glsl: make use of active_shader_mask when building resource list

[mesa.git] / src / compiler / glsl / float64.glsl

diff --git a/src/compiler/glsl/float64.glsl b/src/compiler/glsl/float64.glsl
index c6ba37221af7e2d690d19720f6a27aab0b7ba48d..0433d925a3958289e040781c094c63e5b1f4334e 100644 (file)
--- a/src/compiler/glsl/float64.glsl
+++ b/src/compiler/glsl/float64.glsl
@@ -218,6 +218,18 @@ __fge64(uint64_t a, uint64_t b)
    return !__flt64_nonnan(a, b);
 }
 
+uint64_t
+__fsat64(uint64_t __a)
+{
+   if (__flt64(__a, 0ul))
+      return 0ul;
+
+   if (__fge64(__a, 0x3FF0000000000000ul /* 1.0 */))
+      return 0x3FF0000000000000ul;
+
+   return __a;
+}
+
 /* Adds the 64-bit value formed by concatenating `a0' and `a1' to the 64-bit
  * value formed by concatenating `b0' and `b1'.  Addition is modulo 2^64, so
  * any carry out is lost.  The result is broken into two 32-bit pieces which
@@ -483,6 +495,42 @@ __roundAndPackUInt64(uint zSign, uint zFrac0, uint zFrac1, uint zFrac2)
               (zSign !=0u && (zFrac0 | zFrac1) != 0u));
 }
 
+int64_t
+__roundAndPackInt64(uint zSign, uint zFrac0, uint zFrac1, uint zFrac2)
+{
+   bool roundNearestEven;
+   bool increment;
+   int64_t default_NegNaN = -0x7FFFFFFFFFFFFFFEL;
+   int64_t default_PosNaN = 0xFFFFFFFFFFFFFFFFL;
+
+   roundNearestEven = FLOAT_ROUNDING_MODE == FLOAT_ROUND_NEAREST_EVEN;
+
+   if (zFrac2 >= 0x80000000u)
+      increment = false;
+
+   if (!roundNearestEven) {
+      if (zSign != 0u) {
+         increment = ((FLOAT_ROUNDING_MODE == FLOAT_ROUND_DOWN) &&
+            (zFrac2 != 0u));
+      } else {
+         increment = (FLOAT_ROUNDING_MODE == FLOAT_ROUND_UP) &&
+            (zFrac2 != 0u);
+      }
+   }
+
+   if (increment) {
+      __add64(zFrac0, zFrac1, 0u, 1u, zFrac0, zFrac1);
+      if ((zFrac0 | zFrac1) != 0u)
+         zFrac1 &= ~(1u) + uint(zFrac2 == 0u) & uint(roundNearestEven);
+   }
+
+   int64_t absZ = mix(int64_t(packUint2x32(uvec2(zFrac1, zFrac0))),
+                      -int64_t(packUint2x32(uvec2(zFrac1, zFrac0))),
+                      (zSign != 0u));
+   int64_t nan = mix(default_PosNaN, default_NegNaN, bool(zSign));
+   return mix(absZ, nan, bool(zSign ^ uint(absZ < 0)) && bool(absZ));
+}
+
 /* Returns the number of leading 0 bits before the most-significant 1 bit of
  * `a'.  If `a' is zero, 32 is returned.
  */
@@ -937,6 +985,122 @@ __uint64_to_fp64(uint64_t a)
    }
 }
 
+uint64_t
+__fp64_to_uint64(uint64_t a)
+{
+   uint aFracLo = __extractFloat64FracLo(a);
+   uint aFracHi = __extractFloat64FracHi(a);
+   int aExp = __extractFloat64Exp(a);
+   uint aSign = __extractFloat64Sign(a);
+   uint zFrac2 = 0u;
+   uint64_t default_nan = 0xFFFFFFFFFFFFFFFFUL;
+
+   aFracHi = mix(aFracHi, aFracHi | 0x00100000u, aExp != 0);
+   int shiftCount = 0x433 - aExp;
+
+   if ( shiftCount <= 0 ) {
+      if (shiftCount < -11 && aExp == 0x7FF) {
+         if ((aFracHi | aFracLo) != 0u)
+            return __propagateFloat64NaN(a, a);
+         return mix(default_nan, a, aSign == 0u);
+      }
+      __shortShift64Left(aFracHi, aFracLo, -shiftCount, aFracHi, aFracLo);
+   } else {
+      __shift64ExtraRightJamming(aFracHi, aFracLo, zFrac2, shiftCount,
+                                 aFracHi, aFracLo, zFrac2);
+   }
+   return __roundAndPackUInt64(aSign, aFracHi, aFracLo, zFrac2);
+}
+
+int64_t
+__fp64_to_int64(uint64_t a)
+{
+   uint zFrac2 = 0u;
+   uint aFracLo = __extractFloat64FracLo(a);
+   uint aFracHi = __extractFloat64FracHi(a);
+   int aExp = __extractFloat64Exp(a);
+   uint aSign = __extractFloat64Sign(a);
+   int64_t default_NegNaN = -0x7FFFFFFFFFFFFFFEL;
+   int64_t default_PosNaN = 0xFFFFFFFFFFFFFFFFL;
+
+   aFracHi = mix(aFracHi, aFracHi | 0x00100000u, aExp != 0);
+   int shiftCount = 0x433 - aExp;
+
+   if (shiftCount <= 0) {
+      if (shiftCount < -11 && aExp == 0x7FF) {
+         if ((aFracHi | aFracLo) != 0u)
+            return default_NegNaN;
+         return mix(default_NegNaN, default_PosNaN, aSign == 0u);
+      }
+      __shortShift64Left(aFracHi, aFracLo, -shiftCount, aFracHi, aFracLo);
+   } else {
+      __shift64ExtraRightJamming(aFracHi, aFracLo, zFrac2, shiftCount,
+                                 aFracHi, aFracLo, zFrac2);
+   }
+
+   return __roundAndPackInt64(aSign, aFracHi, aFracLo, zFrac2);
+}
+
+uint64_t
+__fp32_to_uint64(float f)
+{
+   uint a = floatBitsToUint(f);
+   uint aFrac = a & 0x007FFFFFu;
+   int aExp = int((a>>23) & 0xFFu);
+   uint aSign = a>>31;
+   uint zFrac0 = 0u;
+   uint zFrac1 = 0u;
+   uint zFrac2 = 0u;
+   uint64_t default_nan = 0xFFFFFFFFFFFFFFFFUL;
+   int shiftCount = 0xBE - aExp;
+
+   if (shiftCount <0) {
+      if (aExp == 0xFF)
+         return default_nan;
+   }
+
+   aFrac = mix(aFrac, aFrac | 0x00800000u, aExp != 0);
+   __shortShift64Left(aFrac, 0, 40, zFrac0, zFrac1);
+
+   if (shiftCount != 0) {
+      __shift64ExtraRightJamming(zFrac0, zFrac1, zFrac2, shiftCount,
+                                 zFrac0, zFrac1, zFrac2);
+   }
+
+   return __roundAndPackUInt64(aSign, zFrac0, zFrac1, zFrac2);
+}
+
+int64_t
+__fp32_to_int64(float f)
+{
+   uint a = floatBitsToUint(f);
+   uint aFrac = a & 0x007FFFFFu;
+   int aExp = int((a>>23) & 0xFFu);
+   uint aSign = a>>31;
+   uint zFrac0 = 0u;
+   uint zFrac1 = 0u;
+   uint zFrac2 = 0u;
+   int64_t default_NegNaN = -0x7FFFFFFFFFFFFFFEL;
+   int64_t default_PosNaN = 0xFFFFFFFFFFFFFFFFL;
+   int shiftCount = 0xBE - aExp;
+
+   if (shiftCount <0) {
+      if (aExp == 0xFF && aFrac != 0u)
+         return default_NegNaN;
+      return mix(default_NegNaN, default_PosNaN, aSign == 0u);
+   }
+
+   aFrac = mix(aFrac, aFrac | 0x00800000u, aExp != 0);
+   __shortShift64Left(aFrac, 0, 40, zFrac0, zFrac1);
+
+   if (shiftCount != 0) {
+      __shift64ExtraRightJamming(zFrac0, zFrac1, zFrac2, shiftCount,
+                                 zFrac0, zFrac1, zFrac2);
+   }
+
+   return __roundAndPackInt64(aSign, zFrac0, zFrac1, zFrac2);
+}
+
 uint64_t
 __int64_to_fp64(int64_t a)
 {
@@ -1529,17 +1693,22 @@ __fround64(uint64_t __a)
 
    if (unbiasedExp < 20) {
       if (unbiasedExp < 0) {
+         if ((aHi & 0x80000000u) != 0u && aLo == 0u) {
+            return 0;
+         }
          aHi &= 0x80000000u;
-         if (unbiasedExp == -1 && aLo != 0u)
-            aHi |= (1023u << 20);
+         if ((a.y & 0x000FFFFFu) == 0u && a.x == 0u) {
+            aLo = 0u;
+            return packUint2x32(uvec2(aLo, aHi));
+         }
+         aHi = mix(aHi, (aHi | 0x3FF00000u), unbiasedExp == -1);
          aLo = 0u;
       } else {
          uint maskExp = 0x000FFFFFu >> unbiasedExp;
-         /* a is an integral value */
-         if (((aHi & maskExp) == 0u) && (aLo == 0u))
-            return __a;
-
+         uint lastBit = maskExp + 1;
          aHi += 0x00080000u >> unbiasedExp;
+         if ((aHi & maskExp) == 0u)
+            aHi &= ~lastBit;
          aHi &= ~maskExp;
          aLo = 0u;
       }
@@ -1556,9 +1725,7 @@ __fround64(uint64_t __a)
       aLo &= ~maskExp;
    }
 
-   a.x = aLo;
-   a.y = aHi;
-   return packUint2x32(a);
+   return packUint2x32(uvec2(aLo, aHi));
 }
 
 uint64_t