Skip to content

Fix SGEMV on POWER8 by reverting to the non-vectorized earlier code #5125

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
martin-frbg merged 2 commits into from
Feb 12, 2025
Merged

Fix SGEMV on POWER8 by reverting to the non-vectorized earlier code #5125

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
98b5ef9
Restore the non-vectorized code from before PR4880 for POWER8
martin-frbg Feb 12, 2025
81eed86
Restore the non-vectorized code from before PR4880 for POWER8
martin-frbg Feb 12, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 19 additions & 4 deletions kernel/power/sgemv_t.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -78,7 +78,17 @@ static void sgemv_kernel_4x8(BLASLONG n, BLASLONG lda, FLOAT *ap, FLOAT *x, FLOA
temp7 += v_x[i] * va7[i];
}


#if defined(POWER8)
y[0] += alpha * (temp0[0] + temp0[1]+temp0[2] + temp0[3]);
y[1] += alpha * (temp1[0] + temp1[1]+temp1[2] + temp1[3]);
y[2] += alpha * (temp2[0] + temp2[1]+temp2[2] + temp2[3]);
y[3] += alpha * (temp3[0] + temp3[1]+temp3[2] + temp3[3]);

y[4] += alpha * (temp4[0] + temp4[1]+temp4[2] + temp4[3]);
y[5] += alpha * (temp5[0] + temp5[1]+temp5[2] + temp5[3]);
y[6] += alpha * (temp6[0] + temp6[1]+temp6[2] + temp6[3]);
y[7] += alpha * (temp7[0] + temp7[1]+temp7[2] + temp7[3]);
#else
register __vector float t0, t1, t2, t3;
register __vector float a = { alpha, alpha, alpha, alpha };
__vector float *v_y = (__vector float*) y;
Expand All @@ -105,7 +115,7 @@ static void sgemv_kernel_4x8(BLASLONG n, BLASLONG lda, FLOAT *ap, FLOAT *x, FLOA

v_y[0] += a * temp0;
v_y[1] += a * temp4;

#endif
}


Expand All @@ -132,7 +142,12 @@ static void sgemv_kernel_4x4(BLASLONG n, BLASLONG lda, FLOAT *ap, FLOAT *x, FLOA
temp2 += v_x[i] * va2[i];
temp3 += v_x[i] * va3[i];
}

#if defined(POWER8)
y[0] += alpha * (temp0[0] + temp0[1]+temp0[2] + temp0[3]);
y[1] += alpha * (temp1[0] + temp1[1]+temp1[2] + temp1[3]);
y[2] += alpha * (temp2[0] + temp2[1]+temp2[2] + temp2[3]);
y[3] += alpha * (temp3[0] + temp3[1]+temp3[2] + temp3[3]);
#else
register __vector float t0, t1, t2, t3;
register __vector float a = { alpha, alpha, alpha, alpha };
__vector float *v_y = (__vector float*) y;
Expand All @@ -148,7 +163,7 @@ static void sgemv_kernel_4x4(BLASLONG n, BLASLONG lda, FLOAT *ap, FLOAT *x, FLOA
temp0 += temp1 + temp2 + temp3;

v_y[0] += a * temp0;

#endif
}


Expand Down
24 changes: 20 additions & 4 deletions kernel/power/sgemv_t_8.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -99,7 +99,17 @@ static void sgemv_kernel_8x8(BLASLONG n, BLASLONG lda, FLOAT *ap, FLOAT *x, FLOA
temp7 += vx1* va7_1 + vx2 * va7_2;
}


#if defined(POWER8)
y[0] += alpha * (temp0[0] + temp0[1]+temp0[2] + temp0[3]);
y[1] += alpha * (temp1[0] + temp1[1]+temp1[2] + temp1[3]);
y[2] += alpha * (temp2[0] + temp2[1]+temp2[2] + temp2[3]);
y[3] += alpha * (temp3[0] + temp3[1]+temp3[2] + temp3[3]);

y[4] += alpha * (temp4[0] + temp4[1]+temp4[2] + temp4[3]);
y[5] += alpha * (temp5[0] + temp5[1]+temp5[2] + temp5[3]);
y[6] += alpha * (temp6[0] + temp6[1]+temp6[2] + temp6[3]);
y[7] += alpha * (temp7[0] + temp7[1]+temp7[2] + temp7[3]);
#else
register __vector float t0, t1, t2, t3;
register __vector float a = { alpha, alpha, alpha, alpha };
__vector float *v_y = (__vector float*) y;
Expand All @@ -126,7 +136,7 @@ static void sgemv_kernel_8x8(BLASLONG n, BLASLONG lda, FLOAT *ap, FLOAT *x, FLOA

v_y[0] += a * temp0;
v_y[1] += a * temp4;

#endif
}


Expand All @@ -153,7 +163,13 @@ static void sgemv_kernel_8x4(BLASLONG n, BLASLONG lda, FLOAT *ap, FLOAT *x, FLOA
temp2 += v_x[i] * va2[i] + v_x[i+1] * va2[i+1];
temp3 += v_x[i] * va3[i] + v_x[i+1] * va3[i+1];
}


#if defined(POWER8)
y[0] += alpha * (temp0[0] + temp0[1]+temp0[2] + temp0[3]);
y[1] += alpha * (temp1[0] + temp1[1]+temp1[2] + temp1[3]);
y[2] += alpha * (temp2[0] + temp2[1]+temp2[2] + temp2[3]);
y[3] += alpha * (temp3[0] + temp3[1]+temp3[2] + temp3[3]);
#else
register __vector float t0, t1, t2, t3;
register __vector float a = { alpha, alpha, alpha, alpha };
__vector float *v_y = (__vector float*) y;
Expand All @@ -169,7 +185,7 @@ static void sgemv_kernel_8x4(BLASLONG n, BLASLONG lda, FLOAT *ap, FLOAT *x, FLOA
temp0 += temp1 + temp2 + temp3;

v_y[0] += a * temp0;

#endif
}


Expand Down
Loading