From: "Joseph S. Myers" <joseph@codesourcery.com>
To: <libc-ports@sourceware.org>
Subject: Make soft-float ARM use soft-fp fma/fmaf
Date: Tue, 02 Jul 2013 19:52:00 -0000 [thread overview]
Message-ID: <Pine.LNX.4.64.1307021951410.28677@digraph.polyomino.org.uk> (raw)
I've committed this patch to make soft-float ARM glibc use the soft-fp
versions of fma/fmaf, instead of the versions that rely on rounding
modes and exceptions that aren't supported for ARM soft-float and so
produce inaccurate results (even in round-to-nearest, the only
rounding mode that otherwise works properly for ARM soft-float).
(The sfp-machine.h doesn't attempt any run-time integration with
hardware exceptions and rounding modes if present, as there doesn't
seem to be much point without a proper fix for bug 10064 to make
libgcc __aeabi_* functions respect rounding modes and exceptions via
STT_GNU_IFUNC if VFP hardware is present.)
2013-07-02 Joseph Myers <joseph@codesourcery.com>
[BZ #13304]
* sysdeps/arm/s_fma.c: New file.
* sysdeps/arm/s_fmaf.c: New file.
* sysdeps/arm/soft-fp/sfp-machine.h: New file. Based on powerpc
version.
* sysdeps/arm/Implies: Add arm/soft-fp.
diff --git a/ports/sysdeps/arm/Implies b/ports/sysdeps/arm/Implies
index 780c4e2..10bd93a 100644
--- a/ports/sysdeps/arm/Implies
+++ b/ports/sysdeps/arm/Implies
@@ -1,3 +1,4 @@
+arm/soft-fp
wordsize-32
ieee754/flt-32
ieee754/dbl-64
diff --git a/ports/sysdeps/arm/s_fma.c b/ports/sysdeps/arm/s_fma.c
new file mode 100644
index 0000000..dc4e27b
--- /dev/null
+++ b/ports/sysdeps/arm/s_fma.c
@@ -0,0 +1,5 @@
+#ifdef __SOFTFP__
+# include <soft-fp/fmadf4.c>
+#else
+# include <sysdeps/ieee754/dbl-64/s_fma.c>
+#endif
diff --git a/ports/sysdeps/arm/s_fmaf.c b/ports/sysdeps/arm/s_fmaf.c
new file mode 100644
index 0000000..550d8b8
--- /dev/null
+++ b/ports/sysdeps/arm/s_fmaf.c
@@ -0,0 +1,5 @@
+#ifdef __SOFTFP__
+# include <soft-fp/fmasf4.c>
+#else
+# include <sysdeps/ieee754/dbl-64/s_fmaf.c>
+#endif
diff --git a/ports/sysdeps/arm/soft-fp/sfp-machine.h b/ports/sysdeps/arm/soft-fp/sfp-machine.h
new file mode 100644
index 0000000..eba6e35
--- /dev/null
+++ b/ports/sysdeps/arm/soft-fp/sfp-machine.h
@@ -0,0 +1,49 @@
+#define _FP_W_TYPE_SIZE 32
+#define _FP_W_TYPE unsigned long
+#define _FP_WS_TYPE signed long
+#define _FP_I_TYPE long
+
+#define _FP_MUL_MEAT_S(R,X,Y) \
+ _FP_MUL_MEAT_1_wide(_FP_WFRACBITS_S,R,X,Y,umul_ppmm)
+#define _FP_MUL_MEAT_D(R,X,Y) \
+ _FP_MUL_MEAT_2_wide(_FP_WFRACBITS_D,R,X,Y,umul_ppmm)
+#define _FP_MUL_MEAT_Q(R,X,Y) \
+ _FP_MUL_MEAT_4_wide(_FP_WFRACBITS_Q,R,X,Y,umul_ppmm)
+
+#define _FP_MUL_MEAT_DW_S(R,X,Y) \
+ _FP_MUL_MEAT_DW_1_wide(_FP_WFRACBITS_S,R,X,Y,umul_ppmm)
+#define _FP_MUL_MEAT_DW_D(R,X,Y) \
+ _FP_MUL_MEAT_DW_2_wide(_FP_WFRACBITS_D,R,X,Y,umul_ppmm)
+#define _FP_MUL_MEAT_DW_Q(R,X,Y) \
+ _FP_MUL_MEAT_DW_4_wide(_FP_WFRACBITS_Q,R,X,Y,umul_ppmm)
+
+#define _FP_DIV_MEAT_S(R,X,Y) _FP_DIV_MEAT_1_loop(S,R,X,Y)
+#define _FP_DIV_MEAT_D(R,X,Y) _FP_DIV_MEAT_2_udiv(D,R,X,Y)
+#define _FP_DIV_MEAT_Q(R,X,Y) _FP_DIV_MEAT_4_udiv(Q,R,X,Y)
+
+#define _FP_NANFRAC_S ((_FP_QNANBIT_S << 1) - 1)
+#define _FP_NANFRAC_D ((_FP_QNANBIT_D << 1) - 1), -1
+#define _FP_NANFRAC_Q ((_FP_QNANBIT_Q << 1) - 1), -1, -1, -1
+#define _FP_NANSIGN_S 0
+#define _FP_NANSIGN_D 0
+#define _FP_NANSIGN_Q 0
+
+#define _FP_KEEPNANFRACP 1
+#define _FP_QNANNEGATEDP 0
+
+/* Someone please check this. */
+#define _FP_CHOOSENAN(fs, wc, R, X, Y, OP) \
+ do { \
+ if ((_FP_FRAC_HIGH_RAW_##fs(X) & _FP_QNANBIT_##fs) \
+ && !(_FP_FRAC_HIGH_RAW_##fs(Y) & _FP_QNANBIT_##fs)) \
+ { \
+ R##_s = Y##_s; \
+ _FP_FRAC_COPY_##wc(R,Y); \
+ } \
+ else \
+ { \
+ R##_s = X##_s; \
+ _FP_FRAC_COPY_##wc(R,X); \
+ } \
+ R##_c = FP_CLS_NAN; \
+ } while (0)
--
Joseph S. Myers
joseph@codesourcery.com
reply other threads:[~2013-07-02 19:52 UTC|newest]
Thread overview: [no followups] expand[flat|nested] mbox.gz Atom feed
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=Pine.LNX.4.64.1307021951410.28677@digraph.polyomino.org.uk \
--to=joseph@codesourcery.com \
--cc=libc-ports@sourceware.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).