On April 11, 2013 04:24:26 Simon Peyton-Jones wrote:
It would be fantastic if someone could investigate Levent's suggestion "Of course, implementations can take advantage of the underlying CPU's native floating-point abs/sign functions if available as well, avoiding explicit tests at the Haskell code; based on the underlying platform"
Compiling double test(double value) { return fabs(value) } with -O2 -S in gcc gives you test: .LFB3: .cfi_startproc movsd .LC0(%rip), %xmm1 andpd %xmm1, %xmm0 ret .cfi_endproc .LFE3: .size test, .-test .section .rodata.cst16,"aM",@progbits,16 .align 16 .LC0: .long 4294967295 .long 2147483647 .long 0 .long 0 .ident "GCC: (Debian 4.7.2-5) 4.7.2" .section .note.GNU-stack,"",@progbits That is, it does an andpd (and packed double) operation against the constant 2^63-1 (all 1 except the top bit) to just mask out the sign bit. This should work with NaN and Inf too as the former ignores the sign bit and the later interpets it in the expectd manner. Cheers! -Tyson