diff options
| author | Adhemerval Zanella Netto <adhemerval.zanella@linaro.org> | 2023-03-20 13:01:17 -0300 |
|---|---|---|
| committer | Adhemerval Zanella <adhemerval.zanella@linaro.org> | 2023-04-03 16:45:18 -0300 |
| commit | cf9cf33199fdd6550920ad43f19ad8b2435fc0c6 (patch) | |
| tree | e03b3fb7d42424e4b7095a01b334e8db8296c1e8 /math | |
| parent | 34b9f8bc170810c44184ad57ecf1800587e752a6 (diff) | |
| download | glibc-cf9cf33199fdd6550920ad43f19ad8b2435fc0c6.tar.gz glibc-cf9cf33199fdd6550920ad43f19ad8b2435fc0c6.tar.xz glibc-cf9cf33199fdd6550920ad43f19ad8b2435fc0c6.zip | |
math: Improve fmodf
This uses a new algorithm similar to already proposed earlier [1].
With x = mx * 2^ex and y = my * 2^ey (mx, my, ex, ey being integers),
the simplest implementation is:
mx * 2^ex == 2 * mx * 2^(ex - 1)
while (ex > ey)
{
mx *= 2;
--ex;
mx %= my;
}
With mx/my being mantissa of double floating pointer, on each step the
argument reduction can be improved 8 (which is sizeof of uint32_t minus
MANTISSA_WIDTH plus the signal bit):
while (ex > ey)
{
mx << 8;
ex -= 8;
mx %= my;
} */
The implementation uses builtin clz and ctz, along with shifts to
convert hx/hy back to doubles. Different than the original patch,
this path assume modulo/divide operation is slow, so use multiplication
with invert values.
I see the following performance improvements using fmod benchtests
(result only show the 'mean' result):
Architecture | Input | master | patch
-----------------|-----------------|----------|--------
x86_64 (Ryzen 9) | subnormals | 17.2549 | 12.0318
x86_64 (Ryzen 9) | normal | 85.4096 | 49.9641
x86_64 (Ryzen 9) | close-exponents | 19.1072 | 15.8224
aarch64 (N1) | subnormal | 10.2182 | 6.81778
aarch64 (N1) | normal | 60.0616 | 20.3667
aarch64 (N1) | close-exponents | 11.5256 | 8.39685
I also see similar improvements on arm-linux-gnueabihf when running on
the N1 aarch64 chips, where it a lot of soft-fp implementation (for
modulo, and multiplication):
Architecture | Input | master | patch
-----------------|-----------------|----------|--------
armhf (N1) | subnormal | 11.6662 | 10.8955
armhf (N1) | normal | 69.2759 | 34.1524
armhf (N1) | close-exponents | 13.6472 | 18.2131
Instead of using the math_private.h definitions, I used the
math_config.h instead which is used on newer math implementations.
Co-authored-by: kirill <kirill.okhotnikov@gmail.com>
[1] https://sourceware.org/pipermail/libc-alpha/2020-November/119794.html
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
Diffstat (limited to 'math')
| -rw-r--r-- | math/libm-test-fmod.inc | 4 |
1 files changed, 4 insertions, 0 deletions
diff --git a/math/libm-test-fmod.inc b/math/libm-test-fmod.inc index 8841c1332c..43376c3df2 100644 --- a/math/libm-test-fmod.inc +++ b/math/libm-test-fmod.inc @@ -213,6 +213,10 @@ static const struct test_ff_f_data fmod_test_data[] = TEST_ff_f (fmod, -0x1p127L, -0x3p-148L, -0x1p-147L, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmod, -0x1p127L, 0x3p-126L, -0x1p-125L, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmod, -0x1p127L, -0x3p-126L, -0x1p-125L, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmod, 0x1.3a3e6p-127, 0x1.8b8338p-128, 0x1.d1f31p-129, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmod, 0x1.3a3e6p-127, -0x1.8b8338p-128, 0x1.d1f31p-129, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmod, -0x1.3a3e6p-127, 0x1.8b8338p-128, -0x1.d1f31p-129, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), + TEST_ff_f (fmod, -0x1.3a3e6p-127, -0x1.8b8338p-128, -0x1.d1f31p-129, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), #if !TEST_COND_binary32 TEST_ff_f (fmod, 0x1p1023L, 0x3p-1074L, 0x1p-1073L, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), TEST_ff_f (fmod, 0x1p1023L, -0x3p-1074L, 0x1p-1073L, NO_INEXACT_EXCEPTION|ERRNO_UNCHANGED), |