libgcc _BitInt helper documentation [PR102989]

On Mon, Aug 21, 2023 at 05:32:04PM +0000, Joseph Myers wrote:
> I think the libgcc functions (i.e. those exported by libgcc, to which
> references are generated by the compiler) need documenting in libgcc.texi.
> Internal functions or macros in the libgcc patch need appropriate comments
> specifying their semantics; especially FP_TO_BITINT and FP_FROM_BITINT
> which have a lot of arguments and no comments saying what the semantics of
> the macros and their arguments are supposed to me.

Here is an incremental patch which does that.

2023-09-06  Jakub Jelinek  <jakub@redhat.com>

	PR c/102989
gcc/
	* doc/libgcc.texi (Bit-precise integer arithmetic functions):
	Document general rules for _BitInt support library functions
	and document __mulbitint3 and __divmodbitint4.
	(Conversion functions): Document __fix{s,d,x,t}fbitint,
	__floatbitint{s,d,x,t,h,b}f, __bid_fix{s,d,t}dbitint and
	__bid_floatbitint{s,d,t}d.
libgcc/
	* libgcc2.c (bitint_negate): Add function comment.
	* soft-fp/bitint.h (bitint_negate): Add function comment.
	(FP_TO_BITINT, FP_FROM_BITINT): Add comment explaining the macros.

3 files changed, 111 insertions, 0 deletions

diff --git a/gcc/doc/libgcc.texi b/gcc/doc/libgcc.texi
index 73aa803..08219bf 100644
--- a/gcc/doc/libgcc.texi
+++ b/gcc/doc/libgcc.texi
@@ -218,6 +218,51 @@ These functions return the number of bits set in @var{a}.
 These functions return the @var{a} byteswapped.
 @end deftypefn
 
+@subsection Bit-precise integer arithmetic functions
+
+@code{_BitInt(@var{n})} library functions operate on arrays of limbs, where
+each limb has @code{__LIBGCC_BITINT_LIMB_WIDTH__} bits and the limbs are
+ordered according to @code{__LIBGCC_BITINT_ORDER__} ordering.  The most
+significant limb if @var{n} is not divisible by
+@code{__LIBGCC_BITINT_LIMB_WIDTH__} contains padding bits which should be
+ignored on read (sign or zero extended), but extended on write.  For the
+library functions, all bit-precise integers regardless of @var{n} are
+represented like that, even when the target ABI says that for some small
+@var{n} they should be represented differently in memory.  A pointer
+to the array of limbs argument is always accompanied with a bit size
+argument.  If that argument is positive, it is number of bits and the
+number is assumed to be zero-extended to infinite precision, if that
+argument is negative, it is negated number of bits above which all bits
+are assumed to be sign-extended to infinite precision.  These number of bits
+arguments don't need to match actual @var{n} for the operation used in the
+source, they could be lowered because of sign or zero extensions on the
+input or because value-range optimization figures value will need certain
+lower number of bits.  For big-endian ordering of limbs, when lowering
+the bit size argument the pointer argument needs to be adjusted as well.
+Negative bit size argument should be always smaller or equal to @code{-2},
+because @code{signed _BitInt(1)} is not valid.
+For output arguments, either the corresponding bit size argument should
+be always positive (for multiplication and division), or is negative when
+the output of conversion from floating-point value is signed and positive
+when unsigned.  The arrays of limbs output arguments point to should not
+overlap any inputs, while input arrays of limbs can overlap.
+@code{UBILtype} below stands for unsigned integer type with
+@code{__LIBGCC_BITINT_LIMB_WIDTH__} bit precision.
+
+@deftypefn {Runtime Function} void __mulbitint3 (@code{UBILtype} *@var{ret}, int32_t @var{retprec}, const @code{UBILtype} *u, int32_t @var{uprec}, const @code{UBILtype} *v, int32_t @var{vprec})
+This function multiplies bit-precise integer operands @var{u} and @var{v} and stores
+result into @var{retprec} precision bit-precise integer result @var{ret}.
+@end deftypefn
+
+@deftypefn {Runtime Function} void __divmodbitint4 (@code{UBILtype} *@var{q}, int32_t @var{qprec}, @code{UBILtype} *@var{r}, int32_t @var{rprec},  const @code{UBILtype} *u, int32_t @var{uprec}, const @code{UBILtype} *v, int32_t @var{vprec})
+This function divides bit-precise integer operands @var{u} and @var{v} and stores
+quotient into @var{qprec} precision bit-precise integer result @var{q}
+(unless @var{q} is @code{NULL} and @var{qprec} is 0, in that case quotient
+is not stored anywhere) and remainder into @var{rprec} precision bit-precise
+integer result @var{r} (similarly, unless @var{r} is @code{NULL} and @var{rprec}
+is 0).
+@end deftypefn
+
 @node Soft float library routines
 @section Routines for floating point emulation
 @cindex soft float library
@@ -384,6 +429,27 @@ These functions convert @var{i}, an unsigned long, to floating point.
 These functions convert @var{i}, an unsigned long long, to floating point.
 @end deftypefn
 
+@deftypefn {Runtime Function} void __fixsfbitint (@code{UBILtype} *@var{r}, int32_t @var{rprec}, float @var{a})
+@deftypefnx {Runtime Function} void __fixdfbitint (@code{UBILtype} *@var{r}, int32_t @var{rprec}, double @var{a})
+@deftypefnx {Runtime Function} void __fixxfbitint (@code{UBILtype} *@var{r}, int32_t @var{rprec}, __float80 @var{a})
+@deftypefnx {Runtime Function} void __fixtfbitint (@code{UBILtype} *@var{r}, int32_t @var{rprec}, _Float128 @var{a})
+These functions convert @var{a} to bit-precise integer @var{r}, rounding toward zero.
+If @var{rprec} is positive, it converts to unsigned bit-precise integer and
+negative values all become zero, if @var{rprec} is negative, it converts
+to signed bit-precise integer.
+@end deftypefn
+
+@deftypefn {Runtime Function} float __floatbitintsf (@code{UBILtype} *@var{i}, int32_t @var{iprec})
+@deftypefnx {Runtime Function} double __floatbitintdf (@code{UBILtype} *@var{i}, int32_t @var{iprec})
+@deftypefnx {Runtime Function} __float80 __floatbitintxf (@code{UBILtype} *@var{i}, int32_t @var{iprec})
+@deftypefnx {Runtime Function} _Float128 __floatbitinttf (@code{UBILtype} *@var{i}, int32_t @var{iprec})
+@deftypefnx {Runtime Function} _Float16 __floatbitinthf (@code{UBILtype} *@var{i}, int32_t @var{iprec})
+@deftypefnx {Runtime Function} __bf16 __floatbitintbf (@code{UBILtype} *@var{i}, int32_t @var{iprec})
+These functions convert bit-precise integer @var{i} to floating point.  If
+@var{iprec} is positive, it is conversion from unsigned bit-precise integer,
+otherwise from signed bit-precise integer.
+@end deftypefn
+
 @subsection Comparison functions
 
 There are two sets of basic comparison functions.
@@ -707,6 +773,23 @@ These functions convert @var{i}, an unsigned integer, to decimal floating point.
 These functions convert @var{i}, an unsigned long, to decimal floating point.
 @end deftypefn
 
+@deftypefn {Runtime Function} void __bid_fixsdbitint (@code{UBILtype} *@var{r}, int32_t @var{rprec}, _Decimal32 @var{a})
+@deftypefnx {Runtime Function} void __bid_fixddbitint (@code{UBILtype} *@var{r}, int32_t @var{rprec}, _Decimal64 @var{a})
+@deftypefnx {Runtime Function} void __bid_fixtdbitint (@code{UBILtype} *@var{r}, int32_t @var{rprec}, _Decimal128 @var{a})
+These functions convert @var{a} to bit-precise integer @var{r}, rounding toward zero.
+If @var{rprec} is positive, it converts to unsigned bit-precise integer and
+negative values all become zero, if @var{rprec} is negative, it converts
+to signed bit-precise integer.  So far implemented for BID format only.
+@end deftypefn
+
+@deftypefn {Runtime Function} _Decimal32 __bid_floatbitintsd (@code{UBILtype} *@var{i}, int32_t @var{iprec})
+@deftypefnx {Runtime Function} _Decimal64 __bid_floatbitintdd (@code{UBILtype} *@var{i}, int32_t @var{iprec})
+@deftypefnx {Runtime Function} _Decimal128 __bid_floatbitinttd (@code{UBILtype} *@var{i}, int32_t @var{iprec})
+These functions convert bit-precise integer @var{i} to decimal floating point.  If
+@var{iprec} is positive, it is conversion from unsigned bit-precise integer,
+otherwise from signed bit-precise integer.  So far implemented for BID format only.
+@end deftypefn
+
 @subsection Comparison functions
 
 @deftypefn {Runtime Function} int __dpd_unordsd2 (_Decimal32 @var{a}, _Decimal32 @var{b})
diff --git a/libgcc/libgcc2.c b/libgcc/libgcc2.c
index d217e33..8b42210 100644
--- a/libgcc/libgcc2.c
+++ b/libgcc/libgcc2.c
@@ -1640,6 +1640,8 @@ __mulbitint3 (UWtype *ret, SItype retprec,
 #endif
 
 #ifdef L_divmodbitint4
+/* D = -S.  */
+
 static void
 bitint_negate (UWtype *d, const UWtype *s, SItype n)
 {
diff --git a/libgcc/soft-fp/bitint.h b/libgcc/soft-fp/bitint.h
index 5e8b764..20cd41b 100644
--- a/libgcc/soft-fp/bitint.h
+++ b/libgcc/soft-fp/bitint.h
@@ -160,6 +160,9 @@ bitint_reduce_prec (const UBILtype **p, SItype prec)
 # define BITINT_END(be, le) (le)
 #endif
 
+/* Negate N limbs from S into D.  D and S should point to
+   the least significant limb.  */
+
 static inline __attribute__((__always_inline__)) void
 bitint_negate (UBILtype *d, const UBILtype *s, SItype n)
 {
@@ -175,6 +178,19 @@ bitint_negate (UBILtype *d, const UBILtype *s, SItype n)
   while (--n);
 }
 
+/* Common final part of __fix?fbitint conversion functions.
+   The A floating point value should have been converted using
+   soft-fp macros into RV, U##DI##type DI##_BITS precise normal
+   integral type and SHIFT, how many bits should that value be
+   shifted to the left.  R is pointer to limbs array passed to the
+   function, RN number of limbs in it, ARPREC absolute value of
+   RPREC argument passed to it, RSIZE number of significant bits in RV.
+   RSIGNED is non-zero if the result is signed bit-precise integer,
+   otherwise zero.  If OVF is true, instead of storing RV shifted left
+   by SHIFT bits and zero or sign extended store minimum or maximum
+   of the signed or unsigned bit-precise integer type or zero depending on if
+   RV contains the minimum or maximum signed or unsigned value or zero.  */
+
 #define FP_TO_BITINT(r, rn, arprec, shift, rv, rsize, rsigned, ovf, DI) \
   if (ovf)								\
     {									\
@@ -232,6 +248,16 @@ bitint_negate (UBILtype *d, const UBILtype *s, SItype n)
 			  * sizeof (UBILtype));				\
     }
 
+/* Common initial part of __floatbitint?f conversion functions.
+   I and IPREC are arguments passed to those functions, convert that
+   into a pair of DI##type IV integer and SHIFT, such that converting
+   IV to floating point and multiplicating that by pow (2, SHIFT)
+   gives the expected result.  IV size needs to be chosen such that
+   it is larger than number of bits in floating-point mantissa and
+   contains there even at least a two bits below the mantissa for
+   rounding purposes.  If any of the SHIFT bits shifted out is non-zero,
+   the least significant bit should be non-zero.  */
+
 #define FP_FROM_BITINT(i, iprec, iv, shift, DI)				\
   do									\
     {									\