Re: [PATCH 9/12] libgcc _BitInt support [PR102989]
On Fri, Sep 01, 2023 at 09:48:22PM +, Joseph Myers wrote:
> This patch is OK with these fixes.
Thanks, here is an updated patch, thanks for catching the _Decimal128
bug. Will post testsuite additions/adjustment patch as follow-up on Monday.
2023-09-02 Jakub Jelinek
PR c/102989
libgcc/
* config/aarch64/t-softfp (softfp_extras): Use += rather than :=.
* config/i386/64/t-softfp (softfp_extras): Likewise.
* config/i386/libgcc-glibc.ver (GCC_14.0.0): Export _BitInt support
routines.
* config/i386/t-softfp (softfp_extras): Add fixxfbitint and
bf, hf and xf mode floatbitint.
(CFLAGS-floatbitintbf.c, CFLAGS-floatbitinthf.c): Add -msse2.
* config/riscv/t-softfp32 (softfp_extras): Use += rather than :=.
* config/rs6000/t-e500v1-fp (softfp_extras): Likewise.
* config/rs6000/t-e500v2-fp (softfp_extras): Likewise.
* config/t-softfp (softfp_floatbitint_funcs): New.
(softfp_bid_list): New.
(softfp_func_list): Add sf and df mode from and to _BitInt libcalls.
(softfp_bid_file_list): New.
(LIB2ADD_ST): Add $(softfp_bid_file_list).
* config/t-softfp-sfdftf (softfp_extras): Add fixtfbitint and
floatbitinttf.
* config/t-softfp-tf (softfp_extras): Likewise.
* libgcc2.c (bitint_reduce_prec): New inline function.
(BITINT_INC, BITINT_END): Define.
(bitint_mul_1, bitint_addmul_1): New helper functions.
(__mulbitint3): New function.
(bitint_negate, bitint_submul_1): New helper functions.
(__divmodbitint4): New function.
* libgcc2.h (LIBGCC2_UNITS_PER_WORD): When building _BitInt support
libcalls, redefine depending on __LIBGCC_BITINT_LIMB_WIDTH__.
(__mulbitint3, __divmodbitint4): Declare.
* libgcc-std.ver.in (GCC_14.0.0): Export _BitInt support routines.
* Makefile.in (lib2funcs): Add _mulbitint3.
(LIB2_DIVMOD_FUNCS): Add _divmodbitint4.
* soft-fp/bitint.h: New file.
* soft-fp/fixdfbitint.c: New file.
* soft-fp/fixsfbitint.c: New file.
* soft-fp/fixtfbitint.c: New file.
* soft-fp/fixxfbitint.c: New file.
* soft-fp/floatbitintbf.c: New file.
* soft-fp/floatbitintdf.c: New file.
* soft-fp/floatbitinthf.c: New file.
* soft-fp/floatbitintsf.c: New file.
* soft-fp/floatbitinttf.c: New file.
* soft-fp/floatbitintxf.c: New file.
* soft-fp/op-common.h (_FP_FROM_INT): Add support for rsize up to
4 * _FP_W_TYPE_SIZE rather than just 2 * _FP_W_TYPE_SIZE.
* soft-fp/bitintpow10.c: New file.
* soft-fp/fixsdbitint.c: New file.
* soft-fp/fixddbitint.c: New file.
* soft-fp/fixtdbitint.c: New file.
* soft-fp/floatbitintsd.c: New file.
* soft-fp/floatbitintdd.c: New file.
* soft-fp/floatbitinttd.c: New file.
--- libgcc/config/aarch64/t-softfp.jj 2023-08-08 15:54:35.737595343 +0200
+++ libgcc/config/aarch64/t-softfp 2023-08-08 16:12:02.346939560 +0200
@@ -3,7 +3,7 @@ softfp_int_modes := si di ti
softfp_extensions := sftf dftf hftf bfsf
softfp_truncations := tfsf tfdf tfhf tfbf dfbf sfbf hfbf
softfp_exclude_libgcc2 := n
-softfp_extras := fixhfti fixunshfti floattihf floatuntihf \
+softfp_extras += fixhfti fixunshfti floattihf floatuntihf \
floatdibf floatundibf floattibf floatuntibf
TARGET_LIBGCC2_CFLAGS += -Wno-missing-prototypes
--- libgcc/config/i386/64/t-softfp.jj 2023-08-08 15:54:35.766594936 +0200
+++ libgcc/config/i386/64/t-softfp 2023-08-08 16:12:02.346939560 +0200
@@ -1,4 +1,4 @@
-softfp_extras := fixhfti fixunshfti floattihf floatuntihf \
+softfp_extras += fixhfti fixunshfti floattihf floatuntihf \
floattibf floatuntibf
CFLAGS-fixhfti.c += -msse2
--- libgcc/config/i386/libgcc-glibc.ver.jj 2023-08-08 15:54:35.831594026
+0200
+++ libgcc/config/i386/libgcc-glibc.ver 2023-08-08 16:12:02.347939546 +0200
@@ -226,3 +226,13 @@ GCC_13.0.0 {
__truncxfbf2
__trunchfbf2
}
+
+%inherit GCC_14.0.0 GCC_13.0.0
+GCC_14.0.0 {
+ __PFX__fixxfbitint
+ __PFX__fixtfbitint
+ __PFX__floatbitintbf
+ __PFX__floatbitinthf
+ __PFX__floatbitintxf
+ __PFX__floatbitinttf
+}
--- libgcc/config/i386/t-softfp.jj 2023-08-08 15:55:09.819118062 +0200
+++ libgcc/config/i386/t-softfp 2023-08-08 16:12:02.347939546 +0200
@@ -10,7 +10,7 @@ softfp_extensions := hfsf hfdf hftf hfxf
softfp_truncations := tfhf xfhf dfhf sfhf tfsf dfsf tfdf tfxf \
tfbf xfbf dfbf sfbf hfbf
-softfp_extras += eqhf2
+softfp_extras += eqhf2 fixxfbitint $(foreach m,hf bf xf,floatbitint$(m))
CFLAGS-extendhfsf2.c += -msse2
CFLAGS-extendhfdf2.c += -msse2
@@ -28,6 +28,9 @@ CFLAGS-truncxfbf2.c += -msse2
CFLAGS-trunctfbf2.c += -msse2
CFLAGS-trunchfbf2.c += -msse2
+CFLAGS-floatbitintbf.c += -msse2
+CFLAGS-floatbitinthf.c += -msse2
+
CFLAGS-eqhf2.c += -msse2
CFLAGS-_divhc3.c += -msse2
Re: [PATCH 9/12] libgcc _BitInt support [PR102989]
On Wed, 9 Aug 2023, Jakub Jelinek via Gcc-patches wrote:
> I know that soft-fp is owned by glibc and I think the op-common.h change
> should be propagated there, but the bitint stuff is really GCC specific
> and IMHO doesn't belong into the glibc copy.
The op-common.h change is OK for glibc.
Some additional tests I think should be added to the testsuite for
floating-point functionality in this patch, that I didn't spot in the
testsuite patches - if any of these aren't included initially, there
should at least be bugs filed in Bugzilla for the omissions:
1. Test overflowing conversions to integers (including from inf or NaN)
raise FE_INVALID. (Note: it's not specified in the standard whether
inexact conversions to integers raise FE_INEXACT or not, so testing that
seems less important.)
2. Test conversions from integers to floating point raise FE_INEXACT when
inexact, together with FE_OVERFLOW when overflowing (while exact
conversions don't raise exceptions).
3. Test conversions from integers to floating point respect the rounding
mode.
4. Test converting floating-point values in the range (-1.0, 0.0] to both
unsigned and signed _BitInt; I didn't see such tests for binary floating
types, only for decimal types, and the decimal tests didn't include tests
of negative zero itself as the value converted to _BitInt.
5. Test conversions of noncanonical BID zero to integers (these tests
would be specific to BID). See below for a bug in this area.
For points 2 and 3 above, it's probably appropriate to test only for
binary floating point, to avoid any issues with the separate DFP rounding
mode and with DFP arithmetic operations not necessarily working correctly
with exceptions - but then a bug should be filed in Bugzilla noting the
omission of such tests for DFP.
For points 1, 2 and 3 above, if the conversions for types such as
_BitInt(32) might end up using the same conversions as for types such as
int, then tests for such types should probably be omitted (again with a
bug filed) given the range of known bugs about exceptions from such
operations with types such as int.
> +__bid_fixtdbitint (UBILtype *r, SItype rprec, _Decimal128 a)
> +{
> + FP_DECL_EX;
> + USItype arprec = rprec < 0 ? -rprec : rprec;
> + USItype rn = (arprec + BIL_TYPE_SIZE - 1) / BIL_TYPE_SIZE;
> + union { _Decimal128 d; UDItype u[2]; } u;
> + UDItype mantissahi, mantissalo, t;
> + SItype sgn;
> + SItype exponent;
> + USItype exp_bits, mant_bits;
> + UBILtype *pow10v, *resv;
> + USItype pow10_limbs, res_limbs, min_limbs, mant_limbs, low_zeros;
> +
> + FP_INIT_EXCEPTIONS;
> + u.d = a;
> + mantissahi = u.u[__FLOAT_WORD_ORDER__ != __ORDER_BIG_ENDIAN__];
> + mantissalo = u.u[__FLOAT_WORD_ORDER__ == __ORDER_BIG_ENDIAN__];
> + t = mantissahi >> 47;
> + sgn = (DItype) mantissahi < 0;
> + if ((t & (3 << 14)) != (3 << 14))
> +{
> + mantissahi &= UDItype) 1) << 49) - 1);
> + exponent = (t >> 2) & 0x3fff;
Overflow (thus producing a noncanonical zero) is possible in this case for
TDmode. An appropriate test of a noncanonical zero that goes through this
case should thus be added to the testsuite.
> +}
> + else if ((t & (3 << 12)) != (3 << 12))
> +{
> + mantissahi &= UDItype) 1) << 47) - 1);
> + mantissahi |= ((UDItype) 1) << 49;
> + exponent = t & 0x3fff;
> + if (mantissahi > (UDItype) 0x1ed09bead87c0
> + || (mantissahi == (UDItype) 0x1ed09bead87c0
> + && mantissalo > 0x378d8e63))
> + {
> + mantissahi = 0;
> + mantissalo = 0;
> + }
And in this case, overflow is guaranteed; the check for the overflow
threshold should thus move to the previous case.
This patch is OK with these fixes.
Note for powerpc architecture maintainers: adding _BitInt support on that
architecture will mean, as well as adding support for the conversions
to/from DPD (if S/390 doesn't get there first), also adding support for
conversions to/from IBM long double.
--
Joseph S. Myers
jos...@codesourcery.com
