Search results

AOCL erfc / erfcf accuracy issue (maximum 3-ULP error)

Status: Open.

#57 In amd/aocl-libm-ose;

· leekillough opened on Jun 22, 2026

Vectorised log gives wrong results for +0 and positive subnormal inputs due to fallback conditions

Status: Open.

#56 In amd/aocl-libm-ose;

· rangure opened on Jun 17, 2026

Meta-Issue: Summary of UCRT vs AOCL and recommendations

Status: Open.

#55 In amd/aocl-libm-ose;

· leekillough opened on Jun 14, 2026

amd_coshf returns results up to 1 ULP from correctly-rounded value on some inputs

Status: Open.

#54 In amd/aocl-libm-ose;

· leekillough opened on Jun 14, 2026

Miscellaneous analysis of AOCL perfomance deficiencies with respect to Intel

Status: Open.

#53 In amd/aocl-libm-ose;

· leekillough opened on Jun 13, 2026

AOCL amd_tan() is 2.4x slower than Intel tan()

Status: Open.

#52 In amd/aocl-libm-ose;

· leekillough opened on Jun 13, 2026

Intel atan2()/atan2f() are 2-3x faster than AOCL

Status: Open.

#51 In amd/aocl-libm-ose;

· leekillough opened on Jun 13, 2026

Intel Hyperbolic Functions 1.5x faster than AOCL

Status: Open.

#50 In amd/aocl-libm-ose;

· leekillough opened on Jun 13, 2026

amd_sin() and amd_cos() are 2.3x slower than Intel sin() and cos()

Status: Open.

#49 In amd/aocl-libm-ose;

· leekillough opened on Jun 13, 2026

amd_cbrt() is 4.36 times slower than Intel cbrt()

Status: Open.

#48 In amd/aocl-libm-ose;

· leekillough opened on Jun 13, 2026

vrs8_cosf 12.7 times worse than the claimed max ULP

Status: Open.

#47 In amd/aocl-libm-ose;

· rangure opened on Jun 10, 2026

amd_ilogbf() 42% slower than UCRT ilogbf()

Status: Open.

#46 In amd/aocl-libm-ose;

· leekillough opened on Jun 9, 2026