caio/clang-p2996
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[libclc] Move step to the CLC library; add missing half variants (#140936)

Browse Source 
The half variants were missing but are trivial to implement. There were
some incorrect mixed type overloads (step(float, double)) which aren't
in the OpenCL specification and so have been removed.

Like certain other builtins the CLC step function only deals with
identical types. The OpenCL layer is responsible for casting the scalar
argument to a vector.

This commit also trivially vectorizes the CLC function, generating
better bytecode.
This commit is contained in:
Fraser Cormack
2025-05-22 09:54:27 +01:00
committed by GitHub
parent 52698a13c6
commit 9fa81a486e
8 changed files with 70 additions and 67 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
libclc/clc/include/clc/clcmacro.h
View File

@@ -73,39 +73,6 @@
FUNCTION(x.sf, y.sf)); \
}
#define _CLC_V_S_V_VECTORIZE(DECLSPEC, RET_TYPE, FUNCTION, ARG1_TYPE, \
ARG2_TYPE) \
DECLSPEC RET_TYPE##2 FUNCTION(ARG1_TYPE x, ARG2_TYPE##2 y) { \
return (RET_TYPE##2)(FUNCTION(x, y.s0), FUNCTION(x, y.s1)); \
} \
\
DECLSPEC RET_TYPE##3 FUNCTION(ARG1_TYPE x, ARG2_TYPE##3 y) { \
return (RET_TYPE##3)(FUNCTION(x, y.s0), FUNCTION(x, y.s1), \
FUNCTION(x, y.s2)); \
} \
\
DECLSPEC RET_TYPE##4 FUNCTION(ARG1_TYPE x, ARG2_TYPE##4 y) { \
return (RET_TYPE##4)(FUNCTION(x, y.s0), FUNCTION(x, y.s1), \
FUNCTION(x, y.s2), FUNCTION(x, y.s3)); \
} \
\
DECLSPEC RET_TYPE##8 FUNCTION(ARG1_TYPE x, ARG2_TYPE##8 y) { \
return (RET_TYPE##8)(FUNCTION(x, y.s0), FUNCTION(x, y.s1), \
FUNCTION(x, y.s2), FUNCTION(x, y.s3), \
FUNCTION(x, y.s4), FUNCTION(x, y.s5), \
FUNCTION(x, y.s6), FUNCTION(x, y.s7)); \
} \
\
DECLSPEC RET_TYPE##16 FUNCTION(ARG1_TYPE x, ARG2_TYPE##16 y) { \
return (RET_TYPE##16)( \
FUNCTION(x, y.s0), FUNCTION(x, y.s1), FUNCTION(x, y.s2), \
FUNCTION(x, y.s3), FUNCTION(x, y.s4), FUNCTION(x, y.s5), \
FUNCTION(x, y.s6), FUNCTION(x, y.s7), FUNCTION(x, y.s8), \
FUNCTION(x, y.s9), FUNCTION(x, y.sa), FUNCTION(x, y.sb), \
FUNCTION(x, y.sc), FUNCTION(x, y.sd), FUNCTION(x, y.se), \
FUNCTION(x, y.sf)); \
}
#define _CLC_TERNARY_VECTORIZE(DECLSPEC, RET_TYPE, FUNCTION, ARG1_TYPE, \
ARG2_TYPE, ARG3_TYPE) \
DECLSPEC RET_TYPE##2 FUNCTION(ARG1_TYPE##2 x, ARG2_TYPE##2 y, \

19
libclc/clc/include/clc/common/clc_step.h Normal file
View File

@@ -0,0 +1,19 @@
//===----------------------------------------------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
#ifndef __CLC_COMMON_CLC_STEP_H__
#define __CLC_COMMON_CLC_STEP_H__
#define __CLC_FUNCTION __clc_step
#define __CLC_BODY <clc/shared/binary_decl.inc>
#include <clc/math/gentype.inc>
#undef __CLC_FUNCTION
#endif // __CLC_COMMON_CLC_STEP_H__

1
libclc/clc/lib/generic/SOURCES
View File

@@ -2,6 +2,7 @@ common/clc_degrees.cl
common/clc_radians.cl
common/clc_sign.cl
common/clc_smoothstep.cl
common/clc_step.cl
geometric/clc_cross.cl
geometric/clc_distance.cl
geometric/clc_dot.cl

12
libclc/clc/lib/generic/common/clc_step.cl Normal file
View File

@@ -0,0 +1,12 @@
//===----------------------------------------------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
#include <clc/clcmacro.h>
#define __CLC_BODY <clc_step.inc>
#include <clc/math/gentype.inc>

12
libclc/clc/lib/generic/common/clc_step.inc Normal file
View File

@@ -0,0 +1,12 @@
//===----------------------------------------------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE __clc_step(__CLC_GENTYPE edge,
__CLC_GENTYPE x) {
return x < edge ? __CLC_FP_LIT(0.0) : __CLC_FP_LIT(1.0);
}

6
libclc/opencl/include/clc/opencl/common/step.inc
View File

@@ -7,8 +7,8 @@
//===----------------------------------------------------------------------===//
_CLC_OVERLOAD _CLC_DECL __CLC_GENTYPE step(__CLC_GENTYPE edge, __CLC_GENTYPE x);
_CLC_OVERLOAD _CLC_DECL __CLC_GENTYPE step(float edge, __CLC_GENTYPE x);
#ifdef cl_khr_fp64
_CLC_OVERLOAD _CLC_DECL __CLC_GENTYPE step(double edge, __CLC_GENTYPE x);
#ifndef __CLC_SCALAR
_CLC_OVERLOAD _CLC_DECL __CLC_GENTYPE step(__CLC_SCALAR_GENTYPE edge,
__CLC_GENTYPE x);
#endif

34
libclc/opencl/lib/generic/common/step.cl
View File

@@ -6,36 +6,8 @@
//
//===----------------------------------------------------------------------===//
#include <clc/clcmacro.h>
#include <clc/common/clc_step.h>
#include <clc/opencl/clc.h>
_CLC_OVERLOAD _CLC_DEF float step(float edge, float x) {
return x < edge ? 0.0f : 1.0f;
}
_CLC_BINARY_VECTORIZE(_CLC_OVERLOAD _CLC_DEF, float, step, float, float);
_CLC_V_S_V_VECTORIZE(_CLC_OVERLOAD _CLC_DEF, float, step, float, float);
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64 : enable
#define STEP_DEF(edge_type, x_type) \
_CLC_OVERLOAD _CLC_DEF x_type step(edge_type edge, x_type x) { \
return x < edge ? 0.0 : 1.0; \
}
STEP_DEF(double, double);
_CLC_BINARY_VECTORIZE(_CLC_OVERLOAD _CLC_DEF, double, step, double, double);
_CLC_V_S_V_VECTORIZE(_CLC_OVERLOAD _CLC_DEF, double, step, double, double);
#if !defined(CLC_SPIRV)
STEP_DEF(float, double);
STEP_DEF(double, float);
_CLC_V_S_V_VECTORIZE(_CLC_OVERLOAD _CLC_DEF, double, step, float, double);
_CLC_V_S_V_VECTORIZE(_CLC_OVERLOAD _CLC_DEF, float, step, double, float);
#endif
#endif
#define __CLC_BODY <step.inc>
#include <clc/math/gentype.inc>

20
libclc/opencl/lib/generic/common/step.inc Normal file
View File

@@ -0,0 +1,20 @@
//===----------------------------------------------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE step(__CLC_GENTYPE edge, __CLC_GENTYPE x) {
return __clc_step(edge, x);
}
#if !defined(__CLC_SCALAR)
_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE step(__CLC_SCALAR_GENTYPE edge,
__CLC_GENTYPE x) {
return __clc_step((__CLC_GENTYPE)edge, x);
}
#endif