fpu: Add float64_to_int{32,64}_modulo

Add versions of float64_to_int* which do not saturate the result.

Reviewed-by: Christoph Muellner <christoph.muellner@vrull.eu>
Tested-by: Christoph Muellner <christoph.muellner@vrull.eu>
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
Message-Id: <20230527141910.1885950-2-richard.henderson@linaro.org>

2 files changed, 109 insertions, 0 deletions

diff --git a/fpu/softfloat-parts.c.inc b/fpu/softfloat-parts.c.inc
index 247400031c..527e15e6ab 100644
--- a/fpu/softfloat-parts.c.inc
+++ b/fpu/softfloat-parts.c.inc
@@ -1182,6 +1182,84 @@ static uint64_t partsN(float_to_uint)(FloatPartsN *p, FloatRoundMode rmode,
 }
 
 /*
+ * Like partsN(float_to_sint), except do not saturate the result.
+ * Instead, return the rounded unbounded precision two's compliment result,
+ * modulo 2**(bitsm1 + 1).
+ */
+static int64_t partsN(float_to_sint_modulo)(FloatPartsN *p,
+                                            FloatRoundMode rmode,
+                                            int bitsm1, float_status *s)
+{
+    int flags = 0;
+    uint64_t r;
+    bool overflow = false;
+
+    switch (p->cls) {
+    case float_class_snan:
+        flags |= float_flag_invalid_snan;
+        /* fall through */
+    case float_class_qnan:
+        flags |= float_flag_invalid;
+        r = 0;
+        break;
+
+    case float_class_inf:
+        overflow = true;
+        r = 0;
+        break;
+
+    case float_class_zero:
+        return 0;
+
+    case float_class_normal:
+        /* TODO: N - 2 is frac_size for rounding; could use input fmt. */
+        if (parts_round_to_int_normal(p, rmode, 0, N - 2)) {
+            flags = float_flag_inexact;
+        }
+
+        if (p->exp <= DECOMPOSED_BINARY_POINT) {
+            /*
+             * Because we rounded to integral, and exp < 64,
+             * we know frac_low is zero.
+             */
+            r = p->frac_hi >> (DECOMPOSED_BINARY_POINT - p->exp);
+            if (p->exp < bitsm1) {
+                /* Result in range. */
+            } else if (p->exp == bitsm1) {
+                /* The only in-range value is INT_MIN. */
+                overflow = !p->sign || p->frac_hi != DECOMPOSED_IMPLICIT_BIT;
+            } else {
+                overflow = true;
+            }
+        } else {
+            /* Overflow, but there might still be bits to return. */
+            int shl = p->exp - DECOMPOSED_BINARY_POINT;
+            if (shl < N) {
+                frac_shl(p, shl);
+                r = p->frac_hi;
+            } else {
+                r = 0;
+            }
+            overflow = true;
+        }
+
+        if (p->sign) {
+            r = -r;
+        }
+        break;
+
+    default:
+        g_assert_not_reached();
+    }
+
+    if (overflow) {
+        flags = float_flag_invalid | float_flag_invalid_cvti;
+    }
+    float_raise(flags, s);
+    return r;
+}
+
+/*
  * Integer to float conversions
  *
  * Returns the result of converting the two's complement integer `a'
diff --git a/fpu/softfloat.c b/fpu/softfloat.c
index 42e6c188b4..0cc130ae9b 100644
--- a/fpu/softfloat.c
+++ b/fpu/softfloat.c
@@ -852,11 +852,24 @@ static uint64_t parts128_float_to_uint(FloatParts128 *p, FloatRoundMode rmode,
 #define parts_float_to_uint(P, R, Z, M, S) \
     PARTS_GENERIC_64_128(float_to_uint, P)(P, R, Z, M, S)
 
+static int64_t parts64_float_to_sint_modulo(FloatParts64 *p,
+                                            FloatRoundMode rmode,
+                                            int bitsm1, float_status *s);
+static int64_t parts128_float_to_sint_modulo(FloatParts128 *p,
+                                             FloatRoundMode rmode,
+                                             int bitsm1, float_status *s);
+
+#define parts_float_to_sint_modulo(P, R, M, S) \
+    PARTS_GENERIC_64_128(float_to_sint_modulo, P)(P, R, M, S)
+
 static void parts64_sint_to_float(FloatParts64 *p, int64_t a,
                                   int scale, float_status *s);
 static void parts128_sint_to_float(FloatParts128 *p, int64_t a,
                                    int scale, float_status *s);
 
+#define parts_float_to_sint(P, R, Z, MN, MX, S) \
+    PARTS_GENERIC_64_128(float_to_sint, P)(P, R, Z, MN, MX, S)
+
 #define parts_sint_to_float(P, I, Z, S) \
     PARTS_GENERIC_64_128(sint_to_float, P)(P, I, Z, S)
 
@@ -3409,6 +3422,24 @@ int64_t bfloat16_to_int64_round_to_zero(bfloat16 a, float_status *s)
     return bfloat16_to_int64_scalbn(a, float_round_to_zero, 0, s);
 }
 
+int32_t float64_to_int32_modulo(float64 a, FloatRoundMode rmode,
+                                float_status *s)
+{
+    FloatParts64 p;
+
+    float64_unpack_canonical(&p, a, s);
+    return parts_float_to_sint_modulo(&p, rmode, 31, s);
+}
+
+int64_t float64_to_int64_modulo(float64 a, FloatRoundMode rmode,
+                                float_status *s)
+{
+    FloatParts64 p;
+
+    float64_unpack_canonical(&p, a, s);
+    return parts_float_to_sint_modulo(&p, rmode, 63, s);
+}
+
 /*
  * Floating-point to unsigned integer conversions
  */