Addd vaddwod

README.md

@@ -30,18 +30,14 @@ Vector Subtract
 
 Vector Subtract with Width Extension
 
-### vaddwod.h.b/w.h/d.w/q.d
 ### vsubwod.h.b/w.h/d.w/q.d
 
 ### vsubwev.h.bu/w.hu/d.wu/q.du
 
 Vector Subtract with Width Extension Unsigned
 
-### vaddwod.h.bu/w.hu/d.wu/q.du
 ### vsubwod.h.bu/w.hu/d.wu/q.du
 
-### vaddwod.h.bu.b/w.hu.h/q.wu.w/q.du.d
-
 ### vsadd.b/h/w/d
 
 Vector Saturated Add

code/gen_impl.py

@@ -1,4 +1,4 @@
-widths_all = ["b", "bu", "h", "hu", "w", "wu", "d", "du"]
+import os
 
 widths = {
     "b": 8,
@@ -10,6 +10,18 @@
     "d": 64,
     "du": 64,
     "q": 128,
+    "qu": 128,
+}
+
+double_widths = {
+    "b": "h",
+    "bu": "h",
+    "h": "w",
+    "hu": "w",
+    "w": "d",
+    "wu": "d",
+    "d": "q",
+    "du": "q",
 }
 
 members = {
@@ -21,20 +33,66 @@
     "wu": "word",
     "d": "dword",
     "du": "dword",
+    "q": "qword",
+    "qu": "qword",
 }
 
-for width in widths_all:
+for width in ["b", "bu", "h", "hu", "w", "wu", "d", "du"]:
+    w = widths[width]
+    m = members[width]
+    if "u" in width:
+        sign = "u"
+    else:
+        sign = "s"
+    with open(f"vavg_{width}.h", "w") as f:
+        print(f"for (int i = 0;i < {128 // w};i++) {{", file=f)
+        print(
+            f"  dst.{m}[i] = (({sign}{w})a.{m}[i] >> 1) + (({sign}{w})b.{m}[i] >> 1) + (a.{m}[i] & b.{m}[i] & 1);",
+            file=f,
+        )
+        print(f"}}", file=f)
+    with open(f"vavgr_{width}.h", "w") as f:
+        print(f"for (int i = 0;i < {128 // w};i++) {{", file=f)
+        print(
+            f"  dst.{m}[i] = (({sign}{w})a.{m}[i] >> 1) + (({sign}{w})b.{m}[i] >> 1) + ((a.{m}[i] | b.{m}[i]) & 1);",
+            file=f,
+        )
+        print(f"}}", file=f)
+
+for width in ["b", "bu", "h", "hu", "w", "wu", "d", "du"]:
+    double_width = double_widths[width]
     w = widths[width]
     m = members[width]
+    double_w = widths[double_width]
+    double_m = members[double_width]
     if "u" in width:
         sign = "u"
     else:
         sign = "s"
-    with open(f"vavg_{width}.h", 'w') as f:
-        print(f'for (int i = 0;i < {128 // w};i++) {{', file=f)
-        print(f'  dst.{m}[i] = (({sign}{w})a.{m}[i] >> 1) + (({sign}{w})b.{m}[i] >> 1) + (a.{m}[i] & b.{m}[i] & 1);', file=f)
-        print(f'}}', file=f)
-    with open(f"vavgr_{width}.h", 'w') as f:
-        print(f'for (int i = 0;i < {128 // w};i++) {{', file=f)
-        print(f'  dst.{m}[i] = (({sign}{w})a.{m}[i] >> 1) + (({sign}{w})b.{m}[i] >> 1) + ((a.{m}[i] | b.{m}[i]) & 1);', file=f)
-        print(f'}}', file=f)
+    for signed2 in [False, True]:
+        if "u" not in width and not signed2:
+            continue
+        if "u" in width and signed2:
+            suffix = f"_{width[:-1]}"
+        else:
+            suffix = ""
+        if signed2:
+            sign2 = "s"
+        else:
+            sign2 = "u"
+        with open(f"vaddwev_{double_width}_{width}{suffix}.h", "w") as f:
+            print(f"for (int i = 0;i < {128 // double_w};i++) {{", file=f)
+            print(
+                f"  dst.{double_m}[i] = ({sign}{double_w})({sign}{w})a.{m}[2 * i] + ({sign2}{double_w})({sign2}{w})b.{m}[2 * i];",
+                file=f,
+            )
+            print(f"}}", file=f)
+        with open(f"vaddwod_{double_width}_{width}{suffix}.h", "w") as f:
+            print(f"for (int i = 0;i < {128 // double_w};i++) {{", file=f)
+            print(
+                f"  dst.{double_m}[i] = ({sign}{double_w})({sign}{w})a.{m}[2 * i + 1] + ({sign2}{double_w})({sign2}{w})b.{m}[2 * i + 1];",
+                file=f,
+            )
+            print(f"}}", file=f)
+
+os.system("clang-format -i *.cpp *.h")

code/gen_tb.py

@@ -1,8 +1,26 @@
+import os
+
 widths_all = ["b", "bu", "h", "hu", "w", "wu", "d", "du"]
+widths_vaddw = [
+    "h_b",
+    "h_bu",
+    "h_bu_b",
+    "w_h",
+    "w_hu",
+    "w_hu_h",
+    "d_w",
+    "d_wu",
+    "d_wu_w",
+    "q_d",
+    "q_du",
+    "q_du_d",
+]
 
 tb = {
     "vavg": (widths_all, "v128 a, v128 b"),
-    "vavgr": (widths_all, "v128 a, v128 b")
+    "vavgr": (widths_all, "v128 a, v128 b"),
+    "vaddwev": (widths_vaddw, "v128 a, v128 b"),
+    "vaddwod": (widths_vaddw, "v128 a, v128 b"),
 }
 
 for name in tb:
@@ -14,20 +32,22 @@
         inst_name = name + "_" + width
 
         fuzz_args = 0
-        for arg in args.split(', '):
-            if 'v128' in arg:
+        for arg in args.split(", "):
+            if "v128" in arg:
                 fuzz_args += 1
 
-        print(f'Saving {inst_name}.cpp')
-        with open(f'{inst_name}.cpp', 'w') as f:
+        print(f"Saving {inst_name}.cpp")
+        with open(f"{inst_name}.cpp", "w") as f:
             print('#include "common.h"', file=f)
-            print('', file=f)
-            print(f'v128 {inst_name}({args}) {{', file=f)
-            print('  v128 dst;', file=f)
+            print("", file=f)
+            print(f"v128 {inst_name}({args}) {{", file=f)
+            print("  v128 dst;", file=f)
             print(f'#include "{inst_name}.h"', file=f)
-            print('  return dst;', file=f)
-            print('}', file=f)
-            print('', file=f)
-            print('void test() {', file=f)
-            print(f'  FUZZ{fuzz_args}({inst_name});', file=f)
-            print('}', file=f)
+            print("  return dst;", file=f)
+            print("}", file=f)
+            print("", file=f)
+            print("void test() {", file=f)
+            print(f"  FUZZ{fuzz_args}({inst_name});", file=f)
+            print("}", file=f)
+
+os.system("clang-format -i *.cpp *.h")

code/vabsd_b.h

@@ -1,3 +1,4 @@
 for (int i = 0; i < 16; i++) {
-  dst.byte[i] = ((s8)a.byte[i] > (s8)b.byte[i]) ? (a.byte[i] - b.byte[i]) : (b.byte[i] - a.byte[i]);
+  dst.byte[i] = ((s8)a.byte[i] > (s8)b.byte[i]) ? (a.byte[i] - b.byte[i])
+                                                : (b.byte[i] - a.byte[i]);
 }

code/vabsd_bu.h

@@ -1,3 +1,4 @@
 for (int i = 0; i < 16; i++) {
-  dst.byte[i] = ((u8)a.byte[i] > (u8)b.byte[i]) ? (a.byte[i] - b.byte[i]) : (b.byte[i] - a.byte[i]);
+  dst.byte[i] = ((u8)a.byte[i] > (u8)b.byte[i]) ? (a.byte[i] - b.byte[i])
+                                                : (b.byte[i] - a.byte[i]);
 }

code/vabsd_d.h

@@ -1,3 +1,5 @@
 for (int i = 0; i < 2; i++) {
-  dst.dword[i] = ((s64)a.dword[i] > (s64)b.dword[i]) ? (a.dword[i] - b.dword[i]) : (b.dword[i] - a.dword[i]);
+  dst.dword[i] = ((s64)a.dword[i] > (s64)b.dword[i])
+                     ? (a.dword[i] - b.dword[i])
+                     : (b.dword[i] - a.dword[i]);
 }

code/vabsd_du.h

@@ -1,3 +1,5 @@
 for (int i = 0; i < 2; i++) {
-  dst.dword[i] = ((u64)a.dword[i] > (u64)b.dword[i]) ? (a.dword[i] - b.dword[i]) : (b.dword[i] - a.dword[i]);
+  dst.dword[i] = ((u64)a.dword[i] > (u64)b.dword[i])
+                     ? (a.dword[i] - b.dword[i])
+                     : (b.dword[i] - a.dword[i]);
 }

code/vabsd_h.h

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

code/vabsd_hu.h

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

code/vabsd_w.h

@@ -1,3 +1,4 @@
 for (int i = 0; i < 4; i++) {
-  dst.word[i] = ((s32)a.word[i] > (s32)b.word[i]) ? (a.word[i] - b.word[i]) : (b.word[i] - a.word[i]);
+  dst.word[i] = ((s32)a.word[i] > (s32)b.word[i]) ? (a.word[i] - b.word[i])
+                                                  : (b.word[i] - a.word[i]);
 }

code/vabsd_wu.h

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

code/vadda_d.h

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

code/vadda_h.cpp

@@ -7,4 +7,3 @@ v128 vadda_h(v128 a, v128 b) {
 }
 
 void test() { FUZZ2(vadda_h); }
-

code/vadda_h.h

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

code/vadda_w.cpp

@@ -7,4 +7,3 @@ v128 vadda_w(v128 a, v128 b) {
 }
 
 void test() { FUZZ2(vadda_w); }
-

code/vaddwev_d_w.h

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

code/vaddwev_d_wu.h

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

code/vaddwev_d_wu_w.h

@@ -1,3 +1,3 @@
-for (int i = 0;i < 2;i++) {
+for (int i = 0; i < 2; i++) {
   dst.dword[i] = (u64)(u32)a.word[2 * i] + (s64)(s32)b.word[2 * i];
 }

code/vaddwev_h_b.h

@@ -1,3 +1,3 @@
-for (int i = 0;i < 8;i++) {
+for (int i = 0; i < 8; i++) {
   dst.half[i] = (s16)(s8)a.byte[2 * i] + (s16)(s8)b.byte[2 * i];
-}
+}

code/vaddwev_h_bu.h

@@ -1,3 +1,3 @@
-for (int i = 0;i < 8;i++) {
+for (int i = 0; i < 8; i++) {
   dst.half[i] = (u16)(u8)a.byte[2 * i] + (u16)(u8)b.byte[2 * i];
-}
+}

code/vaddwev_h_bu_b.h

@@ -1,3 +1,3 @@
-for (int i = 0;i < 8;i++) {
+for (int i = 0; i < 8; i++) {
   dst.half[i] = (u16)(u8)a.byte[2 * i] + (s16)(s8)b.byte[2 * i];
 }

code/vaddwev_q_d.h

@@ -1 +1,3 @@
-dst.qword[0] = (s128)(s64)a.dword[0] + (s128)(s64)b.dword[0];
+for (int i = 0; i < 1; i++) {
+  dst.qword[i] = (s128)(s64)a.dword[2 * i] + (s128)(s64)b.dword[2 * i];
+}

code/vaddwev_q_du.h

@@ -1 +1,3 @@
-dst.qword[0] = (u128)(u64)a.dword[0] + (u128)(u64)b.dword[0];
+for (int i = 0; i < 1; i++) {
+  dst.qword[i] = (u128)(u64)a.dword[2 * i] + (u128)(u64)b.dword[2 * i];
+}

code/vaddwev_q_du_d.h

@@ -1 +1,3 @@
-dst.qword[0] = (u128)(u64)a.dword[0] + (s128)(s64)b.dword[0];
+for (int i = 0; i < 1; i++) {
+  dst.qword[i] = (u128)(u64)a.dword[2 * i] + (s128)(s64)b.dword[2 * i];
+}

code/vaddwev_w_h.h

@@ -1,3 +1,3 @@
-for (int i = 0;i < 4;i++) {
+for (int i = 0; i < 4; i++) {
   dst.word[i] = (s32)(s16)a.half[2 * i] + (s32)(s16)b.half[2 * i];
 }

code/vaddwev_w_hu.h

@@ -1,3 +1,3 @@
-for (int i = 0;i < 4;i++) {
+for (int i = 0; i < 4; i++) {
   dst.word[i] = (u32)(u16)a.half[2 * i] + (u32)(u16)b.half[2 * i];
 }

code/vaddwev_w_hu_h.h

@@ -1,3 +1,3 @@
-for (int i = 0;i < 4;i++) {
+for (int i = 0; i < 4; i++) {
   dst.word[i] = (u32)(u16)a.half[2 * i] + (s32)(s16)b.half[2 * i];
 }

code/vaddwod_d_w.cpp

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

code/vaddwod_d_w.h

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

code/vaddwod_d_wu.cpp

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

code/vaddwod_d_wu.h

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

code/vaddwod_d_wu_w.cpp

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

code/vaddwod_d_wu_w.h

@@ -0,0 +1,3 @@
+for (int i = 0; i < 2; i++) {
+  dst.dword[i] = (u64)(u32)a.word[2 * i + 1] + (s64)(s32)b.word[2 * i + 1];
+}

code/vaddwod_h_b.cpp

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

code/vaddwod_h_b.h

@@ -0,0 +1,3 @@
+for (int i = 0; i < 8; i++) {
+  dst.half[i] = (s16)(s8)a.byte[2 * i + 1] + (s16)(s8)b.byte[2 * i + 1];
+}

code/vaddwod_h_bu.cpp

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

code/vaddwod_h_bu.h

@@ -0,0 +1,3 @@
+for (int i = 0; i < 8; i++) {
+  dst.half[i] = (u16)(u8)a.byte[2 * i + 1] + (u16)(u8)b.byte[2 * i + 1];
+}