325 lines
9.2 KiB
C++
Executable File
325 lines
9.2 KiB
C++
Executable File
#include "UnaryF32.h"
|
|
#include "Error.h"
|
|
|
|
/* Upper bound of maximum matrix dimension used by Python */
|
|
#define MAXMATRIXDIM 40
|
|
|
|
/*
|
|
|
|
Offset in input test pattern for matrix of dimension d * d.
|
|
Must be coherent with Python script Matrix.py
|
|
|
|
*/
|
|
static int cholesky_offset(int d)
|
|
{
|
|
int offset=14;
|
|
switch (d)
|
|
{
|
|
case 4:
|
|
offset = 14;
|
|
break;
|
|
case 8:
|
|
offset = 79;
|
|
break;
|
|
case 9:
|
|
offset = 143;
|
|
break;
|
|
case 15:
|
|
offset = 224;
|
|
break;
|
|
case 16:
|
|
offset = 449;
|
|
break;
|
|
default:
|
|
offset = 14;
|
|
break;
|
|
}
|
|
|
|
return(offset);
|
|
}
|
|
|
|
|
|
void UnaryF32::test_mat_scale_f32()
|
|
{
|
|
arm_mat_scale_f32(&this->in1,0.5,&this->out);
|
|
}
|
|
|
|
void UnaryF32::test_mat_inverse_f32()
|
|
{
|
|
arm_mat_inverse_f32(&this->in1,&this->out);
|
|
}
|
|
|
|
void UnaryF32::test_mat_trans_f32()
|
|
{
|
|
arm_mat_trans_f32(&this->in1,&this->out);
|
|
}
|
|
|
|
void UnaryF32::test_mat_cmplx_trans_f32()
|
|
{
|
|
arm_mat_cmplx_trans_f32(&this->in1,&this->out);
|
|
}
|
|
|
|
void UnaryF32::test_mat_add_f32()
|
|
{
|
|
arm_mat_add_f32(&this->in1,&this->in1,&this->out);
|
|
}
|
|
|
|
void UnaryF32::test_mat_sub_f32()
|
|
{
|
|
arm_mat_sub_f32(&this->in1,&this->in1,&this->out);
|
|
}
|
|
|
|
void UnaryF32::test_mat_vec_mult_f32()
|
|
{
|
|
arm_mat_vec_mult_f32(&this->in1, vecp, outp);
|
|
}
|
|
|
|
void UnaryF32::test_mat_cholesky_dpo_f32()
|
|
{
|
|
arm_mat_cholesky_f32(&this->in1,&this->out);
|
|
}
|
|
|
|
void UnaryF32::test_solve_upper_triangular_f32()
|
|
{
|
|
arm_mat_solve_upper_triangular_f32(&this->in1,&this->in2,&this->out);
|
|
}
|
|
|
|
void UnaryF32::test_solve_lower_triangular_f32()
|
|
{
|
|
arm_mat_solve_lower_triangular_f32(&this->in1,&this->in2,&this->out);
|
|
}
|
|
|
|
void UnaryF32::test_ldlt_decomposition_f32()
|
|
{
|
|
arm_mat_ldlt_f32(&this->in1,&this->outll,&this->outd,(uint16_t*)outp);
|
|
}
|
|
|
|
|
|
void UnaryF32::setUp(Testing::testID_t id,std::vector<Testing::param_t>& params,Client::PatternMgr *mgr)
|
|
{
|
|
|
|
|
|
std::vector<Testing::param_t>::iterator it = params.begin();
|
|
|
|
this->nbr = *it++;
|
|
this->nbc = *it;
|
|
|
|
|
|
switch(id)
|
|
{
|
|
case TEST_MAT_VEC_MULT_F32_6:
|
|
input1.reload(UnaryF32::INPUTA_F32_ID,mgr,this->nbr*this->nbc);
|
|
vec.reload(UnaryF32::INPUTVEC1_F32_ID,mgr,this->nbc);
|
|
output.create(this->nbr,UnaryF32::OUT_F32_ID,mgr);
|
|
vecp=vec.ptr();
|
|
outp=output.ptr();
|
|
|
|
this->in1.numRows = this->nbr;
|
|
this->in1.numCols = this->nbc;
|
|
this->in1.pData = input1.ptr();
|
|
break;
|
|
case TEST_MAT_TRANS_F32_3:
|
|
input1.reload(UnaryF32::INPUTA_F32_ID,mgr,this->nbr*this->nbc);
|
|
output.create(this->nbr*this->nbc,UnaryF32::OUT_F32_ID,mgr);
|
|
|
|
this->out.numRows = this->nbc;
|
|
this->out.numCols = this->nbr;
|
|
this->out.pData = output.ptr();
|
|
|
|
this->in1.numRows = this->nbr;
|
|
this->in1.numCols = this->nbc;
|
|
this->in1.pData = input1.ptr();
|
|
break;
|
|
case TEST_MAT_CMPLX_TRANS_F32_7:
|
|
input1.reload(UnaryF32::INPUTAC_F32_ID,mgr,2*this->nbr*this->nbc);
|
|
output.create(2*this->nbr*this->nbc,UnaryF32::OUT_F32_ID,mgr);
|
|
|
|
this->out.numRows = this->nbc;
|
|
this->out.numCols = this->nbr;
|
|
this->out.pData = output.ptr();
|
|
|
|
this->in1.numRows = this->nbr;
|
|
this->in1.numCols = this->nbc;
|
|
this->in1.pData = input1.ptr();
|
|
break;
|
|
|
|
case TEST_MAT_CHOLESKY_DPO_F32_8:
|
|
{
|
|
int offset=14;
|
|
float32_t *p;
|
|
float32_t *aPtr;
|
|
input1.reload(UnaryF32::INPUTSCHOLESKY1_DPO_F32_ID,mgr);
|
|
output.create(this->nbc * this->nbr,UnaryF32::OUT_F32_ID,mgr);
|
|
|
|
a.create(this->nbr*this->nbc,UnaryF32::TMPA_F32_ID,mgr);
|
|
|
|
/* Offsets must be coherent with the sizes used in python script
|
|
Matrix.py for pattern generation */
|
|
offset=cholesky_offset(this->nbr);
|
|
|
|
p = input1.ptr();
|
|
aPtr = a.ptr();
|
|
|
|
|
|
|
|
memcpy(aPtr,p + offset,sizeof(float32_t)*this->nbr*this->nbr);
|
|
|
|
this->out.numRows = this->nbr;
|
|
this->out.numCols = this->nbc;
|
|
this->out.pData = output.ptr();
|
|
|
|
this->in1.numRows = this->nbr;
|
|
this->in1.numCols = this->nbc;
|
|
this->in1.pData = aPtr;
|
|
|
|
|
|
|
|
}
|
|
break;
|
|
|
|
case TEST_SOLVE_UPPER_TRIANGULAR_F32_9:
|
|
{
|
|
int offset=14;
|
|
float32_t *p;
|
|
float32_t *aPtr;
|
|
float32_t *bPtr;
|
|
|
|
input1.reload(UnaryF32::INPUT_UT_DPO_F32_ID,mgr);
|
|
input2.reload(UnaryF32::INPUT_RNDA_DPO_F32_ID,mgr);
|
|
output.create(this->nbc * this->nbr,UnaryF32::OUT_F32_ID,mgr);
|
|
|
|
a.create(this->nbr*this->nbc,UnaryF32::TMPA_F32_ID,mgr);
|
|
b.create(this->nbr*this->nbc,UnaryF32::TMPB_F32_ID,mgr);
|
|
|
|
/* Offsets must be coherent with the sizes used in python script
|
|
Matrix.py for pattern generation */
|
|
offset=cholesky_offset(this->nbr);
|
|
|
|
p = input1.ptr();
|
|
aPtr = a.ptr();
|
|
memcpy(aPtr,&p[offset],sizeof(float32_t)*this->nbr*this->nbr);
|
|
|
|
p = input2.ptr();
|
|
bPtr = b.ptr();
|
|
memcpy(bPtr,&p[offset],sizeof(float32_t)*this->nbr*this->nbr);
|
|
|
|
this->out.numRows = this->nbr;
|
|
this->out.numCols = this->nbc;
|
|
this->out.pData = output.ptr();
|
|
|
|
this->in1.numRows = this->nbr;
|
|
this->in1.numCols = this->nbc;
|
|
this->in1.pData = aPtr;
|
|
|
|
this->in2.numRows = this->nbr;
|
|
this->in2.numCols = this->nbc;
|
|
this->in2.pData = bPtr;
|
|
}
|
|
break;
|
|
|
|
case TEST_SOLVE_LOWER_TRIANGULAR_F32_10:
|
|
{
|
|
int offset=14;
|
|
float32_t *p;
|
|
float32_t *aPtr;
|
|
float32_t *bPtr;
|
|
|
|
input1.reload(UnaryF32::INPUT_LT_DPO_F32_ID,mgr);
|
|
input2.reload(UnaryF32::INPUT_RNDA_DPO_F32_ID,mgr);
|
|
output.create(this->nbc * this->nbr,UnaryF32::OUT_F32_ID,mgr);
|
|
|
|
a.create(this->nbr*this->nbc,UnaryF32::TMPA_F32_ID,mgr);
|
|
b.create(this->nbr*this->nbc,UnaryF32::TMPB_F32_ID,mgr);
|
|
|
|
/* Offsets must be coherent with the sizes used in python script
|
|
Matrix.py for pattern generation */
|
|
offset=cholesky_offset(this->nbr);
|
|
|
|
p = input1.ptr();
|
|
aPtr = a.ptr();
|
|
memcpy(aPtr,&p[offset],sizeof(float32_t)*this->nbr*this->nbr);
|
|
|
|
p = input2.ptr();
|
|
bPtr = b.ptr();
|
|
memcpy(bPtr,&p[offset],sizeof(float32_t)*this->nbr*this->nbr);
|
|
|
|
this->out.numRows = this->nbr;
|
|
this->out.numCols = this->nbc;
|
|
this->out.pData = output.ptr();
|
|
|
|
this->in1.numRows = this->nbr;
|
|
this->in1.numCols = this->nbc;
|
|
this->in1.pData = aPtr;
|
|
|
|
this->in2.numRows = this->nbr;
|
|
this->in2.numCols = this->nbc;
|
|
this->in2.pData = bPtr;
|
|
}
|
|
break;
|
|
|
|
case TEST_LDLT_DECOMPOSITION_F32_11:
|
|
{
|
|
float32_t *p, *aPtr;
|
|
|
|
int offset=14;
|
|
input1.reload(UnaryF32::INPUTSCHOLESKY1_DPO_F32_ID,mgr);
|
|
|
|
|
|
outputll.create(this->nbr*this->nbr,UnaryF32::LL_F32_ID,mgr);
|
|
outputd.create(this->nbr*this->nbr,UnaryF32::D_F32_ID,mgr);
|
|
outputp.create(this->nbr,UnaryF32::PERM_S16_ID,mgr);
|
|
|
|
|
|
a.create(MAXMATRIXDIM*MAXMATRIXDIM,UnaryF32::TMPA_F32_ID,mgr);
|
|
|
|
/* Offsets must be coherent with the sizes used in python script
|
|
Matrix.py for pattern generation */
|
|
offset=cholesky_offset(this->nbr);
|
|
|
|
p = input1.ptr();
|
|
aPtr = a.ptr();
|
|
memcpy(aPtr,&p[offset],sizeof(float32_t)*this->nbr*this->nbr);
|
|
|
|
this->in1.numRows = this->nbr;
|
|
this->in1.numCols = this->nbc;
|
|
this->in1.pData = aPtr;
|
|
|
|
this->outll.numRows = this->nbr;
|
|
this->outll.numCols = this->nbc;
|
|
this->outll.pData = outputll.ptr();
|
|
|
|
this->outd.numRows = this->nbr;
|
|
this->outd.numCols = this->nbc;
|
|
this->outd.pData = outputd.ptr();
|
|
|
|
|
|
outpp = outputp.ptr();
|
|
}
|
|
break;
|
|
|
|
default:
|
|
input1.reload(UnaryF32::INPUTA_F32_ID,mgr,this->nbr*this->nbc);
|
|
output.create(this->nbr*this->nbc,UnaryF32::OUT_F32_ID,mgr);
|
|
|
|
this->out.numRows = this->nbr;
|
|
this->out.numCols = this->nbc;
|
|
this->out.pData = output.ptr();
|
|
|
|
this->in1.numRows = this->nbr;
|
|
this->in1.numCols = this->nbc;
|
|
this->in1.pData = input1.ptr();
|
|
break;
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
}
|
|
|
|
void UnaryF32::tearDown(Testing::testID_t id,Client::PatternMgr *mgr)
|
|
{
|
|
}
|