Add vssrani/vssrlni

README.md

@@ -6,18 +6,6 @@ Arranged from QEMU implementation and [GCC Intrinsics](https://gcc.gnu.org/onlin
 
 TODO List:
 
-### vssrln.b.h/h.w/w.d
-
-### vssran.b.h/h.w/w.d
-
-### vssrlrn.b.h/h.w/w.d
-
-### vssrarn.b.h/h.w/w.d
-
-### vssrln.bu.h/hu.w/wu.d
-
-### vssran.bu.h/hu.w/wu.d
-
 ### vssrlrn.bu.h/hu.w/wu.d
 
 ### vssrarn.bu.h/hu.w/wu.d
@@ -118,12 +106,8 @@ TODO List:
 
 ### vsat.b/h/w/d/bu/hu/wu/du
 
-### vssrlni.b.h/h.w/w.d/d.q/bu.h/hu.w/wu.d/du.q
-
 ### vssrlrni.b.h/h.w/w.d/d.q/bu.h/hu.w/wu.d/du.q
 
-### vssrani.b.h/h.w/w.d/d.q/bu.h/hu.w/wu.d/du.q
-
 ### vssrarni.b.h/h.w/w.d/d.q
 
 ### vssrarni.bu.h/hu.w/wu.d/du.q

code/gen_impl.py

@@ -204,6 +204,44 @@
                 print(f"}}", file=f)
                 print(f"}}", file=f)
 
+    for name, shift_sign in [("srl", "u"), ("sra", "s")]:
+        double_width_signed = double_width[:1]
+        with open(f"vs{name}ni_{width}_{double_width_signed}.h", "w") as f:
+            if shift_sign == "u":
+                min = 0
+                if sign == "u":
+                    max = (2**w) - 1
+                else:
+                    max = (2**(w - 1)) - 1
+            else:
+                if sign == "u":
+                    min = 0
+                    max = (2**w) - 1
+                else:
+                    min = -(2 ** (w - 1))
+                    max = (2 ** (w - 1)) - 1
+            print(f"for (int i = 0;i < {128 // w};i++) {{", file=f)
+            print(f"if (i < {64 // w}) {{", file=f)
+            print(
+                f"  {shift_sign}{double_w} temp = ({shift_sign}{double_w})b.{double_m}[i] >> imm;",
+                file=f,
+            )
+            print(
+                f"  dst.{m}[i] = clamp<{shift_sign}{double_w}>(temp, {min}, {max});",
+                file=f,
+            )
+            print(f"}} else {{", file=f)
+            print(
+                f"  {shift_sign}{double_w} temp = ({shift_sign}{double_w})a.{double_m}[i - {64 // w}] >> imm;",
+                file=f,
+            )
+            print(
+                f"  dst.{m}[i] = clamp<{shift_sign}{double_w}>(temp, {min}, {max});",
+                file=f,
+            )
+            print(f"}}", file=f)
+            print(f"}}", file=f)
+
     if width == "d" or width == "du":
         with open(f"vextl_{double_width}_{width}.h", "w") as f:
             print(f"for (int i = 0;i < {128 // double_w};i++) {{", file=f)

code/gen_tb.py

@@ -9,6 +9,7 @@
 widths_vsrln = ["b_h", "h_w", "w_d"]
 widths_vssrln = ["b_h", "bu_h", "h_w", "hu_w", "w_d", "wu_d"]
 widths_vsrlni = ["b_h", "h_w", "w_d", "d_q"]
+widths_vssrlni = ["b_h", "bu_h", "h_w", "hu_w", "w_d", "wu_d", "d_q", "du_q"]
 widths_vaddw = [
     "h_b",
     "h_bu",
@@ -126,8 +127,10 @@
     "vsrlri": (widths_signed, "v128 a, int imm", [0, 7]),
     "vsrlrn": (widths_vsrln, "v128 a, v128 b"),
     "vsrlrni": (widths_vsrlni, "v128 a, v128 b, int imm", [0, 7, 15]),
-    "vssrln": (widths_vssrln, "v128 a, v128 b"),
     "vssran": (widths_vssrln, "v128 a, v128 b"),
+    "vssrani": (widths_vssrlni, "v128 a, v128 b, int imm", [0, 7, 15]),
+    "vssrln": (widths_vssrln, "v128 a, v128 b"),
+    "vssrlni": (widths_vssrlni, "v128 a, v128 b, int imm", [0, 7, 15]),
     "vsub": (widths_signed, "v128 a, v128 b"),
     "vsubwev": (widths_vsubw, "v128 a, v128 b"),
     "vsubwod": (widths_vsubw, "v128 a, v128 b"),

code/vssrani_b_h.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrani_b_h(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrani_b_h.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrani_b_h, 0);
+  FUZZ2(vssrani_b_h, 7);
+  FUZZ2(vssrani_b_h, 15);
+}

code/vssrani_b_h.h

@@ -0,0 +1,9 @@
+for (int i = 0; i < 16; i++) {
+  if (i < 8) {
+    s16 temp = (s16)b.half[i] >> imm;
+    dst.byte[i] = clamp<s16>(temp, -128, 127);
+  } else {
+    s16 temp = (s16)a.half[i - 8] >> imm;
+    dst.byte[i] = clamp<s16>(temp, -128, 127);
+  }
+}

code/vssrani_bu_h.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrani_bu_h(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrani_bu_h.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrani_bu_h, 0);
+  FUZZ2(vssrani_bu_h, 7);
+  FUZZ2(vssrani_bu_h, 15);
+}

code/vssrani_bu_h.h

@@ -0,0 +1,9 @@
+for (int i = 0; i < 16; i++) {
+  if (i < 8) {
+    s16 temp = (s16)b.half[i] >> imm;
+    dst.byte[i] = clamp<s16>(temp, 0, 255);
+  } else {
+    s16 temp = (s16)a.half[i - 8] >> imm;
+    dst.byte[i] = clamp<s16>(temp, 0, 255);
+  }
+}

code/vssrani_d_q.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrani_d_q(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrani_d_q.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrani_d_q, 0);
+  FUZZ2(vssrani_d_q, 7);
+  FUZZ2(vssrani_d_q, 15);
+}

code/vssrani_d_q.h

@@ -0,0 +1,9 @@
+for (int i = 0; i < 2; i++) {
+  if (i < 1) {
+    s128 temp = (s128)b.qword[i] >> imm;
+    dst.dword[i] = clamp<s128>(temp, -9223372036854775808, 9223372036854775807);
+  } else {
+    s128 temp = (s128)a.qword[i - 1] >> imm;
+    dst.dword[i] = clamp<s128>(temp, -9223372036854775808, 9223372036854775807);
+  }
+}

code/vssrani_du_q.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrani_du_q(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrani_du_q.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrani_du_q, 0);
+  FUZZ2(vssrani_du_q, 7);
+  FUZZ2(vssrani_du_q, 15);
+}

code/vssrani_du_q.h

@@ -0,0 +1,9 @@
+for (int i = 0; i < 2; i++) {
+  if (i < 1) {
+    s128 temp = (s128)b.qword[i] >> imm;
+    dst.dword[i] = clamp<s128>(temp, 0, 18446744073709551615);
+  } else {
+    s128 temp = (s128)a.qword[i - 1] >> imm;
+    dst.dword[i] = clamp<s128>(temp, 0, 18446744073709551615);
+  }
+}

code/vssrani_h_w.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrani_h_w(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrani_h_w.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrani_h_w, 0);
+  FUZZ2(vssrani_h_w, 7);
+  FUZZ2(vssrani_h_w, 15);
+}

code/vssrani_h_w.h

@@ -0,0 +1,9 @@
+for (int i = 0; i < 8; i++) {
+  if (i < 4) {
+    s32 temp = (s32)b.word[i] >> imm;
+    dst.half[i] = clamp<s32>(temp, -32768, 32767);
+  } else {
+    s32 temp = (s32)a.word[i - 4] >> imm;
+    dst.half[i] = clamp<s32>(temp, -32768, 32767);
+  }
+}

code/vssrani_hu_w.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrani_hu_w(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrani_hu_w.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrani_hu_w, 0);
+  FUZZ2(vssrani_hu_w, 7);
+  FUZZ2(vssrani_hu_w, 15);
+}

code/vssrani_hu_w.h

@@ -0,0 +1,9 @@
+for (int i = 0; i < 8; i++) {
+  if (i < 4) {
+    s32 temp = (s32)b.word[i] >> imm;
+    dst.half[i] = clamp<s32>(temp, 0, 65535);
+  } else {
+    s32 temp = (s32)a.word[i - 4] >> imm;
+    dst.half[i] = clamp<s32>(temp, 0, 65535);
+  }
+}

code/vssrani_w_d.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrani_w_d(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrani_w_d.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrani_w_d, 0);
+  FUZZ2(vssrani_w_d, 7);
+  FUZZ2(vssrani_w_d, 15);
+}

code/vssrani_w_d.h

@@ -0,0 +1,9 @@
+for (int i = 0; i < 4; i++) {
+  if (i < 2) {
+    s64 temp = (s64)b.dword[i] >> imm;
+    dst.word[i] = clamp<s64>(temp, -2147483648, 2147483647);
+  } else {
+    s64 temp = (s64)a.dword[i - 2] >> imm;
+    dst.word[i] = clamp<s64>(temp, -2147483648, 2147483647);
+  }
+}

code/vssrani_wu_d.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrani_wu_d(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrani_wu_d.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrani_wu_d, 0);
+  FUZZ2(vssrani_wu_d, 7);
+  FUZZ2(vssrani_wu_d, 15);
+}

code/vssrani_wu_d.h

@@ -0,0 +1,9 @@
+for (int i = 0; i < 4; i++) {
+  if (i < 2) {
+    s64 temp = (s64)b.dword[i] >> imm;
+    dst.word[i] = clamp<s64>(temp, 0, 4294967295);
+  } else {
+    s64 temp = (s64)a.dword[i - 2] >> imm;
+    dst.word[i] = clamp<s64>(temp, 0, 4294967295);
+  }
+}

code/vssrlni_b_h.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrlni_b_h(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrlni_b_h.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrlni_b_h, 0);
+  FUZZ2(vssrlni_b_h, 7);
+  FUZZ2(vssrlni_b_h, 15);
+}

code/vssrlni_b_h.h

@@ -0,0 +1,9 @@
+for (int i = 0; i < 16; i++) {
+  if (i < 8) {
+    u16 temp = (u16)b.half[i] >> imm;
+    dst.byte[i] = clamp<u16>(temp, 0, 127);
+  } else {
+    u16 temp = (u16)a.half[i - 8] >> imm;
+    dst.byte[i] = clamp<u16>(temp, 0, 127);
+  }
+}

code/vssrlni_bu_h.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrlni_bu_h(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrlni_bu_h.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrlni_bu_h, 0);
+  FUZZ2(vssrlni_bu_h, 7);
+  FUZZ2(vssrlni_bu_h, 15);
+}

code/vssrlni_bu_h.h

@@ -0,0 +1,9 @@
+for (int i = 0; i < 16; i++) {
+  if (i < 8) {
+    u16 temp = (u16)b.half[i] >> imm;
+    dst.byte[i] = clamp<u16>(temp, 0, 255);
+  } else {
+    u16 temp = (u16)a.half[i - 8] >> imm;
+    dst.byte[i] = clamp<u16>(temp, 0, 255);
+  }
+}

code/vssrlni_d_q.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrlni_d_q(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrlni_d_q.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrlni_d_q, 0);
+  FUZZ2(vssrlni_d_q, 7);
+  FUZZ2(vssrlni_d_q, 15);
+}

code/vssrlni_d_q.h

@@ -0,0 +1,9 @@
+for (int i = 0; i < 2; i++) {
+  if (i < 1) {
+    u128 temp = (u128)b.qword[i] >> imm;
+    dst.dword[i] = clamp<u128>(temp, 0, 9223372036854775807);
+  } else {
+    u128 temp = (u128)a.qword[i - 1] >> imm;
+    dst.dword[i] = clamp<u128>(temp, 0, 9223372036854775807);
+  }
+}

code/vssrlni_du_q.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrlni_du_q(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrlni_du_q.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrlni_du_q, 0);
+  FUZZ2(vssrlni_du_q, 7);
+  FUZZ2(vssrlni_du_q, 15);
+}

code/vssrlni_du_q.h

@@ -0,0 +1,9 @@
+for (int i = 0; i < 2; i++) {
+  if (i < 1) {
+    u128 temp = (u128)b.qword[i] >> imm;
+    dst.dword[i] = clamp<u128>(temp, 0, 18446744073709551615);
+  } else {
+    u128 temp = (u128)a.qword[i - 1] >> imm;
+    dst.dword[i] = clamp<u128>(temp, 0, 18446744073709551615);
+  }
+}

code/vssrlni_h_w.cpp

@@ -0,0 +1,13 @@
+#include "common.h"
+
+v128 vssrlni_h_w(v128 a, v128 b, int imm) {
+  v128 dst;
+#include "vssrlni_h_w.h"
+  return dst;
+}
+
+void test() {
+  FUZZ2(vssrlni_h_w, 0);
+  FUZZ2(vssrlni_h_w, 7);
+  FUZZ2(vssrlni_h_w, 15);
+}