[PATCH v2 10/10] softfloat: Use ppc64 assembly for {add, sub}{192, 256}

[Richard Henderson <richard.henderson@linaro.org>]

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/fpu/softfloat-macros.h | 14 ++++++++++++++
 fpu/softfloat.c                | 27 +++++++++++++++++++++++++++
 2 files changed, 41 insertions(+)

diff --git a/include/fpu/softfloat-macros.h b/include/fpu/softfloat-macros.h
index 969a486fd2..d26cfaf267 100644
--- a/include/fpu/softfloat-macros.h
+++ b/include/fpu/softfloat-macros.h
@@ -449,6 +449,13 @@ static inline void
         : "=&r"(*z0Ptr), "=&r"(*z1Ptr), "=&r"(*z2Ptr)
         : "rZ"(a0), "rZ"(a1), "rZ"(a2), "rZ"(b0), "rZ"(b1), "rZ"(b2)
         : "cc");
+#elif defined(__powerpc64__)
+    asm("addc %2, %5, %8\n\t"
+        "adde %1, %4, %7\n\t"
+        "adde %0, %3, %6"
+        : "=r"(*z0Ptr), "=&r"(*z1Ptr), "=&r"(*z2Ptr)
+        : "r"(a0), "r"(a1), "r"(a2), "r"(b0), "r"(b1), "r"(b2)
+        : "ca");
 #else
     uint64_t z0, z1, z2;
     int8_t carry0, carry1;
@@ -521,6 +528,13 @@ static inline void
         : "=&r"(*z0Ptr), "=&r"(*z1Ptr), "=&r"(*z2Ptr)
         : "rZ"(a0), "rZ"(a1), "rZ"(a2), "rZ"(b0), "rZ"(b1), "rZ"(b2)
         : "cc");
+#elif defined(__powerpc64__)
+    asm("subfc %2, %8, %5\n\t"
+        "subfe %1, %7, %4\n\t"
+        "subfe %0, %6, %3"
+        : "=&r"(*z0Ptr), "=&r"(*z1Ptr), "=&r"(*z2Ptr)
+        : "r"(a0), "r"(a1), "r"(a2), "r"(b0), "r"(b1), "r"(b2)
+        : "ca");
 #else
     uint64_t z0, z1, z2;
     int8_t borrow0, borrow1;
diff --git a/fpu/softfloat.c b/fpu/softfloat.c
index 07dc17caad..9af75b9146 100644
--- a/fpu/softfloat.c
+++ b/fpu/softfloat.c
@@ -7370,6 +7370,18 @@ static void sub256(UInt256 *r, UInt256 *a, UInt256 *b)
           [b0] "rZ"(b->w[0]), [b1] "rZ"(b->w[1]),
           [b2] "rZ"(b->w[2]), [b3] "rZ"(b->w[3])
         : "cc");
+#elif defined(__powerpc64__)
+    asm("subfc %[r3], %[b3], %[a3]\n\t"
+        "subfe %[r2], %[b2], %[a2]\n\t"
+        "subfe %[r1], %[b1], %[a1]\n\t"
+        "subfe %[r0], %[b0], %[a0]"
+        : [r0] "=&r"(r->w[0]), [r1] "=&r"(r->w[1]),
+          [r2] "=&r"(r->w[2]), [r3] "=&r"(r->w[3])
+        : [a0] "r"(a->w[0]), [a1] "r"(a->w[1]),
+          [a2] "r"(a->w[2]), [a3] "r"(a->w[3]),
+          [b0] "r"(b->w[0]), [b1] "r"(b->w[1]),
+          [b2] "r"(b->w[2]), [b3] "r"(b->w[3])
+        : "ca");
 #else
     bool borrow = false;
 
@@ -7407,6 +7419,13 @@ static void neg256(UInt256 *a)
         "ngc  %0, %0"
         : "+r"(a->w[0]), "+r"(a->w[1]), "+r"(a->w[2]), "+r"(a->w[3])
         : : "cc");
+#elif defined(__powerpc64__)
+    asm("subfic %3, %3, 0\n\t"
+        "subfze %2, %2\n\t"
+        "subfze %1, %1\n\t"
+        "subfze %0, %0"
+        : "+r"(a->w[0]), "+r"(a->w[1]), "+r"(a->w[2]), "+r"(a->w[3])
+        : : "ca");
 #else
     /*
      * Recall that -X - 1 = ~X, and that since this is negation,
@@ -7457,6 +7476,14 @@ static void add256(UInt256 *a, UInt256 *b)
         : "+r"(a->w[0]), "+r"(a->w[1]), "+r"(a->w[2]), "+r"(a->w[3])
         : "rZ"(b->w[0]), "rZ"(b->w[1]), "rZ"(b->w[2]), "rZ"(b->w[3])
         : "cc");
+#elif defined(__powerpc64__)
+    asm("addc %3, %3, %7\n\t"
+        "adde %2, %2, %6\n\t"
+        "adde %1, %1, %5\n\t"
+        "adde %0, %0, %4"
+        : "+r"(a->w[0]), "+r"(a->w[1]), "+r"(a->w[2]), "+r"(a->w[3])
+        :  "r"(b->w[0]),  "r"(b->w[1]),  "r"(b->w[2]),  "r"(b->w[3])
+        : "ca");
 #else
     bool carry = false;
 
-- 
2.25.1