arch-power: Add population count instructions

This adds the following instructions.
  * Population Count Bytes (popcntb)
  * Population Count Words (popcntw)
  * Population Count Doubleword (popcntd)

Change-Id: Id15188482b45552735c1d960418d5d6ba1f2ede8
Signed-off-by: Sandipan Das <sandipan@linux.ibm.com>

diff --git a/src/arch/power/isa/decoder.isa b/src/arch/power/isa/decoder.isa
index dfad978747e106462e96ad16b7d8918c6e64b3c5..08b9d10ad2f40145c79f79a8eefd45b547b458e2 100644 (file)
--- a/src/arch/power/isa/decoder.isa
+++ b/src/arch/power/isa/decoder.isa
@@ -515,6 +515,45 @@ decode PO default Unknown::unknown() {
                 Ra = res;
             }});
 
+            122: popcntb({{
+                // Based on "Counting bits set, in parallel"
+                // from https://graphics.stanford.edu/~seander/bithacks.html
+                const uint64_t m1 = 0x5555555555555555ULL;
+                const uint64_t m2 = 0x3333333333333333ULL;
+                const uint64_t m4 = 0x0f0f0f0f0f0f0f0fULL;
+                uint64_t res = Rs;
+                res = (res & m1) + ((res >> 1) & m1);
+                res = (res & m2) + ((res >> 2) & m2);
+                res = (res & m4) + ((res >> 4) & m4);
+                Ra = res;
+            }});
+
+            378: popcntw({{
+            #if defined(__GNUC__) || (defined(__clang__) && \
+                    __has_builtin(__builtin_popcount))
+                uint64_t src = Rs;
+                uint64_t res = __builtin_popcount(src >> 32);
+                res = (res << 32) | __builtin_popcount(src);
+            #else
+                // Based on "Counting bits set, in parallel"
+                // from https://graphics.stanford.edu/~seander/bithacks.html
+                const uint64_t m1 = 0x5555555555555555ULL;
+                const uint64_t m2 = 0x3333333333333333ULL;
+                const uint64_t m4 = 0x0f0f0f0f0f0f0f0fULL;
+                const uint64_t m8 = 0x00ff00ff00ff00ffULL;
+                const uint64_t m16 = 0x0000ffff0000ffffULL;
+                uint64_t res = Rs;
+                res = (res & m1) + ((res >> 1) & m1);
+                res = (res & m2) + ((res >> 2) & m2);
+                res = (res & m4) + ((res >> 4) & m4);
+                res = (res & m8) + ((res >> 8) & m8);
+                res = (res & m16) + ((res >> 16) & m16);
+            #endif
+                Ra = res;
+            }});
+
+            506: popcntd({{ Ra = popCount(Rs); }});
+
             24: slw({{
                 if (Rb & 0x20) {
                     Ra = 0;