6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-12-30 02:32:08 +00:00
Code Issues Projects Releases Wiki Activity
05cd886561
llvm-6502/test/Transforms/MemCpyOpt/memcpy.ll

132 lines
4.5 KiB
LLVM
Raw Normal View History

remove a pointless restriction from memcpyopt. It was refusing to optimize two memcpy's like this: copy A <- B copy C <- A if it couldn't prove that noalias(B,C). We can eliminate the copy by producing a memmove instead of memcpy. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@119694 91177308-0d34-0410-b5e6-96231b3b80d8
2010-11-18 08:00:57 +00:00
; RUN: opt < %s -basicaa -memcpyopt -dse -S | FileCheck %s
Allow GVN to hack on memcpy's, making them open to further optimization. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@46693 91177308-0d34-0410-b5e6-96231b3b80d8
2008-02-04 02:59:58 +00:00
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"
target triple = "i686-apple-darwin9"
filecheckize, this is still not optimal, see PR8643 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@119693 91177308-0d34-0410-b5e6-96231b3b80d8
2010-11-18 07:49:32 +00:00
define void @test1({ x86_fp80, x86_fp80 }* sret %agg.result, x86_fp80 %z.0, x86_fp80 %z.1) nounwind {
Allow GVN to hack on memcpy's, making them open to further optimization. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@46693 91177308-0d34-0410-b5e6-96231b3b80d8
2008-02-04 02:59:58 +00:00
entry:
%tmp2 = alloca { x86_fp80, x86_fp80 } ; <{ x86_fp80, x86_fp80 }*> [#uses=1]
%memtmp = alloca { x86_fp80, x86_fp80 }, align 16 ; <{ x86_fp80, x86_fp80 }*> [#uses=2]
Split the Add, Sub, and Mul instruction opcodes into separate integer and floating-point opcodes, introducing FAdd, FSub, and FMul. For now, the AsmParser, BitcodeReader, and IRBuilder all preserve backwards compatability, and the Core LLVM APIs preserve backwards compatibility for IR producers. Most front-ends won't need to change immediately. This implements the first step of the plan outlined here: http://nondot.org/sabre/LLVMNotes/IntegerOverflow.txt git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@72897 91177308-0d34-0410-b5e6-96231b3b80d8
2009-06-04 22:49:04 +00:00
%tmp5 = fsub x86_fp80 0xK80000000000000000000, %z.1 ; <x86_fp80> [#uses=1]
Allow GVN to hack on memcpy's, making them open to further optimization. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@46693 91177308-0d34-0410-b5e6-96231b3b80d8
2008-02-04 02:59:58 +00:00
call void @ccoshl( { x86_fp80, x86_fp80 }* sret %memtmp, x86_fp80 %tmp5, x86_fp80 %z.0 ) nounwind
%tmp219 = bitcast { x86_fp80, x86_fp80 }* %tmp2 to i8* ; <i8*> [#uses=2]
%memtmp20 = bitcast { x86_fp80, x86_fp80 }* %memtmp to i8* ; <i8*> [#uses=1]
call void @llvm.memcpy.i32( i8* %tmp219, i8* %memtmp20, i32 32, i32 16 )
%agg.result21 = bitcast { x86_fp80, x86_fp80 }* %agg.result to i8* ; <i8*> [#uses=1]
call void @llvm.memcpy.i32( i8* %agg.result21, i8* %tmp219, i32 32, i32 16 )
filecheckize, this is still not optimal, see PR8643 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@119693 91177308-0d34-0410-b5e6-96231b3b80d8
2010-11-18 07:49:32 +00:00
; Check that one of the memcpy's are removed.
;; FIXME: PR 8643 We should be able to eliminate the last memcpy here.
; CHECK: @test1
; CHECK: call void @ccoshl
remove a pointless restriction from memcpyopt. It was refusing to optimize two memcpy's like this: copy A <- B copy C <- A if it couldn't prove that noalias(B,C). We can eliminate the copy by producing a memmove instead of memcpy. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@119694 91177308-0d34-0410-b5e6-96231b3b80d8
2010-11-18 08:00:57 +00:00
; CHECK: call void @llvm.memcpy
filecheckize, this is still not optimal, see PR8643 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@119693 91177308-0d34-0410-b5e6-96231b3b80d8
2010-11-18 07:49:32 +00:00
; CHECK-NOT: llvm.memcpy
; CHECK: ret void
Allow GVN to hack on memcpy's, making them open to further optimization. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@46693 91177308-0d34-0410-b5e6-96231b3b80d8
2008-02-04 02:59:58 +00:00
ret void
}
declare void @ccoshl({ x86_fp80, x86_fp80 }* sret , x86_fp80, x86_fp80) nounwind
declare void @llvm.memcpy.i32(i8*, i8*, i32, i32) nounwind
remove a pointless restriction from memcpyopt. It was refusing to optimize two memcpy's like this: copy A <- B copy C <- A if it couldn't prove that noalias(B,C). We can eliminate the copy by producing a memmove instead of memcpy. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@119694 91177308-0d34-0410-b5e6-96231b3b80d8
2010-11-18 08:00:57 +00:00
; The intermediate alloca and one of the memcpy's should be eliminated, the
; other should be related with a memmove.
define void @test2(i8* %P, i8* %Q) nounwind {
%memtmp = alloca { x86_fp80, x86_fp80 }, align 16
%R = bitcast { x86_fp80, x86_fp80 }* %memtmp to i8*
call void @llvm.memcpy.i32( i8* %R, i8* %P, i32 32, i32 16 )
call void @llvm.memcpy.i32( i8* %Q, i8* %R, i32 32, i32 16 )
ret void
; CHECK: @test2
; CHECK-NEXT: call void @llvm.memmove{{.*}}(i8* %Q, i8* %P
; CHECK-NEXT: ret void
}
@x = external global { x86_fp80, x86_fp80 }
define void @test3({ x86_fp80, x86_fp80 }* noalias sret %agg.result) nounwind {
%x.0 = alloca { x86_fp80, x86_fp80 }
%x.01 = bitcast { x86_fp80, x86_fp80 }* %x.0 to i8*
call void @llvm.memcpy.i32( i8* %x.01, i8* bitcast ({ x86_fp80, x86_fp80 }* @x to i8*), i32 32, i32 16 )
%agg.result2 = bitcast { x86_fp80, x86_fp80 }* %agg.result to i8*
call void @llvm.memcpy.i32( i8* %agg.result2, i8* %x.01, i32 32, i32 16 )
ret void
; CHECK: @test3
; CHECK-NEXT: %agg.result2 = bitcast
; CHECK-NEXT: call void @llvm.memcpy
; CHECK-NEXT: ret void
}
Implement PR8644: forwarding a memcpy value to a byval, allowing the memcpy to be eliminated. Unfortunately, the requirements on byval's without explicit alignment are really weak and impossible to predict in the mid-level optimizer, so this doesn't kick in much with current frontends. The fix is to change clang to set alignment on all byval arguments. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@119916 91177308-0d34-0410-b5e6-96231b3b80d8
2010-11-21 00:28:59 +00:00
; PR8644
define void @test4(i8 *%P) {
%A = alloca {i32, i32}
%a = bitcast {i32, i32}* %A to i8*
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* %P, i64 8, i32 4, i1 false)
call void @test4a(i8* byval align 1 %a)
ret void
; CHECK: @test4
; CHECK-NEXT: call void @test4a(
}
declare void @test4a(i8* byval align 1)
declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind
fix a bozo bug I introduced in r119930, causing a miscompile of 20040709-1.c from the gcc testsuite. I was using the size of a pointer instead of the pointee. This fixes rdar://8713376 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@120519 91177308-0d34-0410-b5e6-96231b3b80d8
2010-12-01 01:24:55 +00:00
%struct.S = type { i128, [4 x i8]}
@sS = external global %struct.S, align 16
declare void @test5a(%struct.S* byval align 16) nounwind ssp
declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind
; rdar://8713376 - This memcpy can't be eliminated.
define i32 @test5(i32 %x) nounwind ssp {
entry:
%y = alloca %struct.S, align 16
%tmp = bitcast %struct.S* %y to i8*
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %tmp, i8* bitcast (%struct.S* @sS to i8*), i64 32, i32 16, i1 false)
%a = getelementptr %struct.S* %y, i64 0, i32 1, i64 0
store i8 4, i8* %a
call void @test5a(%struct.S* byval align 16 %y)
ret i32 0
; CHECK: @test5(
; CHECK: store i8 4
; CHECK: call void @test5a(%struct.S* byval align 16 %y)
}
enhance memcpyopt to zap memcpy's that have the same src/dst. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@121362 91177308-0d34-0410-b5e6-96231b3b80d8
2010-12-09 07:45:45 +00:00
;; Noop memcpy should be zapped.
define void @test6(i8 *%P) {
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %P, i8* %P, i64 8, i32 4, i1 false)
ret void
; CHECK: @test6
; CHECK-NEXT: ret void
}
Teach valuetracking that byval arguments with a specified alignment are aligned. Teach memcpyopt to not give up all hope when confonted with an underaligned memcpy feeding an overaligned byval. If the *source* of the memcpy can be determined to be adequeately aligned, or if it can be forced to be, we can eliminate the memcpy. This addresses PR9794. We now compile the example into: define i32 @f(%struct.p* nocapture byval align 8 %q) nounwind ssp { entry: %call = call i32 @g(%struct.p* byval align 8 %q) nounwind ret i32 %call } in both x86-64 and x86-32 mode. We still don't get a tailcall though, because tailcalls apparently can't handle byval. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@131884 91177308-0d34-0410-b5e6-96231b3b80d8
2011-05-23 00:03:39 +00:00
; PR9794 - Should forward memcpy into byval argument even though the memcpy
; isn't itself 8 byte aligned.
%struct.p = type { i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32 }
define i32 @test7(%struct.p* nocapture byval align 8 %q) nounwind ssp {
entry:
%agg.tmp = alloca %struct.p, align 4
%tmp = bitcast %struct.p* %agg.tmp to i8*
%tmp1 = bitcast %struct.p* %q to i8*
call void @llvm.memcpy.p0i8.p0i8.i64(i8* %tmp, i8* %tmp1, i64 48, i32 4, i1 false)
%call = call i32 @g(%struct.p* byval align 8 %agg.tmp) nounwind
ret i32 %call
; CHECK: @test7
; CHECK: call i32 @g(%struct.p* byval align 8 %q) nounwind
}
declare i32 @g(%struct.p* byval align 8)
Reference in New Issue Copy Permalink