[dev.simd] cmd/compile, simd: add VPLZCNT[DQ]

Change-Id: Ifd6d8c12deac9c41722fdf2511d860a334e83438
Reviewed-on: https://go-review.googlesource.com/c/go/+/701915
Reviewed-by: Cherry Mui <cherryyz@google.com>
TryBot-Bypass: Junyang Shao <shaojunyang@google.com>

4 files changed, 92 insertions, 0 deletions

diff --git a/src/simd/_gen/simdgen/ops/Others/categories.yaml b/src/simd/_gen/simdgen/ops/Others/categories.yaml
new file mode 100644
index 0000000000..4489f4f403
--- /dev/null
+++ b/src/simd/_gen/simdgen/ops/Others/categories.yaml
@@ -0,0 +1,5 @@
+!sum
+- go: LeadingZeros
+  commutative: false
+  documentation: !string |-
+    // NAME counts the leading zeros of each element in x.
diff --git a/src/simd/_gen/simdgen/ops/Others/go.yaml b/src/simd/_gen/simdgen/ops/Others/go.yaml
new file mode 100644
index 0000000000..a4fd87407b
--- /dev/null
+++ b/src/simd/_gen/simdgen/ops/Others/go.yaml
@@ -0,0 +1,8 @@
+!sum
+- go: LeadingZeros
+  asm: "VPLZCNT[DQ]"
+  in:
+  - &any
+    go: $t
+  out:
+  - *any
diff --git a/src/simd/internal/simd_test/simd_test.go b/src/simd/internal/simd_test/simd_test.go
index 1d4311d75c..0ebd10d147 100644
--- a/src/simd/internal/simd_test/simd_test.go
+++ b/src/simd/internal/simd_test/simd_test.go
@@ -540,3 +540,20 @@ func TestClearAVXUpperBits(t *testing.T) {
 	checkSlices[int64](t, r, []int64{11, 22, 33, 44})
 	checkSlices[int64](t, s, []int64{9, 18, 27, 36})
 }
+
+func TestLeadingZeros(t *testing.T) {
+	if !simd.HasAVX512() {
+		t.Skip("Test requires HasAVX512, not available on this hardware")
+		return
+	}
+
+	src := []uint64{0b1111, 0}
+	want := []uint64{60, 64}
+	got := make([]uint64, 2)
+	simd.LoadUint64x2Slice(src).LeadingZeros().StoreSlice(got)
+	for i := range 2 {
+		if want[i] != got[i] {
+			t.Errorf("Result incorrect at %d: want %d, got %d", i, want[i], got[i])
+		}
+	}
+}
diff --git a/src/simd/ops_amd64.go b/src/simd/ops_amd64.go
index 39552131bf..c1d0e8338a 100644
--- a/src/simd/ops_amd64.go
+++ b/src/simd/ops_amd64.go
@@ -3298,6 +3298,68 @@ func (x Float64x4) IsNan(y Float64x4) Mask64x4
 // Asm: VCMPPD, CPU Feature: AVX512
 func (x Float64x8) IsNan(y Float64x8) Mask64x8
 
+/* LeadingZeros */
+
+// LeadingZeros counts the leading zeros of each element in x.
+//
+// Asm: VPLZCNTD, CPU Feature: AVX512
+func (x Int32x4) LeadingZeros() Int32x4
+
+// LeadingZeros counts the leading zeros of each element in x.
+//
+// Asm: VPLZCNTD, CPU Feature: AVX512
+func (x Int32x8) LeadingZeros() Int32x8
+
+// LeadingZeros counts the leading zeros of each element in x.
+//
+// Asm: VPLZCNTD, CPU Feature: AVX512
+func (x Int32x16) LeadingZeros() Int32x16
+
+// LeadingZeros counts the leading zeros of each element in x.
+//
+// Asm: VPLZCNTQ, CPU Feature: AVX512
+func (x Int64x2) LeadingZeros() Int64x2
+
+// LeadingZeros counts the leading zeros of each element in x.
+//
+// Asm: VPLZCNTQ, CPU Feature: AVX512
+func (x Int64x4) LeadingZeros() Int64x4
+
+// LeadingZeros counts the leading zeros of each element in x.
+//
+// Asm: VPLZCNTQ, CPU Feature: AVX512
+func (x Int64x8) LeadingZeros() Int64x8
+
+// LeadingZeros counts the leading zeros of each element in x.
+//
+// Asm: VPLZCNTD, CPU Feature: AVX512
+func (x Uint32x4) LeadingZeros() Uint32x4
+
+// LeadingZeros counts the leading zeros of each element in x.
+//
+// Asm: VPLZCNTD, CPU Feature: AVX512
+func (x Uint32x8) LeadingZeros() Uint32x8
+
+// LeadingZeros counts the leading zeros of each element in x.
+//
+// Asm: VPLZCNTD, CPU Feature: AVX512
+func (x Uint32x16) LeadingZeros() Uint32x16
+
+// LeadingZeros counts the leading zeros of each element in x.
+//
+// Asm: VPLZCNTQ, CPU Feature: AVX512
+func (x Uint64x2) LeadingZeros() Uint64x2
+
+// LeadingZeros counts the leading zeros of each element in x.
+//
+// Asm: VPLZCNTQ, CPU Feature: AVX512
+func (x Uint64x4) LeadingZeros() Uint64x4
+
+// LeadingZeros counts the leading zeros of each element in x.
+//
+// Asm: VPLZCNTQ, CPU Feature: AVX512
+func (x Uint64x8) LeadingZeros() Uint64x8
+
 /* Less */
 
 // Less compares for less than.