diff options
Diffstat (limited to 'src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp')
| -rw-r--r-- | src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp | 84 |
1 files changed, 42 insertions, 42 deletions
diff --git a/src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp b/src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp index da0239cea8..d9cebd1363 100644 --- a/src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp +++ b/src/core/NEON/kernels/arm_conv/pooling/kernels/sve_fp32_nhwc_max_2x2_s1_output2x2_depthfirst/generic.cpp @@ -66,10 +66,10 @@ void sve_fp32_nhwc_max_2x2_s1_output2x2_depthfirst_impl( "ldr x15, [%x[args], %[offsetof_n_channels]]\n" "ldr x21, [%x[args], %[offsetof_outptrs]]\n" "mov x14, #0x0\n" - "whilelt p2.s, x14, x15\n" + "whilelt p0.s, x14, x15\n" "ldr x20, [%x[args], %[offsetof_inptrs]]\n" "ldp x13, x12, [x21, #0x0]\n" - "ptrue p1.b\n" + "ptrue p2.b\n" "mov x11, #0x0\n" "ldp x10, x9, [x21, #0x10]\n" "ldp x28, x27, [x20, #0x0]\n" @@ -77,61 +77,61 @@ void sve_fp32_nhwc_max_2x2_s1_output2x2_depthfirst_impl( "ldp x24, x23, [x20, #0x20]\n" "ldp x22, x21, [x20, #0x30]\n" "ldr x20, [x20, #0x40]\n" - "ld1w { z31.s }, p2/Z, [x27, x14, LSL #2]\n" - "ld1w { z30.s }, p2/Z, [x24, x14, LSL #2]\n" - "ld1w { z29.s }, p2/Z, [x21, x14, LSL #2]\n" - "ld1w { z28.s }, p2/Z, [x25, x14, LSL #2]\n" - "ld1w { z27.s }, p2/Z, [x28, x14, LSL #2]\n" - "ld1w { z26.s }, p2/Z, [x26, x14, LSL #2]\n" - "ld1w { z25.s }, p2/Z, [x23, x14, LSL #2]\n" - "ld1w { z24.s }, p2/Z, [x22, x14, LSL #2]\n" - "ld1w { z23.s }, p2/Z, [x20, x14, LSL #2]\n" + "ld1w { z31.s }, p0/Z, [x27, x14, LSL #2]\n" + "ld1w { z30.s }, p0/Z, [x24, x14, LSL #2]\n" + "ld1w { z29.s }, p0/Z, [x21, x14, LSL #2]\n" + "ld1w { z28.s }, p0/Z, [x25, x14, LSL #2]\n" + "ld1w { z27.s }, p0/Z, [x28, x14, LSL #2]\n" + "ld1w { z26.s }, p0/Z, [x26, x14, LSL #2]\n" + "ld1w { z25.s }, p0/Z, [x23, x14, LSL #2]\n" + "ld1w { z24.s }, p0/Z, [x22, x14, LSL #2]\n" + "ld1w { z23.s }, p0/Z, [x20, x14, LSL #2]\n" "incw x14\n" - "whilelt p2.s, x14, x15\n" + "whilelt p1.s, x14, x15\n" "b.none 2f\n" "1:" // Vector: Loop - "movprfx z22, z31\n fmax z22.s, p1/M, z22.s, z30.s\n" - "movprfx z21, z30\n fmax z21.s, p1/M, z21.s, z29.s\n" - "ld1w { z31.s }, p2/Z, [x27, x14, LSL #2]\n" - "ld1w { z30.s }, p2/Z, [x24, x14, LSL #2]\n" - "movprfx z20, z28\n fmax z20.s, p1/M, z20.s, z27.s\n" - "movprfx z19, z26\n fmax z19.s, p1/M, z19.s, z25.s\n" - "ld1w { z29.s }, p2/Z, [x21, x14, LSL #2]\n" - "ld1w { z27.s }, p2/Z, [x28, x14, LSL #2]\n" - "movprfx z17, z28\n fmax z17.s, p1/M, z17.s, z24.s\n" - "movprfx z18, z25\n fmax z18.s, p1/M, z18.s, z23.s\n" - "ld1w { z28.s }, p2/Z, [x25, x14, LSL #2]\n" - "ld1w { z26.s }, p2/Z, [x26, x14, LSL #2]\n" - "ld1w { z25.s }, p2/Z, [x23, x14, LSL #2]\n" - "ld1w { z24.s }, p2/Z, [x22, x14, LSL #2]\n" + "movprfx z22, z31\n fmax z22.s, p2/M, z22.s, z30.s\n" + "movprfx z21, z30\n fmax z21.s, p2/M, z21.s, z29.s\n" + "ld1w { z31.s }, p1/Z, [x27, x14, LSL #2]\n" + "ld1w { z30.s }, p1/Z, [x24, x14, LSL #2]\n" + "movprfx z20, z28\n fmax z20.s, p2/M, z20.s, z27.s\n" + "movprfx z19, z26\n fmax z19.s, p2/M, z19.s, z25.s\n" + "ld1w { z29.s }, p1/Z, [x21, x14, LSL #2]\n" + "ld1w { z27.s }, p1/Z, [x28, x14, LSL #2]\n" + "movprfx z17, z28\n fmax z17.s, p2/M, z17.s, z24.s\n" + "movprfx z18, z25\n fmax z18.s, p2/M, z18.s, z23.s\n" + "ld1w { z28.s }, p1/Z, [x25, x14, LSL #2]\n" + "ld1w { z26.s }, p1/Z, [x26, x14, LSL #2]\n" + "ld1w { z25.s }, p1/Z, [x23, x14, LSL #2]\n" + "ld1w { z24.s }, p1/Z, [x22, x14, LSL #2]\n" "whilelt p0.s, x11, x15\n" - "movprfx z16, z22\n fmax z16.s, p1/M, z16.s, z20.s\n" - "ld1w { z23.s }, p2/Z, [x20, x14, LSL #2]\n" + "movprfx z16, z22\n fmax z16.s, p2/M, z16.s, z20.s\n" + "ld1w { z23.s }, p1/Z, [x20, x14, LSL #2]\n" "incw x14\n" - "whilelt p2.s, x14, x15\n" + "whilelt p1.s, x14, x15\n" "st1w { z16.s }, p0, [x13, x11, LSL #2]\n" - "movprfx z16, z19\n fmax z16.s, p1/M, z16.s, z22.s\n" - "fmax z17.s, p1/M, z17.s, z21.s\n" + "movprfx z16, z19\n fmax z16.s, p2/M, z16.s, z22.s\n" + "fmax z17.s, p2/M, z17.s, z21.s\n" "st1w { z16.s }, p0, [x12, x11, LSL #2]\n" - "movprfx z16, z18\n fmax z16.s, p1/M, z16.s, z21.s\n" + "movprfx z16, z21\n fmax z16.s, p2/M, z16.s, z18.s\n" "st1w { z17.s }, p0, [x10, x11, LSL #2]\n" "st1w { z16.s }, p0, [x9, x11, LSL #2]\n" "incw x11\n" "b.any 1b\n" "2:" // Vector: Tail - "movprfx z22, z31\n fmax z22.s, p1/M, z22.s, z30.s\n" - "movprfx z21, z30\n fmax z21.s, p1/M, z21.s, z29.s\n" - "movprfx z20, z28\n fmax z20.s, p1/M, z20.s, z27.s\n" - "movprfx z19, z26\n fmax z19.s, p1/M, z19.s, z25.s\n" - "movprfx z17, z28\n fmax z17.s, p1/M, z17.s, z24.s\n" - "movprfx z18, z25\n fmax z18.s, p1/M, z18.s, z23.s\n" + "movprfx z22, z31\n fmax z22.s, p2/M, z22.s, z30.s\n" + "movprfx z21, z30\n fmax z21.s, p2/M, z21.s, z29.s\n" + "movprfx z20, z28\n fmax z20.s, p2/M, z20.s, z27.s\n" + "movprfx z19, z26\n fmax z19.s, p2/M, z19.s, z25.s\n" + "movprfx z17, z28\n fmax z17.s, p2/M, z17.s, z24.s\n" + "movprfx z18, z25\n fmax z18.s, p2/M, z18.s, z23.s\n" "whilelt p0.s, x11, x15\n" - "movprfx z16, z22\n fmax z16.s, p1/M, z16.s, z20.s\n" + "movprfx z16, z22\n fmax z16.s, p2/M, z16.s, z20.s\n" "st1w { z16.s }, p0, [x13, x11, LSL #2]\n" - "movprfx z16, z19\n fmax z16.s, p1/M, z16.s, z22.s\n" - "fmax z17.s, p1/M, z17.s, z21.s\n" + "movprfx z16, z19\n fmax z16.s, p2/M, z16.s, z22.s\n" + "fmax z17.s, p2/M, z17.s, z21.s\n" "st1w { z16.s }, p0, [x12, x11, LSL #2]\n" - "movprfx z16, z18\n fmax z16.s, p1/M, z16.s, z21.s\n" + "movprfx z16, z21\n fmax z16.s, p2/M, z16.s, z18.s\n" "st1w { z17.s }, p0, [x10, x11, LSL #2]\n" "st1w { z16.s }, p0, [x9, x11, LSL #2]\n" : |