Switched order of concatonation for NEON UMAXP instruction to match H…

…ardware.
UoB-HPC · Nov 6, 2024 · 5945bae · 5945bae
1 parent 32948cf
commit 5945bae
Show file tree

Hide file tree

Showing 2 changed files with 3 additions and 3 deletions.
diff --git a/src/include/simeng/arch/aarch64/helpers/neon.hh b/src/include/simeng/arch/aarch64/helpers/neon.hh
@@ -570,8 +570,8 @@ RegisterValue vecUMaxP(srcValContainer& sourceValues) {
 
   // Concatenate the vectors
   T temp[2 * I];
-  memcpy(temp, m, sizeof(T) * I);
-  memcpy(temp + (sizeof(T) * I), n, sizeof(T) * I);
+  memcpy(temp, n, sizeof(T) * I);
+  memcpy(temp + (sizeof(T) * I), m, sizeof(T) * I);
   // Compare each adjacent pair of elements
   T out[I];
   for (int i = 0; i < I; i++) {

diff --git a/test/regression/aarch64/instructions/neon.cc b/test/regression/aarch64/instructions/neon.cc
@@ -2738,7 +2738,7 @@ TEST_P(InstNeon, umaxp) {
 
     ldr q0, [x0]
     ldr q1, [x0, #16]
-    umaxp v2.16b, v0.16b, v1.16b
+    umaxp v2.16b, v1.16b, v0.16b
 
   )");
   CHECK_NEON(2, uint8_t,