fri, 1 Nov 2024 09:27:03 +0800
Subject: [PATCH 09/11] libc may be different
---
llvm/docs/LangRef.rst | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index 1d6b34fc4a12b0..be1e24cc9315ee 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -16589,9 +16589,9 @@ attribute on the intrinsic call.
Some architectures, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, have instructions that match
these semantics exactly; thus it is quite simple for these architectures.
-Historically, libc returns NUM for NUM vs (sNaN or qNaN), and may return
-sNaN for qNaN vs sNaN. With the recent libc versions, libc follows IEEE754-2008:
-NUM vs sNaN -> qNaN; NUM vs qNaN -> NUM; qNaN vs sNaN -> qNaN; sNaN vs sNaN -> qNaN.
+In the real libc worlds, the bebhaviors of fmin may be quite different on sNaN and signed zero behaviors,
+even in the same release of a single libm implemention. Such as in glibc 2.24, the Arm64 asm implemention
+has different behaviour with the generic C implemention.
Note that that arithmetic on an sNaN doesn't consistently produce a qNaN,
so arithmetic feeding into a minnum can produce inconsistent results.
@@ -16648,9 +16648,9 @@ attribute on the intrinsic call.
Some architectures, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, have instructions that match
these semantics exactly; thus it is quite simple for these architectures.
-Historically, libc returns NUM for NUM vs (sNaN or qNaN), and may return
-sNaN for qNaN vs sNaN. With the recent libc versions, libc follows IEEE754-2008:
-NUM vs sNaN -> qNaN; NUM vs qNaN -> NUM; qNaN vs sNaN -> qNaN; sNaN vs sNaN -> qNaN.
+In the real libc worlds, the bebhaviors of fmin may be quite different on sNaN and signed zero behaviors,
+even in the same release of a single libm implemention. Such as in glibc 2.24, the Arm64 asm implemention
+has different behaviour with the generic C implemention.
Note that that arithmetic on an sNaN doesn't consistently produce a qNaN,
so arithmetic feeding into a maxnum can produce inconsistent results.
>From d7c85cea4eb834566430e675390af50ed0d24443 Mon Sep 17 00:00:00 2001
From: YunQiang Su <yunqiang at isrc.iscas.ac.cn>
Date: Tue, 3 Dec 2024 09:34:42 +0800
Subject: [PATCH 10/11] Update
---
llvm/docs/LangRef.rst | 47 ++++++++++++++++----------
llvm/include/llvm/CodeGen/ISDOpcodes.h | 3 ++
2 files changed, 32 insertions(+), 18 deletions(-)
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index be1e24cc9315ee..d9c37cc2551628 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -16577,26 +16577,32 @@ Semantics:
Follows the IEEE-754 semantics for minNum, except that -0.0 < +0.0 for the purposes
of this intrinsic. As for signaling NaNs, per the IEEE-754 semantics, if either operand
is an sNaN, the result is always a qNaN. This matches the recommended behavior for the libm
-function fmin, although not all implementations have implemented these recommended behaviors.
+function `fmin`, although not all implementations have implemented these recommended behaviors.
If either operand is a qNaN, returns the other non-NaN operand. Returns
-NaN only if both operands are NaN or either operand is sNaN.
+NaN only if both operands are NaN or if either operand is sNaN.
This behavior is more strict than the definition in C and IEEE 754, where either zero may be returned.
To achieve the same permissiveness, the backend may implement the nsz attribute, and one may use the nsz
attribute on the intrinsic call.
-Some architectures, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, have instructions that match
-these semantics exactly; thus it is quite simple for these architectures.
+If the intrinsic is marked with the nsz attribute, then the effect is as in the definition in C
+and IEEE 754: the result of minnum(-0.0, +0.0) may be either -0.0 or +0.0.
+
+Some architectures, such as ARMv8 (FMINNM), LoongArch (fmin), MIPSr6 (min.fmt), PowerPC/VSX (xsmindp),
+have instructions that match these semantics exactly; thus it is quite simple for these architectures.
+Some architectures have similiar while they are not exact equivalent. Such as x86 implements `MINPS`,
+which implements the semantics of C code `a<b?a:b`: NUM vs qNaN always return qNaN. `MINPS` can be used
+if `nsz` and `nnan` are given.
+
In the real libc worlds, the bebhaviors of fmin may be quite different on sNaN and signed zero behaviors,
-even in the same release of a single libm implemention. Such as in glibc 2.24, the Arm64 asm implemention
-has different behaviour with the generic C implemention.
+even in the same release of a single libm implemention.
-Note that that arithmetic on an sNaN doesn't consistently produce a qNaN,
+Note that arithmetic on an sNaN doesn't consistently produce a qNaN,
so arithmetic feeding into a minnum can produce inconsistent results.
-Such as `fmin(sNaN+0.0, 1.0)` can produce qNaN or 1.0 depending on whether `+0.0`
-is optimized out.
+For example, `maxnum(fadd(sNaN, 0.0), 1.0)` can produce qNaN or 1.0 depending on whether `fadd`
+is folded.
.. _i_maxnum:
@@ -16636,26 +16642,31 @@ Semantics:
Follows the IEEE-754 semantics for minNum, except that -0.0 < +0.0 for the purposes
of this intrinsic. As for signaling NaNs, per the IEEE-754 semantics, if either operand
is an sNaN, the result is always a qNaN. This matches the recommended behavior for the libm
-function fmin, although not all implementations have implemented these recommended behaviors.
+function `fmax`, although not all implementations have implemented these recommended behaviors.
If either operand is a qNaN, returns the other non-NaN operand. Returns
-NaN only if both operands are NaN or either operand is sNaN.
+NaN only if both operands are NaN or if either operand is sNaN.
This behavior is more strict than the definition in C and IEEE 754, where either zero may be returned.
To achieve the same permissiveness, the backend may implement the nsz attribute, and one may use the nsz
attribute on the intrinsic call.
-Some architectures, such as ARMv8, LoongArch, MIPSr6, PowerPC/VSX, have instructions that match
-these semantics exactly; thus it is quite simple for these architectures.
+If the intrinsic is marked with the nsz attribute, then the effect is as in the definition in C
+and IEEE 754: the result of maxnum(-0.0, +0.0) may be either -0.0 or +0.0.
+
+Some architectures, such as ARMv8 (FMAXNM), LoongArch (fmax), MIPSr6 (max.fmt), PowerPC/VSX (xsmaxdp),
+have instructions that match these semantics exactly; thus it is quite simple for these architectures.
+Some architectures have similiar while they are not exact equivalent. Such as x86 implements `MAXPS`,
+which implements the semantics of C code `a>b?a:b`: NUM vs qNaN always return qNaN. `MAXPS` can be used
+if `nsz` and `nnan` are given.
In the real libc worlds, the bebhaviors of fmin may be quite different on sNaN and signed zero behaviors,
-even in the same release of a single libm implemention. Such as in glibc 2.24, the Arm64 asm implemention
-has different behaviour with the generic C implemention.
+even in the same release of a single libm implemention.
-Note that that arithmetic on an sNaN doesn't consistently produce a qNaN,
+Note that arithmetic on an sNaN doesn't consistently produce a qNaN,
so arithmetic feeding into a maxnum can produce inconsistent results.
-Such as `fmax(sNaN+0.0, 1.0)` can produce qNaN or 1.0 depending on whether `+0.0`
-is optimized out.
+For example, `maxnum(fadd(sNaN, 0.0), 1.0)` can produce qNaN or 1.0 depending on whether `fadd`
+is folded.
.. _i_minimum:
diff --git a/llvm/include/llvm/CodeGen/ISDOpcodes.h b/llvm/include/llvm/CodeGen/ISDOpcodes.h
index e70118f8ce1d2d..c3187bcb070fe6 100644
--- a/llvm/include/llvm/CodeGen/ISDOpcodes.h
+++ b/llvm/include/llvm/CodeGen/ISDOpcodes.h
@@ -1048,6 +1048,9 @@ enum NodeType {
///
/// These treat -0 as ordered less than +0, matching the behavior of IEEE-754
/// 2019's minimumNumber/maximumNumber.
+ ///
+ /// Deprecated, and will be removed soon, as FMINNUM/FMAXNUM have the same
+ /// semantics now.
FMINNUM_IEEE,
FMAXNUM_IEEE,
>From ee4f97c0315f06085bb198ee163343dfe85dde06 Mon Sep 17 00:00:00 2001
From: YunQiang Su <yunqiang at isrc.iscas.ac.cn>
Date: Wed, 4 Dec 2024 10:33:47 +0800
Subject: [PATCH 11/11] highlight IEEE-754-2008
---
llvm/docs/LangRef.rst | 32 ++++++++++++++++----------------
1 file changed, 16 insertions(+), 16 deletions(-)
diff --git a/llvm/docs/LangRef.rst b/llvm/docs/LangRef.rst
index d9c37cc2551628..995c1b6e35d2d5 100644
--- a/llvm/docs/LangRef.rst
+++ b/llvm/docs/LangRef.rst
@@ -16574,20 +16574,20 @@ type.
Semantics:
""""""""""
-Follows the IEEE-754 semantics for minNum, except that -0.0 < +0.0 for the purposes
-of this intrinsic. As for signaling NaNs, per the IEEE-754 semantics, if either operand
+Follows the semantics of minNum in IEEE-754-2008, except that -0.0 < +0.0 for the purposes
+of this intrinsic. As for signaling NaNs, per the minNum semantics, if either operand
is an sNaN, the result is always a qNaN. This matches the recommended behavior for the libm
function `fmin`, although not all implementations have implemented these recommended behaviors.
If either operand is a qNaN, returns the other non-NaN operand. Returns
NaN only if both operands are NaN or if either operand is sNaN.
-This behavior is more strict than the definition in C and IEEE 754, where either zero may be returned.
-To achieve the same permissiveness, the backend may implement the nsz attribute, and one may use the nsz
-attribute on the intrinsic call.
+This behavior is stricter than minNum in IEEE-754-2008, where either zero may be returned.
+To achieve the same permissiveness, the backend may implement the nsz attribute, and one may use
+the nsz attribute.
If the intrinsic is marked with the nsz attribute, then the effect is as in the definition in C
-and IEEE 754: the result of minnum(-0.0, +0.0) may be either -0.0 or +0.0.
+and IEEE-754-2008: the result of minnum(-0.0, +0.0) may be either -0.0 or +0.0.
Some architectures, such as ARMv8 (FMINNM), LoongArch (fmin), MIPSr6 (min.fmt), PowerPC/VSX (xsmindp),
have instructions that match these semantics exactly; thus it is quite simple for these architectures.
@@ -16639,20 +16639,20 @@ type.
Semantics:
""""""""""
-Follows the IEEE-754 semantics for minNum, except that -0.0 < +0.0 for the purposes
-of this intrinsic. As for signaling NaNs, per the IEEE-754 semantics, if either operand
+Follows the semantics of maxNum in IEEE-754-2008, except that -0.0 < +0.0 for the purposes
+of this intrinsic. As for signaling NaNs, per the maxNum semantics, if either operand
is an sNaN, the result is always a qNaN. This matches the recommended behavior for the libm
function `fmax`, although not all implementations have implemented these recommended behaviors.
If either operand is a qNaN, returns the other non-NaN operand. Returns
NaN only if both operands are NaN or if either operand is sNaN.
-This behavior is more strict than the definition in C and IEEE 754, where either zero may be returned.
-To achieve the same permissiveness, the backend may implement the nsz attribute, and one may use the nsz
-attribute on the intrinsic call.
+This behavior is stricter than maxNum in IEEE-754-2008, where either zero may be returned.
+To achieve the same permissiveness, the backend may implement the nsz attribute, and one may use
+the nsz attribute.
If the intrinsic is marked with the nsz attribute, then the effect is as in the definition in C
-and IEEE 754: the result of maxnum(-0.0, +0.0) may be either -0.0 or +0.0.
+and IEEE-754-2008: the result of maxnum(-0.0, +0.0) may be either -0.0 or +0.0.
Some architectures, such as ARMv8 (FMAXNM), LoongArch (fmax), MIPSr6 (max.fmt), PowerPC/VSX (xsmaxdp),
have instructions that match these semantics exactly; thus it is quite simple for these architectures.
@@ -21849,7 +21849,7 @@ This is an overloaded intrinsic.
Overview:
"""""""""
-Predicated floating-point IEEE-754 minNum of two vectors of floating-point values.
+Predicated floating-point IEEE-754-2008 minNum of two vectors of floating-point values.
Arguments:
@@ -21898,7 +21898,7 @@ This is an overloaded intrinsic.
Overview:
"""""""""
-Predicated floating-point IEEE-754 maxNum of two vectors of floating-point values.
+Predicated floating-point IEEE-754-2008 maxNum of two vectors of floating-point values.
Arguments:
@@ -27942,7 +27942,7 @@ The third argument specifies the exception behavior as described above.
Semantics:
""""""""""
-This function follows the IEEE-754 semantics for maxNum.
+This function follows the IEEE-754-2008 semantics for maxNum.
'``llvm.experimental.constrained.minnum``' Intrinsic
@@ -27974,7 +27974,7 @@ The third argument specifies the exception behavior as described above.
Semantics:
""""""""""
-This function follows the IEEE-754 semantics for minNum.
+This function follows the IEEE-754-2008 semantics for minNum.
'``llvm.experimental.constrained.maximum``' Intrinsic
More information about the llvm-commits
mailing list