Use double-precision routines from arm-optimized-routines
This patch ues exp, exp2, log, log2, and pow from arm optimized
routines. For pow on x86_64, although slight slower it simplifies
the code required on both bionic and arm-optimized-routines (so
there is no need to select and export the symbol based on
architecture).
Performance-wise the improvements are:
x86_64 throughput latency
exp 1.16x 1.16x
log 1.08x 0.95x
exp2 1.27x 1.55x
log2 1.40x 1.43x
pow 0.77x 0.89x *
* I tried to check if AVX2/FMA but without success.
aarch64 throughput latency
exp 2.33x 2.16x
exp2 1.99x 1.50x
log 1.79x 1.43x
log2 2.15x 1.80x
pow 3.81x 3.07x
Test: ran bionic tests on static mode.
Change-Id: Ib16bf3280c5329fd257a3b3f0b6c4f2f3cb34deb
diff --git a/libm/Android.bp b/libm/Android.bp
index fee2bac..da13ab1 100644
--- a/libm/Android.bp
+++ b/libm/Android.bp
@@ -29,7 +29,6 @@
"upstream-freebsd/lib/msun/src/e_atanhf.c",
"upstream-freebsd/lib/msun/src/e_cosh.c",
"upstream-freebsd/lib/msun/src/e_coshf.c",
- "upstream-freebsd/lib/msun/src/e_exp.c",
"upstream-freebsd/lib/msun/src/e_fmod.c",
"upstream-freebsd/lib/msun/src/e_fmodf.c",
"upstream-freebsd/lib/msun/src/e_gamma.c",
@@ -50,9 +49,6 @@
"upstream-freebsd/lib/msun/src/e_lgamma_r.c",
"upstream-freebsd/lib/msun/src/e_log10.c",
"upstream-freebsd/lib/msun/src/e_log10f.c",
- "upstream-freebsd/lib/msun/src/e_log2.c",
- "upstream-freebsd/lib/msun/src/e_log.c",
- "upstream-freebsd/lib/msun/src/e_pow.c",
"upstream-freebsd/lib/msun/src/e_remainder.c",
"upstream-freebsd/lib/msun/src/e_remainderf.c",
"upstream-freebsd/lib/msun/src/e_rem_pio2.c",
@@ -115,7 +111,6 @@
"upstream-freebsd/lib/msun/src/s_ctanhf.c",
"upstream-freebsd/lib/msun/src/s_erf.c",
"upstream-freebsd/lib/msun/src/s_erff.c",
- "upstream-freebsd/lib/msun/src/s_exp2.c",
"upstream-freebsd/lib/msun/src/s_expm1.c",
"upstream-freebsd/lib/msun/src/s_expm1f.c",
"upstream-freebsd/lib/msun/src/s_fdim.c",
@@ -349,11 +344,8 @@
"x86/e_asin.S",
"x86/e_atan2.S",
"x86/e_cosh.S",
- "x86/e_exp.S",
"x86/e_hypot.S",
"x86/e_log10.S",
- "x86/e_log.S",
- "x86/e_pow.S",
"x86/e_sinh.S",
"x86/libm_reduce_pi04l.S",
"x86/libm_sincos_huge.S",
@@ -374,11 +366,8 @@
"upstream-freebsd/lib/msun/src/e_asin.c",
"upstream-freebsd/lib/msun/src/e_atan2.c",
"upstream-freebsd/lib/msun/src/e_cosh.c",
- "upstream-freebsd/lib/msun/src/e_exp.c",
"upstream-freebsd/lib/msun/src/e_hypot.c",
- "upstream-freebsd/lib/msun/src/e_log.c",
"upstream-freebsd/lib/msun/src/e_log10.c",
- "upstream-freebsd/lib/msun/src/e_pow.c",
"upstream-freebsd/lib/msun/src/e_sinh.c",
"upstream-freebsd/lib/msun/src/e_sqrt.c",
"upstream-freebsd/lib/msun/src/e_sqrtf.c",
@@ -430,11 +419,8 @@
"x86_64/e_asin.S",
"x86_64/e_atan2.S",
"x86_64/e_cosh.S",
- "x86_64/e_exp.S",
"x86_64/e_hypot.S",
"x86_64/e_log10.S",
- "x86_64/e_log.S",
- "x86_64/e_pow.S",
"x86_64/e_sinh.S",
"x86_64/lrint.S",
"x86_64/lrintf.S",
@@ -452,11 +438,8 @@
"upstream-freebsd/lib/msun/src/e_asin.c",
"upstream-freebsd/lib/msun/src/e_atan2.c",
"upstream-freebsd/lib/msun/src/e_cosh.c",
- "upstream-freebsd/lib/msun/src/e_exp.c",
"upstream-freebsd/lib/msun/src/e_hypot.c",
- "upstream-freebsd/lib/msun/src/e_log.c",
"upstream-freebsd/lib/msun/src/e_log10.c",
- "upstream-freebsd/lib/msun/src/e_pow.c",
"upstream-freebsd/lib/msun/src/e_sinh.c",
"upstream-freebsd/lib/msun/src/e_sqrt.c",
"upstream-freebsd/lib/msun/src/e_sqrtf.c",
diff --git a/libm/fake_long_double.c b/libm/fake_long_double.c
index d81fa39..ef031cc 100644
--- a/libm/fake_long_double.c
+++ b/libm/fake_long_double.c
@@ -45,3 +45,12 @@
// FreeBSD doesn't have an ld128 implementations of tgammal, so both LP32 and LP64 need this.
long double tgammal(long double x) { return tgamma(x); }
+
+// external/arm-optimized-routines does not provide the long double
+// wrappers for the routines it implements.
+#if (LDBL_MANT_DIG == 53)
+long double expl(long double x) { return exp(x); }
+long double exp2l(long double x) { return exp2(x); }
+long double logl(long double x) { return log(x); }
+long double log2l(long double x) { return log2(x); }
+#endif