mirror of
https://github.com/tinygo-org/tinygo.git
synced 2026-08-05 03:27:48 +00:00
compiler: implement copy directly
The compiler now implements the copy builtin directly instead of calling sliceCopy. The length is calculated with the llvm.umax.* intrinsics, and the move is performed by llvm.memmove.*. Both of these operations are easily understood by LLVM's optimization passes. The type's alignment is also provided to llvm.memmove.*, which is useful when rewriting the move. Interp no longer needs to reimplement sliceCopy. Some edge case handling was implemented by sliceCopy but not llvm.memmove.*/llvm.memcpy.*. I copied this over, so copies of external slices should work now. Volatile moves/copies are now run at runtime by interp. There is a 4-byte size increase due to some confusing length logic in sendUSBPacket. I will look at sendUSBPacket in a future PR.
This commit is contained in:
Nia Waldvogel
Vendored
+68
@@ -0,0 +1,68 @@
|
||||
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
|
||||
target triple = "x86_64--linux"
|
||||
|
||||
@string = internal unnamed_addr constant [3 x i8] c"foo"
|
||||
@moveDst = global [3 x i8] zeroinitializer
|
||||
@copyDst = global [3 x i8] zeroinitializer
|
||||
|
||||
@externalSrc = external global [2 x i8]
|
||||
@moveExternalDst = global [2 x i8] zeroinitializer
|
||||
|
||||
@moveEscapedSrc = global [4 x i8] c"abcd"
|
||||
@moveEscapedDst = global [4 x i8] zeroinitializer
|
||||
|
||||
@volatileSrc = global [2 x i8] c"xy"
|
||||
@volatileDst = global [2 x i8] zeroinitializer
|
||||
|
||||
declare void @use(ptr)
|
||||
|
||||
define void @runtime.initAll() {
|
||||
call void @main.init()
|
||||
ret void
|
||||
}
|
||||
|
||||
define internal void @main.init() {
|
||||
call void @testMove()
|
||||
call void @testCopy()
|
||||
call void @testMoveExternal()
|
||||
call void @testMoveEscaped()
|
||||
call void @testVolatileCopy()
|
||||
ret void
|
||||
}
|
||||
|
||||
; Test a simple memmove between globals.
|
||||
define internal void @testMove() {
|
||||
call void @llvm.memmove.p0.p0.i64(ptr @moveDst, ptr @string, i64 3, i1 false)
|
||||
ret void
|
||||
}
|
||||
|
||||
; Test a simple memcpy between globals.
|
||||
define internal void @testCopy() {
|
||||
call void @llvm.memcpy.p0.p0.i64(ptr @copyDst, ptr @string, i64 3, i1 false)
|
||||
ret void
|
||||
}
|
||||
|
||||
; Test a memmove from an external global.
|
||||
; This should be run at runtime.
|
||||
define internal void @testMoveExternal() {
|
||||
call void @llvm.memmove.p0.p0.i64(ptr @moveExternalDst, ptr @externalSrc, i64 2, i1 false)
|
||||
ret void
|
||||
}
|
||||
|
||||
; Test a memmove from an escaped (and potentially modified) source buffer.
|
||||
define internal void @testMoveEscaped() {
|
||||
call void @use(ptr @moveEscapedSrc)
|
||||
call void @llvm.memmove.p0.p0.i64(ptr @moveEscapedDst, ptr @moveEscapedSrc, i64 4, i1 false)
|
||||
ret void
|
||||
}
|
||||
|
||||
; Test a volatile memcpy.
|
||||
; This should always be run at runtime.
|
||||
define internal void @testVolatileCopy() {
|
||||
call void @llvm.memcpy.p0.p0.i64(ptr @volatileDst, ptr @volatileSrc, i64 2, i1 true)
|
||||
ret void
|
||||
}
|
||||
|
||||
declare void @llvm.memmove.p0.p0.i64(ptr, ptr, i64, i1)
|
||||
|
||||
declare void @llvm.memcpy.p0.p0.i64(ptr, ptr, i64, i1)
|
||||
Vendored
+27
@@ -0,0 +1,27 @@
|
||||
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
|
||||
target triple = "x86_64--linux"
|
||||
|
||||
@moveDst = local_unnamed_addr global [3 x i8] c"foo"
|
||||
@copyDst = local_unnamed_addr global [3 x i8] c"foo"
|
||||
@externalSrc = external local_unnamed_addr global [2 x i8]
|
||||
@moveExternalDst = local_unnamed_addr global [2 x i8] zeroinitializer
|
||||
@moveEscapedSrc = global [4 x i8] c"abcd"
|
||||
@moveEscapedDst = local_unnamed_addr global [4 x i8] zeroinitializer
|
||||
@volatileSrc = global [2 x i8] c"xy"
|
||||
@volatileDst = global [2 x i8] zeroinitializer
|
||||
|
||||
declare void @use(ptr) local_unnamed_addr
|
||||
|
||||
define void @runtime.initAll() local_unnamed_addr {
|
||||
call void @llvm.memmove.p0.p0.i64(ptr @moveExternalDst, ptr @externalSrc, i64 2, i1 false)
|
||||
call void @use(ptr @moveEscapedSrc)
|
||||
call void @llvm.memmove.p0.p0.i64(ptr @moveEscapedDst, ptr @moveEscapedSrc, i64 4, i1 false)
|
||||
call void @llvm.memcpy.p0.p0.i64(ptr @volatileDst, ptr @volatileSrc, i64 2, i1 true)
|
||||
ret void
|
||||
}
|
||||
|
||||
declare void @llvm.memmove.p0.p0.i64(ptr nocapture writeonly, ptr nocapture readonly, i64, i1 immarg) #0
|
||||
|
||||
declare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #0
|
||||
|
||||
attributes #0 = { nocallback nofree nounwind willreturn memory(argmem: readwrite) }
|
||||
Vendored
+52
@@ -0,0 +1,52 @@
|
||||
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
|
||||
target triple = "x86_64--linux"
|
||||
|
||||
@uminResult = global i32 0
|
||||
@sminResult = global i32 0
|
||||
@umaxResult = global i32 0
|
||||
@smaxResult = global i32 0
|
||||
|
||||
define void @runtime.initAll() {
|
||||
call void @main.init()
|
||||
ret void
|
||||
}
|
||||
|
||||
define internal void @main.init() {
|
||||
call void @testUMin()
|
||||
call void @testSMin()
|
||||
call void @testUMax()
|
||||
call void @testSMax()
|
||||
ret void
|
||||
}
|
||||
|
||||
define internal void @testUMin() {
|
||||
%umin = call i32 @llvm.umin.i32(i32 12, i32 -1)
|
||||
store i32 %umin, ptr @uminResult
|
||||
ret void
|
||||
}
|
||||
|
||||
declare i32 @llvm.umin.i32(i32, i32)
|
||||
|
||||
define internal void @testSMin() {
|
||||
%smin = call i32 @llvm.smin.i32(i32 12, i32 -1)
|
||||
store i32 %smin, ptr @sminResult
|
||||
ret void
|
||||
}
|
||||
|
||||
declare i32 @llvm.smin.i32(i32, i32)
|
||||
|
||||
define internal void @testUMax() {
|
||||
%umax = call i32 @llvm.umax.i32(i32 12, i32 -1)
|
||||
store i32 %umax, ptr @umaxResult
|
||||
ret void
|
||||
}
|
||||
|
||||
declare i32 @llvm.umax.i32(i32, i32)
|
||||
|
||||
define internal void @testSMax() {
|
||||
%smax = call i32 @llvm.smax.i32(i32 12, i32 -1)
|
||||
store i32 %smax, ptr @smaxResult
|
||||
ret void
|
||||
}
|
||||
|
||||
declare i32 @llvm.smax.i32(i32, i32)
|
||||
Vendored
+11
@@ -0,0 +1,11 @@
|
||||
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
|
||||
target triple = "x86_64--linux"
|
||||
|
||||
@uminResult = local_unnamed_addr global i32 12
|
||||
@sminResult = local_unnamed_addr global i32 -1
|
||||
@umaxResult = local_unnamed_addr global i32 -1
|
||||
@smaxResult = local_unnamed_addr global i32 12
|
||||
|
||||
define void @runtime.initAll() local_unnamed_addr {
|
||||
ret void
|
||||
}
|
||||
Vendored
-124
@@ -1,124 +0,0 @@
|
||||
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
|
||||
target triple = "x86_64--linux"
|
||||
|
||||
@main.uint8SliceSrc.buf = internal global [2 x i8] c"\03d"
|
||||
@main.uint8SliceSrc = internal unnamed_addr global { ptr, i64, i64 } { ptr @main.uint8SliceSrc.buf, i64 2, i64 2 }
|
||||
@main.uint8SliceDst = internal unnamed_addr global { ptr, i64, i64 } zeroinitializer
|
||||
@main.int16SliceSrc.buf = internal global [3 x i16] [i16 5, i16 123, i16 1024]
|
||||
@main.int16SliceSrc = internal unnamed_addr global { ptr, i64, i64 } { ptr @main.int16SliceSrc.buf, i64 3, i64 3 }
|
||||
@main.int16SliceDst = internal unnamed_addr global { ptr, i64, i64 } zeroinitializer
|
||||
@main.sliceSrcUntaint.buf = internal global [2 x i8] c"ab"
|
||||
@main.sliceDstUntaint.buf = internal global [2 x i8] zeroinitializer
|
||||
@main.sliceSrcTaint.buf = internal global [2 x i8] c"cd"
|
||||
@main.sliceDstTaint.buf = internal global [2 x i8] zeroinitializer
|
||||
@main.sliceSrcExternal1.buf = external global [2 x i8]
|
||||
@main.sliceDstExternal1.buf = internal global [2 x i8] zeroinitializer
|
||||
@main.sliceSrcExternal2.buf = internal global [2 x i8] zeroinitializer
|
||||
@main.sliceDstExternal2.buf = external global [2 x i8]
|
||||
|
||||
declare i64 @runtime.sliceCopy(ptr %dst, ptr %src, i64 %dstLen, i64 %srcLen, i64 %elemSize) unnamed_addr
|
||||
|
||||
declare ptr @runtime.alloc(i64, ptr) unnamed_addr
|
||||
|
||||
declare void @runtime.printuint8(i8)
|
||||
|
||||
declare void @runtime.printint16(i16)
|
||||
|
||||
declare void @use(ptr)
|
||||
|
||||
define void @runtime.initAll() unnamed_addr {
|
||||
entry:
|
||||
call void @main.init()
|
||||
ret void
|
||||
}
|
||||
|
||||
define void @main() unnamed_addr {
|
||||
entry:
|
||||
; print(uintSliceSrc[0])
|
||||
%uint8SliceSrc.buf = load ptr, ptr @main.uint8SliceSrc
|
||||
%uint8SliceSrc.val = load i8, ptr %uint8SliceSrc.buf
|
||||
call void @runtime.printuint8(i8 %uint8SliceSrc.val)
|
||||
|
||||
; print(uintSliceDst[0])
|
||||
%uint8SliceDst.buf = load ptr, ptr @main.uint8SliceDst
|
||||
%uint8SliceDst.val = load i8, ptr %uint8SliceDst.buf
|
||||
call void @runtime.printuint8(i8 %uint8SliceDst.val)
|
||||
|
||||
; print(int16SliceSrc[0])
|
||||
%int16SliceSrc.buf = load ptr, ptr @main.int16SliceSrc
|
||||
%int16SliceSrc.val = load i16, ptr %int16SliceSrc.buf
|
||||
call void @runtime.printint16(i16 %int16SliceSrc.val)
|
||||
|
||||
; print(int16SliceDst[0])
|
||||
%int16SliceDst.buf = load ptr, ptr @main.int16SliceDst
|
||||
%int16SliceDst.val = load i16, ptr %int16SliceDst.buf
|
||||
call void @runtime.printint16(i16 %int16SliceDst.val)
|
||||
|
||||
; print(sliceDstUntaint[0])
|
||||
%sliceDstUntaint.val = load i8, ptr getelementptr inbounds (i8, ptr @main.sliceDstUntaint.buf, i32 0)
|
||||
call void @runtime.printuint8(i8 %sliceDstUntaint.val)
|
||||
|
||||
; print(sliceDstTaint[0])
|
||||
%sliceDstTaint.val = load i8, ptr getelementptr inbounds (i8, ptr @main.sliceDstTaint.buf, i32 0)
|
||||
call void @runtime.printuint8(i8 %sliceDstTaint.val)
|
||||
|
||||
; print(sliceDstExternal1[0])
|
||||
%sliceDstExternal1.val = load i8, ptr getelementptr inbounds (i8, ptr @main.sliceDstExternal1.buf, i32 0)
|
||||
call void @runtime.printuint8(i8 %sliceDstExternal1.val)
|
||||
|
||||
; print(sliceDstExternal2[0])
|
||||
%sliceDstExternal2.val = load i8, ptr getelementptr inbounds (i8, ptr @main.sliceDstExternal2.buf, i32 0)
|
||||
call void @runtime.printuint8(i8 %sliceDstExternal2.val)
|
||||
|
||||
ret void
|
||||
}
|
||||
|
||||
define internal void @main.init() unnamed_addr {
|
||||
entry:
|
||||
; equivalent of:
|
||||
; uint8SliceDst = make([]uint8, len(uint8SliceSrc))
|
||||
%uint8SliceSrc = load { ptr, i64, i64 }, ptr @main.uint8SliceSrc
|
||||
%uint8SliceSrc.len = extractvalue { ptr, i64, i64 } %uint8SliceSrc, 1
|
||||
%uint8SliceDst.buf = call ptr @runtime.alloc(i64 %uint8SliceSrc.len, ptr null)
|
||||
%0 = insertvalue { ptr, i64, i64 } undef, ptr %uint8SliceDst.buf, 0
|
||||
%1 = insertvalue { ptr, i64, i64 } %0, i64 %uint8SliceSrc.len, 1
|
||||
%2 = insertvalue { ptr, i64, i64 } %1, i64 %uint8SliceSrc.len, 2
|
||||
store { ptr, i64, i64 } %2, ptr @main.uint8SliceDst
|
||||
|
||||
; equivalent of:
|
||||
; copy(uint8SliceDst, uint8SliceSrc)
|
||||
%uint8SliceSrc.buf = extractvalue { ptr, i64, i64 } %uint8SliceSrc, 0
|
||||
%copy.n = call i64 @runtime.sliceCopy(ptr %uint8SliceDst.buf, ptr %uint8SliceSrc.buf, i64 %uint8SliceSrc.len, i64 %uint8SliceSrc.len, i64 1)
|
||||
|
||||
; equivalent of:
|
||||
; int16SliceDst = make([]int16, len(int16SliceSrc))
|
||||
%int16SliceSrc = load { ptr, i64, i64 }, ptr @main.int16SliceSrc
|
||||
%int16SliceSrc.len = extractvalue { ptr, i64, i64 } %int16SliceSrc, 1
|
||||
%int16SliceSrc.len.bytes = mul i64 %int16SliceSrc.len, 2
|
||||
%int16SliceDst.buf = call ptr @runtime.alloc(i64 %int16SliceSrc.len.bytes, ptr null)
|
||||
%3 = insertvalue { ptr, i64, i64 } undef, ptr %int16SliceDst.buf, 0
|
||||
%4 = insertvalue { ptr, i64, i64 } %3, i64 %int16SliceSrc.len, 1
|
||||
%5 = insertvalue { ptr, i64, i64 } %4, i64 %int16SliceSrc.len, 2
|
||||
store { ptr, i64, i64 } %5, ptr @main.int16SliceDst
|
||||
|
||||
; equivalent of:
|
||||
; copy(int16SliceDst, int16SliceSrc)
|
||||
%int16SliceSrc.buf = extractvalue { ptr, i64, i64 } %int16SliceSrc, 0
|
||||
%copy.n2 = call i64 @runtime.sliceCopy(ptr %int16SliceDst.buf, ptr %int16SliceSrc.buf, i64 %int16SliceSrc.len, i64 %int16SliceSrc.len, i64 2)
|
||||
|
||||
; Copy slice that has a known value.
|
||||
%copy.n3 = call i64 @runtime.sliceCopy(ptr @main.sliceDstUntaint.buf, ptr @main.sliceSrcUntaint.buf, i64 2, i64 2, i64 1)
|
||||
|
||||
; Copy slice that might have been modified by the external @use call.
|
||||
; This is a fix for https://github.com/tinygo-org/tinygo/issues/3890.
|
||||
call void @use(ptr @main.sliceSrcTaint.buf)
|
||||
%copy.n4 = call i64 @runtime.sliceCopy(ptr @main.sliceDstTaint.buf, ptr @main.sliceSrcTaint.buf, i64 2, i64 2, i64 1)
|
||||
|
||||
; Test that copying from or into external buffers works correctly.
|
||||
; These copy operations must be done at runtime.
|
||||
; https://github.com/tinygo-org/tinygo/issues/4895
|
||||
%copy.n5 = call i64 @runtime.sliceCopy(ptr @main.sliceDstExternal1.buf, ptr @main.sliceSrcExternal1.buf, i64 2, i64 2, i64 1)
|
||||
%copy.n6 = call i64 @runtime.sliceCopy(ptr @main.sliceDstExternal2.buf, ptr @main.sliceSrcExternal2.buf, i64 2, i64 2, i64 1)
|
||||
|
||||
ret void
|
||||
}
|
||||
Vendored
-42
@@ -1,42 +0,0 @@
|
||||
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
|
||||
target triple = "x86_64--linux"
|
||||
|
||||
@main.sliceSrcTaint.buf = internal global [2 x i8] c"cd"
|
||||
@main.sliceDstTaint.buf = internal global [2 x i8] zeroinitializer
|
||||
@main.sliceSrcExternal1.buf = external global [2 x i8]
|
||||
@main.sliceDstExternal1.buf = internal global [2 x i8] zeroinitializer
|
||||
@main.sliceSrcExternal2.buf = internal global [2 x i8] zeroinitializer
|
||||
@main.sliceDstExternal2.buf = external global [2 x i8]
|
||||
|
||||
declare i64 @runtime.sliceCopy(ptr, ptr, i64, i64, i64) unnamed_addr
|
||||
|
||||
declare void @runtime.printuint8(i8) local_unnamed_addr
|
||||
|
||||
declare void @runtime.printint16(i16) local_unnamed_addr
|
||||
|
||||
declare void @use(ptr) local_unnamed_addr
|
||||
|
||||
define void @runtime.initAll() unnamed_addr {
|
||||
entry:
|
||||
call void @use(ptr @main.sliceSrcTaint.buf)
|
||||
%copy.n4 = call i64 @runtime.sliceCopy(ptr @main.sliceDstTaint.buf, ptr @main.sliceSrcTaint.buf, i64 2, i64 2, i64 1)
|
||||
%copy.n5 = call i64 @runtime.sliceCopy(ptr @main.sliceDstExternal1.buf, ptr @main.sliceSrcExternal1.buf, i64 2, i64 2, i64 1)
|
||||
%copy.n6 = call i64 @runtime.sliceCopy(ptr @main.sliceDstExternal2.buf, ptr @main.sliceSrcExternal2.buf, i64 2, i64 2, i64 1)
|
||||
ret void
|
||||
}
|
||||
|
||||
define void @main() unnamed_addr {
|
||||
entry:
|
||||
call void @runtime.printuint8(i8 3)
|
||||
call void @runtime.printuint8(i8 3)
|
||||
call void @runtime.printint16(i16 5)
|
||||
call void @runtime.printint16(i16 5)
|
||||
call void @runtime.printuint8(i8 97)
|
||||
%sliceDstTaint.val = load i8, ptr @main.sliceDstTaint.buf, align 1
|
||||
call void @runtime.printuint8(i8 %sliceDstTaint.val)
|
||||
%sliceDstExternal1.val = load i8, ptr @main.sliceDstExternal1.buf, align 1
|
||||
call void @runtime.printuint8(i8 %sliceDstExternal1.val)
|
||||
%sliceDstExternal2.val = load i8, ptr @main.sliceDstExternal2.buf, align 1
|
||||
call void @runtime.printuint8(i8 %sliceDstExternal2.val)
|
||||
ret void
|
||||
}
|
||||
Reference in New Issue
Block a user