forks/go-ethereum
1
0
Fork 1
mirror of https://github.com/ethereum/go-ethereum.git synced 2026-07-23 05:06:43 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

chore: update fast_XOR/AND/OR_Bytes optimization

Browse source
This commit is contained in:
Sahil-4555 2025-10-12 13:25:53 +05:30
parent bceb7fabef
commit 55fdee27cd
1 changed files with 78 additions and 87 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

149
common/bitutil/bitutil.go
View file

@ -27,36 +27,32 @@ func XORBytes(dst, a, b []byte) int {
// fastXORBytes xors in bulk. It only works on architectures that support
// unaligned read/writes.
func fastXORBytes(dst, a, b []byte) int {
n := min(len(b), len(a))
n := min(len(a), len(b), len(dst))
if n == 0 {
return 0
}
wordSize := int(unsafe.Sizeof(uintptr(0)))
w := n / wordSize
i := 0
for ; i < n && uintptr(unsafe.Pointer(&a[i]))%uintptr(wordSize) != 0; i++ {
dst[i] = a[i] ^ b[i]
}
if w > 0 {
dw := *(*[]uintptr)(unsafe.Pointer(&dst))
aw := *(*[]uintptr)(unsafe.Pointer(&a))
bw := *(*[]uintptr)(unsafe.Pointer(&b))
w := (n - i) / wordSize
if w > 0 {
ad := unsafe.Slice((*uintptr)(unsafe.Pointer(&a[i])), w)
bd := unsafe.Slice((*uintptr)(unsafe.Pointer(&b[i])), w)
dd := unsafe.Slice((*uintptr)(unsafe.Pointer(&dst[i])), w)
for j := 0; j < w; j++ {
dd[j] = ad[j] ^ bd[j]
}
i += w * wordSize
}
i := 0
for i <= w-4 {
dw[i] = aw[i] ^ bw[i]
dw[i+1] = aw[i+1] ^ bw[i+1]
dw[i+2] = aw[i+2] ^ bw[i+2]
dw[i+3] = aw[i+3] ^ bw[i+3]
i += 4
}
// Handle remaining words
for i < w {
dw[i] = aw[i] ^ bw[i]
i++
}
}
// Handle remaining bytes
start := w * wordSize
for i := 0; i < n&(wordSize-1); i++ {
dst[start+i] = a[start+i] ^ b[start+i]
}
return n
for ; i < n; i++ {
dst[i] = a[i] ^ b[i]
}
return n
}
// safeXORBytes xors one by one. It works on all architectures, independent if
@ -81,35 +77,32 @@ func ANDBytes(dst, a, b []byte) int {
// fastANDBytes ands in bulk. It only works on architectures that support
// unaligned read/writes.
func fastANDBytes(dst, a, b []byte) int {
n := min(len(b), len(a))
w := n / wordSize
n := min(len(a), len(b), len(dst))
if n == 0 {
return 0
}
wordSize := int(unsafe.Sizeof(uintptr(0)))
if w > 0 {
dw := *(*[]uintptr)(unsafe.Pointer(&dst))
aw := *(*[]uintptr)(unsafe.Pointer(&a))
bw := *(*[]uintptr)(unsafe.Pointer(&b))
i := 0
for ; i < n && uintptr(unsafe.Pointer(&a[i]))%uintptr(wordSize) != 0; i++ {
dst[i] = a[i] & b[i]
}
i := 0
for i <= w-4 {
dw[i] = aw[i] & bw[i]
dw[i+1] = aw[i+1] & bw[i+1]
dw[i+2] = aw[i+2] & bw[i+2]
dw[i+3] = aw[i+3] & bw[i+3]
i += 4
}
// Handle remaining words
for i < w {
dw[i] = aw[i] & bw[i]
i++
}
}
w := (n - i) / wordSize
if w > 0 {
ad := unsafe.Slice((*uintptr)(unsafe.Pointer(&a[i])), w)
bd := unsafe.Slice((*uintptr)(unsafe.Pointer(&b[i])), w)
dd := unsafe.Slice((*uintptr)(unsafe.Pointer(&dst[i])), w)
for j := 0; j < w; j++ {
dd[j] = ad[j] & bd[j]
}
i += w * wordSize
}
// Handle remaining bytes
start := w * wordSize
for i := 0; i < n&(wordSize-1); i++ {
dst[start+i] = a[start+i] & b[start+i]
}
return n
for ; i < n; i++ {
dst[i] = a[i] & b[i]
}
return n
}
// safeANDBytes ands one by one. It works on all architectures, independent if
@ -134,34 +127,32 @@ func ORBytes(dst, a, b []byte) int {
// fastORBytes ors in bulk. It only works on architectures that support
// unaligned read/writes.
func fastORBytes(dst, a, b []byte) int {
n := min(len(b), len(a))
w := n / wordSize
if w > 0 {
dw := *(*[]uintptr)(unsafe.Pointer(&dst))
aw := *(*[]uintptr)(unsafe.Pointer(&a))
bw := *(*[]uintptr)(unsafe.Pointer(&b))
n := min(len(a), len(b), len(dst))
if n == 0 {
return 0
}
wordSize := int(unsafe.Sizeof(uintptr(0)))
i := 0
for i <= w-4 {
dw[i] = aw[i] | bw[i]
dw[i+1] = aw[i+1] | bw[i+1]
dw[i+2] = aw[i+2] | bw[i+2]
dw[i+3] = aw[i+3] | bw[i+3]
i += 4
}
// Handle remaining words
for i < w {
dw[i] = aw[i] | bw[i]
i++
}
}
i := 0
for ; i < n && uintptr(unsafe.Pointer(&a[i]))%uintptr(wordSize) != 0; i++ {
dst[i] = a[i] | b[i]
}
// Handle remaining bytes
start := w * wordSize
for i := 0; i < n&(wordSize-1); i++ {
dst[start+i] = a[start+i] | b[start+i]
}
return n
w := (n - i) / wordSize
if w > 0 {
ad := unsafe.Slice((*uintptr)(unsafe.Pointer(&a[i])), w)
bd := unsafe.Slice((*uintptr)(unsafe.Pointer(&b[i])), w)
dd := unsafe.Slice((*uintptr)(unsafe.Pointer(&dst[i])), w)
for j := 0; j < w; j++ {
dd[j] = ad[j] | bd[j]
}
i += w * wordSize
}
for ; i < n; i++ {
dst[i] = a[i] | b[i]
}
return n
}
// safeORBytes ors one by one. It works on all architectures, independent if