diff options
author | bde <bde@FreeBSD.org> | 2008-02-07 03:17:05 +0000 |
---|---|---|
committer | bde <bde@FreeBSD.org> | 2008-02-07 03:17:05 +0000 |
commit | 22e608f1ceb21dc3bb9bf5cda8f34642cc6e623f (patch) | |
tree | 39de11ac1fa020a02f07f29ebaac4e42ac04c4d5 /lib/msun/ld128 | |
parent | 67c8e0948c3ec70e3bdfebe1cf5c79cf617f726b (diff) | |
download | FreeBSD-src-22e608f1ceb21dc3bb9bf5cda8f34642cc6e623f.zip FreeBSD-src-22e608f1ceb21dc3bb9bf5cda8f34642cc6e623f.tar.gz |
Use a better method of scaling by 2**k. Instead of adding to the
exponent bits of the reduced result, construct 2**k (hopefully in
parallel with the construction of the reduced result) and multiply by
it. This tends to be much faster if the construction of 2**k is
actually in parallel, and might be faster even with no parallelism
since adjustment of the exponent requires a read-modify-wrtite at an
unfortunate time for pipelines.
In some cases involving exp2* on amd64 (A64), this change saves about
40 cycles or 30%. I think it is inherently only about 12 cycles faster
in these cases and the rest of the speedup is from partly-accidentally
avoiding compiler pessimizations (the construction of 2**k is now
manually scheduled for good results, and -O2 doesn't always mess this
up). In most cases on amd64 (A64) and i386 (A64) the speedup is about
20 cycles. The worst case that I found is expf on ia64 where this
change is a pessimization of about 10 cycles or 5%. The manual
scheduling for plain exp[f] is harder and not as tuned.
This change ld128/s_exp2l.c has not been tested.
Diffstat (limited to 'lib/msun/ld128')
-rw-r--r-- | lib/msun/ld128/s_exp2l.c | 26 |
1 files changed, 15 insertions, 11 deletions
diff --git a/lib/msun/ld128/s_exp2l.c b/lib/msun/ld128/s_exp2l.c index 28a6935..a94b892 100644 --- a/lib/msun/ld128/s_exp2l.c +++ b/lib/msun/ld128/s_exp2l.c @@ -356,8 +356,8 @@ static const float eps[TBLSIZE] = { long double exp2l(long double x) { - union IEEEl2bits u; - long double r, t, z; + union IEEEl2bits u, v; + long double r, t, twopk, twopkp10000, z; uint32_t hx, ix, i0; int k; @@ -403,6 +403,15 @@ exp2l(long double x) i0 = i0 & (TBLSIZE - 1); u.e -= redux; z = x - u.e; + v.xbits.manh = 0; + v.xbits.manl = 0; + if (k >= LDBL_MIN_EXP) { + v.xbits.expsign = LDBL_MAX_EXP - 1 + k; + twopk = v.e; + } else { + v.xbits.expsign = LDBL_MAX_EXP - 1 + k + 10000; + twopkp10000 = v.e; + } /* Compute r = exp2(y) = exp2t[i0] * p(z - eps[i]). */ t = tbl[i0]; /* exp2t[i0] */ @@ -412,15 +421,10 @@ exp2l(long double x) /* Scale by 2**k. */ if(k >= LDBL_MIN_EXP) { - if (k != 0) { - u.e = r; - u.xbits.expsign += k; - r = u.e; - } - return (r); + if (k == LDBL_MAX_EXP) + return (r * 2.0 * 0x1p16383L); + return (r * twopk); } else { - u.e = r; - u.xbits.expsign += k + 10000; - return (u.e * twom10000); + return (r * twopkp10000 * twom10000); } } |