Enable PLD prefetch version of a75 microkernel for aarch32
on Cortex A57, Cortex A72, Cortex A73, Exynos M1 and Exynos M2

A57 Nexus 6P
Was
MobileNetV1/T:1/real_time          164837 us       163904 us            8 Freq=1.9584G
MobileNetV2/T:1/real_time           97915 us        97336 us           13 Freq=1.9584G
MobileNetV3Large/T:1/real_time      76266 us        75770 us           18 Freq=1.9584G
MobileNetV3Small/T:1/real_time      24066 us        23902 us           59 Freq=1.9584G

Now
MobileNetV1/T:1/real_time          133361 us       132568 us            9 Freq=1.9584G
MobileNetV2/T:1/real_time           85308 us        84727 us           14 Freq=1.9584G
MobileNetV3Large/T:1/real_time      71922 us        71471 us           20 Freq=1.9584G
MobileNetV3Small/T:1/real_time      23726 us        23569 us           59 Freq=1.9584G

A72
Was
MobileNetV1/T:1/real_time          150548 us       150546 us           11 Freq=2.516G
MobileNetV2/T:1/real_time           66435 us        66430 us           21 Freq=2.516G
MobileNetV3Large/T:1/real_time      51950 us        51949 us           27 Freq=2.516G
MobileNetV3Small/T:1/real_time      17160 us        17159 us           82 Freq=2.516G

Now
MobileNetV1/T:1/real_time          113302 us       113293 us           13 Freq=2.516G
MobileNetV2/T:1/real_time           64918 us        64914 us           22 Freq=2.516G
MobileNetV3Large/T:1/real_time      51061 us        51024 us           28 Freq=2.516G
MobileNetV3Small/T:1/real_time      17124 us        17122 us           83 Freq=2.516G

A73 Pixel 2
Was
MobileNetV1/T:1/real_time          120831 us       120080 us           10 Freq=2.4576G
MobileNetV2/T:1/real_time           65185 us        64704 us           21 Freq=2.4576G
MobileNetV3Large/T:1/real_time      51117 us        50699 us           27 Freq=2.4576G
MobileNetV3Small/T:1/real_time      16777 us        16610 us           83 Freq=2.4576G

Now
MobileNetV1/T:1/real_time          116305 us       115524 us           11 Freq=2.4576G
MobileNetV2/T:1/real_time           64558 us        64091 us           22 Freq=2.4576G
MobileNetV3Large/T:1/real_time      51401 us        51020 us           27 Freq=2.4576G
MobileNetV3Small/T:1/real_time      16762 us        16630 us           84 Freq=2.4576G

PiperOrigin-RevId: 285288947
diff --git a/src/init.c b/src/init.c
index 5e9218f..98ffe80 100644
--- a/src/init.c
+++ b/src/init.c
@@ -144,6 +144,20 @@
             .nr = 8,
           };
           break;
+
+        case cpuinfo_uarch_cortex_a57:
+        case cpuinfo_uarch_cortex_a72:
+        case cpuinfo_uarch_cortex_a73:
+          xnn_params.f32.gemm = (struct gemm_parameters) {
+            .gemm = (xnn_gemm_ukernel_function) xnn_f32_gemm_ukernel_4x8__aarch32_neon_pld_cortex_a75,
+            .igemm = (xnn_igemm_ukernel_function) xnn_f32_igemm_ukernel_4x8__neon_lane_ld128,
+            .gemm1 = (xnn_gemm_ukernel_function) xnn_f32_gemm_ukernel_1x8__neon_lane_ld64,
+            .igemm1 = (xnn_igemm_ukernel_function) xnn_f32_igemm_ukernel_1x8__neon_lane_ld64,
+            .mr = 4,
+            .nr = 8,
+          };
+          break;
+
         default:
           xnn_params.f32.gemm = (struct gemm_parameters) {
             .gemm = (xnn_gemm_ukernel_function) xnn_f32_gemm_ukernel_4x8__aarch32_neon_cortex_a75,