NCG lowering of sqrt

[Ryan Yates]

Hi Kavon,

I looked a bit and it does not appear that there is an SSE sqrt in the
native code gen.  It should be easy to add (see a similar addition here:
https://phabricator.haskell.org/D3265).  The x87 version was available for
32-bit.  I think if you use the LLVM backend it will give you the SSE sqrt.

Ryan

On Fri, Apr 28, 2017 at 9:27 AM, Kavon Farvardin <kavon at farvard.in> wrote:

> Given a Cmm expression such as
>
>     (_c8Gq::F64) = call MO_F64_Sqrt(_s8oX::F64);   // CmmUnsafeForeignCall
>
> the native code generator produces an actual call to the sqrt C function,
> which has the side-effect of causing all floating-point registers to be
> dumped as they are caller-saved. In the nbody benchmark, this is
> particularly bad for a rather hot piece of code (see below).
>
> Ideally the NCG would recognize this foreign call and instead use the
> `sqrtsd` SSE instruction when targeting x86-64.
>
> Does anyone know if the NCG can produce this instruction? I think it would
> be beneficial, as the below would turn into one or two instructions.
>
> Other math functions such as sin/cos require x87 FPU instructions, which
> as far as I know we're not using.
>
>
> ;;;;;;;;;;;
> ; NCG generates this in parts of the nbody benchmark
> ; to compute the sqrt
> ;
>     subq $8,%rsp
>     movsd %xmm9,176(%rsp)  ; all floating-point registers
>     movsd %xmm1,184(%rsp)  ; are caller-saved in SysV ABI
>     movsd %xmm2,192(%rsp)
>     movsd %xmm3,200(%rsp)
>     movq %rdi,208(%rsp)
>     movq %rcx,216(%rsp)
>     movq %rsi,224(%rsp)
>     movsd %xmm4,232(%rsp)
>     movsd %xmm5,240(%rsp)
>     movsd %xmm6,248(%rsp)
>     movsd %xmm7,256(%rsp)
>     movsd %xmm8,264(%rsp)
>     movsd %xmm11,272(%rsp)
>     call _sqrt
>     ;; the loads
>     ;; below are interleaved
>     ;; with computations
>     addq $8,%rsp
>     movsd 264(%rsp),%xmm1
>     movsd 240(%rsp),%xmm2
>     movsd 224(%rsp),%xmm2
>     movsd 232(%rsp),%xmm4
>     movq 200(%rsp),%rax
>     movsd 248(%rsp),%xmm4
>     movsd 256(%rsp),%xmm4
>     movq 216(%rsp),%rcx
>     movsd 192(%rsp),%xmm2
>
>
> ~kavon
>
> _______________________________________________
> ghc-devs mailing list
> ghc-devs at haskell.org
> http://mail.haskell.org/cgi-bin/mailman/listinfo/ghc-devs
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/ghc-devs/attachments/20170428/725f650b/attachment.html>