Add vbitset

code/gen_impl.py

@@ -112,5 +112,19 @@
             file=f,
         )
         print(f"}}", file=f)
+    with open(f"vbitset_{width}.h", "w") as f:
+        print(f"for (int i = 0;i < {128 // w};i++) {{", file=f)
+        print(
+            f"  dst.{m}[i] = a.{m}[i] | ((u{w})1 << (b.{m}[i] % {w}));",
+            file=f,
+        )
+        print(f"}}", file=f)
+    with open(f"vbitseti_{width}.h", "w") as f:
+        print(f"for (int i = 0;i < {128 // w};i++) {{", file=f)
+        print(
+            f"  dst.{m}[i] = a.{m}[i] | ((u{w})1 << imm);",
+            file=f,
+        )
+        print(f"}}", file=f)
 
 os.system("clang-format -i *.cpp *.h")

code/gen_tb.py

@@ -25,6 +25,8 @@
     "vaddwod": (widths_vaddw, "v128 a, v128 b"),
     "vbitclr": (widths_signed, "v128 a, v128 b"),
     "vbitclri": (widths_signed, "v128 a, int imm", [0, 3, 7]),
+    "vbitset": (widths_signed, "v128 a, v128 b"),
+    "vbitseti": (widths_signed, "v128 a, int imm", [0, 3, 7]),
 }
 
 for name in tb:

code/vbitset_b.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v128 vbitset_b(v128 a, v128 b) {
+  v128 dst;
+#include "vbitset_b.h"
+  return dst;
+}
+
+void test() { FUZZ2(vbitset_b); }

code/vbitset_b.h

@@ -0,0 +1,3 @@
+for (int i = 0; i < 16; i++) {
+  dst.byte[i] = a.byte[i] | ((u8)1 << (b.byte[i] % 8));
+}

code/vbitset_d.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v128 vbitset_d(v128 a, v128 b) {
+  v128 dst;
+#include "vbitset_d.h"
+  return dst;
+}
+
+void test() { FUZZ2(vbitset_d); }

code/vbitset_d.h

@@ -0,0 +1,3 @@
+for (int i = 0; i < 2; i++) {
+  dst.dword[i] = a.dword[i] | ((u64)1 << (b.dword[i] % 64));
+}

code/vbitset_h.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v128 vbitset_h(v128 a, v128 b) {
+  v128 dst;
+#include "vbitset_h.h"
+  return dst;
+}
+
+void test() { FUZZ2(vbitset_h); }

code/vbitset_h.h

@@ -0,0 +1,3 @@
+for (int i = 0; i < 8; i++) {
+  dst.half[i] = a.half[i] | ((u16)1 << (b.half[i] % 16));
+}

code/vbitset_w.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v128 vbitset_w(v128 a, v128 b) {
+  v128 dst;
+#include "vbitset_w.h"
+  return dst;
+}
+
+void test() { FUZZ2(vbitset_w); }

code/vbitset_w.h

@@ -0,0 +1,3 @@
+for (int i = 0; i < 4; i++) {
+  dst.word[i] = a.word[i] | ((u32)1 << (b.word[i] % 32));
+}

code/vbitseti_b.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vbitseti_b(v128 a, int imm) {
+  v128 dst;
+#include "vbitseti_b.h"
+  return dst;
+}
+
+void test() {
+  FUZZ1(vbitseti_b, 0);
+  FUZZ1(vbitseti_b, 3);
+  FUZZ1(vbitseti_b, 7);
+}

code/vbitseti_b.h

@@ -0,0 +1,3 @@
+for (int i = 0; i < 16; i++) {
+  dst.byte[i] = a.byte[i] | ((u8)1 << imm);
+}

code/vbitseti_d.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vbitseti_d(v128 a, int imm) {
+  v128 dst;
+#include "vbitseti_d.h"
+  return dst;
+}
+
+void test() {
+  FUZZ1(vbitseti_d, 0);
+  FUZZ1(vbitseti_d, 3);
+  FUZZ1(vbitseti_d, 7);
+}

code/vbitseti_d.h

@@ -0,0 +1,3 @@
+for (int i = 0; i < 2; i++) {
+  dst.dword[i] = a.dword[i] | ((u64)1 << imm);
+}

code/vbitseti_h.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vbitseti_h(v128 a, int imm) {
+  v128 dst;
+#include "vbitseti_h.h"
+  return dst;
+}
+
+void test() {
+  FUZZ1(vbitseti_h, 0);
+  FUZZ1(vbitseti_h, 3);
+  FUZZ1(vbitseti_h, 7);
+}

code/vbitseti_h.h

@@ -0,0 +1,3 @@
+for (int i = 0; i < 8; i++) {
+  dst.half[i] = a.half[i] | ((u16)1 << imm);
+}

code/vbitseti_w.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vbitseti_w(v128 a, int imm) {
+  v128 dst;
+#include "vbitseti_w.h"
+  return dst;
+}
+
+void test() {
+  FUZZ1(vbitseti_w, 0);
+  FUZZ1(vbitseti_w, 3);
+  FUZZ1(vbitseti_w, 7);
+}

code/vbitseti_w.h

@@ -0,0 +1,3 @@
+for (int i = 0; i < 4; i++) {
+  dst.word[i] = a.word[i] | ((u32)1 << imm);
+}

docs/lsx_bitops/vbitwise.md

@@ -29,4 +29,14 @@ Compute bitwise selection: for each bit position, if the bit in `c` equals to on
 {{ vbitclri('b') }}
 {{ vbitclri('h') }}
 {{ vbitclri('w') }}
-{{ vbitclri('d') }}
+{{ vbitclri('d') }}
+
+{{ vbitset('b') }}
+{{ vbitset('h') }}
+{{ vbitset('w') }}
+{{ vbitset('d') }}
+
+{{ vbitseti('b') }}
+{{ vbitseti('h') }}
+{{ vbitseti('w') }}
+{{ vbitseti('d') }}

main.py

@@ -175,6 +175,25 @@ def vbitclri(name):
             desc=f"Clear the bit specified by `imm` from {width}-bit elements in `a`, save the result in `dst`.",
         )
 
+    @env.macro
+    def vbitset(name):
+        width = widths[name]
+        return instruction(
+            intrinsic=f"__m128i __lsx_vbitset_{name} (__m128i a, __m128i b)",
+            instr=f"vbitset.{name} vr, vr, vr",
+            desc=f"Set the bit specified by elements in `b` from {width}-bit elements in `a`, save the result in `dst`.",
+        )
+
+    @env.macro
+    def vbitseti(name):
+        width = widths[name]
+        imm_upper = width - 1
+        return instruction(
+            intrinsic=f"__m128i __lsx_vbitseti_{name} (__m128i a, imm0_{imm_upper} imm)",
+            instr=f"vbitseti.{name} vr, vr, imm",
+            desc=f"Set the bit specified by `imm` from {width}-bit elements in `a`, save the result in `dst`.",
+        )
+
     @env.macro
     def vshuf_hwd(name):
         width = widths[name]