mindspore-Ecosystem
/
mindspore
21
0
Fork 27
Code Issues Pull Requests 6 Projects Releases Wiki Activity

!33283 [MSLITE] vs build bug 

Merge pull request !33283 from ling/op
This commit is contained in:
i-robot 2022-04-19 11:27:36 +00:00 committed by Gitee
parent f735be0e6b 3fa86aa040
commit a7a9486d32
No known key found for this signature in database
GPG Key ID: 173E9B9CA92EEF8F
2 changed files with 9 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
mindspore/ccsrc/plugin/device/cpu/kernel/nnacl/experimental/conv_fp32_nchwx_avx512.c
View File

@ -21,7 +21,7 @@
static const int UNIT_LEN = 512; // register length
static const int UNIT_NR = 512 / sizeof(float);
#define UNIT_NR 128 // (512 / sizeof(float))
static const int TILE_ROW;
int conv2d_prepare_fp32_nchwx_avx512(struct KernelBase *self) {
KConv2d *conv = (KConv2d *)self;
@ -36,7 +36,7 @@ int conv2d_prepare_fp32_nchwx_avx512(struct KernelBase *self) {
int kw = weight->shape_[kNCHW_W];
size_t lineLen = cin * kw * kh;
conv->packedWeight = malloc(lineLen * UP_ROUND_DIV(cout, UNIT_NR) * sizeof(float)); // allocate packed weight buf
conv->packedWeight = malloc(lineLen * UP_DIV(cout, UNIT_NR) * sizeof(float)); // allocate packed weight buf
float *rpos[16] = {0};
float *data = (float *)weight->data_;
@ -154,12 +154,10 @@ int conv2d_compute_fp32_nchwx_avx512(struct KernelBase *self) {
#ifdef VECTORIZE_OPTIMIZE
// use AVX2 instruction to optimize gemm
#else
int InputRegNr = 16;
int WeightRegNr = 16;
int OutputRegNr = 16;
float intputReg[InputRegNr][UNIT_NR];
float weightReg[WeightRegNr][UNIT_NR];
float outputReg[OutputRegNr][UNIT_NR];
float intputReg[C16NUM][UNIT_NR];
float weightReg[C16NUM][UNIT_NR];
float outputReg[C16NUM][UNIT_NR];
memset(outputReg, 0, sizeof(outputReg));
int lpos = 0;
int rpos = 0;
@ -192,7 +190,7 @@ int conv2d_compute_fp32_nchwx_avx512(struct KernelBase *self) {
tilePos += n;
}
// flush outputReg to output tensor memory
memcpy(out->data_ + outOffset, outputReg, m * k * sizeof(float));
memcpy((float *)out->data_ + outOffset, outputReg, m * k * sizeof(float));
outOffset += m * k * sizeof(float);
y += m;
}
@ -201,6 +199,7 @@ int conv2d_compute_fp32_nchwx_avx512(struct KernelBase *self) {
#endif
return 0;
}
int conv2d_infershape_fp32_nchwx_avx512(struct KernelBase *self) {
return Conv2dInferShape((const struct TensorC *const *)self->in, self->insize, self->out, self->outsize, self->param);
}
@ -220,7 +219,7 @@ int conv2d_resize_fp32_nchwx_avx512(struct KernelBase *self, TensorC *inputs[],
self->inferShape(self);
out->format_ = Format_NC16HW16;
out->shape_[1] = UP_ROUND_DIV(out->shape_[1], C16NUM);
out->shape_[1] = UP_DIV(out->shape_[1], C16NUM);
out->shape_[4] = 16;
out->shape_size_ = 5;

1
mindspore/ccsrc/plugin/device/cpu/kernel/nnacl/op_base.h
View File

@ -57,7 +57,6 @@
#define UP_DIV(x, y) (((x) + (y) - (1)) / (y))
#define UP_ROUND(x, y) (((x) + (y) - (1)) / (y) * (y))
#define UP_ROUND_DIV(x, y) (x % y == 0 ? (x / y) : (x / y) + 1)
#define DOWN_DIV(x, y) ((x) / (y))
#define DOWN_ROUND(x, y) ((x) / (y) * (y))