mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-30 23:42:52 +01:00
34658dca78
- 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.
llvm-svn: 164281
139 lines
4.2 KiB
LLVM
139 lines
4.2 KiB
LLVM
; RUN: llc < %s -mcpu=generic -march=x86 -mattr=+cmov | FileCheck %s
|
|
|
|
target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
|
|
|
|
define void @func(i32 %argc, i8** %argv) nounwind {
|
|
entry:
|
|
%argc.addr = alloca i32 ; <i32*> [#uses=1]
|
|
%argv.addr = alloca i8** ; <i8***> [#uses=1]
|
|
%val1 = alloca i32 ; <i32*> [#uses=2]
|
|
%val2 = alloca i32 ; <i32*> [#uses=15]
|
|
%andt = alloca i32 ; <i32*> [#uses=2]
|
|
%ort = alloca i32 ; <i32*> [#uses=2]
|
|
%xort = alloca i32 ; <i32*> [#uses=2]
|
|
%old = alloca i32 ; <i32*> [#uses=18]
|
|
%temp = alloca i32 ; <i32*> [#uses=2]
|
|
%temp64 = alloca i64
|
|
store i32 %argc, i32* %argc.addr
|
|
store i8** %argv, i8*** %argv.addr
|
|
store i32 0, i32* %val1
|
|
store i32 31, i32* %val2
|
|
store i32 3855, i32* %andt
|
|
store i32 3855, i32* %ort
|
|
store i32 3855, i32* %xort
|
|
store i32 4, i32* %temp
|
|
%tmp = load i32* %temp
|
|
; CHECK: lock
|
|
; CHECK: xaddl
|
|
%0 = atomicrmw add i32* %val1, i32 %tmp monotonic
|
|
store i32 %0, i32* %old
|
|
; CHECK: lock
|
|
; CHECK: xaddl
|
|
%1 = atomicrmw sub i32* %val2, i32 30 monotonic
|
|
store i32 %1, i32* %old
|
|
; CHECK: lock
|
|
; CHECK: xaddl
|
|
%2 = atomicrmw add i32* %val2, i32 1 monotonic
|
|
store i32 %2, i32* %old
|
|
; CHECK: lock
|
|
; CHECK: xaddl
|
|
%3 = atomicrmw sub i32* %val2, i32 1 monotonic
|
|
store i32 %3, i32* %old
|
|
; CHECK: andl
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%4 = atomicrmw and i32* %andt, i32 4080 monotonic
|
|
store i32 %4, i32* %old
|
|
; CHECK: orl
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%5 = atomicrmw or i32* %ort, i32 4080 monotonic
|
|
store i32 %5, i32* %old
|
|
; CHECK: xorl
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%6 = atomicrmw xor i32* %xort, i32 4080 monotonic
|
|
store i32 %6, i32* %old
|
|
; CHECK: cmov
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%7 = atomicrmw min i32* %val2, i32 16 monotonic
|
|
store i32 %7, i32* %old
|
|
%neg = sub i32 0, 1 ; <i32> [#uses=1]
|
|
; CHECK: cmov
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%8 = atomicrmw min i32* %val2, i32 %neg monotonic
|
|
store i32 %8, i32* %old
|
|
; CHECK: cmov
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%9 = atomicrmw max i32* %val2, i32 1 monotonic
|
|
store i32 %9, i32* %old
|
|
; CHECK: cmov
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%10 = atomicrmw max i32* %val2, i32 0 monotonic
|
|
store i32 %10, i32* %old
|
|
; CHECK: cmov
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%11 = atomicrmw umax i32* %val2, i32 65535 monotonic
|
|
store i32 %11, i32* %old
|
|
; CHECK: cmov
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%12 = atomicrmw umax i32* %val2, i32 10 monotonic
|
|
store i32 %12, i32* %old
|
|
; CHECK: cmov
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%13 = atomicrmw umin i32* %val2, i32 1 monotonic
|
|
store i32 %13, i32* %old
|
|
; CHECK: cmov
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%14 = atomicrmw umin i32* %val2, i32 10 monotonic
|
|
store i32 %14, i32* %old
|
|
; CHECK: xchgl %{{.*}}, {{.*}}(%esp)
|
|
%15 = atomicrmw xchg i32* %val2, i32 1976 monotonic
|
|
store i32 %15, i32* %old
|
|
%neg1 = sub i32 0, 10 ; <i32> [#uses=1]
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%16 = cmpxchg i32* %val2, i32 %neg1, i32 1 monotonic
|
|
store i32 %16, i32* %old
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl
|
|
%17 = cmpxchg i32* %val2, i32 1976, i32 1 monotonic
|
|
store i32 %17, i32* %old
|
|
; CHECK: movl [[R17atomic:.*]], %eax
|
|
; CHECK: movl $1401, %[[R17mask:[a-z]*]]
|
|
; CHECK: andl %eax, %[[R17mask]]
|
|
; CHECK: notl %[[R17mask]]
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl %[[R17mask]], [[R17atomic]]
|
|
; CHECK: jne
|
|
; CHECK: movl %eax,
|
|
%18 = atomicrmw nand i32* %val2, i32 1401 monotonic
|
|
store i32 %18, i32* %old
|
|
; CHECK: andl
|
|
; CHECK: andl
|
|
; CHECK: notl
|
|
; CHECK: notl
|
|
; CHECK: lock
|
|
; CHECK: cmpxchg8b
|
|
%19 = atomicrmw nand i64* %temp64, i64 17361641481138401520 monotonic
|
|
store i64 %19, i64* %temp64
|
|
ret void
|
|
}
|
|
|
|
define void @test2(i32 addrspace(256)* nocapture %P) nounwind {
|
|
entry:
|
|
; CHECK: lock
|
|
; CHECK: cmpxchgl %{{.*}}, %gs:(%{{.*}})
|
|
|
|
%0 = cmpxchg i32 addrspace(256)* %P, i32 0, i32 1 monotonic
|
|
ret void
|
|
}
|