Skip to content

Commit 1393aa4

Browse files
Rob Hessjeffdonahue
Rob Hess
authored and
jeffdonahue
committed
Fix compilation errors.
1 parent 6db9c8c commit 1393aa4

File tree

3 files changed

+39
-100
lines changed

3 files changed

+39
-100
lines changed

‎include/caffe/device.hpp‎

Lines changed: 8 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -81,8 +81,8 @@ class Device {
8181
virtual void dot(const int N, const Dtype* x, const Dtype* y, Dtype* out) {
8282
NOT_IMPLEMENTED; }
8383

84-
virtual void hamming_distance(const int N, const Dtype* x, const Dtype* y,
85-
uint32_t* out) { NOT_IMPLEMENTED; }
84+
virtual uint32_t hamming_distance(const int N, const Dtype* x,
85+
const Dtype* y) { NOT_IMPLEMENTED; return 0; }
8686

8787
// Returns the sum of the absolute values of the elements of vector x
8888
virtual void asum(const int N, const Dtype* x, Dtype* y) { NOT_IMPLEMENTED; }
@@ -102,7 +102,7 @@ class Device {
102102
const int stride, Dtype* data_col) { NOT_IMPLEMENTED; }
103103

104104
virtual void col2im(const Dtype* data_col, const int channels,
105-
const int height, const int width, const int psize, const int pad,
105+
const int height, const int width, const int ksize, const int pad,
106106
const int stride, Dtype* data_im) { NOT_IMPLEMENTED; }
107107
};
108108

@@ -160,8 +160,8 @@ class CPUDevice : public Device<Dtype> {
160160

161161
virtual void dot(const int N, const Dtype* x, const Dtype* y, Dtype* out);
162162

163-
virtual void hamming_distance(const int N, const Dtype* x, const Dtype* y,
164-
uint32_t* out);
163+
virtual uint32_t hamming_distance(const int N, const Dtype* x,
164+
const Dtype* y);
165165

166166
// Returns the sum of the absolute values of the elements of vector x
167167
virtual void asum(const int N, const Dtype* x, Dtype* y);
@@ -179,7 +179,7 @@ class CPUDevice : public Device<Dtype> {
179179
const int stride, Dtype* data_col);
180180

181181
virtual void col2im(const Dtype* data_col, const int channels,
182-
const int height, const int width, const int psize, const int pad,
182+
const int height, const int width, const int ksize, const int pad,
183183
const int stride, Dtype* data_im);
184184
};
185185

@@ -235,8 +235,8 @@ class GPUDevice : public Device<Dtype> {
235235

236236
virtual void dot(const int N, const Dtype* x, const Dtype* y, Dtype* out);
237237

238-
virtual void hamming_distance(const int N, const Dtype* x, const Dtype* y,
239-
uint32_t* out);
238+
virtual uint32_t hamming_distance(const int N, const Dtype* x,
239+
const Dtype* y);
240240

241241
// Returns the sum of the absolute values of the elements of vector x
242242
virtual void asum(const int N, const Dtype* x, Dtype* y);

‎src/caffe/devices/cpu_device.cpp‎

Lines changed: 28 additions & 89 deletions
Original file line numberDiff line numberDiff line change
@@ -93,39 +93,38 @@ void CPUDevice<double>::copy(const int N, const double *X, double *Y) {
9393

9494
template<typename Dtype>
9595
void CPUDevice<Dtype>::copy_from_cpu(const int N, const Dtype *X, Dtype *Y) {
96-
copy<Dtype>(N, X, Y);
96+
this->copy(N, X, Y);
9797
}
9898

9999
template<typename Dtype>
100100
void CPUDevice<Dtype>::set(const int N, const Dtype alpha, Dtype *X) {
101101
if (alpha == 0) {
102-
memset(Y, 0, sizeof(Dtype) * N);
102+
memset(X, 0, sizeof(Dtype) * N);
103103
return;
104104
}
105105
for (int i = 0; i < N; ++i) {
106-
Y[i] = alpha;
106+
X[i] = alpha;
107107
}
108108
}
109109

110110
template<>
111-
void CPUDevice<int>::set(const int N, const int alpha, int *X);
112-
template<>
113-
void CPUDevice<float>::set(const int N, const float alpha, float *X);
114-
template<>
115-
void CPUDevice<double>::set(const int N, const double alpha, double *X);
111+
void CPUDevice<int>::set(const int N, const int alpha, int *X) {
112+
if (alpha == 0) {
113+
memset(X, 0, sizeof(int) * N);
114+
return;
115+
}
116+
for (int i = 0; i < N; ++i) {
117+
X[i] = alpha;
118+
}
119+
}
116120

117121
template<typename Dtype>
118122
void CPUDevice<Dtype>::add_scalar(const int N, const Dtype alpha, Dtype *X) {
119123
for (int i = 0; i < N; ++i) {
120-
Y[i] += alpha;
124+
X[i] += alpha;
121125
}
122126
}
123127

124-
template<>
125-
void CPUDevice<float>::add_scalar(const int N, const float alpha, float *X);
126-
template<>
127-
void CPUDevice<double>::add_scalar(const int N, const double alpha, double *X);
128-
129128
template<>
130129
void CPUDevice<float>::scal(const int N, const float alpha, float *X) {
131130
cblas_sscal(N, alpha, X, 1);
@@ -138,12 +137,12 @@ void CPUDevice<double>::scal(const int N, const double alpha, double *X) {
138137

139138
template<>
140139
void CPUDevice<float>::sqr(const int N, const float* a, float* y) {
141-
vsSqr(n, a, y);
140+
vsSqr(N, a, y);
142141
}
143142

144143
template<>
145144
void CPUDevice<double>::sqr(const int N, const double* a, double* y) {
146-
vdSqr(n, a, y);
145+
vdSqr(N, a, y);
147146
}
148147

149148
template<>
@@ -195,13 +194,13 @@ void CPUDevice<double>::div(const int N, const double* a, const double* b,
195194
}
196195

197196
template<>
198-
void CPUDevice<float>::powx(const int N, const float* a, const float* b,
197+
void CPUDevice<float>::powx(const int N, const float* a, const float b,
199198
float* y) {
200199
vsPowx(N, a, b, y);
201200
}
202201

203202
template<>
204-
void CPUDevice<double>::powx(const int N, const double* a, const double* b,
203+
void CPUDevice<double>::powx(const int N, const double* a, const double b,
205204
double* y) {
206205
vdPowx(N, a, b, y);
207206
}
@@ -212,9 +211,6 @@ static Dtype _nextafter(const Dtype b) {
212211
b, std::numeric_limits<Dtype>::max());
213212
}
214213

215-
template static float _nextafter(const float b);
216-
template static double _nextafter(const double b);
217-
218214
template<typename Dtype>
219215
void CPUDevice<Dtype>::rng_uniform(const int N, const Dtype a,
220216
const Dtype b, Dtype* r) {
@@ -229,13 +225,6 @@ void CPUDevice<Dtype>::rng_uniform(const int N, const Dtype a,
229225
}
230226
}
231227

232-
template<>
233-
void CPUDevice<float>::rng_uniform(const int N, const float a, const float b,
234-
float* r);
235-
template<>
236-
void CPUDevice<double>::rng_uniform(const int N, const double a, const double b,
237-
double* r);
238-
239228
template<typename Dtype>
240229
void CPUDevice<Dtype>::rng_gaussian(const int N, const Dtype mu,
241230
const Dtype sigma, Dtype* r) {
@@ -250,13 +239,6 @@ void CPUDevice<Dtype>::rng_gaussian(const int N, const Dtype mu,
250239
}
251240
}
252241

253-
template<>
254-
void CPUDevice<float>::rng_gaussian(const int N, const float mu,
255-
const float sigma, float* r);
256-
template<>
257-
void CPUDevice<double>::rng_gaussian(const int N, const double mu,
258-
const double sigma, double* r);
259-
260242
template<typename Dtype>
261243
void CPUDevice<Dtype>::rng_bernoulli(const int N, const Dtype p, int* r) {
262244
CHECK_GE(N, 0);
@@ -271,11 +253,6 @@ void CPUDevice<Dtype>::rng_bernoulli(const int N, const Dtype p, int* r) {
271253
}
272254
}
273255

274-
template<>
275-
void CPUDevice<float>::rng_bernoulli(const int N, const float p, int* r);
276-
template<>
277-
void CPUDevice<double>::rng_bernoulli(const int N, const double p, int* r);
278-
279256
template<typename Dtype>
280257
void CPUDevice<Dtype>::rng_bernoulli(const int N, const Dtype p,
281258
unsigned int* r) {
@@ -291,13 +268,6 @@ void CPUDevice<Dtype>::rng_bernoulli(const int N, const Dtype p,
291268
}
292269
}
293270

294-
template<>
295-
void CPUDevice<float>::rng_bernoulli(const int N, const float p,
296-
unsigned int* r);
297-
template<>
298-
void CPUDevice<double>::rng_bernoulli(const int N, const double p,
299-
unsigned int* r);
300-
301271
template<>
302272
void CPUDevice<float>::exp(const int N, const float* a, float* y) {
303273
vsExp(N, a, y);
@@ -321,25 +291,25 @@ void CPUDevice<double>::dot(const int N, const double* x, const double* y,
321291
}
322292

323293
template<>
324-
void CPUDevice<float>::hamming_distance(const int N, const float* x,
325-
const float* y, uint32_t* out) {
326-
int dist = 0;
327-
for (int i = 0; i < n; ++i) {
294+
uint32_t CPUDevice<float>::hamming_distance(const int N, const float* x,
295+
const float* y) {
296+
uint32_t dist = 0;
297+
for (int i = 0; i < N; ++i) {
328298
dist += __builtin_popcount(static_cast<uint32_t>(x[i]) ^
329299
static_cast<uint32_t>(y[i]));
330300
}
331-
*out = dist;
301+
return dist;
332302
}
333303

334304
template<>
335-
void CPUDevice<double>::hamming_distance(const int N, const double* x,
336-
const double* y, uint64_t* out) {
337-
int dist = 0;
338-
for (int i = 0; i < n; ++i) {
305+
uint32_t CPUDevice<double>::hamming_distance(const int N, const double* x,
306+
const double* y) {
307+
uint32_t dist = 0;
308+
for (int i = 0; i < N; ++i) {
339309
dist += __builtin_popcount(static_cast<uint64_t>(x[i]) ^
340310
static_cast<uint64_t>(y[i]));
341311
}
342-
*out = dist;
312+
return dist;
343313
}
344314

345315
template<>
@@ -371,11 +341,6 @@ void CPUDevice<Dtype>::sign(const int N, const Dtype* x, Dtype* y) {
371341
}
372342
}
373343

374-
template<>
375-
void CPUDevice<float>::sign(const int N, const float* x, float* y);
376-
template<>
377-
void CPUDevice<double>::sign(const int N, const double* x, double* y);
378-
379344
// This returns a nonzero value if the input has its sign bit set.
380345
// The name sngbit is meant to avoid conflicts with std::signbit in the macro
381346
template<typename Dtype>
@@ -388,11 +353,6 @@ void CPUDevice<Dtype>::sgnbit(const int N, const Dtype* x, Dtype* y) {
388353
}
389354
}
390355

391-
template<>
392-
void CPUDevice<float>::sgnbit(const int N, const float* x, float* y);
393-
template<>
394-
void CPUDevice<double>::sgnbit(const int N, const double* x, double* y);
395-
396356
template<typename Dtype>
397357
void CPUDevice<Dtype>::fabs(const int N, const Dtype* x, Dtype* y) {
398358
CHECK_GT(N, 0);
@@ -403,11 +363,6 @@ void CPUDevice<Dtype>::fabs(const int N, const Dtype* x, Dtype* y) {
403363
}
404364
}
405365

406-
template<>
407-
void CPUDevice<float>::fabs(const int N, const float* x, float* y);
408-
template<>
409-
void CPUDevice<double>::fabs(const int N, const double* x, double* y);
410-
411366
template<>
412367
void CPUDevice<float>::scale(const int N, const float alpha, const float *x,
413368
float* y) {
@@ -447,18 +402,9 @@ void CPUDevice<Dtype>::im2col(const Dtype* data_im, const int channels,
447402
}
448403
}
449404

450-
template<>
451-
void CPUDevice<float>::im2col(const float* data_im, const int channels,
452-
const int height, const int width, const int ksize, const int pad,
453-
const int stride, float* data_col);
454-
template<>
455-
void CPUDevice<double>::im2col(const double* data_im, const int channels,
456-
const int height, const int width, const int ksize, const int pad,
457-
const int stride, double* data_col);
458-
459405
template<typename Dtype>
460406
void CPUDevice<Dtype>::col2im(const Dtype* data_col, const int channels,
461-
const int height, const int width, const int psize, const int pad,
407+
const int height, const int width, const int ksize, const int pad,
462408
const int stride, Dtype* data_im) {
463409
memset(data_im, 0, sizeof(Dtype) * height * width * channels);
464410
int height_col = (height + 2 * pad - ksize) / stride + 1;
@@ -480,13 +426,6 @@ void CPUDevice<Dtype>::col2im(const Dtype* data_col, const int channels,
480426
}
481427
}
482428

483-
void CPUDevice<float>::col2im(const float* data_col, const int channels,
484-
const int height, const int width, const int psize, const int pad,
485-
const int stride, float* data_im);
486-
void CPUDevice<double>::col2im(const double* data_col, const int channels,
487-
const int height, const int width, const int psize, const int pad,
488-
const int stride, double* data_im);
489-
490429
INSTANTIATE_CLASS(CPUDevice);
491430

492431
} // namespace caffe

‎src/caffe/devices/gpu_device.cpp‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -125,9 +125,9 @@ void GPUDevice<Dtype>::dot(const int N, const Dtype* x, const Dtype* y,
125125
}
126126

127127
template<typename Dtype>
128-
void GPUDevice<Dtype>::hamming_distance(const int N, const Dtype* x,
129-
const Dtype* y, uint32_t* out) {
130-
*out = caffe_gpu_hamming_distance<Dtype>(N, x, y);
128+
uint32_t GPUDevice<Dtype>::hamming_distance(const int N, const Dtype* x,
129+
const Dtype* y) {
130+
return caffe_gpu_hamming_distance<Dtype>(N, x, y);
131131
}
132132

133133
template<typename Dtype>

0 commit comments

Comments
 (0)