[Bug target/109463] suboptimal sequence for converting 64-bit unsigned int to float

[elronnd at elronnd dot net via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109463

--- Comment #3 from elronnd at elronnd dot net ---
Yes, I think the gcc approach of branching is definitely better.  But it's
still a good idea to optimise for size in the cold path.

[Bug c/109463] New: suboptimal sequence for converting 64-bit unsigned int to float

[elronnd at elronnd dot net via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109463

Bug ID: 109463
   Summary: suboptimal sequence for converting 64-bit unsigned int
to float
   Product: gcc
   Version: 12.2.0
Status: UNCONFIRMED
  Severity: normal
  Priority: P3
 Component: c
  Assignee: unassigned at gcc dot gnu.org
  Reporter: elronnd at elronnd dot net
  Target Milestone: ---

double f(uint64_t x) { return x; } gives:

test   rdi,rdi
js 10 
pxor   xmm0,xmm0
cvtsi2sd xmm0,rdi
ret
nop
10:
movrax,rdi
andedi,0x1
pxor   xmm0,xmm0
shrrax,1
or rax,rdi
cvtsi2sd xmm0,rax
addsd  xmm0,xmm0
ret

In particular, the sequence:

movrax,rdi
andedi,0x1
shrrax,1
or rax,rdi
cvtsi2sd xmm0,rax

Can be replaced with:

movzx  eax,dil
shrrdi,1
or rdi,rax
cvtsi2sd xmm0,rdi

Since all 9 low bits of rdi are below the sticky bit, oring them together in
any order suffices to round correctly.

Alternatively, in order to avoid clobbering rdi, use the following sequence:

movrax,rdi
shrrax,1
or al,dil
cvtsi2sd xmm0,rax

(The penalty for partial register access appears to be very cheap or
nonexistent on recent uarchs.)

[Bug c/103113] New: Bad error message with multiply indirect pointer to struct

[elronnd at elronnd dot net via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103113

Bug ID: 103113
   Summary: Bad error message with multiply indirect pointer to
struct
   Product: gcc
   Version: 11.1.0
Status: UNCONFIRMED
  Severity: normal
  Priority: P3
 Component: c
  Assignee: unassigned at gcc dot gnu.org
  Reporter: elronnd at elronnd dot net
  Target Milestone: ---

$ cat q.c
int f() {
struct { int m; } **x;
return x->m;
}
$ gcc -c q.c
q.c: In function ‘f’:
q.c:3:17: error: ‘*x’ is a pointer; did you mean to use ‘->’?
3 | return x->m;
  | ^~
  | ->

[Bug rtl-optimization/97607] New: Spurious sign extension

[elronnd at elronnd dot net via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97607

Bug ID: 97607
   Summary: Spurious sign extension
   Product: gcc
   Version: 11.0
Status: UNCONFIRMED
  Severity: normal
  Priority: P3
 Component: rtl-optimization
  Assignee: unassigned at gcc dot gnu.org
  Reporter: elronnd at elronnd dot net
  Target Milestone: ---

The following code

extern void fun(char);  
void wrapper(char x) { fun(x); }

Should compile 'wrapper' to a single jump, but instead it does this (on amd64):

wrapper:
movsx   edi, dil
jmp fun

Presumably 'x' is getting promoted to int in wrapper, and the promotion never
gets removed.