floraachy
/
OpenBLAS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Remove more duplicate.

This commit is contained in:
Chip Kerchner 2024-10-01 11:23:32 -05:00
parent e238a68c03
commit 7cc00f68c9
2 changed files with 4 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
kernel/power/sbgemv_common.c
View File

@ -111,11 +111,6 @@ FORCEINLINE vec_f32 vec_loadNHi_mult(vec_bf16 *in, vec_f32 v_inp0, BLASLONG n, v
return (v_inp0 * v_in00);
}
FORCEINLINE vec_f32 vec_loadNHi_vec(vec_bf16 *in, BLASLONG i, BLASLONG n, vec_bf16 zero)
{
return vec_loadNHi(&in[i], n, zero);
}
FORCEINLINE void copy_x(BLASLONG n, IFLOAT *src, IFLOAT *dest, BLASLONG inc_src)
{
for (BLASLONG i = 0; i < n; i++) {

8
kernel/power/sbgemv_t_vsx.c
View File

@ -66,7 +66,7 @@ static void BF16GEMV_T_VSX_1(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL
temp0 += vec_loadN_mult(&va0[i], inp, n, zero);
} else if (n) {
inp[0] = vec_loadNHi_vec(v_x, i, n, zero);
inp[0] = vec_loadNHi(&v_x[i], n, zero);
temp0 += vec_loadNHi_mult(&va0[i], inp[0], n, zero);
}
@ -105,7 +105,7 @@ static void BF16GEMV_T_VSX_2(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL
temp0 += vec_loadN_mult(&va0[i], inp, n, zero);
temp1 += vec_loadN_mult(&va1[i], inp, n, zero);
} else if (n) {
inp[0] = vec_loadNHi_vec(v_x, i, n, zero);
inp[0] = vec_loadNHi(&v_x[i], n, zero);
temp0 += vec_loadNHi_mult(&va0[i], inp[0], n, zero);
temp1 += vec_loadNHi_mult(&va1[i], inp[0], n, zero);
@ -156,7 +156,7 @@ static void BF16GEMV_T_VSX_4(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL
temp2 += vec_loadN_mult(&va2[i], inp, n, zero);
temp3 += vec_loadN_mult(&va3[i], inp, n, zero);
} else if (n) {
inp[0] = vec_loadNHi_vec(v_x, i, n, zero);
inp[0] = vec_loadNHi(&v_x[i], n, zero);
temp0 += vec_loadNHi_mult(&va0[i], inp[0], n, zero);
temp1 += vec_loadNHi_mult(&va1[i], inp[0], n, zero);
@ -245,7 +245,7 @@ static void BF16GEMV_T_VSX_8(BLASLONG n, BLASLONG lda, IFLOAT *ap, IFLOAT *x, FL
temp6 += vec_loadN_mult(&va6[i], inp, n, zero);
temp7 += vec_loadN_mult(&va7[i], inp, n, zero);
} else if (n) {
inp[0] = vec_loadNHi_vec(v_x, i, n, zero);
inp[0] = vec_loadNHi(&v_x[i], n, zero);
temp0 += vec_loadNHi_mult(&va0[i], inp[0], n, zero);
temp1 += vec_loadNHi_mult(&va1[i], inp[0], n, zero);