[Issue 17965] Unexplained usage of the FPU while function result already in right XMM registers

[d-bugmail--- via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=17965

Walter Bright  changed:

   What|Removed |Added

   Keywords||backend
 CC||bugzi...@digitalmars.com

--

[Issue 17965] Unexplained usage of the FPU while function result already in right XMM registers

[d-bugmail--- via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=17965

Citrus  changed:

   What|Removed |Added

 CC||bitter.ta...@gmx.com

--- Comment #4 from Citrus  ---
This is a side-effect of a silly optimization that turns Point into a complex
number instead of a double[2] as one would expect.
You can see by yourself how the codegen improves if `TYcdouble` is replaced by
`TYdouble` in `elstruct` (in the `if (I64 && targ1 && targ2)` branch):

---
   33090:   55  push   rbp
   33091:   48 8b ecmovrbp,rsp
   33094:   48 8d 05 b5 bf 02 00learax,[rip+0x2bfb5]# 5f050
<_D3foo5Point6__initZ>
   3309b:   66 0f 28 00 movapd xmm0,XMMWORD PTR [rax]
   3309f:   5d  poprbp
   330a0:   c3  ret
---

But I'm pretty sure this change may wreak havoc due to various concerns
regarding the alignment and scalar broadcasting.

The bottom line here is:

> You can avoid such FPU <=> XMM round trips by using the LDC compiler.

Or GDC, everything but DMD is fine.

--

[Issue 17965] Unexplained usage of the FPU while function result already in right XMM registers

[d-bugmail--- via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=17965

--- Comment #3 from ponce  ---
@Basile: I also find that unsettling.

--

[Issue 17965] Unexplained usage of the FPU while function result already in right XMM registers

[d-bugmail--- via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=17965

b2.t...@gmx.com changed:

   What|Removed |Added

   Keywords||wrong-code

--- Comment #2 from b2.t...@gmx.com ---
(In reply to ponce from comment #1)
> You can avoid such FPU <=> XMM round trips by using the LDC compiler. The
> DMD backend tend to generate those.

I know p0nce, but this is really a "pathological case" to me. The "trip", as
you say, that happens at the end is a bug, not just something that could be
better.

--

[Issue 17965] Unexplained usage of the FPU while function result already in right XMM registers

[d-bugmail--- via Digitalmars-d-bugs]

https://issues.dlang.org/show_bug.cgi?id=17965

ponce  changed:

   What|Removed |Added

 CC||alil...@gmail.com

--- Comment #1 from ponce  ---
You can avoid such FPU <=> XMM round trips by using the LDC compiler. The DMD
backend tend to generate those.

--