mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-11-15 20:06:46 +00:00
b118a073d7
- Rewrite/merge pseudo-atomic instruction emitters to address the
following issue:
* Reduce one unnecessary load in spin-loop
previously the spin-loop looks like
thisMBB:
newMBB:
ld t1 = [bitinstr.addr]
op t2 = t1, [bitinstr.val]
not t3 = t2 (if Invert)
mov EAX = t1
lcs dest = [bitinstr.addr], t3 [EAX is implicit]
bz newMBB
fallthrough -->nextMBB
the 'ld' at the beginning of newMBB should be lift out of the loop
as lcs (or CMPXCHG on x86) will load the current memory value into
EAX. This loop is refined as:
thisMBB:
EAX = LOAD [MI.addr]
mainMBB:
t1 = OP [MI.val], EAX
LCMPXCHG [MI.addr], t1, [EAX is implicitly used & defined]
JNE mainMBB
sinkMBB:
* Remove immopc as, so far, all pseudo-atomic instructions has
all-register form only, there is no immedidate operand.
* Remove unnecessary attributes/modifiers in pseudo-atomic instruction
td
* Fix issues in PR13458
- Add comprehensive tests on atomic ops on various data types.
NOTE: Some of them are turned off due to missing functionality.
- Revise tests due to the new spin-loop generated.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@164281 91177308-0d34-0410-b5e6-96231b3b80d8
217 lines
4.6 KiB
LLVM
217 lines
4.6 KiB
LLVM
; RUN: llc < %s -O0 -march=x86-64 -mcpu=corei7 | FileCheck %s --check-prefix X64
|
|
|
|
@sc64 = external global i64
|
|
|
|
define void @atomic_fetch_add64() nounwind {
|
|
; X64: atomic_fetch_add64
|
|
entry:
|
|
%t1 = atomicrmw add i64* @sc64, i64 1 acquire
|
|
; X64: lock
|
|
; X64: incq
|
|
%t2 = atomicrmw add i64* @sc64, i64 3 acquire
|
|
; X64: lock
|
|
; X64: addq $3
|
|
%t3 = atomicrmw add i64* @sc64, i64 5 acquire
|
|
; X64: lock
|
|
; X64: xaddq
|
|
%t4 = atomicrmw add i64* @sc64, i64 %t3 acquire
|
|
; X64: lock
|
|
; X64: addq
|
|
ret void
|
|
; X64: ret
|
|
}
|
|
|
|
define void @atomic_fetch_sub64() nounwind {
|
|
; X64: atomic_fetch_sub64
|
|
%t1 = atomicrmw sub i64* @sc64, i64 1 acquire
|
|
; X64: lock
|
|
; X64: decq
|
|
%t2 = atomicrmw sub i64* @sc64, i64 3 acquire
|
|
; X64: lock
|
|
; X64: subq $3
|
|
%t3 = atomicrmw sub i64* @sc64, i64 5 acquire
|
|
; X64: lock
|
|
; X64: xaddq
|
|
%t4 = atomicrmw sub i64* @sc64, i64 %t3 acquire
|
|
; X64: lock
|
|
; X64: subq
|
|
ret void
|
|
; X64: ret
|
|
}
|
|
|
|
define void @atomic_fetch_and64() nounwind {
|
|
; X64: atomic_fetch_and64
|
|
%t1 = atomicrmw and i64* @sc64, i64 3 acquire
|
|
; X64: lock
|
|
; X64: andq $3
|
|
%t2 = atomicrmw and i64* @sc64, i64 5 acquire
|
|
; X64: andq
|
|
; X64: lock
|
|
; X64: cmpxchgq
|
|
%t3 = atomicrmw and i64* @sc64, i64 %t2 acquire
|
|
; X64: lock
|
|
; X64: andq
|
|
ret void
|
|
; X64: ret
|
|
}
|
|
|
|
define void @atomic_fetch_or64() nounwind {
|
|
; X64: atomic_fetch_or64
|
|
%t1 = atomicrmw or i64* @sc64, i64 3 acquire
|
|
; X64: lock
|
|
; X64: orq $3
|
|
%t2 = atomicrmw or i64* @sc64, i64 5 acquire
|
|
; X64: orq
|
|
; X64: lock
|
|
; X64: cmpxchgq
|
|
%t3 = atomicrmw or i64* @sc64, i64 %t2 acquire
|
|
; X64: lock
|
|
; X64: orq
|
|
ret void
|
|
; X64: ret
|
|
}
|
|
|
|
define void @atomic_fetch_xor64() nounwind {
|
|
; X64: atomic_fetch_xor64
|
|
%t1 = atomicrmw xor i64* @sc64, i64 3 acquire
|
|
; X64: lock
|
|
; X64: xorq $3
|
|
%t2 = atomicrmw xor i64* @sc64, i64 5 acquire
|
|
; X64: xorq
|
|
; X64: lock
|
|
; X64: cmpxchgq
|
|
%t3 = atomicrmw xor i64* @sc64, i64 %t2 acquire
|
|
; X64: lock
|
|
; X64: xorq
|
|
ret void
|
|
; X64: ret
|
|
}
|
|
|
|
define void @atomic_fetch_nand64(i64 %x) nounwind {
|
|
; X64: atomic_fetch_nand64
|
|
; X32: atomic_fetch_nand64
|
|
%t1 = atomicrmw nand i64* @sc64, i64 %x acquire
|
|
; X64: andq
|
|
; X64: notq
|
|
; X64: lock
|
|
; X64: cmpxchgq
|
|
; X32: andl
|
|
; X32: andl
|
|
; X32: notl
|
|
; X32: notl
|
|
; X32: lock
|
|
; X32: cmpxchg8b
|
|
ret void
|
|
; X64: ret
|
|
; X32: ret
|
|
}
|
|
|
|
define void @atomic_fetch_max64(i64 %x) nounwind {
|
|
%t1 = atomicrmw max i64* @sc64, i64 %x acquire
|
|
; X64: cmpq
|
|
; X64: cmov
|
|
; X64: lock
|
|
; X64: cmpxchgq
|
|
|
|
; X32: cmpl
|
|
; X32: cmpl
|
|
; X32: cmov
|
|
; X32: cmov
|
|
; X32: cmov
|
|
; X32: lock
|
|
; X32: cmpxchg8b
|
|
ret void
|
|
; X64: ret
|
|
; X32: ret
|
|
}
|
|
|
|
define void @atomic_fetch_min64(i64 %x) nounwind {
|
|
%t1 = atomicrmw min i64* @sc64, i64 %x acquire
|
|
; X64: cmpq
|
|
; X64: cmov
|
|
; X64: lock
|
|
; X64: cmpxchgq
|
|
|
|
; X32: cmpl
|
|
; X32: cmpl
|
|
; X32: cmov
|
|
; X32: cmov
|
|
; X32: cmov
|
|
; X32: lock
|
|
; X32: cmpxchg8b
|
|
ret void
|
|
; X64: ret
|
|
; X32: ret
|
|
}
|
|
|
|
define void @atomic_fetch_umax64(i64 %x) nounwind {
|
|
%t1 = atomicrmw umax i64* @sc64, i64 %x acquire
|
|
; X64: cmpq
|
|
; X64: cmov
|
|
; X64: lock
|
|
; X64: cmpxchgq
|
|
|
|
; X32: cmpl
|
|
; X32: cmpl
|
|
; X32: cmov
|
|
; X32: cmov
|
|
; X32: cmov
|
|
; X32: lock
|
|
; X32: cmpxchg8b
|
|
ret void
|
|
; X64: ret
|
|
; X32: ret
|
|
}
|
|
|
|
define void @atomic_fetch_umin64(i64 %x) nounwind {
|
|
%t1 = atomicrmw umin i64* @sc64, i64 %x acquire
|
|
; X64: cmpq
|
|
; X64: cmov
|
|
; X64: lock
|
|
; X64: cmpxchgq
|
|
|
|
; X32: cmpl
|
|
; X32: cmpl
|
|
; X32: cmov
|
|
; X32: cmov
|
|
; X32: cmov
|
|
; X32: lock
|
|
; X32: cmpxchg8b
|
|
ret void
|
|
; X64: ret
|
|
; X32: ret
|
|
}
|
|
|
|
define void @atomic_fetch_cmpxchg64() nounwind {
|
|
%t1 = cmpxchg i64* @sc64, i64 0, i64 1 acquire
|
|
; X64: lock
|
|
; X64: cmpxchgq
|
|
; X32: lock
|
|
; X32: cmpxchg8b
|
|
ret void
|
|
; X64: ret
|
|
; X32: ret
|
|
}
|
|
|
|
define void @atomic_fetch_store64(i64 %x) nounwind {
|
|
store atomic i64 %x, i64* @sc64 release, align 8
|
|
; X64-NOT: lock
|
|
; X64: movq
|
|
; X32: lock
|
|
; X32: cmpxchg8b
|
|
ret void
|
|
; X64: ret
|
|
; X32: ret
|
|
}
|
|
|
|
define void @atomic_fetch_swap64(i64 %x) nounwind {
|
|
%t1 = atomicrmw xchg i64* @sc64, i64 %x acquire
|
|
; X64-NOT: lock
|
|
; X64: xchgq
|
|
; X32: lock
|
|
; X32: xchg8b
|
|
ret void
|
|
; X64: ret
|
|
; X32: ret
|
|
}
|