From 4ad3b6ef764b917367284d087267ff053e3dac60 Mon Sep 17 00:00:00 2001
From: Finn Wilkinson <finnwilkinson5@gmail.com>
Date: Mon, 14 Oct 2024 10:47:36 +0100
Subject: [PATCH] NEON instruction logic fixes.

---
 .../simeng/arch/aarch64/helpers/neon.hh        | 14 ++++++++++++--
 src/lib/arch/aarch64/Instruction_execute.cc    |  4 ++--
 test/regression/aarch64/instructions/neon.cc   | 18 +++++++++---------
 3 files changed, 23 insertions(+), 13 deletions(-)

diff --git a/src/include/simeng/arch/aarch64/helpers/neon.hh b/src/include/simeng/arch/aarch64/helpers/neon.hh
index 0fcf04f03f..79be89bce1 100644
--- a/src/include/simeng/arch/aarch64/helpers/neon.hh
+++ b/src/include/simeng/arch/aarch64/helpers/neon.hh
@@ -568,9 +568,14 @@ RegisterValue vecUMaxP(srcValContainer& sourceValues) {
   const T* n = sourceValues[0].getAsVector<T>();
   const T* m = sourceValues[1].getAsVector<T>();
 
+  // Concatenate the vectors
+  T temp[2 * I];
+  memcpy(temp, m, sizeof(T) * I);
+  memcpy(temp + (sizeof(T) * I), n, sizeof(T) * I);
+  // Compare each adjacent pair of elements
   T out[I];
   for (int i = 0; i < I; i++) {
-    out[i] = std::max(n[i], m[i]);
+    out[i] = std::max(temp[2 * i], temp[2 * i + 1]);
   }
   return {out, 256};
 }
@@ -585,9 +590,14 @@ RegisterValue vecUMinP(srcValContainer& sourceValues) {
   const T* n = sourceValues[0].getAsVector<T>();
   const T* m = sourceValues[1].getAsVector<T>();
 
+  // Concatenate the vectors
+  T temp[2 * I];
+  memcpy(temp, m, sizeof(T) * I);
+  memcpy(temp + (sizeof(T) * I), n, sizeof(T) * I);
+
   T out[I];
   for (int i = 0; i < I; i++) {
-    out[i] = std::min(n[i], m[i]);
+    out[i] = std::min(temp[2 * i], temp[2 * i + 1]);
   }
   return {out, 256};
 }
diff --git a/src/lib/arch/aarch64/Instruction_execute.cc b/src/lib/arch/aarch64/Instruction_execute.cc
index 9d6cc78f9d..7e2478b06a 100644
--- a/src/lib/arch/aarch64/Instruction_execute.cc
+++ b/src/lib/arch/aarch64/Instruction_execute.cc
@@ -838,9 +838,9 @@ void Instruction::execute() {
         break;
       }
       case Opcode::AArch64_CMHSv16i8: {  // cmhs vd.16b, vn.16b, vm.16b
-        results_[0] = vecCompare<int8_t, 16>(
+        results_[0] = vecCompare<uint8_t, 16>(
             sourceValues_, false,
-            [](int8_t x, int8_t y) -> bool { return (x >= y); });
+            [](uint8_t x, uint8_t y) -> bool { return (x >= y); });
         break;
       }
       case Opcode::AArch64_CMPEQ_PPzZI_B: {  // cmpeq pd.b, pg/z, zn.b, #imm
diff --git a/test/regression/aarch64/instructions/neon.cc b/test/regression/aarch64/instructions/neon.cc
index e23573be9f..093be25bf0 100644
--- a/test/regression/aarch64/instructions/neon.cc
+++ b/test/regression/aarch64/instructions/neon.cc
@@ -726,8 +726,8 @@ TEST_P(InstNeon, cmhs) {
   heap[1] = 0x7F;
   heap[2] = INT8_MAX;
   heap[3] = 1;
-  heap[4] = -128;
-  heap[5] = -1;
+  heap[4] = 128;
+  heap[5] = 1;
   heap[6] = 0xAA;
   heap[7] = 0xBB;
   heap[8] = 0xCC;
@@ -743,7 +743,7 @@ TEST_P(InstNeon, cmhs) {
   heap[16] = INT8_MAX;
   heap[17] = 0x7F;
   heap[18] = 0;
-  heap[19] = -128;
+  heap[19] = 128;
   heap[20] = 1;
   heap[21] = 0;
   heap[22] = 0xAA;
@@ -771,10 +771,10 @@ TEST_P(InstNeon, cmhs) {
   )");
 
   CHECK_NEON(2, uint8_t,
-             {0x00, 0xFF, 0xFF, 0xFF, 0x00, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
+             {0x00, 0xFF, 0xFF, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF,
               0xFF, 0xFF, 0xFF, 0xFF, 0xFF});
   CHECK_NEON(3, uint8_t,
-             {0xFF, 0xFF, 0x00, 0x00, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x00, 0x00,
+             {0xFF, 0xFF, 0x00, 0xFF, 0x00, 0x00, 0xFF, 0xFF, 0xFF, 0x00, 0x00,
               0x00, 0xFF, 0xFF, 0xFF, 0xFF});
 }
 
@@ -2683,8 +2683,8 @@ TEST_P(InstNeon, uminp) {
 
   )");
   CHECK_NEON(2, uint8_t,
-             {0x00, 0x00, 0xEE, 0x11, 0x22, 0x33, 0x44, 0x55, 0x01, 0x02, 0x03,
-              0x04, 0x05, 0x06, 0x07, 0x08});
+             {0x00, 0x11, 0x22, 0x44, 0xEE, 0xCC, 0xAA, 0x88, 0x00, 0xAA, 0xBB,
+              0xDD, 0x01, 0x03, 0x05, 0x07});
 }
 TEST_P(InstNeon, umaxp) {
   // umaxp vd.16b vn.16b vm.16b
@@ -2741,8 +2741,8 @@ TEST_P(InstNeon, umaxp) {
 
   )");
   CHECK_NEON(2, uint8_t,
-             {0x01, 0x00, 0xFF, 0xAA, 0xBB, 0xCC, 0xDD, 0xEE, 0xFF, 0xEE, 0xDD,
-              0xCC, 0xBB, 0xAA, 0x99, 0x88});
+             {0x00, 0xEE, 0x33, 0x55, 0xFF, 0xDD, 0xBB, 0x99, 0x01, 0xFF, 0xCC,
+              0xEE, 0x02, 0x04, 0x06, 0x08});
 }
 
 TEST_P(InstNeon, smax) {