Skip to content

Commit ce73cbb

Browse files
arsenmarsenm
authored
clang: Use generic builtins in cuda complex builtins header (#171106)
There's no reason to use the ocml or nv prefixed functions and maintain this list of alias macros. I left these macros in for NVPTX in the scalbn and logb case, since those have a special case hack in the AMDGPU codegen and probably do not work on ptx.
1 parent cc19f42 commit ce73cbb

File tree

5 files changed

+156
-212
lines changed

5 files changed

+156
-212
lines changed

‎clang/lib/Headers/__clang_cuda_complex_builtins.h‎

Lines changed: 87 additions & 143 deletions
Original file line numberDiff line numberDiff line change
@@ -23,63 +23,17 @@
2323
#define __DEVICE__ __device__ inline
2424
#endif
2525

26-
// To make the algorithms available for C and C++ in CUDA and OpenMP we select
27-
// different but equivalent function versions. TODO: For OpenMP we currently
28-
// select the native builtins as the overload support for templates is lacking.
29-
#if !defined(__OPENMP_NVPTX__) && !defined(__OPENMP_AMDGCN__)
30-
#define _ISNANd std::isnan
31-
#define _ISNANf std::isnan
32-
#define _ISINFd std::isinf
33-
#define _ISINFf std::isinf
34-
#define _ISFINITEd std::isfinite
35-
#define _ISFINITEf std::isfinite
36-
#define _COPYSIGNd std::copysign
37-
#define _COPYSIGNf std::copysign
38-
#define _SCALBNd std::scalbn
39-
#define _SCALBNf std::scalbn
40-
#define _ABSd std::abs
41-
#define _ABSf std::abs
42-
#define _LOGBd std::logb
43-
#define _LOGBf std::logb
44-
// Rather than pulling in std::max from algorithm everytime, use available ::max.
45-
#define _fmaxd max
46-
#define _fmaxf max
47-
#else
48-
#ifdef __AMDGCN__
49-
#define _ISNANd __ocml_isnan_f64
50-
#define _ISNANf __ocml_isnan_f32
51-
#define _ISINFd __ocml_isinf_f64
52-
#define _ISINFf __ocml_isinf_f32
53-
#define _ISFINITEd __ocml_isfinite_f64
54-
#define _ISFINITEf __ocml_isfinite_f32
55-
#define _COPYSIGNd __ocml_copysign_f64
56-
#define _COPYSIGNf __ocml_copysign_f32
57-
#define _SCALBNd __ocml_scalbn_f64
58-
#define _SCALBNf __ocml_scalbn_f32
59-
#define _ABSd __ocml_fabs_f64
60-
#define _ABSf __ocml_fabs_f32
61-
#define _LOGBd __ocml_logb_f64
62-
#define _LOGBf __ocml_logb_f32
63-
#define _fmaxd __ocml_fmax_f64
64-
#define _fmaxf __ocml_fmax_f32
65-
#else
66-
#define _ISNANd __nv_isnand
67-
#define _ISNANf __nv_isnanf
68-
#define _ISINFd __nv_isinfd
69-
#define _ISINFf __nv_isinff
70-
#define _ISFINITEd __nv_isfinited
71-
#define _ISFINITEf __nv_finitef
72-
#define _COPYSIGNd __nv_copysign
73-
#define _COPYSIGNf __nv_copysignf
26+
#ifdef __NVPTX__
27+
// FIXME: NVPTX should use generic builtins.
7428
#define _SCALBNd __nv_scalbn
7529
#define _SCALBNf __nv_scalbnf
76-
#define _ABSd __nv_fabs
77-
#define _ABSf __nv_fabsf
7830
#define _LOGBd __nv_logb
7931
#define _LOGBf __nv_logbf
80-
#define _fmaxd __nv_fmax
81-
#define _fmaxf __nv_fmaxf
82-
#endif
32+
#else
33+
#define _SCALBNd __builtin_scalbn
34+
#define _SCALBNf __builtin_scalbnf
35+
#define _LOGBd __builtin_logb
36+
#define _LOGBf __builtin_logbf
8337
#endif
8438

8539
#if defined(__cplusplus)
@@ -95,36 +49,36 @@ __DEVICE__ double _Complex __muldc3(double __a, double __b, double __c,
9549
double _Complex z;
9650
__real__(z) = __ac - __bd;
9751
__imag__(z) = __ad + __bc;
98-
if (_ISNANd(__real__(z)) && _ISNANd(__imag__(z))) {
52+
if (__builtin_isnan(__real__(z)) && __builtin_isnan(__imag__(z))) {
9953
int __recalc = 0;
100-
if (_ISINFd(__a) || _ISINFd(__b)) {
101-
__a = _COPYSIGNd(_ISINFd(__a) ? 1 : 0, __a);
102-
__b = _COPYSIGNd(_ISINFd(__b) ? 1 : 0, __b);
103-
if (_ISNANd(__c))
104-
__c = _COPYSIGNd(0, __c);
105-
if (_ISNANd(__d))
106-
__d = _COPYSIGNd(0, __d);
54+
if (__builtin_isinf(__a) || __builtin_isinf(__b)) {
55+
__a = __builtin_copysign(__builtin_isinf(__a) ? 1 : 0, __a);
56+
__b = __builtin_copysign(__builtin_isinf(__b) ? 1 : 0, __b);
57+
if (__builtin_isnan(__c))
58+
__c = __builtin_copysign(0, __c);
59+
if (__builtin_isnan(__d))
60+
__d = __builtin_copysign(0, __d);
10761
__recalc = 1;
10862
}
109-
if (_ISINFd(__c) || _ISINFd(__d)) {
110-
__c = _COPYSIGNd(_ISINFd(__c) ? 1 : 0, __c);
111-
__d = _COPYSIGNd(_ISINFd(__d) ? 1 : 0, __d);
112-
if (_ISNANd(__a))
113-
__a = _COPYSIGNd(0, __a);
114-
if (_ISNANd(__b))
115-
__b = _COPYSIGNd(0, __b);
63+
if (__builtin_isinf(__c) || __builtin_isinf(__d)) {
64+
__c = __builtin_copysign(__builtin_isinf(__c) ? 1 : 0, __c);
65+
__d = __builtin_copysign(__builtin_isinf(__d) ? 1 : 0, __d);
66+
if (__builtin_isnan(__a))
67+
__a = __builtin_copysign(0, __a);
68+
if (__builtin_isnan(__b))
69+
__b = __builtin_copysign(0, __b);
11670
__recalc = 1;
11771
}
118-
if (!__recalc &&
119-
(_ISINFd(__ac) || _ISINFd(__bd) || _ISINFd(__ad) || _ISINFd(__bc))) {
120-
if (_ISNANd(__a))
121-
__a = _COPYSIGNd(0, __a);
122-
if (_ISNANd(__b))
123-
__b = _COPYSIGNd(0, __b);
124-
if (_ISNANd(__c))
125-
__c = _COPYSIGNd(0, __c);
126-
if (_ISNANd(__d))
127-
__d = _COPYSIGNd(0, __d);
72+
if (!__recalc && (__builtin_isinf(__ac) || __builtin_isinf(__bd) ||
73+
__builtin_isinf(__ad) || __builtin_isinf(__bc))) {
74+
if (__builtin_isnan(__a))
75+
__a = __builtin_copysign(0, __a);
76+
if (__builtin_isnan(__b))
77+
__b = __builtin_copysign(0, __b);
78+
if (__builtin_isnan(__c))
79+
__c = __builtin_copysign(0, __c);
80+
if (__builtin_isnan(__d))
81+
__d = __builtin_copysign(0, __d);
12882
__recalc = 1;
12983
}
13084
if (__recalc) {
@@ -145,36 +99,36 @@ __DEVICE__ float _Complex __mulsc3(float __a, float __b, float __c, float __d) {
14599
float _Complex z;
146100
__real__(z) = __ac - __bd;
147101
__imag__(z) = __ad + __bc;
148-
if (_ISNANf(__real__(z)) && _ISNANf(__imag__(z))) {
102+
if (__builtin_isnan(__real__(z)) && __builtin_isnan(__imag__(z))) {
149103
int __recalc = 0;
150-
if (_ISINFf(__a) || _ISINFf(__b)) {
151-
__a = _COPYSIGNf(_ISINFf(__a) ? 1 : 0, __a);
152-
__b = _COPYSIGNf(_ISINFf(__b) ? 1 : 0, __b);
153-
if (_ISNANf(__c))
154-
__c = _COPYSIGNf(0, __c);
155-
if (_ISNANf(__d))
156-
__d = _COPYSIGNf(0, __d);
104+
if (__builtin_isinf(__a) || __builtin_isinf(__b)) {
105+
__a = __builtin_copysignf(__builtin_isinf(__a) ? 1 : 0, __a);
106+
__b = __builtin_copysignf(__builtin_isinf(__b) ? 1 : 0, __b);
107+
if (__builtin_isnan(__c))
108+
__c = __builtin_copysignf(0, __c);
109+
if (__builtin_isnan(__d))
110+
__d = __builtin_copysignf(0, __d);
157111
__recalc = 1;
158112
}
159-
if (_ISINFf(__c) || _ISINFf(__d)) {
160-
__c = _COPYSIGNf(_ISINFf(__c) ? 1 : 0, __c);
161-
__d = _COPYSIGNf(_ISINFf(__d) ? 1 : 0, __d);
162-
if (_ISNANf(__a))
163-
__a = _COPYSIGNf(0, __a);
164-
if (_ISNANf(__b))
165-
__b = _COPYSIGNf(0, __b);
113+
if (__builtin_isinf(__c) || __builtin_isinf(__d)) {
114+
__c = __builtin_copysignf(__builtin_isinf(__c) ? 1 : 0, __c);
115+
__d = __builtin_copysignf(__builtin_isinf(__d) ? 1 : 0, __d);
116+
if (__builtin_isnan(__a))
117+
__a = __builtin_copysignf(0, __a);
118+
if (__builtin_isnan(__b))
119+
__b = __builtin_copysignf(0, __b);
166120
__recalc = 1;
167121
}
168-
if (!__recalc &&
169-
(_ISINFf(__ac) || _ISINFf(__bd) || _ISINFf(__ad) || _ISINFf(__bc))) {
170-
if (_ISNANf(__a))
171-
__a = _COPYSIGNf(0, __a);
172-
if (_ISNANf(__b))
173-
__b = _COPYSIGNf(0, __b);
174-
if (_ISNANf(__c))
175-
__c = _COPYSIGNf(0, __c);
176-
if (_ISNANf(__d))
177-
__d = _COPYSIGNf(0, __d);
122+
if (!__recalc && (__builtin_isinf(__ac) || __builtin_isinf(__bd) ||
123+
__builtin_isinf(__ad) || __builtin_isinf(__bc))) {
124+
if (__builtin_isnan(__a))
125+
__a = __builtin_copysignf(0, __a);
126+
if (__builtin_isnan(__b))
127+
__b = __builtin_copysignf(0, __b);
128+
if (__builtin_isnan(__c))
129+
__c = __builtin_copysignf(0, __c);
130+
if (__builtin_isnan(__d))
131+
__d = __builtin_copysignf(0, __d);
178132
__recalc = 1;
179133
}
180134
if (__recalc) {
@@ -191,8 +145,9 @@ __DEVICE__ double _Complex __divdc3(double __a, double __b, double __c,
191145
// Can't use std::max, because that's defined in <algorithm>, and we don't
192146
// want to pull that in for every compile. The CUDA headers define
193147
// ::max(float, float) and ::max(double, double), which is sufficient for us.
194-
double __logbw = _LOGBd(_fmaxd(_ABSd(__c), _ABSd(__d)));
195-
if (_ISFINITEd(__logbw)) {
148+
double __logbw =
149+
_LOGBd(__builtin_fmax(__builtin_fabs(__c), __builtin_fabs(__d)));
150+
if (__builtin_isfinite(__logbw)) {
196151
__ilogbw = (int)__logbw;
197152
__c = _SCALBNd(__c, -__ilogbw);
198153
__d = _SCALBNd(__d, -__ilogbw);
@@ -201,20 +156,20 @@ __DEVICE__ double _Complex __divdc3(double __a, double __b, double __c,
201156
double _Complex z;
202157
__real__(z) = _SCALBNd((__a * __c + __b * __d) / __denom, -__ilogbw);
203158
__imag__(z) = _SCALBNd((__b * __c - __a * __d) / __denom, -__ilogbw);
204-
if (_ISNANd(__real__(z)) && _ISNANd(__imag__(z))) {
205-
if ((__denom == 0.0) && (!_ISNANd(__a) || !_ISNANd(__b))) {
206-
__real__(z) = _COPYSIGNd(__builtin_huge_val(), __c) * __a;
207-
__imag__(z) = _COPYSIGNd(__builtin_huge_val(), __c) * __b;
208-
} else if ((_ISINFd(__a) || _ISINFd(__b)) && _ISFINITEd(__c) &&
209-
_ISFINITEd(__d)) {
210-
__a = _COPYSIGNd(_ISINFd(__a) ? 1.0 : 0.0, __a);
211-
__b = _COPYSIGNd(_ISINFd(__b) ? 1.0 : 0.0, __b);
159+
if (__builtin_isnan(__real__(z)) && __builtin_isnan(__imag__(z))) {
160+
if ((__denom == 0.0) && (!__builtin_isnan(__a) || !__builtin_isnan(__b))) {
161+
__real__(z) = __builtin_copysign(__builtin_huge_val(), __c) * __a;
162+
__imag__(z) = __builtin_copysign(__builtin_huge_val(), __c) * __b;
163+
} else if ((__builtin_isinf(__a) || __builtin_isinf(__b)) &&
164+
__builtin_isfinite(__c) && __builtin_isfinite(__d)) {
165+
__a = __builtin_copysign(__builtin_isinf(__a) ? 1.0 : 0.0, __a);
166+
__b = __builtin_copysign(__builtin_isinf(__b) ? 1.0 : 0.0, __b);
212167
__real__(z) = __builtin_huge_val() * (__a * __c + __b * __d);
213168
__imag__(z) = __builtin_huge_val() * (__b * __c - __a * __d);
214-
} else if (_ISINFd(__logbw) && __logbw > 0.0 && _ISFINITEd(__a) &&
215-
_ISFINITEd(__b)) {
216-
__c = _COPYSIGNd(_ISINFd(__c) ? 1.0 : 0.0, __c);
217-
__d = _COPYSIGNd(_ISINFd(__d) ? 1.0 : 0.0, __d);
169+
} else if (__builtin_isinf(__logbw) && __logbw > 0.0 &&
170+
__builtin_isfinite(__a) && __builtin_isfinite(__b)) {
171+
__c = __builtin_copysign(__builtin_isinf(__c) ? 1.0 : 0.0, __c);
172+
__d = __builtin_copysign(__builtin_isinf(__d) ? 1.0 : 0.0, __d);
218173
__real__(z) = 0.0 * (__a * __c + __b * __d);
219174
__imag__(z) = 0.0 * (__b * __c - __a * __d);
220175
}
@@ -224,8 +179,9 @@ __DEVICE__ double _Complex __divdc3(double __a, double __b, double __c,
224179

225180
__DEVICE__ float _Complex __divsc3(float __a, float __b, float __c, float __d) {
226181
int __ilogbw = 0;
227-
float __logbw = _LOGBf(_fmaxf(_ABSf(__c), _ABSf(__d)));
228-
if (_ISFINITEf(__logbw)) {
182+
float __logbw =
183+
_LOGBf(__builtin_fmaxf(__builtin_fabsf(__c), __builtin_fabsf(__d)));
184+
if (__builtin_isfinite(__logbw)) {
229185
__ilogbw = (int)__logbw;
230186
__c = _SCALBNf(__c, -__ilogbw);
231187
__d = _SCALBNf(__d, -__ilogbw);
@@ -234,20 +190,20 @@ __DEVICE__ float _Complex __divsc3(float __a, float __b, float __c, float __d) {
234190
float _Complex z;
235191
__real__(z) = _SCALBNf((__a * __c + __b * __d) / __denom, -__ilogbw);
236192
__imag__(z) = _SCALBNf((__b * __c - __a * __d) / __denom, -__ilogbw);
237-
if (_ISNANf(__real__(z)) && _ISNANf(__imag__(z))) {
238-
if ((__denom == 0) && (!_ISNANf(__a) || !_ISNANf(__b))) {
239-
__real__(z) = _COPYSIGNf(__builtin_huge_valf(), __c) * __a;
240-
__imag__(z) = _COPYSIGNf(__builtin_huge_valf(), __c) * __b;
241-
} else if ((_ISINFf(__a) || _ISINFf(__b)) && _ISFINITEf(__c) &&
242-
_ISFINITEf(__d)) {
243-
__a = _COPYSIGNf(_ISINFf(__a) ? 1 : 0, __a);
244-
__b = _COPYSIGNf(_ISINFf(__b) ? 1 : 0, __b);
193+
if (__builtin_isnan(__real__(z)) && __builtin_isnan(__imag__(z))) {
194+
if ((__denom == 0) && (!__builtin_isnan(__a) || !__builtin_isnan(__b))) {
195+
__real__(z) = __builtin_copysignf(__builtin_huge_valf(), __c) * __a;
196+
__imag__(z) = __builtin_copysignf(__builtin_huge_valf(), __c) * __b;
197+
} else if ((__builtin_isinf(__a) || __builtin_isinf(__b)) &&
198+
__builtin_isfinite(__c) && __builtin_isfinite(__d)) {
199+
__a = __builtin_copysignf(__builtin_isinf(__a) ? 1 : 0, __a);
200+
__b = __builtin_copysignf(__builtin_isinf(__b) ? 1 : 0, __b);
245201
__real__(z) = __builtin_huge_valf() * (__a * __c + __b * __d);
246202
__imag__(z) = __builtin_huge_valf() * (__b * __c - __a * __d);
247-
} else if (_ISINFf(__logbw) && __logbw > 0 && _ISFINITEf(__a) &&
248-
_ISFINITEf(__b)) {
249-
__c = _COPYSIGNf(_ISINFf(__c) ? 1 : 0, __c);
250-
__d = _COPYSIGNf(_ISINFf(__d) ? 1 : 0, __d);
203+
} else if (__builtin_isinf(__logbw) && __logbw > 0 &&
204+
__builtin_isfinite(__a) && __builtin_isfinite(__b)) {
205+
__c = __builtin_copysignf(__builtin_isinf(__c) ? 1 : 0, __c);
206+
__d = __builtin_copysignf(__builtin_isinf(__d) ? 1 : 0, __d);
251207
__real__(z) = 0 * (__a * __c + __b * __d);
252208
__imag__(z) = 0 * (__b * __c - __a * __d);
253209
}
@@ -259,22 +215,10 @@ __DEVICE__ float _Complex __divsc3(float __a, float __b, float __c, float __d) {
259215
} // extern "C"
260216
#endif
261217

262-
#undef _ISNANd
263-
#undef _ISNANf
264-
#undef _ISINFd
265-
#undef _ISINFf
266-
#undef _COPYSIGNd
267-
#undef _COPYSIGNf
268-
#undef _ISFINITEd
269-
#undef _ISFINITEf
270218
#undef _SCALBNd
271219
#undef _SCALBNf
272-
#undef _ABSd
273-
#undef _ABSf
274220
#undef _LOGBd
275221
#undef _LOGBf
276-
#undef _fmaxd
277-
#undef _fmaxf
278222

279223
#if defined(__OPENMP_NVPTX__) || defined(__OPENMP_AMDGCN__)
280224
#pragma omp end declare target

‎clang/test/Headers/amdgcn-openmp-device-math-complex.c‎

Lines changed: 18 additions & 18 deletions
Original file line numberDiff line numberDiff line change
@@ -14,17 +14,17 @@ void test_complex_f64(double _Complex a) {
1414
}
1515

1616
// CHECK: define weak {{.*}} @__divdc3
17-
// CHECK-DAG: call double @__ocml_fabs_f64(
18-
// CHECK-DAG: call i32 @__ocml_isnan_f64(
19-
// CHECK-DAG: call i32 @__ocml_isfinite_f64(
20-
// CHECK-DAG: call double @__ocml_copysign_f64(
21-
// CHECK-DAG: call double @__ocml_scalbn_f64(
22-
// CHECK-DAG: call double @__ocml_logb_f64(
17+
// CHECK-DAG: call double @llvm.fabs.f64(
18+
// CHECK-DAG: call i1 @llvm.is.fpclass.f64(double %{{.+}}, i32 3)
19+
// CHECK-DAG: call i1 @llvm.is.fpclass.f64(double %{{.+}}, i32 504)
20+
// CHECK-DAG: call double @llvm.copysign.f64(
21+
// CHECK-DAG: call double @llvm.ldexp.f64.i32(
22+
// CHECK-DAG: call { double, i32 } @llvm.frexp.f64.i32
2323

2424
// CHECK: define weak {{.*}} @__muldc3
25-
// CHECK-DAG: call i32 @__ocml_isnan_f64(
26-
// CHECK-DAG: call i32 @__ocml_isinf_f64(
27-
// CHECK-DAG: call double @__ocml_copysign_f64(
25+
// CHECK-DAG: call i1 @llvm.is.fpclass.f64(double %{{.+}}, i32 3)
26+
// CHECK-DAG: call i1 @llvm.is.fpclass.f64(double %{{.+}}, i32 516)
27+
// CHECK-DAG: call double @llvm.copysign.f64(
2828

2929
void test_complex_f32(float _Complex a) {
3030
// CHECK-LABEL: define {{.*}}test_complex_f32
@@ -37,14 +37,14 @@ void test_complex_f32(float _Complex a) {
3737
}
3838

3939
// CHECK: define weak {{.*}} @__divsc3
40-
// CHECK-DAG: call float @__ocml_fabs_f32(
41-
// CHECK-DAG: call i32 @__ocml_isnan_f32(
42-
// CHECK-DAG: call i32 @__ocml_isfinite_f32(
43-
// CHECK-DAG: call float @__ocml_copysign_f32(
44-
// CHECK-DAG: call float @__ocml_scalbn_f32(
45-
// CHECK-DAG: call float @__ocml_logb_f32(
40+
// CHECK-DAG: call float @llvm.fabs.f32(
41+
// CHECK-DAG: call i1 @llvm.is.fpclass.f32(float %{{.+}}, i32 3)
42+
// CHECK-DAG: call i1 @llvm.is.fpclass.f32(float %{{.+}}, i32 516)
43+
// CHECK-DAG: call float @llvm.copysign.f32(
44+
// CHECK-DAG: call float @llvm.ldexp.f32.i32(
45+
// CHECK-DAG: call { float, i32 } @llvm.frexp.f32.i32
4646

4747
// CHECK: define weak {{.*}} @__mulsc3
48-
// CHECK-DAG: call i32 @__ocml_isnan_f32(
49-
// CHECK-DAG: call i32 @__ocml_isinf_f32(
50-
// CHECK-DAG: call float @__ocml_copysign_f32(
48+
// CHECK-DAG: call i1 @llvm.is.fpclass.f32(float %{{.+}}, i32 3)
49+
// CHECK-DAG: call i1 @llvm.is.fpclass.f32(float %{{.+}}, i32 516)
50+
// CHECK-DAG: call float @llvm.copysign.f32(

0 commit comments

Comments
 (0)