Add missing tests, fix vext2xv

code/gen_impl.py

@@ -1625,7 +1625,7 @@
             else:
                 sign = "s"
             with open(f"{prefix}ext2xv_{width}_{width2}.h", "w") as f:
-                print(f"for (int i = 0;i < {vlen // w};i++) {{", file=f)
+                print(f"for (int i = 0;i < {256 // w};i++) {{", file=f)
                 print(
                     f"  dst.{m}[i] = ({sign}{w})({sign}{w2})a.{m2}[i];",
                     file=f,

code/gen_tb.py

@@ -67,6 +67,7 @@
         "v128 a, v128 b, int imm",
         [0, 3, 7, 15, 16, 32, 64, 128, 255],
     ),
+    "vext2xv": (widths_vsllwil, "v128 a"),
     "vfrstp": (widths_vfrstp, "v128 a, v128 b, v128 c"),
     "vfrstpi": (widths_vfrstp, "v128 a, v128 b, int imm", [0, 4, 31]),
     "vhaddw": (widths_vexth, "v128 a, v128 b"),
@@ -115,12 +116,11 @@
     "vsat": (widths_all, "v128 a, int imm", [0, 7]),
     "vseq": (widths_signed, "v128 a, v128 b"),
     "vseqi": (widths_signed, "v128 a, int imm", [-16, 0, 15]),
-    "vshuf4i": (["b", "h", "w"], "v128 a, int imm", [0, 13, 100, 128, 255]),
+    "vshuf": (widths_signed, "v128 a, v128 b, v128 c"),
+    "vshuf4i": (["b", "h", "w", "d"], "v128 a, int imm", [0, 13, 100, 128, 255]),
     "vsigncov": (widths_signed, "v128 a, v128 b"),
     "vsllwil": (widths_vsllwil, "v128 a, int imm", [0, 7]),
     "vssub": (widths_all, "v128 a, v128 b"),
-    "vsub": (widths_signed + ["q"], "v128 a, v128 b"),
-    "vsubi": (widths_unsigned, "v128 a, int imm", [0, 31]),
     "vsll": (widths_signed, "v128 a, v128 b"),
     "vslli": (widths_signed, "v128 a, int imm", [0, 7]),
     "vslt": (widths_all, "v128 a, v128 b"),
@@ -151,7 +151,8 @@
     "vssrlni": (widths_vssrlni, "v128 a, v128 b, int imm", [0, 7, 15]),
     "vssrlrn": (widths_vssrln, "v128 a, v128 b"),
     "vssrlrni": (widths_vssrlni, "v128 a, v128 b, int imm", [0, 7, 15]),
-    "vsub": (widths_signed, "v128 a, v128 b"),
+    "vsub": (widths_signed + ["q"], "v128 a, v128 b"),
+    "vsubi": (widths_unsigned, "v128 a, int imm", [0, 31]),
     "vsubwev": (widths_vsubw, "v128 a, v128 b"),
     "vsubwod": (widths_vsubw, "v128 a, v128 b"),
     "vxor": (["v"], "v128 a, v128 b"),
@@ -171,9 +172,15 @@
             # skip xvinsgr2vr_b/h
             if inst_name in ["xvinsgr2vr_b", "xvinsgr2vr_h"]:
                 continue
+            # skip xvext2xv
+            if inst_name.startswith("xvext2xv"):
+                continue
             # skip xvreplvei
             if inst_name.startswith("xvreplvei"):
                 continue
+            # skip xv/vshuf4i.d due to extra argument
+            if inst_name == "vshuf4i_d" or inst_name == "xvshuf4i_d":
+                continue
 
             fuzz_args = 0
             for arg in args.split(", "):
@@ -182,7 +189,8 @@
 
             print(f"Saving {inst_name}.cpp")
             with open(f"{inst_name}.cpp", "w") as f:
-                if prefix == "":
+                # vext2xv is in lasx
+                if prefix == "" and not inst_name.startswith("vext2xv"):
                     vtype = "v128"
                     fuzz = "FUZZ"
                 else:

code/vext2xv_d_b.h

@@ -1,3 +1,3 @@
-for (int i = 0; i < 2; i++) {
+for (int i = 0; i < 4; i++) {
   dst.dword[i] = (s64)(s8)a.byte[i];
 }

code/vext2xv_d_h.h

@@ -1,3 +1,3 @@
-for (int i = 0; i < 2; i++) {
+for (int i = 0; i < 4; i++) {
   dst.dword[i] = (s64)(s16)a.half[i];
 }

code/vext2xv_d_w.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v256 vext2xv_d_w(v256 a) {
+  v256 dst;
+#include "vext2xv_d_w.h"
+  return dst;
+}
+
+void test() { XFUZZ1(vext2xv_d_w); }

code/vext2xv_d_w.h

@@ -1,3 +1,3 @@
-for (int i = 0; i < 2; i++) {
+for (int i = 0; i < 4; i++) {
   dst.dword[i] = (s64)(s32)a.word[i];
 }

code/vext2xv_du_bu.h

@@ -1,3 +1,3 @@
-for (int i = 0; i < 2; i++) {
+for (int i = 0; i < 4; i++) {
   dst.dword[i] = (u64)(u8)a.byte[i];
 }

code/vext2xv_du_hu.h

@@ -1,3 +1,3 @@
-for (int i = 0; i < 2; i++) {
+for (int i = 0; i < 4; i++) {
   dst.dword[i] = (u64)(u16)a.half[i];
 }

code/vext2xv_du_wu.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v256 vext2xv_du_wu(v256 a) {
+  v256 dst;
+#include "vext2xv_du_wu.h"
+  return dst;
+}
+
+void test() { XFUZZ1(vext2xv_du_wu); }

code/vext2xv_du_wu.h

@@ -1,3 +1,3 @@
-for (int i = 0; i < 2; i++) {
+for (int i = 0; i < 4; i++) {
   dst.dword[i] = (u64)(u32)a.word[i];
 }

code/vext2xv_h_b.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v256 vext2xv_h_b(v256 a) {
+  v256 dst;
+#include "vext2xv_h_b.h"
+  return dst;
+}
+
+void test() { XFUZZ1(vext2xv_h_b); }

code/vext2xv_h_b.h

@@ -1,3 +1,3 @@
-for (int i = 0; i < 8; i++) {
+for (int i = 0; i < 16; i++) {
   dst.half[i] = (s16)(s8)a.byte[i];
 }

code/vext2xv_hu_bu.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v256 vext2xv_hu_bu(v256 a) {
+  v256 dst;
+#include "vext2xv_hu_bu.h"
+  return dst;
+}
+
+void test() { XFUZZ1(vext2xv_hu_bu); }

code/vext2xv_hu_bu.h

@@ -1,3 +1,3 @@
-for (int i = 0; i < 8; i++) {
+for (int i = 0; i < 16; i++) {
   dst.half[i] = (u16)(u8)a.byte[i];
 }

code/vext2xv_w_b.h

@@ -1,3 +1,3 @@
-for (int i = 0; i < 4; i++) {
+for (int i = 0; i < 8; i++) {
   dst.word[i] = (s32)(s8)a.byte[i];
 }

code/vext2xv_w_h.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v256 vext2xv_w_h(v256 a) {
+  v256 dst;
+#include "vext2xv_w_h.h"
+  return dst;
+}
+
+void test() { XFUZZ1(vext2xv_w_h); }

code/vext2xv_w_h.h

@@ -1,3 +1,3 @@
-for (int i = 0; i < 4; i++) {
+for (int i = 0; i < 8; i++) {
   dst.word[i] = (s32)(s16)a.half[i];
 }

code/vext2xv_wu_bu.h

@@ -1,3 +1,3 @@
-for (int i = 0; i < 4; i++) {
+for (int i = 0; i < 8; i++) {
   dst.word[i] = (u32)(u8)a.byte[i];
 }

code/vext2xv_wu_hu.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v256 vext2xv_wu_hu(v256 a) {
+  v256 dst;
+#include "vext2xv_wu_hu.h"
+  return dst;
+}
+
+void test() { XFUZZ1(vext2xv_wu_hu); }

code/vext2xv_wu_hu.h

@@ -1,3 +1,3 @@
-for (int i = 0; i < 4; i++) {
+for (int i = 0; i < 8; i++) {
   dst.word[i] = (u32)(u16)a.half[i];
 }

code/xvperm_w.h

@@ -1,3 +1,3 @@
-for (int i = 0;i < 8;i++) {
+for (int i = 0; i < 8; i++) {
   dst.word[i] = a.word[b.word[i] & 0x8];
 }

code/xvshuf4i_d.cpp

@@ -0,0 +1,15 @@
+#include "common.h"
+
+v256 xvshuf4i_d(v256 a, v256 b, int imm) {
+  v256 dst;
+#include "xvshuf4i_d.h"
+  return dst;
+}
+
+void test() {
+  XFUZZ2(xvshuf4i_d, 0);
+  XFUZZ2(xvshuf4i_d, 13);
+  XFUZZ2(xvshuf4i_d, 100);
+  XFUZZ2(xvshuf4i_d, 128);
+  XFUZZ2(xvshuf4i_d, 255);
+}

code/xvshuf_b.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v256 xvshuf_b(v256 a, v256 b, v256 c) {
+  v256 dst;
+#include "xvshuf_b.h"
+  return dst;
+}
+
+void test() { XFUZZ3(xvshuf_b); }

code/xvshuf_d.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v256 xvshuf_d(v256 a, v256 b, v256 c) {
+  v256 dst;
+#include "xvshuf_d.h"
+  return dst;
+}
+
+void test() { XFUZZ3(xvshuf_d); }

code/xvshuf_h.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v256 xvshuf_h(v256 a, v256 b, v256 c) {
+  v256 dst;
+#include "xvshuf_h.h"
+  return dst;
+}
+
+void test() { XFUZZ3(xvshuf_h); }

code/xvshuf_w.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v256 xvshuf_w(v256 a, v256 b, v256 c) {
+  v256 dst;
+#include "xvshuf_w.h"
+  return dst;
+}
+
+void test() { XFUZZ3(xvshuf_w); }

code/xvsub_q.cpp

@@ -0,0 +1,9 @@
+#include "common.h"
+
+v256 xvsub_q(v256 a, v256 b) {
+  v256 dst;
+#include "xvsub_q.h"
+  return dst;
+}
+
+void test() { XFUZZ2(xvsub_q); }