aaron/ispc
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Saturation arithmetic mul and div for int8/int16/int32 and div for int64 was added

Browse Source
This commit is contained in:
Vsevolod Livinskij
2014-02-18 02:07:13 +04:00
parent f5508db24f
commit 735e6a8ab3
29 changed files with 931 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

210
stdlib.ispc
View File

@@ -4612,6 +4612,216 @@ static inline varying unsigned int64 saturating_sub(varying unsigned int64 a,
result &= (-(varying int64)(result <= a));
return result;
}
static inline uniform int8 saturating_div(uniform int8 a, uniform int8 b) {
/* Only one way to overflow, so test for and prevent it. */
a += !((b + 1) | ((uniform unsigned int8) a + INT8_MIN));
return a / b;
}
static inline varying int8 saturating_div(varying int8 a, varying int8 b) {
/* Only one way to overflow, so test for and prevent it. */
a += !((b + 1) | ((varying unsigned int8) a + INT8_MIN));
return a / b;
}
static inline uniform int16 saturating_div(uniform int16 a, uniform int16 b) {
/* Only one way to overflow, so test for and prevent it. */
a += !((b + 1) | ((uniform unsigned int16) a + INT16_MIN));
return a / b;
}
static inline varying int16 saturating_div(varying int16 a, varying int16 b) {
/* Only one way to overflow, so test for and prevent it. */
a += !((b + 1) | ((varying unsigned int16) a + INT16_MIN));
return a / b;
}
static inline uniform int32 saturating_div(uniform int32 a, uniform int32 b) {
/* Only one way to overflow, so test for and prevent it. */
a += !((b + 1) | ((uniform unsigned int32) a + INT32_MIN));
return a / b;
}
static inline varying int32 saturating_div(varying int32 a, varying int32 b) {
/* Only one way to overflow, so test for and prevent it. */
a += !((b + 1) | ((varying unsigned int32) a + INT32_MIN));
return a / b;
}
static inline uniform int64 saturating_div(uniform int64 a, uniform int64 b) {
/* Only one way to overflow, so test for and prevent it. */
a += !((b + 1) | ((uniform unsigned int64) a + INT64_MIN));
return a / b;
}
static inline varying int64 saturating_div(varying int64 a, varying int64 b) {
/* Only one way to overflow, so test for and prevent it. */
a += !((b + 1) | ((varying unsigned int64) a + INT64_MIN));
return a / b;
}
static inline uniform unsigned int8 saturating_div(uniform unsigned int8 a,
uniform unsigned int8 b) {
/* No overflow possible */
return a / b;
}
static inline varying unsigned int8 saturating_div(varying unsigned int8 a,
varying unsigned int8 b) {
/* No overflow possible */
return a / b;
}
static inline uniform unsigned int16 saturating_div(uniform unsigned int16 a,
uniform unsigned int16 b) {
/* No overflow possible */
return a / b;
}
static inline varying unsigned int16 saturating_div(varying unsigned int16 a,
varying unsigned int16 b) {
/* No overflow possible */
return a / b;
}
static inline uniform unsigned int32 saturating_div(uniform unsigned int32 a,
uniform unsigned int32 b) {
/* No overflow possible */
return a / b;
}
static inline varying unsigned int32 saturating_div(varying unsigned int32 a,
varying unsigned int32 b) {
/* No overflow possible */
return a / b;
}
static inline uniform unsigned int64 saturating_div(uniform unsigned int64 a,
uniform unsigned int64 b) {
/* No overflow possible */
return a / b;
}
static inline varying unsigned int64 saturating_div(varying unsigned int64 a,
varying unsigned int64 b) {
/* No overflow possible */
return a / b;
}
static inline uniform int8 saturating_mul(uniform int8 a, uniform int8 b) {
uniform int16 result = (uniform int16) a * (uniform int16) b;
uniform unsigned int8 result2 = ((uniform unsigned int8) (a ^ b) >> 7) + INT8_MAX;
uniform int8 hi = result >> 8;
uniform int8 lo = result;
if (hi != (lo >> 7))
result = result2;
return result;
}
static inline varying int8 saturating_mul(varying int8 a, varying int8 b) {
varying int16 result = (varying int16) a * (varying int16) b;
varying unsigned int8 result2 = ((varying unsigned int8) (a ^ b) >> 7) + INT8_MAX;
varying int8 hi = result >> 8;
varying int8 lo = result;
if (hi != (lo >> 7))
result = result2;
return result;
}
static inline uniform int16 saturating_mul(uniform int16 a, uniform int16 b) {
uniform int32 result = (uniform int32) a * (uniform int32) b;
uniform unsigned int16 result2 = ((uniform unsigned int16) (a ^ b) >> 15) + INT16_MAX;
uniform int16 hi = result >> 16;
uniform int16 lo = result;
if (hi != (lo >> 15))
result = result2;
return result;
}
static inline varying int16 saturating_mul(varying int16 a, varying int16 b) {
varying int32 result = (varying int32) a * (varying int32) b;
varying unsigned int16 result2 = ((varying unsigned int16) (a ^ b) >> 15) + INT16_MAX;
varying int16 hi = result >> 16;
varying int16 lo = result;
if (hi != (lo >> 15))
result = result2;
return result;
}
static inline uniform int32 saturating_mul(uniform int32 a, uniform int32 b) {
uniform int64 result = (uniform int64) a * (uniform int64) b;
uniform unsigned int32 result2 = ((uniform unsigned int32) (a ^ b) >> 31) + INT32_MAX;
uniform int32 hi = result >> 32;
uniform int32 lo = result;
if (hi != (lo >> 31))
result = result2;
return result;
}
static inline varying int32 saturating_mul(varying int32 a, varying int32 b) {
varying int64 result = (varying int64) a * (varying int64) b;
varying unsigned int32 result2 = ((varying unsigned int32) (a ^ b) >> 31) + INT32_MAX;
varying int32 hi = result >> 32;
varying int32 lo = result;
if (hi != (lo >> 31))
result = result2;
return result;
}
static inline uniform unsigned int8 saturating_mul(uniform unsigned int8 a,
uniform unsigned int8 b) {
uniform unsigned int16 result = (uniform unsigned int16) a *
(uniform unsigned int16) b;
uniform unsigned int8 hi = result >> 8;
uniform unsigned int8 lo = result;
return lo | - (uniform int8) !! hi;
}
static inline varying unsigned int8 saturating_mul(varying unsigned int8 a,
varying unsigned int8 b) {
varying unsigned int16 result = (varying unsigned int16) a *
(varying unsigned int16) b;
varying unsigned int8 hi = result >> 8;
varying unsigned int8 lo = result;
return lo | - (varying int8) !! hi;
}
static inline uniform unsigned int16 saturating_mul(uniform unsigned int16 a,
uniform unsigned int16 b) {
uniform unsigned int32 result = (uniform unsigned int32) a *
(uniform unsigned int32) b;
uniform unsigned int16 hi = result >> 16;
uniform unsigned int16 lo = result;
return lo | - (uniform int16) !! hi;
}
static inline varying unsigned int16 saturating_mul(varying unsigned int16 a,
varying unsigned int16 b) {
varying unsigned int32 result = (varying unsigned int32) a *
(varying unsigned int32) b;
varying unsigned int16 hi = result >> 16;
varying unsigned int16 lo = result;
return lo | - (varying int16) !! hi;
}
static inline uniform unsigned int32 saturating_mul(uniform unsigned int32 a,
uniform unsigned int32 b) {
uniform unsigned int64 result = (uniform unsigned int64) a *
(uniform unsigned int64) b;
uniform unsigned int32 hi = result >> 32;
uniform unsigned int32 lo = result;
return lo | - (uniform int32) !! hi;
}
static inline varying unsigned int32 saturating_mul(varying unsigned int32 a,
varying unsigned int32 b) {
varying unsigned int64 result = (varying unsigned int64) a *
(varying unsigned int64) b;
varying unsigned int32 hi = result >> 32;
varying unsigned int32 lo = result;
return lo | - (varying int32) !! hi;
}
///////////////////////////////////////////////////////////////////////////
// rdrand