metal : minor (readibility)
This commit is contained in:
parent
0c17b08cc7
commit
211d82a800
1 changed files with 3 additions and 5 deletions
|
@ -213,10 +213,9 @@ kernel void kernel_diag_mask_inf_8(
|
||||||
dst[i+0] = src0[i+0];
|
dst[i+0] = src0[i+0];
|
||||||
dst[i+1] = src0[i+1];
|
dst[i+1] = src0[i+1];
|
||||||
int64_t i4 = 4*i;
|
int64_t i4 = 4*i;
|
||||||
const int64_t i02 = i4/(ne00*ne01);
|
const int64_t i02 = i4/(ne00*ne01); i4 -= i02*ne00*ne01;
|
||||||
i4 -= i02*ne00*ne01;
|
const int64_t i01 = i4/(ne00); i4 -= i01*ne00;
|
||||||
const int64_t i01 = i4/ne00;
|
const int64_t i00 = i4;
|
||||||
const int64_t i00 = i4 - i01*ne00;
|
|
||||||
for (int k = 3; k >= 0; --k) {
|
for (int k = 3; k >= 0; --k) {
|
||||||
if (i00 + 4 + k <= n_past + i01) {
|
if (i00 + 4 + k <= n_past + i01) {
|
||||||
break;
|
break;
|
||||||
|
@ -678,7 +677,6 @@ kernel void kernel_mul_mat_f16_f32_l4(
|
||||||
device const half4 * x4 = (device const half4 *) (src0 + r0*nb01 + im/(ne12/ne02)*nb02);
|
device const half4 * x4 = (device const half4 *) (src0 + r0*nb01 + im/(ne12/ne02)*nb02);
|
||||||
|
|
||||||
for (int r1 = 0; r1 < nrows; ++r1) {
|
for (int r1 = 0; r1 < nrows; ++r1) {
|
||||||
|
|
||||||
device const float4 * y4 = (device const float4 *) (src1 + r1*nb11 + im*nb12);
|
device const float4 * y4 = (device const float4 *) (src1 + r1*nb11 + im*nb12);
|
||||||
|
|
||||||
float sumf = 0;
|
float sumf = 0;
|
||||||
|
|
Loading…
Add table
Add a link
Reference in a new issue