diff options
Diffstat (limited to 'src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_generic_depthfirst/generic.cpp')
| -rw-r--r-- | src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_generic_depthfirst/generic.cpp | 32 |
1 files changed, 16 insertions, 16 deletions
diff --git a/src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_generic_depthfirst/generic.cpp b/src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_generic_depthfirst/generic.cpp index 775595f899..8166379ce4 100644 --- a/src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_generic_depthfirst/generic.cpp +++ b/src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_generic_depthfirst/generic.cpp @@ -84,33 +84,33 @@ void sve_fp32_nhwc_max_generic_depthfirst_impl( "subs x24, x24, #0x1\n" "movprfx z23, z1\n fmax z23.s, p4/M, z23.s, z0.s\n" "ldp x21, x20, [x19, #0x10]\n" - "movprfx z18, z31\n fmax z18.s, p4/M, z18.s, z30.s\n" "add x19, x19, #0x20\n" - "fmax z22.s, p4/M, z22.s, z29.s\n" + "movprfx z18, z31\n fmax z18.s, p4/M, z18.s, z30.s\n" "ld1w { z3.s }, p3/Z, [x23, x28, LSL #2]\n" + "fmax z22.s, p4/M, z22.s, z29.s\n" "movprfx z17, z28\n fmax z17.s, p4/M, z17.s, z27.s\n" - "fmax z21.s, p4/M, z21.s, z26.s\n" "ld1w { z2.s }, p3/Z, [x22, x28, LSL #2]\n" - "fmax z16.s, p4/M, z16.s, z25.s\n" + "fmax z21.s, p4/M, z21.s, z26.s\n" "ld1w { z1.s }, p3/Z, [x21, x28, LSL #2]\n" - "fmax z20.s, p4/M, z20.s, z24.s\n" + "fmax z16.s, p4/M, z16.s, z25.s\n" "ld1w { z0.s }, p3/Z, [x20, x28, LSL #2]\n" - "fmax z19.s, p4/M, z19.s, z23.s\n" + "fmax z20.s, p4/M, z20.s, z24.s\n" "ld1w { z31.s }, p2/Z, [x23, x27, LSL #2]\n" - "fmax z18.s, p4/M, z18.s, z22.s\n" + "fmax z19.s, p4/M, z19.s, z23.s\n" "ld1w { z30.s }, p2/Z, [x22, x27, LSL #2]\n" - "fmax z17.s, p4/M, z17.s, z21.s\n" + "fmax z18.s, p4/M, z18.s, z22.s\n" "ld1w { z22.s }, p2/Z, [x21, x27, LSL #2]\n" - "fmax z16.s, p4/M, z16.s, z20.s\n" + "fmax z17.s, p4/M, z17.s, z21.s\n" "ld1w { z29.s }, p2/Z, [x20, x27, LSL #2]\n" - "fmax z7.s, p4/M, z7.s, z19.s\n" + "fmax z16.s, p4/M, z16.s, z20.s\n" "ld1w { z28.s }, p1/Z, [x23, x26, LSL #2]\n" - "fmax z6.s, p4/M, z6.s, z18.s\n" + "fmax z7.s, p4/M, z7.s, z19.s\n" "ld1w { z27.s }, p1/Z, [x22, x26, LSL #2]\n" - "fmax z5.s, p4/M, z5.s, z17.s\n" + "fmax z6.s, p4/M, z6.s, z18.s\n" "ld1w { z21.s }, p1/Z, [x21, x26, LSL #2]\n" - "fmax z4.s, p4/M, z4.s, z16.s\n" + "fmax z5.s, p4/M, z5.s, z17.s\n" "ld1w { z26.s }, p1/Z, [x20, x26, LSL #2]\n" + "fmax z4.s, p4/M, z4.s, z16.s\n" "ld1w { z16.s }, p0/Z, [x23, x25, LSL #2]\n" "ld1w { z25.s }, p0/Z, [x22, x25, LSL #2]\n" "ld1w { z20.s }, p0/Z, [x21, x25, LSL #2]\n" @@ -168,9 +168,9 @@ void sve_fp32_nhwc_max_generic_depthfirst_impl( "lsr x24, %x[n_valid_cells], #0x2\n" "cbz x24, 11f\n" "ldp x23, x22, [x19, #0x0]\n" - "subs x24, x24, #0x1\n" "ldp x21, x20, [x19, #0x10]\n" "add x19, x19, #0x20\n" + "subs x24, x24, #0x1\n" "ld1w { z3.s }, p3/Z, [x23, x28, LSL #2]\n" "ld1w { z2.s }, p3/Z, [x22, x28, LSL #2]\n" "ld1w { z1.s }, p3/Z, [x21, x28, LSL #2]\n" @@ -182,11 +182,11 @@ void sve_fp32_nhwc_max_generic_depthfirst_impl( "subs x24, x24, #0x1\n" "movprfx z23, z1\n fmax z23.s, p4/M, z23.s, z0.s\n" "ldp x21, x20, [x19, #0x10]\n" + "add x19, x19, #0x20\n" "fmax z19.s, p4/M, z19.s, z23.s\n" "ld1w { z3.s }, p3/Z, [x23, x28, LSL #2]\n" - "fmax z7.s, p4/M, z7.s, z19.s\n" "ld1w { z2.s }, p3/Z, [x22, x28, LSL #2]\n" - "add x19, x19, #0x20\n" + "fmax z7.s, p4/M, z7.s, z19.s\n" "ld1w { z1.s }, p3/Z, [x21, x28, LSL #2]\n" "ld1w { z0.s }, p3/Z, [x20, x28, LSL #2]\n" "bgt 9b\n" |