compiler: remove old atomics workaround for AVR

The bug should have been fixed in any LLVM version that we currently
support.
This commit is contained in:
Ayke van Laethem
2024-06-22 17:13:04 +02:00
committed by Ron Evans
parent 1270a50104
commit 6abaee4640
-17
View File
@@ -1,9 +1,6 @@
package compiler
import (
"fmt"
"strings"
"tinygo.org/x/go-llvm"
)
@@ -15,20 +12,6 @@ func (b *builder) createAtomicOp(name string) llvm.Value {
case "AddInt32", "AddInt64", "AddUint32", "AddUint64", "AddUintptr":
ptr := b.getValue(b.fn.Params[0], getPos(b.fn))
val := b.getValue(b.fn.Params[1], getPos(b.fn))
if strings.HasPrefix(b.Triple, "avr") {
// AtomicRMW does not work on AVR as intended:
// - There are some register allocation issues (fixed by https://reviews.llvm.org/D97127 which is not yet in a usable LLVM release)
// - The result is the new value instead of the old value
vType := val.Type()
name := fmt.Sprintf("__sync_fetch_and_add_%d", vType.IntTypeWidth()/8)
fn := b.mod.NamedFunction(name)
if fn.IsNil() {
fn = llvm.AddFunction(b.mod, name, llvm.FunctionType(vType, []llvm.Type{ptr.Type(), vType}, false))
}
oldVal := b.createCall(fn.GlobalValueType(), fn, []llvm.Value{ptr, val}, "")
// Return the new value, not the original value returned.
return b.CreateAdd(oldVal, val, "")
}
oldVal := b.CreateAtomicRMW(llvm.AtomicRMWBinOpAdd, ptr, val, llvm.AtomicOrderingSequentiallyConsistent, true)
// Return the new value, not the original value returned by atomicrmw.
return b.CreateAdd(oldVal, val, "")