#include "UnaryTestsF32.h" #include "Error.h" #define SNR_THRESHOLD 120 /* Reference patterns are generated with a double precision computation. */ #define REL_ERROR (1.0e-6) #define ABS_ERROR (1.0e-5) /* Comparisons for inverse */ /* Not very accurate for big matrix. But big matrix needed for checking the vectorized code */ #define SNR_THRESHOLD_INV 70 #define REL_ERROR_INV (1.0e-3) #define ABS_ERROR_INV (1.0e-3) /* Upper bound of maximum matrix dimension used by Python */ #define MAXMATRIXDIM 40 #define LOADDATA2() \ const float32_t *inp1=input1.ptr(); \ const float32_t *inp2=input2.ptr(); \ \ float32_t *ap=a.ptr(); \ float32_t *bp=b.ptr(); \ \ float32_t *outp=output.ptr(); \ int16_t *dimsp = dims.ptr(); \ int nbMatrixes = dims.nbSamples() >> 1;\ int rows,columns; \ int i; #define LOADDATA1() \ const float32_t *inp1=input1.ptr(); \ \ float32_t *ap=a.ptr(); \ \ float32_t *outp=output.ptr(); \ int16_t *dimsp = dims.ptr(); \ int nbMatrixes = dims.nbSamples() >> 1;\ int rows,columns; \ int i; #define PREPAREDATA2() \ in1.numRows=rows; \ in1.numCols=columns; \ memcpy((void*)ap,(const void*)inp1,sizeof(float32_t)*rows*columns);\ in1.pData = ap; \ \ in2.numRows=rows; \ in2.numCols=columns; \ memcpy((void*)bp,(const void*)inp2,sizeof(float32_t)*rows*columns);\ in2.pData = bp; \ \ out.numRows=rows; \ out.numCols=columns; \ out.pData = outp; #define PREPAREDATA1(TRANSPOSED) \ in1.numRows=rows; \ in1.numCols=columns; \ memcpy((void*)ap,(const void*)inp1,sizeof(float32_t)*rows*columns);\ in1.pData = ap; \ \ if (TRANSPOSED) \ { \ out.numRows=columns; \ out.numCols=rows; \ } \ else \ { \ out.numRows=rows; \ out.numCols=columns; \ } \ out.pData = outp; void UnaryTestsF32::test_mat_add_f32() { LOADDATA2(); for(i=0;i < nbMatrixes ; i ++) { rows = *dimsp++; columns = *dimsp++; PREPAREDATA2(); arm_mat_add_f32(&this->in1,&this->in2,&this->out); outp += (rows * columns); } ASSERT_EMPTY_TAIL(output); ASSERT_SNR(output,ref,(float32_t)SNR_THRESHOLD); ASSERT_CLOSE_ERROR(output,ref,ABS_ERROR,REL_ERROR); } void UnaryTestsF32::test_mat_sub_f32() { LOADDATA2(); for(i=0;i < nbMatrixes ; i ++) { rows = *dimsp++; columns = *dimsp++; PREPAREDATA2(); arm_mat_sub_f32(&this->in1,&this->in2,&this->out); outp += (rows * columns); } ASSERT_EMPTY_TAIL(output); ASSERT_SNR(output,ref,(float32_t)SNR_THRESHOLD); ASSERT_CLOSE_ERROR(output,ref,ABS_ERROR,REL_ERROR); } void UnaryTestsF32::test_mat_scale_f32() { LOADDATA1(); for(i=0;i < nbMatrixes ; i ++) { rows = *dimsp++; columns = *dimsp++; PREPAREDATA1(false); arm_mat_scale_f32(&this->in1,0.5f,&this->out); outp += (rows * columns); } ASSERT_EMPTY_TAIL(output); ASSERT_SNR(output,ref,(float32_t)SNR_THRESHOLD); ASSERT_CLOSE_ERROR(output,ref,ABS_ERROR,REL_ERROR); } void UnaryTestsF32::test_mat_trans_f32() { LOADDATA1(); for(i=0;i < nbMatrixes ; i ++) { rows = *dimsp++; columns = *dimsp++; PREPAREDATA1(true); arm_mat_trans_f32(&this->in1,&this->out); outp += (rows * columns); } ASSERT_EMPTY_TAIL(output); ASSERT_SNR(output,ref,(float32_t)SNR_THRESHOLD); ASSERT_CLOSE_ERROR(output,ref,ABS_ERROR,REL_ERROR); } void UnaryTestsF32::test_mat_inverse_f32() { const float32_t *inp1=input1.ptr(); float32_t *ap=a.ptr(); float32_t *outp=output.ptr(); int16_t *dimsp = dims.ptr(); int nbMatrixes = dims.nbSamples(); int rows,columns; int i; arm_status status; for(i=0;i < nbMatrixes ; i ++) { rows = *dimsp++; columns = rows; PREPAREDATA1(false); status=arm_mat_inverse_f32(&this->in1,&this->out); //ASSERT_TRUE(status==ARM_MATH_SUCCESS); outp += (rows * columns); inp1 += (rows * columns); } ASSERT_EMPTY_TAIL(output); //ASSERT_SNR(output,ref,(float32_t)SNR_THRESHOLD_INV); ASSERT_CLOSE_ERROR(output,ref,ABS_ERROR_INV,REL_ERROR_INV); } void UnaryTestsF32::setUp(Testing::testID_t id,std::vector& params,Client::PatternMgr *mgr) { switch(id) { case TEST_MAT_ADD_F32_1: input1.reload(UnaryTestsF32::INPUTS1_F32_ID,mgr); input2.reload(UnaryTestsF32::INPUTS2_F32_ID,mgr); dims.reload(UnaryTestsF32::DIMSUNARY1_S16_ID,mgr); ref.reload(UnaryTestsF32::REFADD1_F32_ID,mgr); output.create(ref.nbSamples(),UnaryTestsF32::OUT_F32_ID,mgr); a.create(MAXMATRIXDIM*MAXMATRIXDIM,UnaryTestsF32::TMPA_F32_ID,mgr); b.create(MAXMATRIXDIM*MAXMATRIXDIM,UnaryTestsF32::TMPB_F32_ID,mgr); break; case TEST_MAT_SUB_F32_2: input1.reload(UnaryTestsF32::INPUTS1_F32_ID,mgr); input2.reload(UnaryTestsF32::INPUTS2_F32_ID,mgr); dims.reload(UnaryTestsF32::DIMSUNARY1_S16_ID,mgr); ref.reload(UnaryTestsF32::REFSUB1_F32_ID,mgr); output.create(ref.nbSamples(),UnaryTestsF32::OUT_F32_ID,mgr); a.create(MAXMATRIXDIM*MAXMATRIXDIM,UnaryTestsF32::TMPA_F32_ID,mgr); b.create(MAXMATRIXDIM*MAXMATRIXDIM,UnaryTestsF32::TMPB_F32_ID,mgr); break; case TEST_MAT_SCALE_F32_3: input1.reload(UnaryTestsF32::INPUTS1_F32_ID,mgr); dims.reload(UnaryTestsF32::DIMSUNARY1_S16_ID,mgr); ref.reload(UnaryTestsF32::REFSCALE1_F32_ID,mgr); output.create(ref.nbSamples(),UnaryTestsF32::OUT_F32_ID,mgr); a.create(MAXMATRIXDIM*MAXMATRIXDIM,UnaryTestsF32::TMPA_F32_ID,mgr); break; case TEST_MAT_TRANS_F32_4: input1.reload(UnaryTestsF32::INPUTS1_F32_ID,mgr); dims.reload(UnaryTestsF32::DIMSUNARY1_S16_ID,mgr); ref.reload(UnaryTestsF32::REFTRANS1_F32_ID,mgr); output.create(ref.nbSamples(),UnaryTestsF32::OUT_F32_ID,mgr); a.create(MAXMATRIXDIM*MAXMATRIXDIM,UnaryTestsF32::TMPA_F32_ID,mgr); break; case TEST_MAT_INVERSE_F32_5: input1.reload(UnaryTestsF32::INPUTSINV_F32_ID,mgr); dims.reload(UnaryTestsF32::DIMSINVERT1_S16_ID,mgr); ref.reload(UnaryTestsF32::REFINV1_F32_ID,mgr); output.create(ref.nbSamples(),UnaryTestsF32::OUT_F32_ID,mgr); a.create(MAXMATRIXDIM*MAXMATRIXDIM,UnaryTestsF32::TMPA_F32_ID,mgr); break; } } void UnaryTestsF32::tearDown(Testing::testID_t id,Client::PatternMgr *mgr) { output.dump(mgr); }