Changing 'op' here caused us to not enter the store into a map, causing
reemission of the code!! In practice, a simple loop like this:
no_exit: ; preds = %no_exit, %entry
%indvar = phi uint [ %indvar.next, %no_exit ], [ 0, %entry ] ; <uint> [#uses=3]
%tmp.4 = getelementptr "complex long double"* %P, uint %indvar, uint 0 ; <double*> [#uses=1]
store double 0.000000e+00, double* %tmp.4
%indvar.next = add uint %indvar, 1 ; <uint> [#uses=2]
%exitcond = seteq uint %indvar.next, %N ; <bool> [#uses=1]
br bool %exitcond, label %return, label %no_exit
was being code gen'd to:
.LBBtest_1: # no_exit
movl %edx, %esi
shll $4, %esi
movl $0, 4(%eax,%esi)
movl $0, (%eax,%esi)
incl %edx
movl $0, (%eax,%esi)
movl $0, 4(%eax,%esi)
cmpl %ecx, %edx
jne .LBBtest_1 # no_exit
Note that we are doing 4 32-bit stores instead of 2. Now we generate:
.LBBtest_1: # no_exit
movl %edx, %esi
incl %esi
shll $4, %edx
movl $0, (%eax,%edx)
movl $0, 4(%eax,%edx)
cmpl %ecx, %esi
movl %esi, %edx
jne .LBBtest_1 # no_exit
This is much happier, though it would be even better if the increment of ESI
was scheduled after the compare :-/
llvm-svn: 20265
select operations or to shifts that are by a constant. This automatically
implements (with no special code) all of the special cases for shift by 32,
shift by < 32 and shift by > 32.
llvm-svn: 19679
do it. This results in better code on X86 for floats (because if strict
precision is not required, we can elide some more expensive double -> float
conversions like the old isel did), and allows other targets to emit
CopyFromRegs that are not legal for arguments.
llvm-svn: 19668
X86/reg-pressure.ll again, and allows us to do nice things in other cases.
For example, we now codegen this sort of thing:
int %loadload(int *%X, int* %Y) {
%Z = load int* %Y
%Y = load int* %X ;; load between %Z and store
%Q = add int %Z, 1
store int %Q, int* %Y
ret int %Y
}
Into this:
loadload:
mov %EAX, DWORD PTR [%ESP + 4]
mov %EAX, DWORD PTR [%EAX]
mov %ECX, DWORD PTR [%ESP + 8]
inc DWORD PTR [%ECX]
ret
where we weren't able to form the 'inc [mem]' before. This also lets the
instruction selector emit loads in any order it wants to, which can be good
for register pressure as well.
llvm-svn: 19644
the basic block that uses them if possible. This is a big win on X86, as it
lets us fold the argument loads into instructions and reduce register pressure
(by not loading all of the arguments in the entry block).
For this (contrived to show the optimization) testcase:
int %argtest(int %A, int %B) {
%X = sub int 12345, %A
br label %L
L:
%Y = add int %X, %B
ret int %Y
}
we used to produce:
argtest:
mov %ECX, DWORD PTR [%ESP + 4]
mov %EAX, 12345
sub %EAX, %ECX
mov %EDX, DWORD PTR [%ESP + 8]
.LBBargtest_1: # L
add %EAX, %EDX
ret
now we produce:
argtest:
mov %EAX, 12345
sub %EAX, DWORD PTR [%ESP + 4]
.LBBargtest_1: # L
add %EAX, DWORD PTR [%ESP + 8]
ret
This also fixes the FIXME in the code.
BTW, this occurs in real code. 164.gzip shrinks from 8623 to 8608 lines of
.s file. The stack frame in huft_build shrinks from 1644->1628 bytes,
inflate_codes shrinks from 116->108 bytes, and inflate_block from 2620->2612,
due to fewer spills.
Take that alkis. :-)
llvm-svn: 19639
track of how to deal with it, and provide the target with a hook that they
can use to legalize arbitrary operations in arbitrary ways.
Implement custom lowering for a couple of ops, implement promotion for select
operations (which x86 needs).
llvm-svn: 19613
Chris Lattner