Files
kelindar-bitmap/bitmap_amd64.go
T
dualfroz 4c67b4458d fix: union/xor of different-sized bitmaps drops bits in variadic path
Or and Xor with extra bitmaps used _or_many/_xor_many, which apply a single
word count to every input. The count came from pointersOf, whose length is
the maximum over the extra slice only and ignores other, so a longer other
lost its tail words. More generally a uniform count cannot combine operands
of different lengths: a longer source is truncated and a shorter one is read
past its end.

Fold each source in with the single-operand SIMD primitive over its own
length, matching the scalar reference in simd_generic.go. Fixes the amd64
accelerated and AVX-512 branches and the arm64 branch.
2026-09-06 00:42:23 +02:00

175 lines
4.7 KiB
Go

// Copyright (c) Roman Atachiants and contributors. All rights reserved.
// Licensed under the MIT license. See LICENSE file in the project root for details.
//go:build !noasm && amd64
package bitmap
import "unsafe"
// And computes the intersection between two bitmaps and stores the result in the current bitmap
func (dst *Bitmap) And(other Bitmap, extra ...Bitmap) {
max := minlen(*dst, other, extra)
dst.shrink(max)
if max == 0 {
return
}
switch hardware {
case isAccelerated:
switch len(extra) {
case 0:
_and(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&other[0]), uint64(max))
default:
vx, _ := pointersOf(other, extra)
_and_many(unsafe.Pointer(&(*dst)[0]), vx, dimensionsOf(max, len(extra)+1))
}
case isAVX512:
switch len(extra) {
case 0:
_and_avx512(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&other[0]), uint64(max))
default:
vx, _ := pointersOf(other, extra)
_and_many_avx512(unsafe.Pointer(&(*dst)[0]), vx, dimensionsOf(max, len(extra)+1))
}
default:
and(*dst, max, other, extra)
return
}
}
// AndNot computes the difference between two bitmaps and stores the result in the current bitmap.
// Operation works as set subtract: dst - b
func (dst *Bitmap) AndNot(other Bitmap, extra ...Bitmap) {
max := minlen(*dst, other, extra)
if max == 0 {
return
}
switch hardware {
case isAccelerated:
switch len(extra) {
case 0:
_andn(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&other[0]), uint64(max))
default:
vx, _ := pointersOf(other, extra)
_andn_many(unsafe.Pointer(&(*dst)[0]), vx, dimensionsOf(max, len(extra)+1))
}
case isAVX512:
switch len(extra) {
case 0:
_andn_avx512(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&other[0]), uint64(max))
default:
vx, _ := pointersOf(other, extra)
_andn_many_avx512(unsafe.Pointer(&(*dst)[0]), vx, dimensionsOf(max, len(extra)+1))
}
default:
andn(*dst, max, other, extra)
return
}
}
// Or computes the union between two bitmaps and stores the result in the current bitmap
func (dst *Bitmap) Or(other Bitmap, extra ...Bitmap) {
max := maxlen(*dst, other, extra)
if max == 0 {
return
}
dst.grow(max - 1)
switch hardware {
case isAccelerated:
switch len(extra) {
case 0:
_or(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&other[0]), uint64(len(other)))
default:
// _or_many uses one word count for every input, so it cannot union
// bitmaps of different lengths. Union each source over its own.
if len(other) > 0 {
_or(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&other[0]), uint64(len(other)))
}
for i := range extra {
if len(extra[i]) > 0 {
_or(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&extra[i][0]), uint64(len(extra[i])))
}
}
}
case isAVX512:
switch len(extra) {
case 0:
_or_avx512(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&other[0]), uint64(len(other)))
default:
if len(other) > 0 {
_or_avx512(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&other[0]), uint64(len(other)))
}
for i := range extra {
if len(extra[i]) > 0 {
_or_avx512(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&extra[i][0]), uint64(len(extra[i])))
}
}
}
default:
or(*dst, other, extra)
}
}
// Xor computes the symmetric difference between two bitmaps and stores the result in the current bitmap
func (dst *Bitmap) Xor(other Bitmap, extra ...Bitmap) {
max := maxlen(*dst, other, extra)
if max == 0 {
return
}
dst.grow(max - 1)
switch hardware {
case isAccelerated:
switch len(extra) {
case 0:
_xor(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&other[0]), uint64(len(other)))
default:
// _xor_many uses one word count for every input, so it cannot combine
// bitmaps of different lengths. Fold in each source over its own.
if len(other) > 0 {
_xor(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&other[0]), uint64(len(other)))
}
for i := range extra {
if len(extra[i]) > 0 {
_xor(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&extra[i][0]), uint64(len(extra[i])))
}
}
}
case isAVX512:
switch len(extra) {
case 0:
_xor_avx512(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&other[0]), uint64(len(other)))
default:
if len(other) > 0 {
_xor_avx512(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&other[0]), uint64(len(other)))
}
for i := range extra {
if len(extra[i]) > 0 {
_xor_avx512(unsafe.Pointer(&(*dst)[0]), unsafe.Pointer(&extra[i][0]), uint64(len(extra[i])))
}
}
}
default:
xor(*dst, other, extra)
}
}
// Count returns the number of elements in this bitmap
func (dst Bitmap) Count() int {
if len(dst) == 0 {
return 0
}
switch hardware {
case isAccelerated:
var res uint64
_count(unsafe.Pointer(&dst[0]), uint64(len(dst)), unsafe.Pointer(&res))
return int(res)
default:
return count(dst)
}
}