[libclc] Move fp32 sincos helpers to CLC library (#132753)

This commit moves most of the sincos helper functions to the CLC
library. It simultaneously vectorizes them with the aim to increase
performance for vector types by avoiding scalarization.

Some helpers for double types remain as they use various features not
yet ready, like 'fract' which in turn relies on 'fmin'; neither of these
are in the CLC library. They also use table lookups and type punning
which don't translate well to vector versions.

As a proof of concept, float and half versions of the sin and cos
builtins are now vectorized and use the CLC helpers to do so. They
remain in the OpenCL layer but will be simpler to move to the CLC
library when the double versions are ready.

Author: frasercrmck

libclc/clc/include/clc/math/clc_sincos_helpers.h

@@ -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
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef __CLC_MATH_CLC_SINCOS_HELPERS_H__
+#define __CLC_MATH_CLC_SINCOS_HELPERS_H__
+
+#define __FLOAT_ONLY
+#define __CLC_BODY <clc/math/clc_sincos_helpers.inc>
+
+#include <clc/math/gentype.inc>
+
+#undef __CLC_BODY
+#undef __FLOAT_ONLY
+
+#endif // __CLC_MATH_CLC_SINCOS_HELPERS_H__

libclc/clc/include/clc/math/clc_sincos_helpers.inc

@@ -0,0 +1,16 @@
+//===----------------------------------------------------------------------===//
+//
+// 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_DECL _CLC_OVERLOAD __CLC_FLOATN __clc_sinf_piby4(__CLC_FLOATN x,
+                                                      __CLC_FLOATN y);
+_CLC_DECL _CLC_OVERLOAD __CLC_FLOATN __clc_cosf_piby4(__CLC_FLOATN x,
+                                                      __CLC_FLOATN y);
+
+_CLC_DECL _CLC_OVERLOAD __CLC_INTN __clc_argReductionS(private __CLC_FLOATN *r,
+                                                       private __CLC_FLOATN *rr,
+                                                       __CLC_FLOATN x);

libclc/clc/include/clc/math/gentype.inc

@@ -29,6 +29,10 @@
 #define __CLC_UINTN __CLC_XCONCAT(uint, __CLC_VECSIZE)
 #define __CLC_ULONGN __CLC_XCONCAT(ulong, __CLC_VECSIZE)
 
+#define __CLC_AS_HALFN __CLC_XCONCAT(__clc_as_, __CLC_HALFN)
+#define __CLC_AS_FLOATN __CLC_XCONCAT(__clc_as_, __CLC_FLOATN)
+#define __CLC_AS_DOUBLEN __CLC_XCONCAT(__clc_as_, __CLC_DOUBLEN)
+
 #define __CLC_AS_CHARN __CLC_XCONCAT(__clc_as_, __CLC_CHARN)
 #define __CLC_AS_SHORTN __CLC_XCONCAT(__clc_as_, __CLC_SHORTN)
 #define __CLC_AS_INTN __CLC_XCONCAT(__clc_as_, __CLC_INTN)
@@ -67,6 +71,7 @@
 #define __CLC_SCALAR_GENTYPE float
 #define __CLC_FPSIZE 32
 #define __CLC_FP_LIT(x) x##F
+#define __CLC_GENTYPE_NAN FLT_NAN
 
 #define __CLC_S_GENTYPE __CLC_XCONCAT(int, __CLC_VECSIZE)
 #define __CLC_U_GENTYPE __CLC_XCONCAT(uint, __CLC_VECSIZE)
@@ -123,6 +128,7 @@
 
 #undef __CLC_U_GENTYPE
 #undef __CLC_S_GENTYPE
+#undef __CLC_GENTYPE_NAN
 #undef __CLC_FP_LIT
 #undef __CLC_FPSIZE
 #undef __CLC_SCALAR_GENTYPE
@@ -134,6 +140,7 @@
 #define __CLC_SCALAR_GENTYPE double
 #define __CLC_FPSIZE 64
 #define __CLC_FP_LIT(x) (x)
+#define __CLC_GENTYPE_NAN DBL_NAN
 
 #define __CLC_S_GENTYPE __CLC_XCONCAT(long, __CLC_VECSIZE)
 #define __CLC_U_GENTYPE __CLC_XCONCAT(ulong, __CLC_VECSIZE)
@@ -190,6 +197,7 @@
 
 #undef __CLC_U_GENTYPE
 #undef __CLC_S_GENTYPE
+#undef __CLC_GENTYPE_NAN
 #undef __CLC_FP_LIT
 #undef __CLC_FPSIZE
 #undef __CLC_SCALAR_GENTYPE
@@ -203,6 +211,7 @@
 #define __CLC_SCALAR_GENTYPE half
 #define __CLC_FPSIZE 16
 #define __CLC_FP_LIT(x) x##H
+#define __CLC_GENTYPE_NAN HALF_NAN
 
 #define __CLC_S_GENTYPE __CLC_XCONCAT(short, __CLC_VECSIZE)
 #define __CLC_U_GENTYPE __CLC_XCONCAT(ushort, __CLC_VECSIZE)
@@ -259,6 +268,7 @@
 
 #undef __CLC_U_GENTYPE
 #undef __CLC_S_GENTYPE
+#undef __CLC_GENTYPE_NAN
 #undef __CLC_FP_LIT
 #undef __CLC_FPSIZE
 #undef __CLC_SCALAR_GENTYPE
@@ -278,6 +288,10 @@
 #undef __CLC_AS_INTN
 #undef __CLC_AS_LONGN
 
+#undef __CLC_AS_HALFN
+#undef __CLC_AS_FLOATN
+#undef __CLC_AS_DOUBLEN
+
 #undef __CLC_AS_UCHARN
 #undef __CLC_AS_USHORTN
 #undef __CLC_AS_UINTN

libclc/clc/lib/generic/SOURCES

@@ -35,6 +35,7 @@ math/clc_nextafter.cl
 math/clc_rint.cl
 math/clc_round.cl
 math/clc_rsqrt.cl
+math/clc_sincos_helpers.cl
 math/clc_sqrt.cl
 math/clc_sw_fma.cl
 math/clc_trunc.cl

libclc/clc/lib/generic/math/clc_sincos_helpers.cl

@@ -0,0 +1,32 @@
+//===----------------------------------------------------------------------===//
+//
+// 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/clc_convert.h>
+#include <clc/integer/clc_clz.h>
+#include <clc/integer/clc_mul_hi.h>
+#include <clc/internal/clc.h>
+#include <clc/math/clc_fma.h>
+#include <clc/math/clc_mad.h>
+#include <clc/math/clc_trunc.h>
+#include <clc/math/math.h>
+
+#define bitalign(hi, lo, shift) ((hi) << (32 - (shift))) | ((lo) >> (shift));
+
+#define FULL_MUL(A, B, HI, LO)                                                 \
+  LO = A * B;                                                                  \
+  HI = __clc_mul_hi(A, B)
+
+#define FULL_MAD(A, B, C, HI, LO)                                              \
+  LO = ((A) * (B) + (C));                                                      \
+  HI = __clc_mul_hi(A, B);                                                     \
+  HI += LO < C ? 1U : 0U;
+
+#define __FLOAT_ONLY
+#define __CLC_BODY <clc_sincos_helpers.inc>
+
+#include <clc/math/gentype.inc>