diff options
Diffstat (limited to 'src/f32-gemm/gen/6x8-aarch64-neonfma-ld64.S')
-rw-r--r-- | src/f32-gemm/gen/6x8-aarch64-neonfma-ld64.S | 4 |
1 files changed, 1 insertions, 3 deletions
diff --git a/src/f32-gemm/gen/6x8-aarch64-neonfma-ld64.S b/src/f32-gemm/gen/6x8-aarch64-neonfma-ld64.S index 4f869950b..d9460005f 100644 --- a/src/f32-gemm/gen/6x8-aarch64-neonfma-ld64.S +++ b/src/f32-gemm/gen/6x8-aarch64-neonfma-ld64.S @@ -126,7 +126,7 @@ BEGIN_FUNCTION xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64 # Is there at least 2 floats (8 bytes) for main loop? SUBS x0, x2, 8 // k = kc - 8 - B.LO 2f + B.LO 4f # Main loop - 2 floats of A (8 bytes) # 24 FMA + 6 LD64 A + 2 LDP B @@ -167,7 +167,6 @@ BEGIN_FUNCTION xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64 FMLA v31.4s, v19.4s, v5.s[1] B.HS 1b -2: # Is there a remainder?- 1 floats of A (4 bytes) TBNZ x0, 2, 4f 3: @@ -215,7 +214,6 @@ BEGIN_FUNCTION xnn_f32_gemm_ukernel_6x8__aarch64_neonfma_ld64 SUB x4, x4, x2 // a5 -= kc B.HI 0b - RET 4: |