shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity

[libclc] use clc functions in clspv/shared/vstore_half.cl (#171770)

Browse Source
This commit is contained in:
Romaric Jodin 2025-12-11 08:41:55 +01:00 committed by GitHub
parent 2ce17ba347
commit aa31efcee1
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 18 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
libclc/opencl/lib/clspv/shared/vstore_half.cl
View File

@ -8,7 +8,13 @@
#include <clc/clc_as_type.h>
#include <clc/float/definitions.h>
#include <clc/math/clc_copysign.h>
#include <clc/math/clc_fabs.h>
#include <clc/math/clc_nextafter.h>
#include <clc/opencl/opencl-base.h>
#include <clc/relational/clc_isinf.h>
#include <clc/relational/clc_isnan.h>
#include <clc/shared/clc_min.h>
#pragma OPENCL EXTENSION cl_khr_byte_addressable_store : enable
@ -48,32 +54,32 @@
_CLC_DEF _CLC_OVERLOAD float __clc_rtz(float x) {
/* Handle nan corner case */
if (isnan(x))
if (__clc_isnan(x))
return x;
/* RTZ does not produce Inf for large numbers */
if (fabs(x) > 65504.0f && !isinf(x))
return copysign(65504.0f, x);
if (__clc_fabs(x) > 65504.0f && !__clc_isinf(x))
return __clc_copysign(65504.0f, x);
const int exp = (__clc_as_uint(x) >> 23 & 0xff) - 127;
/* Manage range rounded to +- zero explicitely */
if (exp < -24)
return copysign(0.0f, x);
return __clc_copysign(0.0f, x);
/* Remove lower 13 bits to make sure the number is rounded down */
int mask = 0xffffe000;
/* Denormals cannot be flushed, and they use different bit for rounding */
if (exp < -14)
mask <<= min(-(exp + 14), 10);
mask <<= __clc_min(-(exp + 14), 10);
return __clc_as_float(__clc_as_uint(x) & mask);
}
_CLC_DEF _CLC_OVERLOAD float __clc_rti(float x) {
/* Handle nan corner case */
if (isnan(x))
if (__clc_isnan(x))
return x;
const float inf = copysign(INFINITY, x);
const float inf = __clc_copysign(INFINITY, x);
uint ux = __clc_as_uint(x);
/* Manage +- infinity explicitely */
@ -82,23 +88,23 @@ _CLC_DEF _CLC_OVERLOAD float __clc_rti(float x) {
}
/* Manage +- zero explicitely */
if ((ux & 0x7fffffff) == 0) {
return copysign(0.0f, x);
return __clc_copysign(0.0f, x);
}
const int exp = (__clc_as_uint(x) >> 23 & 0xff) - 127;
/* Manage range rounded to smallest half denormal explicitely */
if (exp < -24) {
return copysign(0x1.0p-24f, x);
return __clc_copysign(0x1.0p-24f, x);
}
/* Set lower 13 bits */
int mask = (1 << 13) - 1;
/* Denormals cannot be flushed, and they use different bit for rounding */
if (exp < -14) {
mask = (1 << (13 + min(-(exp + 14), 10))) - 1;
mask = (1 << (13 + __clc_min(-(exp + 14), 10))) - 1;
}
const float next = nextafter(__clc_as_float(ux | mask), inf);
const float next = __clc_nextafter(__clc_as_float(ux | mask), inf);
return ((ux & mask) == 0) ? __clc_as_float(ux) : next;
}
_CLC_DEF _CLC_OVERLOAD float __clc_rtn(float x) {
@ -116,7 +122,7 @@ _CLC_DEF _CLC_OVERLOAD float __clc_rte(float x) {
/* The default assumes lower 13 bits are rounded,
* but it might be more for denormals.
* Shifting beyond last == 0b, and qr == 00b is not necessary */
shift += min(-(exp + 14), 15);
shift += __clc_min(-(exp + 14), 15);
}
int mask = (1 << shift) - 1;
const uint grs = mantissa & mask;