OSchip
/
OpenBLAS
Not watched
1
0
Fork 0
Code
Releases 66 Wiki evaluate Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain HPC
You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
Tree: 6d8a273cca
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '6d8a273cca'
${ noResults }
OpenBLAS/kernel/riscv64/zaxpby_vector.c

273 lines
12 kB
Raw Blame History 

  1. /***************************************************************************

  2. Copyright (c) 2020, The OpenBLAS Project

  3. All rights reserved.

  4. Redistribution and use in source and binary forms, with or without

  5. modification, are permitted provided that the following conditions are

  6. met:

  7. 1. Redistributions of source code must retain the above copyright

  8. notice, this list of conditions and the following disclaimer.

  9. 2. Redistributions in binary form must reproduce the above copyright

  10. notice, this list of conditions and the following disclaimer in

  11. the documentation and/or other materials provided with the

  12. distribution.

  13. 3. Neither the name of the OpenBLAS project nor the names of

  14. its contributors may be used to endorse or promote products

  15. derived from this software without specific prior written permission.

  16. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"

  17. AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE

  18. IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE

  19. ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE

  20. LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL

  21. DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR

  22. SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER

  23. CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,

  24. OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE

  25. USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

  26. *****************************************************************************/

  27. 

  28. #include "common.h"

  29. 

  30. #if !defined(DOUBLE)

  31. #define VSETVL(n) RISCV_RVV(vsetvl_e32m4)(n)

  32. #define FLOAT_V_T vfloat32m4_t

  33. #define VLSEV_FLOAT RISCV_RVV(vlse32_v_f32m4)

  34. #define VSSEV_FLOAT RISCV_RVV(vsse32_v_f32m4)

  35. #define VFMACCVF_FLOAT RISCV_RVV(vfmacc_vf_f32m4)

  36. #define VFMVVF_FLOAT RISCV_RVV(vfmv_v_f_f32m4)

  37. #define VFMULVF_FLOAT RISCV_RVV(vfmul_vf_f32m4)

  38. #define VFMSACVF_FLOAT RISCV_RVV(vfmsac_vf_f32m4)

  39. #define VFNMSACVF_FLOAT RISCV_RVV(vfnmsac_vf_f32m4)

  40. #else

  41. #define VSETVL(n) RISCV_RVV(vsetvl_e64m4)(n)

  42. #define FLOAT_V_T vfloat64m4_t

  43. #define VLSEV_FLOAT RISCV_RVV(vlse64_v_f64m4)

  44. #define VSSEV_FLOAT RISCV_RVV(vsse64_v_f64m4)

  45. #define VFMACCVF_FLOAT RISCV_RVV(vfmacc_vf_f64m4)

  46. #define VFMVVF_FLOAT RISCV_RVV(vfmv_v_f_f64m4)

  47. #define VFMULVF_FLOAT RISCV_RVV(vfmul_vf_f64m4)

  48. #define VFMSACVF_FLOAT RISCV_RVV(vfmsac_vf_f64m4)

  49. #define VFNMSACVF_FLOAT RISCV_RVV(vfnmsac_vf_f64m4)

  50. #endif

  51. 

  52. int CNAME(BLASLONG n, FLOAT alpha_r, FLOAT alpha_i, FLOAT *x, BLASLONG inc_x, FLOAT beta_r, FLOAT beta_i, FLOAT *y, BLASLONG inc_y)

  53. {

  54. 	if (n <= 0)  return(0);

  55. 

  56. 	BLASLONG i=0, j=0;

  57. 	unsigned int gvl = 0;

  58. 	FLOAT_V_T vx0, vx1;

  59.         FLOAT_V_T vy0, vy1;

  60. 

  61. 	BLASLONG stride_x, stride_y, ix = 0, iy = 0;

  62.         stride_x = inc_x * 2 * sizeof(FLOAT);

  63.         stride_y = inc_y * 2 * sizeof(FLOAT);

  64. 

  65. 	if (inc_x == 0 || inc_y == 0) {

  66. 

  67. 	FLOAT temp;

  68. 	BLASLONG inc_x2, inc_y2;

  69. 

  70. 	inc_x2 = 2 * inc_x;

  71. 	inc_y2 = 2 * inc_y;

  72. 

  73. 	if ( beta_r == 0.0 && beta_i == 0.0)

  74. 	{

  75. 		if ( alpha_r == 0.0 && alpha_i == 0.0 )

  76. 		{

  77. 

  78. 			while(i < n)

  79. 			{

  80. 				y[iy]   = 0.0 ;

  81. 				y[iy+1] = 0.0 ;

  82. 				iy += inc_y2 ;

  83. 				i++ ;

  84. 			}

  85. 

  86. 		}

  87. 		else

  88. 		{

  89. 

  90. 			while(i < n)

  91. 			{

  92. 				y[iy]   = ( alpha_r * x[ix]   - alpha_i * x[ix+1] ) ;

  93. 				y[iy+1] = ( alpha_r * x[ix+1] + alpha_i * x[ix]   ) ;

  94. 				ix += inc_x2 ;

  95. 				iy += inc_y2 ;

  96. 				i++ ;

  97. 			}

  98. 

  99. 

  100. 		}

  101. 

  102. 	}

  103. 	else

  104. 	{

  105. 		if ( alpha_r == 0.0 && alpha_i == 0.0 )

  106. 		{

  107. 

  108. 			while(i < n)

  109. 			{

  110. 				temp    = ( beta_r * y[iy]   - beta_i * y[iy+1] ) ;

  111. 				y[iy+1] = ( beta_r * y[iy+1] + beta_i * y[iy]   ) ;

  112. 				y[iy]   = temp;

  113. 				iy += inc_y2 ;

  114. 				i++ ;

  115. 			}

  116. 

  117. 		}

  118. 		else

  119. 		{

  120. 

  121. 			while(i < n)

  122. 			{

  123. 				temp    = ( alpha_r * x[ix]   - alpha_i * x[ix+1] ) + ( beta_r * y[iy]   - beta_i * y[iy+1] ) ;

  124. 				y[iy+1] = ( alpha_r * x[ix+1] + alpha_i * x[ix]   ) + ( beta_r * y[iy+1] + beta_i * y[iy]   ) ;

  125. 				y[iy]   = temp;

  126. 				ix += inc_x2 ;

  127. 				iy += inc_y2 ;

  128. 				i++ ;

  129. 			}

  130. 

  131. 

  132. 		}

  133. 

  134. 

  135. 

  136. 	}

  137. 	return(0);

  138. 

  139. 	} else {

  140. 

  141.         if(beta_r == 0.0 && beta_i == 0.0){

  142.                 if(alpha_r == 0.0 && alpha_i == 0.0){

  143.                         if(inc_y == 1){

  144.                                 memset(&y[0], 0, 2 * n * sizeof(FLOAT));

  145.                         }else{

  146.                                 gvl = VSETVL(n);

  147.                                 if(gvl <= n/2){

  148.                                         vy0 = VFMVVF_FLOAT(0.0, gvl);

  149.                                         BLASLONG inc_yv = inc_y * gvl * 2;

  150.                                         for(i=0,j=0;i<n/(gvl*2);i++){

  151.                                                 VSSEV_FLOAT(&y[iy], stride_y, vy0, gvl);

  152.                                                 VSSEV_FLOAT(&y[iy+1], stride_y, vy0, gvl);

  153.                                                 VSSEV_FLOAT(&y[iy+inc_yv], stride_y, vy0, gvl);

  154.                                                 VSSEV_FLOAT(&y[iy+1+inc_yv], stride_y, vy0, gvl);

  155.                                                 j += gvl * 2;

  156.                                                 iy += inc_yv * 2;

  157.                                         }

  158.                                 }

  159.                                 for(;j<n;){

  160.                                         gvl = VSETVL(n-j);

  161.                                         vy0 = VFMVVF_FLOAT(0.0, gvl);

  162.                                         VSSEV_FLOAT(&y[iy], stride_y, vy0, gvl);

  163.                                         VSSEV_FLOAT(&y[iy+1], stride_y, vy0, gvl);

  164.                                         j += gvl;

  165.                                         iy += inc_y * gvl * 2;

  166.                                 }

  167.                         }

  168. 		}else{

  169.                         gvl = VSETVL(n);

  170.                         BLASLONG inc_xv = inc_x * gvl * 2;

  171.                         BLASLONG inc_yv = inc_y * gvl * 2;

  172.                         for(i=0,j=0; i<n/gvl; i++){

  173.                                 vx0 = VLSEV_FLOAT(&x[ix], stride_x, gvl);

  174.                                 vx1 = VLSEV_FLOAT(&x[ix+1], stride_x, gvl);

  175.                                 vy0 = VFMULVF_FLOAT(vx1, alpha_i, gvl);

  176.                                 vy0 = VFMSACVF_FLOAT(vy0, alpha_r, vx0, gvl);

  177.                                 VSSEV_FLOAT(&y[iy], stride_y, vy0, gvl);

  178.                                 vy1 = VFMULVF_FLOAT(vx1, alpha_r, gvl);

  179.                                 vy1 = VFMACCVF_FLOAT(vy1, alpha_i, vx0, gvl);

  180.                                 VSSEV_FLOAT(&y[iy+1], stride_y, vy1, gvl);

  181. 

  182.                                 j += gvl;

  183.                                 ix += inc_xv;

  184.                                 iy += inc_yv;

  185.                         }

  186.                         if(j<n){

  187.                                 gvl = VSETVL(n-j);

  188.                                 vx0 = VLSEV_FLOAT(&x[ix], stride_x, gvl);

  189.                                 vx1 = VLSEV_FLOAT(&x[ix+1], stride_x, gvl);

  190.                                 vy0 = VFMULVF_FLOAT(vx1, alpha_i, gvl);

  191.                                 vy0 = VFMSACVF_FLOAT(vy0, alpha_r, vx0, gvl);

  192.                                 VSSEV_FLOAT(&y[iy], stride_y, vy0, gvl);

  193.                                 vy1 = VFMULVF_FLOAT(vx1, alpha_r, gvl);

  194.                                 vy1 = VFMACCVF_FLOAT(vy1, alpha_i, vx0, gvl);

  195.                                 VSSEV_FLOAT(&y[iy+1], stride_y, vy1, gvl);

  196.                         }

  197.                 }

  198.         }else{

  199. 	        FLOAT_V_T v0, v1;

  200.                 if(alpha_r == 0.0 && alpha_i == 0.0){

  201.                         gvl = VSETVL(n);

  202.                         BLASLONG inc_yv = inc_y * gvl * 2;

  203.                         for(i=0,j=0;i<n/gvl;i++){

  204.                                 vy0 = VLSEV_FLOAT(&y[iy], stride_y, gvl);

  205.                                 vy1 = VLSEV_FLOAT(&y[iy+1], stride_y, gvl);

  206.                                 v0 = VFMULVF_FLOAT(vy1, beta_i, gvl);

  207.                                 v0 = VFMSACVF_FLOAT(v0, beta_r, vy0, gvl);

  208.                                 VSSEV_FLOAT(&y[iy], stride_y, v0, gvl);

  209.                                 v1 = VFMULVF_FLOAT(vy1, beta_r, gvl);

  210.                                 v1 = VFMACCVF_FLOAT(v1, beta_i, vy0, gvl);

  211.                                 VSSEV_FLOAT(&y[iy+1], stride_y, v1, gvl);

  212.                                 j += gvl;

  213.                                 iy += inc_yv;

  214.                         }

  215.                         if(j<n){

  216.                                 gvl = VSETVL(n-j);

  217.                                 vy0 = VLSEV_FLOAT(&y[iy], stride_y, gvl);

  218.                                 vy1 = VLSEV_FLOAT(&y[iy+1], stride_y, gvl);

  219.                                 v0 = VFMULVF_FLOAT(vy1, beta_i, gvl);

  220.                                 v0 = VFMSACVF_FLOAT(v0, beta_r, vy0, gvl);

  221.                                 VSSEV_FLOAT(&y[iy], stride_y, v0, gvl);

  222.                                 v1 = VFMULVF_FLOAT(vy1, beta_r, gvl);

  223.                                 v1 = VFMACCVF_FLOAT(v1, beta_i, vy0, gvl);

  224.                                 VSSEV_FLOAT(&y[iy+1], stride_y, v1, gvl);

  225.                         }

  226. 		}else{

  227.                         gvl = VSETVL(n);

  228.                         BLASLONG inc_xv = inc_x * gvl * 2;

  229.                         BLASLONG inc_yv = inc_y * gvl * 2;

  230.                         for(i=0,j=0; i<n/gvl; i++){

  231.                                 vx0 = VLSEV_FLOAT(&x[ix], stride_x, gvl);

  232.                                 vx1 = VLSEV_FLOAT(&x[ix+1], stride_x, gvl);

  233.                                 vy0 = VLSEV_FLOAT(&y[iy], stride_y, gvl);

  234.                                 vy1 = VLSEV_FLOAT(&y[iy+1], stride_y, gvl);

  235.                                 v0 = VFMULVF_FLOAT(vx0, alpha_r, gvl);

  236.                                 v0 = VFNMSACVF_FLOAT(v0, alpha_i, vx1, gvl);

  237.                                 v0 = VFMACCVF_FLOAT(v0, beta_r, vy0, gvl);

  238.                                 v0 = VFNMSACVF_FLOAT(v0, beta_i, vy1, gvl);

  239.                                 VSSEV_FLOAT(&y[iy], stride_y, v0, gvl);

  240.                                 v1 = VFMULVF_FLOAT(vx1, alpha_r, gvl);

  241.                                 v1 = VFMACCVF_FLOAT(v1, alpha_i, vx0, gvl);

  242.                                 v1 = VFMACCVF_FLOAT(v1, beta_r, vy1, gvl);

  243.                                 v1 = VFMACCVF_FLOAT(v1, beta_i, vy0, gvl);

  244.                                 VSSEV_FLOAT(&y[iy+1], stride_y, v1, gvl);

  245. 

  246.                                 j += gvl;

  247.                                 ix += inc_xv;

  248.                                 iy += inc_yv;

  249.                         }

  250.                         if(j<n){

  251.                                 gvl = VSETVL(n-j);

  252.                                 vx0 = VLSEV_FLOAT(&x[ix], stride_x, gvl);

  253.                                 vx1 = VLSEV_FLOAT(&x[ix+1], stride_x, gvl);

  254.                                 vy0 = VLSEV_FLOAT(&y[iy], stride_y, gvl);

  255.                                 vy1 = VLSEV_FLOAT(&y[iy+1], stride_y, gvl);

  256.                                 v0 = VFMULVF_FLOAT(vx0, alpha_r, gvl);

  257.                                 v0 = VFNMSACVF_FLOAT(v0, alpha_i, vx1, gvl);

  258.                                 v0 = VFMACCVF_FLOAT(v0, beta_r, vy0, gvl);

  259.                                 v0 = VFNMSACVF_FLOAT(v0, beta_i, vy1, gvl);

  260.                                 VSSEV_FLOAT(&y[iy], stride_y, v0, gvl);

  261.                                 v1 = VFMULVF_FLOAT(vx1, alpha_r, gvl);

  262.                                 v1 = VFMACCVF_FLOAT(v1, alpha_i, vx0, gvl);

  263.                                 v1 = VFMACCVF_FLOAT(v1, beta_r, vy1, gvl);

  264.                                 v1 = VFMACCVF_FLOAT(v1, beta_i, vy0, gvl);

  265.                                 VSSEV_FLOAT(&y[iy+1], stride_y, v1, gvl);

  266.                         }

  267.                 }

  268.         }

  269. 	return(0);

  270. 	}

  271. }