Browse Source

small gemm kernel packing modifications

tags/v0.3.30
Deeksha Goplani 8 months ago
parent
commit
d1bfa979f7
4 changed files with 4 additions and 4 deletions
  1. +1
    -1
      kernel/arm64/dgemm_small_kernel_tn_sve.c
  2. +1
    -1
      kernel/arm64/dgemm_small_kernel_tt_sve.c
  3. +1
    -1
      kernel/arm64/sgemm_small_kernel_tn_sve.c
  4. +1
    -1
      kernel/arm64/sgemm_small_kernel_tt_sve.c

+ 1
- 1
kernel/arm64/dgemm_small_kernel_tn_sve.c View File

@@ -213,7 +213,7 @@ CNAME(BLASLONG M,
const BLASLONG n2 = N & -2; const BLASLONG n2 = N & -2;
const BLASLONG n8 = N & -8; const BLASLONG n8 = N & -8;


const int pack_a = M >= v_size2 && N >= 8 && K >= 8 ? 1 : 0;
const int pack_a = M >= v_size2 && N >= 8 ? 1 : 0;
FLOAT* packed_a = FLOAT* packed_a =
(pack_a) ? packed_a = (FLOAT*)malloc(K * v_size2 * sizeof(FLOAT)) : NULL; (pack_a) ? packed_a = (FLOAT*)malloc(K * v_size2 * sizeof(FLOAT)) : NULL;




+ 1
- 1
kernel/arm64/dgemm_small_kernel_tt_sve.c View File

@@ -219,7 +219,7 @@ CNAME(BLASLONG M,
const BLASLONG n4 = N & -4; const BLASLONG n4 = N & -4;
const BLASLONG n2 = N & -2; const BLASLONG n2 = N & -2;


const int pack_a = M >= v_size2 && N >= 8 && K >= 8 ? 1 : 0;
const int pack_a = M >= v_size2 && N >= 8 ? 1 : 0;
FLOAT* packed_a = FLOAT* packed_a =
(pack_a) ? packed_a = (FLOAT*)malloc(K * v_size2 * sizeof(FLOAT)) : NULL; (pack_a) ? packed_a = (FLOAT*)malloc(K * v_size2 * sizeof(FLOAT)) : NULL;




+ 1
- 1
kernel/arm64/sgemm_small_kernel_tn_sve.c View File

@@ -222,7 +222,7 @@ CNAME(BLASLONG M,
const BLASLONG n8 = N & -8; const BLASLONG n8 = N & -8;
const BLASLONG n4 = N & -4; const BLASLONG n4 = N & -4;


const int pack_a = M >= v_size2 && N >= 8 && K >= 8 ? 1 : 0;
const int pack_a = M >= v_size2 && N >= 8 ? 1 : 0;
FLOAT* packed_a = FLOAT* packed_a =
(pack_a) ? packed_a = (FLOAT*)malloc(K * v_size2 * sizeof(FLOAT)) : NULL; (pack_a) ? packed_a = (FLOAT*)malloc(K * v_size2 * sizeof(FLOAT)) : NULL;




+ 1
- 1
kernel/arm64/sgemm_small_kernel_tt_sve.c View File

@@ -223,7 +223,7 @@ CNAME(BLASLONG M,
const BLASLONG n8 = N & -8; const BLASLONG n8 = N & -8;
const BLASLONG n4 = N & -4; const BLASLONG n4 = N & -4;


const int pack_a = M >= v_size2 && N >= 8 && K >= 8 ? 1 : 0;
const int pack_a = M >= v_size2 && N >= 8 ? 1 : 0;
FLOAT* packed_a = FLOAT* packed_a =
(pack_a) ? packed_a = (FLOAT*)malloc(K * v_size2 * sizeof(FLOAT)) : NULL; (pack_a) ? packed_a = (FLOAT*)malloc(K * v_size2 * sizeof(FLOAT)) : NULL;




Loading…
Cancel
Save