[Qemu-devel] [PATCH 0/4] target/arm: Fix int64_to_float16 double-rounding

[Qemu-devel] [PATCH 0/4] target/arm: Fix int64_to_float16 double-rounding

[Richard Henderson]

In 88808a022c0, I tried to fix an overflow problem that affected float16
scaling by coverting first to float64 and then rounding after that.  

However, Laurent reported that -0x3ff40000000001 converted to float16
resulted in 0xfbfe instead of the expected 0xfbff.  This is caused by
the inexact conversion to float64.

Rather than build more logic into target/arm to compensate, just add
a function that takes a scaling parameter so that the whole thing is
done all at once with only one rounding.

I don't have a failing test case for the float-to-int paths, but it
seemed best to apply the same solution.


r~


Richard Henderson (4):
  softfloat: Add scaling int-to-float routines
  softfloat: Add scaling float-to-int routines
  target/arm: Use the int-to-float-scale softfloat routines
  target/arm: Use the float-to-int-scale softfloat routines

 include/fpu/softfloat.h | 169 ++++++++----
 fpu/softfloat.c         | 579 +++++++++++++++++++++++++++++++---------
 target/arm/helper.c     | 130 ++++-----
 3 files changed, 628 insertions(+), 250 deletions(-)

-- 
2.17.1

[Qemu-devel] [PATCH 1/4] softfloat: Add scaling int-to-float routines

[Richard Henderson]

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/fpu/softfloat.h |  56 ++++++++----
 fpu/softfloat.c         | 188 +++++++++++++++++++++++++++++-----------
 2 files changed, 179 insertions(+), 65 deletions(-)

diff --git a/include/fpu/softfloat.h b/include/fpu/softfloat.h
index 69f4dbc4db..038e375e71 100644
--- a/include/fpu/softfloat.h
+++ b/include/fpu/softfloat.h
@@ -190,22 +190,54 @@ enum {
 /*----------------------------------------------------------------------------
 | Software IEC/IEEE integer-to-floating-point conversion routines.
 *----------------------------------------------------------------------------*/
+
+float16 int16_to_float16_scalbn(int16_t a, int, float_status *status);
+float16 int32_to_float16_scalbn(int32_t a, int, float_status *status);
+float16 int64_to_float16_scalbn(int64_t a, int, float_status *status);
+float16 uint16_to_float16_scalbn(uint16_t a, int, float_status *status);
+float16 uint32_to_float16_scalbn(uint32_t a, int, float_status *status);
+float16 uint64_to_float16_scalbn(uint64_t a, int, float_status *status);
+
+float16 int16_to_float16(int16_t a, float_status *status);
+float16 int32_to_float16(int32_t a, float_status *status);
+float16 int64_to_float16(int64_t a, float_status *status);
+float16 uint16_to_float16(uint16_t a, float_status *status);
+float16 uint32_to_float16(uint32_t a, float_status *status);
+float16 uint64_to_float16(uint64_t a, float_status *status);
+
+float32 int16_to_float32_scalbn(int16_t, int, float_status *status);
+float32 int32_to_float32_scalbn(int32_t, int, float_status *status);
+float32 int64_to_float32_scalbn(int64_t, int, float_status *status);
+float32 uint16_to_float32_scalbn(uint16_t, int, float_status *status);
+float32 uint32_to_float32_scalbn(uint32_t, int, float_status *status);
+float32 uint64_to_float32_scalbn(uint64_t, int, float_status *status);
+
 float32 int16_to_float32(int16_t, float_status *status);
 float32 int32_to_float32(int32_t, float_status *status);
-float64 int16_to_float64(int16_t, float_status *status);
-float64 int32_to_float64(int32_t, float_status *status);
+float32 int64_to_float32(int64_t, float_status *status);
 float32 uint16_to_float32(uint16_t, float_status *status);
 float32 uint32_to_float32(uint32_t, float_status *status);
+float32 uint64_to_float32(uint64_t, float_status *status);
+
+float64 int16_to_float64_scalbn(int16_t, int, float_status *status);
+float64 int32_to_float64_scalbn(int32_t, int, float_status *status);
+float64 int64_to_float64_scalbn(int64_t, int, float_status *status);
+float64 uint16_to_float64_scalbn(uint16_t, int, float_status *status);
+float64 uint32_to_float64_scalbn(uint32_t, int, float_status *status);
+float64 uint64_to_float64_scalbn(uint64_t, int, float_status *status);
+
+float64 int16_to_float64(int16_t, float_status *status);
+float64 int32_to_float64(int32_t, float_status *status);
+float64 int64_to_float64(int64_t, float_status *status);
 float64 uint16_to_float64(uint16_t, float_status *status);
 float64 uint32_to_float64(uint32_t, float_status *status);
-floatx80 int32_to_floatx80(int32_t, float_status *status);
-float128 int32_to_float128(int32_t, float_status *status);
-float32 int64_to_float32(int64_t, float_status *status);
-float64 int64_to_float64(int64_t, float_status *status);
-floatx80 int64_to_floatx80(int64_t, float_status *status);
-float128 int64_to_float128(int64_t, float_status *status);
-float32 uint64_to_float32(uint64_t, float_status *status);
 float64 uint64_to_float64(uint64_t, float_status *status);
+
+floatx80 int32_to_floatx80(int32_t, float_status *status);
+floatx80 int64_to_floatx80(int64_t, float_status *status);
+
+float128 int32_to_float128(int32_t, float_status *status);
+float128 int64_to_float128(int64_t, float_status *status);
 float128 uint64_to_float128(uint64_t, float_status *status);
 
 /*----------------------------------------------------------------------------
@@ -227,12 +259,6 @@ int64_t float16_to_int64(float16, float_status *status);
 uint64_t float16_to_uint64(float16 a, float_status *status);
 int64_t float16_to_int64_round_to_zero(float16, float_status *status);
 uint64_t float16_to_uint64_round_to_zero(float16 a, float_status *status);
-float16 int16_to_float16(int16_t a, float_status *status);
-float16 int32_to_float16(int32_t a, float_status *status);
-float16 int64_to_float16(int64_t a, float_status *status);
-float16 uint16_to_float16(uint16_t a, float_status *status);
-float16 uint32_to_float16(uint32_t a, float_status *status);
-float16 uint64_to_float16(uint64_t a, float_status *status);
 
 /*----------------------------------------------------------------------------
 | Software half-precision operations.
diff --git a/fpu/softfloat.c b/fpu/softfloat.c
index 7d63cffdeb..12f373cbad 100644
--- a/fpu/softfloat.c
+++ b/fpu/softfloat.c
@@ -1603,81 +1603,122 @@ FLOAT_TO_UINT(64, 64)
  * to the IEC/IEEE Standard for Binary Floating-Point Arithmetic.
  */
 
-static FloatParts int_to_float(int64_t a, float_status *status)
+static FloatParts int_to_float(int64_t a, int scale, float_status *status)
 {
-    FloatParts r = {};
+    FloatParts r = { .sign = false };
+
     if (a == 0) {
         r.cls = float_class_zero;
-        r.sign = false;
-    } else if (a == (1ULL << 63)) {
-        r.cls = float_class_normal;
-        r.sign = true;
-        r.frac = DECOMPOSED_IMPLICIT_BIT;
-        r.exp = 63;
     } else {
-        uint64_t f;
-        if (a < 0) {
-            f = -a;
-            r.sign = true;
-        } else {
-            f = a;
-            r.sign = false;
-        }
-        int shift = clz64(f) - 1;
+        uint64_t f = a;
+        int shift;
+
         r.cls = float_class_normal;
-        r.exp = (DECOMPOSED_BINARY_POINT - shift);
-        r.frac = f << shift;
+        if (a < 0) {
+            f = -f;
+            r.sign = true;
+        }
+        shift = clz64(f) - 1;
+        scale = MIN(MAX(scale, -0x10000), 0x10000);
+
+        r.exp = DECOMPOSED_BINARY_POINT - shift + scale;
+        r.frac = (shift < 0 ? DECOMPOSED_IMPLICIT_BIT : f << shift);
     }
 
     return r;
 }
 
+float16 int64_to_float16_scalbn(int64_t a, int scale, float_status *status)
+{
+    FloatParts pa = int_to_float(a, scale, status);
+    return float16_round_pack_canonical(pa, status);
+}
+
+float16 int32_to_float16_scalbn(int32_t a, int scale, float_status *status)
+{
+    return int64_to_float16_scalbn(a, scale, status);
+}
+
+float16 int16_to_float16_scalbn(int16_t a, int scale, float_status *status)
+{
+    return int64_to_float16_scalbn(a, scale, status);
+}
+
 float16 int64_to_float16(int64_t a, float_status *status)
 {
-    FloatParts pa = int_to_float(a, status);
-    return float16_round_pack_canonical(pa, status);
+    return int64_to_float16_scalbn(a, 0, status);
 }
 
 float16 int32_to_float16(int32_t a, float_status *status)
 {
-    return int64_to_float16(a, status);
+    return int64_to_float16_scalbn(a, 0, status);
 }
 
 float16 int16_to_float16(int16_t a, float_status *status)
 {
-    return int64_to_float16(a, status);
+    return int64_to_float16_scalbn(a, 0, status);
+}
+
+float32 int64_to_float32_scalbn(int64_t a, int scale, float_status *status)
+{
+    FloatParts pa = int_to_float(a, scale, status);
+    return float32_round_pack_canonical(pa, status);
+}
+
+float32 int32_to_float32_scalbn(int32_t a, int scale, float_status *status)
+{
+    return int64_to_float32_scalbn(a, scale, status);
+}
+
+float32 int16_to_float32_scalbn(int16_t a, int scale, float_status *status)
+{
+    return int64_to_float32_scalbn(a, scale, status);
 }
 
 float32 int64_to_float32(int64_t a, float_status *status)
 {
-    FloatParts pa = int_to_float(a, status);
-    return float32_round_pack_canonical(pa, status);
+    return int64_to_float32_scalbn(a, 0, status);
 }
 
 float32 int32_to_float32(int32_t a, float_status *status)
 {
-    return int64_to_float32(a, status);
+    return int64_to_float32_scalbn(a, 0, status);
 }
 
 float32 int16_to_float32(int16_t a, float_status *status)
 {
-    return int64_to_float32(a, status);
+    return int64_to_float32_scalbn(a, 0, status);
+}
+
+float64 int64_to_float64_scalbn(int64_t a, int scale, float_status *status)
+{
+    FloatParts pa = int_to_float(a, scale, status);
+    return float64_round_pack_canonical(pa, status);
+}
+
+float64 int32_to_float64_scalbn(int32_t a, int scale, float_status *status)
+{
+    return int64_to_float64_scalbn(a, scale, status);
+}
+
+float64 int16_to_float64_scalbn(int16_t a, int scale, float_status *status)
+{
+    return int64_to_float64_scalbn(a, scale, status);
 }
 
 float64 int64_to_float64(int64_t a, float_status *status)
 {
-    FloatParts pa = int_to_float(a, status);
-    return float64_round_pack_canonical(pa, status);
+    return int64_to_float64_scalbn(a, 0, status);
 }
 
 float64 int32_to_float64(int32_t a, float_status *status)
 {
-    return int64_to_float64(a, status);
+    return int64_to_float64_scalbn(a, 0, status);
 }
 
 float64 int16_to_float64(int16_t a, float_status *status)
 {
-    return int64_to_float64(a, status);
+    return int64_to_float64_scalbn(a, 0, status);
 }
 
 
@@ -1689,73 +1730,120 @@ float64 int16_to_float64(int16_t a, float_status *status)
  * IEC/IEEE Standard for Binary Floating-Point Arithmetic.
  */
 
-static FloatParts uint_to_float(uint64_t a, float_status *status)
+static FloatParts uint_to_float(uint64_t a, int scale, float_status *status)
 {
-    FloatParts r = { .sign = false};
+    FloatParts r = { .sign = false };
 
     if (a == 0) {
         r.cls = float_class_zero;
     } else {
-        int spare_bits = clz64(a) - 1;
+        scale = MIN(MAX(scale, -0x10000), 0x10000);
         r.cls = float_class_normal;
-        r.exp = DECOMPOSED_BINARY_POINT - spare_bits;
-        if (spare_bits < 0) {
-            shift64RightJamming(a, -spare_bits, &a);
+        if ((int64_t)a < 0) {
+            r.exp = DECOMPOSED_BINARY_POINT + 1 + scale;
+            shift64RightJamming(a, 1, &a);
             r.frac = a;
         } else {
-            r.frac = a << spare_bits;
+            int shift = clz64(a) - 1;
+            r.exp = DECOMPOSED_BINARY_POINT - shift + scale;
+            r.frac = a << shift;
         }
     }
 
     return r;
 }
 
+float16 uint64_to_float16_scalbn(uint64_t a, int scale, float_status *status)
+{
+    FloatParts pa = uint_to_float(a, scale, status);
+    return float16_round_pack_canonical(pa, status);
+}
+
+float16 uint32_to_float16_scalbn(uint32_t a, int scale, float_status *status)
+{
+    return uint64_to_float16_scalbn(a, scale, status);
+}
+
+float16 uint16_to_float16_scalbn(uint16_t a, int scale, float_status *status)
+{
+    return uint64_to_float16_scalbn(a, scale, status);
+}
+
 float16 uint64_to_float16(uint64_t a, float_status *status)
 {
-    FloatParts pa = uint_to_float(a, status);
-    return float16_round_pack_canonical(pa, status);
+    return uint64_to_float16_scalbn(a, 0, status);
 }
 
 float16 uint32_to_float16(uint32_t a, float_status *status)
 {
-    return uint64_to_float16(a, status);
+    return uint64_to_float16_scalbn(a, 0, status);
 }
 
 float16 uint16_to_float16(uint16_t a, float_status *status)
 {
-    return uint64_to_float16(a, status);
+    return uint64_to_float16_scalbn(a, 0, status);
+}
+
+float32 uint64_to_float32_scalbn(uint64_t a, int scale, float_status *status)
+{
+    FloatParts pa = uint_to_float(a, scale, status);
+    return float32_round_pack_canonical(pa, status);
+}
+
+float32 uint32_to_float32_scalbn(uint32_t a, int scale, float_status *status)
+{
+    return uint64_to_float32_scalbn(a, scale, status);
+}
+
+float32 uint16_to_float32_scalbn(uint16_t a, int scale, float_status *status)
+{
+    return uint64_to_float32_scalbn(a, scale, status);
 }
 
 float32 uint64_to_float32(uint64_t a, float_status *status)
 {
-    FloatParts pa = uint_to_float(a, status);
-    return float32_round_pack_canonical(pa, status);
+    return uint64_to_float32_scalbn(a, 0, status);
 }
 
 float32 uint32_to_float32(uint32_t a, float_status *status)
 {
-    return uint64_to_float32(a, status);
+    return uint64_to_float32_scalbn(a, 0, status);
 }
 
 float32 uint16_to_float32(uint16_t a, float_status *status)
 {
-    return uint64_to_float32(a, status);
+    return uint64_to_float32_scalbn(a, 0, status);
+}
+
+float64 uint64_to_float64_scalbn(uint64_t a, int scale, float_status *status)
+{
+    FloatParts pa = uint_to_float(a, scale, status);
+    return float64_round_pack_canonical(pa, status);
+}
+
+float64 uint32_to_float64_scalbn(uint32_t a, int scale, float_status *status)
+{
+    return uint64_to_float64_scalbn(a, scale, status);
+}
+
+float64 uint16_to_float64_scalbn(uint16_t a, int scale, float_status *status)
+{
+    return uint64_to_float64_scalbn(a, scale, status);
 }
 
 float64 uint64_to_float64(uint64_t a, float_status *status)
 {
-    FloatParts pa = uint_to_float(a, status);
-    return float64_round_pack_canonical(pa, status);
+    return uint64_to_float64_scalbn(a, 0, status);
 }
 
 float64 uint32_to_float64(uint32_t a, float_status *status)
 {
-    return uint64_to_float64(a, status);
+    return uint64_to_float64_scalbn(a, 0, status);
 }
 
 float64 uint16_to_float64(uint16_t a, float_status *status)
 {
-    return uint64_to_float64(a, status);
+    return uint64_to_float64_scalbn(a, 0, status);
 }
 
 /* Float Min/Max */
-- 
2.17.1

[Qemu-devel] [PATCH 2/4] softfloat: Add scaling float-to-int routines

[Richard Henderson]

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 include/fpu/softfloat.h |  85 ++++++---
 fpu/softfloat.c         | 391 ++++++++++++++++++++++++++++++++--------
 2 files changed, 379 insertions(+), 97 deletions(-)

diff --git a/include/fpu/softfloat.h b/include/fpu/softfloat.h
index 038e375e71..cc1b58b029 100644
--- a/include/fpu/softfloat.h
+++ b/include/fpu/softfloat.h
@@ -243,21 +243,34 @@ float128 uint64_to_float128(uint64_t, float_status *status);
 /*----------------------------------------------------------------------------
 | Software half-precision conversion routines.
 *----------------------------------------------------------------------------*/
+
 float16 float32_to_float16(float32, bool ieee, float_status *status);
 float32 float16_to_float32(float16, bool ieee, float_status *status);
 float16 float64_to_float16(float64 a, bool ieee, float_status *status);
 float64 float16_to_float64(float16 a, bool ieee, float_status *status);
+
+int16_t float16_to_int16_scalbn(float16, int, int, float_status *status);
+int32_t float16_to_int32_scalbn(float16, int, int, float_status *status);
+int64_t float16_to_int64_scalbn(float16, int, int, float_status *status);
+
 int16_t float16_to_int16(float16, float_status *status);
-uint16_t float16_to_uint16(float16 a, float_status *status);
-int16_t float16_to_int16_round_to_zero(float16, float_status *status);
-uint16_t float16_to_uint16_round_to_zero(float16 a, float_status *status);
 int32_t float16_to_int32(float16, float_status *status);
-uint32_t float16_to_uint32(float16 a, float_status *status);
-int32_t float16_to_int32_round_to_zero(float16, float_status *status);
-uint32_t float16_to_uint32_round_to_zero(float16 a, float_status *status);
 int64_t float16_to_int64(float16, float_status *status);
-uint64_t float16_to_uint64(float16 a, float_status *status);
+
+int16_t float16_to_int16_round_to_zero(float16, float_status *status);
+int32_t float16_to_int32_round_to_zero(float16, float_status *status);
 int64_t float16_to_int64_round_to_zero(float16, float_status *status);
+
+uint16_t float16_to_uint16_scalbn(float16 a, int, int, float_status *status);
+uint32_t float16_to_uint32_scalbn(float16 a, int, int, float_status *status);
+uint64_t float16_to_uint64_scalbn(float16 a, int, int, float_status *status);
+
+uint16_t float16_to_uint16(float16 a, float_status *status);
+uint32_t float16_to_uint32(float16 a, float_status *status);
+uint64_t float16_to_uint64(float16 a, float_status *status);
+
+uint16_t float16_to_uint16_round_to_zero(float16 a, float_status *status);
+uint32_t float16_to_uint32_round_to_zero(float16 a, float_status *status);
 uint64_t float16_to_uint64_round_to_zero(float16 a, float_status *status);
 
 /*----------------------------------------------------------------------------
@@ -347,18 +360,31 @@ float16 float16_default_nan(float_status *status);
 /*----------------------------------------------------------------------------
 | Software IEC/IEEE single-precision conversion routines.
 *----------------------------------------------------------------------------*/
+
+int16_t float32_to_int16_scalbn(float32, int, int, float_status *status);
+int32_t float32_to_int32_scalbn(float32, int, int, float_status *status);
+int64_t float32_to_int64_scalbn(float32, int, int, float_status *status);
+
 int16_t float32_to_int16(float32, float_status *status);
-uint16_t float32_to_uint16(float32, float_status *status);
-int16_t float32_to_int16_round_to_zero(float32, float_status *status);
-uint16_t float32_to_uint16_round_to_zero(float32, float_status *status);
 int32_t float32_to_int32(float32, float_status *status);
-int32_t float32_to_int32_round_to_zero(float32, float_status *status);
-uint32_t float32_to_uint32(float32, float_status *status);
-uint32_t float32_to_uint32_round_to_zero(float32, float_status *status);
 int64_t float32_to_int64(float32, float_status *status);
-uint64_t float32_to_uint64(float32, float_status *status);
-uint64_t float32_to_uint64_round_to_zero(float32, float_status *status);
+
+int16_t float32_to_int16_round_to_zero(float32, float_status *status);
+int32_t float32_to_int32_round_to_zero(float32, float_status *status);
 int64_t float32_to_int64_round_to_zero(float32, float_status *status);
+
+uint16_t float32_to_uint16_scalbn(float32, int, int, float_status *status);
+uint32_t float32_to_uint32_scalbn(float32, int, int, float_status *status);
+uint64_t float32_to_uint64_scalbn(float32, int, int, float_status *status);
+
+uint16_t float32_to_uint16(float32, float_status *status);
+uint32_t float32_to_uint32(float32, float_status *status);
+uint64_t float32_to_uint64(float32, float_status *status);
+
+uint16_t float32_to_uint16_round_to_zero(float32, float_status *status);
+uint32_t float32_to_uint32_round_to_zero(float32, float_status *status);
+uint64_t float32_to_uint64_round_to_zero(float32, float_status *status);
+
 float64 float32_to_float64(float32, float_status *status);
 floatx80 float32_to_floatx80(float32, float_status *status);
 float128 float32_to_float128(float32, float_status *status);
@@ -476,18 +502,31 @@ float32 float32_default_nan(float_status *status);
 /*----------------------------------------------------------------------------
 | Software IEC/IEEE double-precision conversion routines.
 *----------------------------------------------------------------------------*/
+
+int16_t float64_to_int16_scalbn(float64, int, int, float_status *status);
+int32_t float64_to_int32_scalbn(float64, int, int, float_status *status);
+int64_t float64_to_int64_scalbn(float64, int, int, float_status *status);
+
 int16_t float64_to_int16(float64, float_status *status);
-uint16_t float64_to_uint16(float64, float_status *status);
-int16_t float64_to_int16_round_to_zero(float64, float_status *status);
-uint16_t float64_to_uint16_round_to_zero(float64, float_status *status);
 int32_t float64_to_int32(float64, float_status *status);
-int32_t float64_to_int32_round_to_zero(float64, float_status *status);
-uint32_t float64_to_uint32(float64, float_status *status);
-uint32_t float64_to_uint32_round_to_zero(float64, float_status *status);
 int64_t float64_to_int64(float64, float_status *status);
+
+int16_t float64_to_int16_round_to_zero(float64, float_status *status);
+int32_t float64_to_int32_round_to_zero(float64, float_status *status);
 int64_t float64_to_int64_round_to_zero(float64, float_status *status);
-uint64_t float64_to_uint64(float64 a, float_status *status);
-uint64_t float64_to_uint64_round_to_zero(float64 a, float_status *status);
+
+uint16_t float64_to_uint16_scalbn(float64, int, int, float_status *status);
+uint32_t float64_to_uint32_scalbn(float64, int, int, float_status *status);
+uint64_t float64_to_uint64_scalbn(float64, int, int, float_status *status);
+
+uint16_t float64_to_uint16(float64, float_status *status);
+uint32_t float64_to_uint32(float64, float_status *status);
+uint64_t float64_to_uint64(float64, float_status *status);
+
+uint16_t float64_to_uint16_round_to_zero(float64, float_status *status);
+uint32_t float64_to_uint32_round_to_zero(float64, float_status *status);
+uint64_t float64_to_uint64_round_to_zero(float64, float_status *status);
+
 float32 float64_to_float32(float64, float_status *status);
 floatx80 float64_to_floatx80(float64, float_status *status);
 float128 float64_to_float128(float64, float_status *status);
diff --git a/fpu/softfloat.c b/fpu/softfloat.c
index 12f373cbad..59ca356d0e 100644
--- a/fpu/softfloat.c
+++ b/fpu/softfloat.c
@@ -1293,19 +1293,23 @@ float32 float64_to_float32(float64 a, float_status *s)
  * Arithmetic.
  */
 
-static FloatParts round_to_int(FloatParts a, int rounding_mode, float_status *s)
+static FloatParts round_to_int(FloatParts a, int rmode,
+                               int scale, float_status *s)
 {
-    if (is_nan(a.cls)) {
-        return return_nan(a, s);
-    }
-
     switch (a.cls) {
+    case float_class_qnan:
+    case float_class_snan:
+        return return_nan(a, s);
+
     case float_class_zero:
     case float_class_inf:
-    case float_class_qnan:
         /* already "integral" */
         break;
+
     case float_class_normal:
+        scale = MIN(MAX(scale, -0x10000), 0x10000);
+        a.exp += scale;
+
         if (a.exp >= DECOMPOSED_BINARY_POINT) {
             /* already integral */
             break;
@@ -1314,7 +1318,7 @@ static FloatParts round_to_int(FloatParts a, int rounding_mode, float_status *s)
             bool one;
             /* all fractional */
             s->float_exception_flags |= float_flag_inexact;
-            switch (rounding_mode) {
+            switch (rmode) {
             case float_round_nearest_even:
                 one = a.exp == -1 && a.frac > DECOMPOSED_IMPLICIT_BIT;
                 break;
@@ -1347,7 +1351,7 @@ static FloatParts round_to_int(FloatParts a, int rounding_mode, float_status *s)
             uint64_t rnd_mask = rnd_even_mask >> 1;
             uint64_t inc;
 
-            switch (rounding_mode) {
+            switch (rmode) {
             case float_round_nearest_even:
                 inc = ((a.frac & rnd_even_mask) != frac_lsbm1 ? frac_lsbm1 : 0);
                 break;
@@ -1387,28 +1391,28 @@ static FloatParts round_to_int(FloatParts a, int rounding_mode, float_status *s)
 float16 float16_round_to_int(float16 a, float_status *s)
 {
     FloatParts pa = float16_unpack_canonical(a, s);
-    FloatParts pr = round_to_int(pa, s->float_rounding_mode, s);
+    FloatParts pr = round_to_int(pa, s->float_rounding_mode, 0, s);
     return float16_round_pack_canonical(pr, s);
 }
 
 float32 float32_round_to_int(float32 a, float_status *s)
 {
     FloatParts pa = float32_unpack_canonical(a, s);
-    FloatParts pr = round_to_int(pa, s->float_rounding_mode, s);
+    FloatParts pr = round_to_int(pa, s->float_rounding_mode, 0, s);
     return float32_round_pack_canonical(pr, s);
 }
 
 float64 float64_round_to_int(float64 a, float_status *s)
 {
     FloatParts pa = float64_unpack_canonical(a, s);
-    FloatParts pr = round_to_int(pa, s->float_rounding_mode, s);
+    FloatParts pr = round_to_int(pa, s->float_rounding_mode, 0, s);
     return float64_round_pack_canonical(pr, s);
 }
 
 float64 float64_trunc_to_int(float64 a, float_status *s)
 {
     FloatParts pa = float64_unpack_canonical(a, s);
-    FloatParts pr = round_to_int(pa, float_round_to_zero, s);
+    FloatParts pr = round_to_int(pa, float_round_to_zero, 0, s);
     return float64_round_pack_canonical(pr, s);
 }
 
@@ -1423,13 +1427,13 @@ float64 float64_trunc_to_int(float64 a, float_status *s)
  * is returned.
 */
 
-static int64_t round_to_int_and_pack(FloatParts in, int rmode,
+static int64_t round_to_int_and_pack(FloatParts in, int rmode, int scale,
                                      int64_t min, int64_t max,
                                      float_status *s)
 {
     uint64_t r;
     int orig_flags = get_float_exception_flags(s);
-    FloatParts p = round_to_int(in, rmode, s);
+    FloatParts p = round_to_int(in, rmode, scale, s);
 
     switch (p.cls) {
     case float_class_snan:
@@ -1469,38 +1473,158 @@ static int64_t round_to_int_and_pack(FloatParts in, int rmode,
     }
 }
 
-#define FLOAT_TO_INT(fsz, isz)                                          \
-int ## isz ## _t float ## fsz ## _to_int ## isz(float ## fsz a,         \
-                                                float_status *s)        \
-{                                                                       \
-    FloatParts p = float ## fsz ## _unpack_canonical(a, s);             \
-    return round_to_int_and_pack(p, s->float_rounding_mode,             \
-                                 INT ## isz ## _MIN, INT ## isz ## _MAX,\
-                                 s);                                    \
-}                                                                       \
-                                                                        \
-int ## isz ## _t float ## fsz ## _to_int ## isz ## _round_to_zero       \
- (float ## fsz a, float_status *s)                                      \
-{                                                                       \
-    FloatParts p = float ## fsz ## _unpack_canonical(a, s);             \
-    return round_to_int_and_pack(p, float_round_to_zero,                \
-                                 INT ## isz ## _MIN, INT ## isz ## _MAX,\
-                                 s);                                    \
+int16_t float16_to_int16_scalbn(float16 a, int rmode, int scale,
+                                float_status *s)
+{
+    return round_to_int_and_pack(float16_unpack_canonical(a, s),
+                                 rmode, scale, INT16_MIN, INT16_MAX, s);
 }
 
-FLOAT_TO_INT(16, 16)
-FLOAT_TO_INT(16, 32)
-FLOAT_TO_INT(16, 64)
+int32_t float16_to_int32_scalbn(float16 a, int rmode, int scale,
+                                float_status *s)
+{
+    return round_to_int_and_pack(float16_unpack_canonical(a, s),
+                                 rmode, scale, INT32_MIN, INT32_MAX, s);
+}
 
-FLOAT_TO_INT(32, 16)
-FLOAT_TO_INT(32, 32)
-FLOAT_TO_INT(32, 64)
+int64_t float16_to_int64_scalbn(float16 a, int rmode, int scale,
+                                float_status *s)
+{
+    return round_to_int_and_pack(float16_unpack_canonical(a, s),
+                                 rmode, scale, INT64_MIN, INT64_MAX, s);
+}
 
-FLOAT_TO_INT(64, 16)
-FLOAT_TO_INT(64, 32)
-FLOAT_TO_INT(64, 64)
+int16_t float32_to_int16_scalbn(float32 a, int rmode, int scale,
+                                float_status *s)
+{
+    return round_to_int_and_pack(float32_unpack_canonical(a, s),
+                                 rmode, scale, INT16_MIN, INT16_MAX, s);
+}
 
-#undef FLOAT_TO_INT
+int32_t float32_to_int32_scalbn(float32 a, int rmode, int scale,
+                                float_status *s)
+{
+    return round_to_int_and_pack(float32_unpack_canonical(a, s),
+                                 rmode, scale, INT32_MIN, INT32_MAX, s);
+}
+
+int64_t float32_to_int64_scalbn(float32 a, int rmode, int scale,
+                                float_status *s)
+{
+    return round_to_int_and_pack(float32_unpack_canonical(a, s),
+                                 rmode, scale, INT64_MIN, INT64_MAX, s);
+}
+
+int16_t float64_to_int16_scalbn(float64 a, int rmode, int scale,
+                                float_status *s)
+{
+    return round_to_int_and_pack(float64_unpack_canonical(a, s),
+                                 rmode, scale, INT16_MIN, INT16_MAX, s);
+}
+
+int32_t float64_to_int32_scalbn(float64 a, int rmode, int scale,
+                                float_status *s)
+{
+    return round_to_int_and_pack(float64_unpack_canonical(a, s),
+                                 rmode, scale, INT32_MIN, INT32_MAX, s);
+}
+
+int64_t float64_to_int64_scalbn(float64 a, int rmode, int scale,
+                                float_status *s)
+{
+    return round_to_int_and_pack(float64_unpack_canonical(a, s),
+                                 rmode, scale, INT64_MIN, INT64_MAX, s);
+}
+
+int16_t float16_to_int16(float16 a, float_status *s)
+{
+    return float16_to_int16_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+int32_t float16_to_int32(float16 a, float_status *s)
+{
+    return float16_to_int32_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+int64_t float16_to_int64(float16 a, float_status *s)
+{
+    return float16_to_int64_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+int16_t float32_to_int16(float32 a, float_status *s)
+{
+    return float32_to_int16_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+int32_t float32_to_int32(float32 a, float_status *s)
+{
+    return float32_to_int32_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+int64_t float32_to_int64(float32 a, float_status *s)
+{
+    return float32_to_int64_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+int16_t float64_to_int16(float64 a, float_status *s)
+{
+    return float64_to_int16_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+int32_t float64_to_int32(float64 a, float_status *s)
+{
+    return float64_to_int32_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+int64_t float64_to_int64(float64 a, float_status *s)
+{
+    return float64_to_int64_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+int16_t float16_to_int16_round_to_zero(float16 a, float_status *s)
+{
+    return float16_to_int16_scalbn(a, float_round_to_zero, 0, s);
+}
+
+int32_t float16_to_int32_round_to_zero(float16 a, float_status *s)
+{
+    return float16_to_int32_scalbn(a, float_round_to_zero, 0, s);
+}
+
+int64_t float16_to_int64_round_to_zero(float16 a, float_status *s)
+{
+    return float16_to_int64_scalbn(a, float_round_to_zero, 0, s);
+}
+
+int16_t float32_to_int16_round_to_zero(float32 a, float_status *s)
+{
+    return float32_to_int16_scalbn(a, float_round_to_zero, 0, s);
+}
+
+int32_t float32_to_int32_round_to_zero(float32 a, float_status *s)
+{
+    return float32_to_int32_scalbn(a, float_round_to_zero, 0, s);
+}
+
+int64_t float32_to_int64_round_to_zero(float32 a, float_status *s)
+{
+    return float32_to_int64_scalbn(a, float_round_to_zero, 0, s);
+}
+
+int16_t float64_to_int16_round_to_zero(float64 a, float_status *s)
+{
+    return float64_to_int16_scalbn(a, float_round_to_zero, 0, s);
+}
+
+int32_t float64_to_int32_round_to_zero(float64 a, float_status *s)
+{
+    return float64_to_int32_scalbn(a, float_round_to_zero, 0, s);
+}
+
+int64_t float64_to_int64_round_to_zero(float64 a, float_status *s)
+{
+    return float64_to_int64_scalbn(a, float_round_to_zero, 0, s);
+}
 
 /*
  *  Returns the result of converting the floating-point value `a' to
@@ -1515,11 +1639,12 @@ FLOAT_TO_INT(64, 64)
  *  flag.
  */
 
-static uint64_t round_to_uint_and_pack(FloatParts in, int rmode, uint64_t max,
-                                       float_status *s)
+static uint64_t round_to_uint_and_pack(FloatParts in, int rmode, int scale,
+                                       uint64_t max, float_status *s)
 {
     int orig_flags = get_float_exception_flags(s);
-    FloatParts p = round_to_int(in, rmode, s);
+    FloatParts p = round_to_int(in, rmode, scale, s);
+    uint64_t r;
 
     switch (p.cls) {
     case float_class_snan:
@@ -1532,8 +1657,6 @@ static uint64_t round_to_uint_and_pack(FloatParts in, int rmode, uint64_t max,
     case float_class_zero:
         return 0;
     case float_class_normal:
-    {
-        uint64_t r;
         if (p.sign) {
             s->float_exception_flags = orig_flags | float_flag_invalid;
             return 0;
@@ -1555,45 +1678,165 @@ static uint64_t round_to_uint_and_pack(FloatParts in, int rmode, uint64_t max,
         if (r > max) {
             s->float_exception_flags = orig_flags | float_flag_invalid;
             return max;
-        } else {
-            return r;
         }
-    }
+        return r;
     default:
         g_assert_not_reached();
     }
 }
 
-#define FLOAT_TO_UINT(fsz, isz) \
-uint ## isz ## _t float ## fsz ## _to_uint ## isz(float ## fsz a,       \
-                                                  float_status *s)      \
-{                                                                       \
-    FloatParts p = float ## fsz ## _unpack_canonical(a, s);             \
-    return round_to_uint_and_pack(p, s->float_rounding_mode,            \
-                                 UINT ## isz ## _MAX, s);               \
-}                                                                       \
-                                                                        \
-uint ## isz ## _t float ## fsz ## _to_uint ## isz ## _round_to_zero     \
- (float ## fsz a, float_status *s)                                      \
-{                                                                       \
-    FloatParts p = float ## fsz ## _unpack_canonical(a, s);             \
-    return round_to_uint_and_pack(p, float_round_to_zero,               \
-                                  UINT ## isz ## _MAX, s);              \
+uint16_t float16_to_uint16_scalbn(float16 a, int rmode, int scale,
+                                  float_status *s)
+{
+    return round_to_uint_and_pack(float16_unpack_canonical(a, s),
+                                  rmode, scale, UINT16_MAX, s);
 }
 
-FLOAT_TO_UINT(16, 16)
-FLOAT_TO_UINT(16, 32)
-FLOAT_TO_UINT(16, 64)
+uint32_t float16_to_uint32_scalbn(float16 a, int rmode, int scale,
+                                  float_status *s)
+{
+    return round_to_uint_and_pack(float16_unpack_canonical(a, s),
+                                  rmode, scale, UINT32_MAX, s);
+}
 
-FLOAT_TO_UINT(32, 16)
-FLOAT_TO_UINT(32, 32)
-FLOAT_TO_UINT(32, 64)
+uint64_t float16_to_uint64_scalbn(float16 a, int rmode, int scale,
+                                  float_status *s)
+{
+    return round_to_uint_and_pack(float16_unpack_canonical(a, s),
+                                  rmode, scale, UINT64_MAX, s);
+}
 
-FLOAT_TO_UINT(64, 16)
-FLOAT_TO_UINT(64, 32)
-FLOAT_TO_UINT(64, 64)
+uint16_t float32_to_uint16_scalbn(float32 a, int rmode, int scale,
+                                  float_status *s)
+{
+    return round_to_uint_and_pack(float32_unpack_canonical(a, s),
+                                  rmode, scale, UINT16_MAX, s);
+}
 
-#undef FLOAT_TO_UINT
+uint32_t float32_to_uint32_scalbn(float32 a, int rmode, int scale,
+                                  float_status *s)
+{
+    return round_to_uint_and_pack(float32_unpack_canonical(a, s),
+                                  rmode, scale, UINT32_MAX, s);
+}
+
+uint64_t float32_to_uint64_scalbn(float32 a, int rmode, int scale,
+                                  float_status *s)
+{
+    return round_to_uint_and_pack(float32_unpack_canonical(a, s),
+                                  rmode, scale, UINT64_MAX, s);
+}
+
+uint16_t float64_to_uint16_scalbn(float64 a, int rmode, int scale,
+                                  float_status *s)
+{
+    return round_to_uint_and_pack(float64_unpack_canonical(a, s),
+                                  rmode, scale, UINT16_MAX, s);
+}
+
+uint32_t float64_to_uint32_scalbn(float64 a, int rmode, int scale,
+                                  float_status *s)
+{
+    return round_to_uint_and_pack(float64_unpack_canonical(a, s),
+                                  rmode, scale, UINT32_MAX, s);
+}
+
+uint64_t float64_to_uint64_scalbn(float64 a, int rmode, int scale,
+                                  float_status *s)
+{
+    return round_to_uint_and_pack(float64_unpack_canonical(a, s),
+                                  rmode, scale, UINT64_MAX, s);
+}
+
+uint16_t float16_to_uint16(float16 a, float_status *s)
+{
+    return float16_to_uint16_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+uint32_t float16_to_uint32(float16 a, float_status *s)
+{
+    return float16_to_uint32_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+uint64_t float16_to_uint64(float16 a, float_status *s)
+{
+    return float16_to_uint64_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+uint16_t float32_to_uint16(float32 a, float_status *s)
+{
+    return float32_to_uint16_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+uint32_t float32_to_uint32(float32 a, float_status *s)
+{
+    return float32_to_uint32_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+uint64_t float32_to_uint64(float32 a, float_status *s)
+{
+    return float32_to_uint64_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+uint16_t float64_to_uint16(float64 a, float_status *s)
+{
+    return float64_to_uint16_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+uint32_t float64_to_uint32(float64 a, float_status *s)
+{
+    return float64_to_uint32_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+uint64_t float64_to_uint64(float64 a, float_status *s)
+{
+    return float64_to_uint64_scalbn(a, s->float_rounding_mode, 0, s);
+}
+
+uint16_t float16_to_uint16_round_to_zero(float16 a, float_status *s)
+{
+    return float16_to_uint16_scalbn(a, float_round_to_zero, 0, s);
+}
+
+uint32_t float16_to_uint32_round_to_zero(float16 a, float_status *s)
+{
+    return float16_to_uint32_scalbn(a, float_round_to_zero, 0, s);
+}
+
+uint64_t float16_to_uint64_round_to_zero(float16 a, float_status *s)
+{
+    return float16_to_uint64_scalbn(a, float_round_to_zero, 0, s);
+}
+
+uint16_t float32_to_uint16_round_to_zero(float32 a, float_status *s)
+{
+    return float32_to_uint16_scalbn(a, float_round_to_zero, 0, s);
+}
+
+uint32_t float32_to_uint32_round_to_zero(float32 a, float_status *s)
+{
+    return float32_to_uint32_scalbn(a, float_round_to_zero, 0, s);
+}
+
+uint64_t float32_to_uint64_round_to_zero(float32 a, float_status *s)
+{
+    return float32_to_uint64_scalbn(a, float_round_to_zero, 0, s);
+}
+
+uint16_t float64_to_uint16_round_to_zero(float64 a, float_status *s)
+{
+    return float64_to_uint16_scalbn(a, float_round_to_zero, 0, s);
+}
+
+uint32_t float64_to_uint32_round_to_zero(float64 a, float_status *s)
+{
+    return float64_to_uint32_scalbn(a, float_round_to_zero, 0, s);
+}
+
+uint64_t float64_to_uint64_round_to_zero(float64 a, float_status *s)
+{
+    return float64_to_uint64_scalbn(a, float_round_to_zero, 0, s);
+}
 
 /*
  * Integer to float conversions
-- 
2.17.1

[Qemu-devel] [PATCH 3/4] target/arm: Use the int-to-float-scale softfloat routines

[Richard Henderson]

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/arm/helper.c | 29 +++++------------------------
 1 file changed, 5 insertions(+), 24 deletions(-)

diff --git a/target/arm/helper.c b/target/arm/helper.c
index 61454a77ec..38439a2ee8 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -11550,12 +11550,7 @@ float32 VFP_HELPER(fcvts, d)(float64 x, CPUARMState *env)
 #define VFP_CONV_FIX_FLOAT(name, p, fsz, isz, itype) \
 float##fsz HELPER(vfp_##name##to##p)(uint##isz##_t  x, uint32_t shift, \
                                      void *fpstp) \
-{ \
-    float_status *fpst = fpstp; \
-    float##fsz tmp; \
-    tmp = itype##_to_##float##fsz(x, fpst); \
-    return float##fsz##_scalbn(tmp, -(int)shift, fpst); \
-}
+{ return itype##_to_##float##fsz##_scalbn(x, -shift, fpstp); }
 
 /* Notice that we want only input-denormal exception flags from the
  * scalbn operation: the other possible flags (overflow+inexact if
@@ -11608,38 +11603,24 @@ VFP_CONV_FIX_A64(uq, s, 32, 64, uint64)
 #undef VFP_CONV_FLOAT_FIX_ROUND
 #undef VFP_CONV_FIX_A64
 
-/* Conversion to/from f16 can overflow to infinity before/after scaling.
- * Therefore we convert to f64, scale, and then convert f64 to f16; or
- * vice versa for conversion to integer.
- *
- * For 16- and 32-bit integers, the conversion to f64 never rounds.
- * For 64-bit integers, any integer that would cause rounding will also
- * overflow to f16 infinity, so there is no double rounding problem.
- */
-
-static float16 do_postscale_fp16(float64 f, int shift, float_status *fpst)
-{
-    return float64_to_float16(float64_scalbn(f, -shift, fpst), true, fpst);
-}
-
 uint32_t HELPER(vfp_sltoh)(uint32_t x, uint32_t shift, void *fpst)
 {
-    return do_postscale_fp16(int32_to_float64(x, fpst), shift, fpst);
+    return int32_to_float16_scalbn(x, -shift, fpst);
 }
 
 uint32_t HELPER(vfp_ultoh)(uint32_t x, uint32_t shift, void *fpst)
 {
-    return do_postscale_fp16(uint32_to_float64(x, fpst), shift, fpst);
+    return uint32_to_float16_scalbn(x, -shift, fpst);
 }
 
 uint32_t HELPER(vfp_sqtoh)(uint64_t x, uint32_t shift, void *fpst)
 {
-    return do_postscale_fp16(int64_to_float64(x, fpst), shift, fpst);
+    return int64_to_float16_scalbn(x, -shift, fpst);
 }
 
 uint32_t HELPER(vfp_uqtoh)(uint64_t x, uint32_t shift, void *fpst)
 {
-    return do_postscale_fp16(uint64_to_float64(x, fpst), shift, fpst);
+    return uint64_to_float16_scalbn(x, -shift, fpst);
 }
 
 static float64 do_prescale_fp16(float16 f, int shift, float_status *fpst)
-- 
2.17.1

[Qemu-devel] [PATCH 4/4] target/arm: Use the float-to-int-scale softfloat routines

[Richard Henderson]

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 target/arm/helper.c | 101 ++++++++++++++++++++++----------------------
 1 file changed, 51 insertions(+), 50 deletions(-)

diff --git a/target/arm/helper.c b/target/arm/helper.c
index 38439a2ee8..e4a7d97805 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -11552,38 +11552,28 @@ float##fsz HELPER(vfp_##name##to##p)(uint##isz##_t  x, uint32_t shift, \
                                      void *fpstp) \
 { return itype##_to_##float##fsz##_scalbn(x, -shift, fpstp); }
 
-/* Notice that we want only input-denormal exception flags from the
- * scalbn operation: the other possible flags (overflow+inexact if
- * we overflow to infinity, output-denormal) aren't correct for the
- * complete scale-and-convert operation.
- */
-#define VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, round) \
-uint##isz##_t HELPER(vfp_to##name##p##round)(float##fsz x, \
-                                             uint32_t shift, \
-                                             void *fpstp) \
-{ \
-    float_status *fpst = fpstp; \
-    int old_exc_flags = get_float_exception_flags(fpst); \
-    float##fsz tmp; \
-    if (float##fsz##_is_any_nan(x)) { \
-        float_raise(float_flag_invalid, fpst); \
-        return 0; \
-    } \
-    tmp = float##fsz##_scalbn(x, shift, fpst); \
-    old_exc_flags |= get_float_exception_flags(fpst) \
-        & float_flag_input_denormal; \
-    set_float_exception_flags(old_exc_flags, fpst); \
-    return float##fsz##_to_##itype##round(tmp, fpst); \
+#define VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, ROUND, suff)   \
+uint##isz##_t HELPER(vfp_to##name##p##suff)(float##fsz x, uint32_t shift, \
+                                            void *fpst)                   \
+{                                                                         \
+    if (unlikely(float##fsz##_is_any_nan(x))) {                           \
+        float_raise(float_flag_invalid, fpst);                            \
+        return 0;                                                         \
+    }                                                                     \
+    return float##fsz##_to_##itype##_scalbn(x, ROUND, shift, fpst);       \
 }
 
 #define VFP_CONV_FIX(name, p, fsz, isz, itype)                   \
 VFP_CONV_FIX_FLOAT(name, p, fsz, isz, itype)                     \
-VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, _round_to_zero) \
-VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, )
+VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype,               \
+                         float_round_to_zero, _round_to_zero)    \
+VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype,               \
+                         get_float_rounding_mode(fpst), )
 
 #define VFP_CONV_FIX_A64(name, p, fsz, isz, itype)               \
 VFP_CONV_FIX_FLOAT(name, p, fsz, isz, itype)                     \
-VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype, )
+VFP_CONV_FLOAT_FIX_ROUND(name, p, fsz, isz, itype,               \
+                         get_float_rounding_mode(fpst), )
 
 VFP_CONV_FIX(sh, d, 64, 64, int16)
 VFP_CONV_FIX(sl, d, 64, 64, int32)
@@ -11623,53 +11613,64 @@ uint32_t HELPER(vfp_uqtoh)(uint64_t x, uint32_t shift, void *fpst)
     return uint64_to_float16_scalbn(x, -shift, fpst);
 }
 
-static float64 do_prescale_fp16(float16 f, int shift, float_status *fpst)
-{
-    if (unlikely(float16_is_any_nan(f))) {
-        float_raise(float_flag_invalid, fpst);
-        return 0;
-    } else {
-        int old_exc_flags = get_float_exception_flags(fpst);
-        float64 ret;
-
-        ret = float16_to_float64(f, true, fpst);
-        ret = float64_scalbn(ret, shift, fpst);
-        old_exc_flags |= get_float_exception_flags(fpst)
-            & float_flag_input_denormal;
-        set_float_exception_flags(old_exc_flags, fpst);
-
-        return ret;
-    }
-}
-
 uint32_t HELPER(vfp_toshh)(uint32_t x, uint32_t shift, void *fpst)
 {
-    return float64_to_int16(do_prescale_fp16(x, shift, fpst), fpst);
+    if (unlikely(float16_is_any_nan(x))) {
+        float_raise(float_flag_invalid, fpst);
+        return 0;
+    }
+    return float16_to_int16_scalbn(x, get_float_rounding_mode(fpst),
+                                   shift, fpst);
 }
 
 uint32_t HELPER(vfp_touhh)(uint32_t x, uint32_t shift, void *fpst)
 {
-    return float64_to_uint16(do_prescale_fp16(x, shift, fpst), fpst);
+    if (unlikely(float16_is_any_nan(x))) {
+        float_raise(float_flag_invalid, fpst);
+        return 0;
+    }
+    return float16_to_uint16_scalbn(x, get_float_rounding_mode(fpst),
+                                    shift, fpst);
 }
 
 uint32_t HELPER(vfp_toslh)(uint32_t x, uint32_t shift, void *fpst)
 {
-    return float64_to_int32(do_prescale_fp16(x, shift, fpst), fpst);
+    if (unlikely(float16_is_any_nan(x))) {
+        float_raise(float_flag_invalid, fpst);
+        return 0;
+    }
+    return float16_to_int32_scalbn(x, get_float_rounding_mode(fpst),
+                                   shift, fpst);
 }
 
 uint32_t HELPER(vfp_toulh)(uint32_t x, uint32_t shift, void *fpst)
 {
-    return float64_to_uint32(do_prescale_fp16(x, shift, fpst), fpst);
+    if (unlikely(float16_is_any_nan(x))) {
+        float_raise(float_flag_invalid, fpst);
+        return 0;
+    }
+    return float16_to_uint32_scalbn(x, get_float_rounding_mode(fpst),
+                                    shift, fpst);
 }
 
 uint64_t HELPER(vfp_tosqh)(uint32_t x, uint32_t shift, void *fpst)
 {
-    return float64_to_int64(do_prescale_fp16(x, shift, fpst), fpst);
+    if (unlikely(float16_is_any_nan(x))) {
+        float_raise(float_flag_invalid, fpst);
+        return 0;
+    }
+    return float16_to_int64_scalbn(x, get_float_rounding_mode(fpst),
+                                   shift, fpst);
 }
 
 uint64_t HELPER(vfp_touqh)(uint32_t x, uint32_t shift, void *fpst)
 {
-    return float64_to_uint64(do_prescale_fp16(x, shift, fpst), fpst);
+    if (unlikely(float16_is_any_nan(x))) {
+        float_raise(float_flag_invalid, fpst);
+        return 0;
+    }
+    return float16_to_uint64_scalbn(x, get_float_rounding_mode(fpst),
+                                    shift, fpst);
 }
 
 /* Set the current fp rounding mode and return the old one.
-- 
2.17.1

Re: [Qemu-devel] [PATCH 0/4] target/arm: Fix int64_to_float16 double-rounding

[Alex Bennée]

Richard Henderson <richard.henderson@linaro.org> writes:

> In 88808a022c0, I tried to fix an overflow problem that affected float16
> scaling by coverting first to float64 and then rounding after that.
>
> However, Laurent reported that -0x3ff40000000001 converted to float16
> resulted in 0xfbfe instead of the expected 0xfbff.  This is caused by
> the inexact conversion to float64.
>
> Rather than build more logic into target/arm to compensate, just add
> a function that takes a scaling parameter so that the whole thing is
> done all at once with only one rounding.
>
> I don't have a failing test case for the float-to-int paths, but it
> seemed best to apply the same solution.

Can't we add the constants to the fcvt test case?

>
>
> r~
>
>
> Richard Henderson (4):
>   softfloat: Add scaling int-to-float routines
>   softfloat: Add scaling float-to-int routines
>   target/arm: Use the int-to-float-scale softfloat routines
>   target/arm: Use the float-to-int-scale softfloat routines
>
>  include/fpu/softfloat.h | 169 ++++++++----
>  fpu/softfloat.c         | 579 +++++++++++++++++++++++++++++++---------
>  target/arm/helper.c     | 130 ++++-----
>  3 files changed, 628 insertions(+), 250 deletions(-)


--
Alex Bennée

Re: [Qemu-devel] [PATCH 0/4] target/arm: Fix int64_to_float16 double-rounding

[Richard Henderson]

On 08/14/2018 01:32 AM, Alex Bennée wrote:
> Can't we add the constants to the fcvt test case?

No, they're all half-to-integer.  This is integer-to-half.

We could write another one, I suppose, but it's not just
an add-one-line kind of thing.


r~

Re: [Qemu-devel] [PATCH 0/4] target/arm: Fix int64_to_float16 double-rounding

[Alex Bennée]

Richard Henderson <richard.henderson@linaro.org> writes:

> On 08/14/2018 01:32 AM, Alex Bennée wrote:
>> Can't we add the constants to the fcvt test case?
>
> No, they're all half-to-integer.  This is integer-to-half.

I'll add the int-to-float conversions, the whole thing could do with a
bit of a re-factor anyway.

>
> We could write another one, I suppose, but it's not just
> an add-one-line kind of thing.
>
>
> r~


--
Alex Bennée

Re: [Qemu-devel] [PATCH 0/4] target/arm: Fix int64_to_float16 double-rounding

[no-reply]

Hi,

This series seems to have some coding style problems. See output below for
more information:

Type: series
Message-id: 20180814002653.12828-1-richard.henderson@linaro.org
Subject: [Qemu-devel] [PATCH 0/4] target/arm: Fix int64_to_float16 double-rounding

=== TEST SCRIPT BEGIN ===
#!/bin/bash

BASE=base
n=1
total=$(git log --oneline $BASE.. | wc -l)
failed=0

git config --local diff.renamelimit 0
git config --local diff.renames True
git config --local diff.algorithm histogram

commits="$(git log --format=%H --reverse $BASE..)"
for c in $commits; do
    echo "Checking PATCH $n/$total: $(git log -n 1 --format=%s $c)..."
    if ! git show $c --format=email | ./scripts/checkpatch.pl --mailback -; then
        failed=1
        echo
    fi
    n=$((n+1))
done

exit $failed
=== TEST SCRIPT END ===

Updating 3c8cf5a9c21ff8782164d1def7f44bd888713384
Switched to a new branch 'test'
709fbe603d target/arm: Use the float-to-int-scale softfloat routines
b158c8d737 target/arm: Use the int-to-float-scale softfloat routines
5f86798067 softfloat: Add scaling float-to-int routines
8ec3fc49ea softfloat: Add scaling int-to-float routines

=== OUTPUT BEGIN ===
Checking PATCH 1/4: softfloat: Add scaling int-to-float routines...
Checking PATCH 2/4: softfloat: Add scaling float-to-int routines...
Checking PATCH 3/4: target/arm: Use the int-to-float-scale softfloat routines...
Checking PATCH 4/4: target/arm: Use the float-to-int-scale softfloat routines...
ERROR: space prohibited before that close parenthesis ')'
#57: FILE: target/arm/helper.c:11531:
+                         get_float_rounding_mode(fpst), )

ERROR: space prohibited before that close parenthesis ')'
#63: FILE: target/arm/helper.c:11536:
+                         get_float_rounding_mode(fpst), )

total: 2 errors, 0 warnings, 142 lines checked

Your patch has style problems, please review.  If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.

=== OUTPUT END ===

Test command exited with code: 1


---
Email generated automatically by Patchew [http://patchew.org/].
Please send your feedback to patchew-devel@redhat.com

Re: [Qemu-devel] [PATCH 0/4] target/arm: Fix int64_to_float16 double-rounding

[Peter Maydell]

On 14 August 2018 at 01:26, Richard Henderson
<richard.henderson@linaro.org> wrote:
> In 88808a022c0, I tried to fix an overflow problem that affected float16
> scaling by coverting first to float64 and then rounding after that.
>
> However, Laurent reported that -0x3ff40000000001 converted to float16
> resulted in 0xfbfe instead of the expected 0xfbff.  This is caused by
> the inexact conversion to float64.
>
> Rather than build more logic into target/arm to compensate, just add
> a function that takes a scaling parameter so that the whole thing is
> done all at once with only one rounding.
>
> I don't have a failing test case for the float-to-int paths, but it
> seemed best to apply the same solution.
>
>
> r~
>
>
> Richard Henderson (4):
>   softfloat: Add scaling int-to-float routines
>   softfloat: Add scaling float-to-int routines
>   target/arm: Use the int-to-float-scale softfloat routines
>   target/arm: Use the float-to-int-scale softfloat routines

series
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>

and applied to target-arm.next.

thanks
-- PMM

Re: [Qemu-devel] [PATCH 0/4] target/arm: Fix int64_to_float16 double-rounding

[no-reply]

Hi,

This series seems to have some coding style problems. See output below for
more information:

Type: series
Message-id: 20180814002653.12828-1-richard.henderson@linaro.org
Subject: [Qemu-devel] [PATCH 0/4] target/arm: Fix int64_to_float16 double-rounding

=== TEST SCRIPT BEGIN ===
#!/bin/bash

BASE=base
n=1
total=$(git log --oneline $BASE.. | wc -l)
failed=0

git config --local diff.renamelimit 0
git config --local diff.renames True
git config --local diff.algorithm histogram

commits="$(git log --format=%H --reverse $BASE..)"
for c in $commits; do
    echo "Checking PATCH $n/$total: $(git log -n 1 --format=%s $c)..."
    if ! git show $c --format=email | ./scripts/checkpatch.pl --mailback -; then
        failed=1
        echo
    fi
    n=$((n+1))
done

exit $failed
=== TEST SCRIPT END ===

Updating 3c8cf5a9c21ff8782164d1def7f44bd888713384
Switched to a new branch 'test'
776a29ed02 target/arm: Use the float-to-int-scale softfloat routines
c08c4abc59 target/arm: Use the int-to-float-scale softfloat routines
71c42653c5 softfloat: Add scaling float-to-int routines
040490a28a softfloat: Add scaling int-to-float routines

=== OUTPUT BEGIN ===
Checking PATCH 1/4: softfloat: Add scaling int-to-float routines...
Checking PATCH 2/4: softfloat: Add scaling float-to-int routines...
Checking PATCH 3/4: target/arm: Use the int-to-float-scale softfloat routines...
Checking PATCH 4/4: target/arm: Use the float-to-int-scale softfloat routines...
ERROR: space prohibited before that close parenthesis ')'
#58: FILE: target/arm/helper.c:11585:
+                         get_float_rounding_mode(fpst), )

ERROR: space prohibited before that close parenthesis ')'
#64: FILE: target/arm/helper.c:11590:
+                         get_float_rounding_mode(fpst), )

total: 2 errors, 0 warnings, 142 lines checked

Your patch has style problems, please review.  If any of these errors
are false positives report them to the maintainer, see
CHECKPATCH in MAINTAINERS.

=== OUTPUT END ===

Test command exited with code: 1


---
Email generated automatically by Patchew [http://patchew.org/].
Please send your feedback to patchew-devel@redhat.com