Use quantized gemm only on aarch64

Summary:
Turns out that __ARM_NEON + armv7/aarch32 does not have same intrinsics as __ARM_NEON + armv8/aarch64

Ran into use of undeclared identifier 'vfmaq_n_f32' and bunch others

Reviewed By: metascroy

Differential Revision: D72413684

Author: kimishpatel

torchao/experimental/kernels/cpu/aarch64/matmul/channelwise_8bit_a_channelwise_8bit_b_1x16x16_f32_smlal-impl.h

@@ -6,7 +6,7 @@
 
 #pragma once
 
-#if defined(__aarch64__) || defined(__ARM_NEON)
+#if defined(__aarch64__) && defined(__ARM_NEON)
 
 #include <algorithm>
 #include <cassert>
@@ -381,4 +381,4 @@ void kernel(
 } // namespace channelwise_8bit_a_channelwise_8bit_b_1x16x16_f32_smlal
 } // namespace torchao::kernels::cpu::aarch64::quantized_matmul
 
-#endif // defined(__aarch64__) || defined(__ARM_NEON)
+#endif // defined(__aarch64__) && defined(__ARM_NEON)

torchao/experimental/kernels/cpu/aarch64/matmul/channelwise_8bit_a_channelwise_8bit_b_1x8x16_f32_neondot-impl.h

@@ -6,7 +6,7 @@
 
 #pragma once
 
-#if defined(__aarch64__) || defined(__ARM_NEON)
+#if defined(__aarch64__) && defined(__ARM_NEON)
 
 #include <algorithm>
 #include <cassert>
@@ -333,4 +333,4 @@ void kernel(
 } // namespace channelwise_8bit_a_channelwise_8bit_b_1x8x16_f32_neondot
 } // namespace torchao::kernels::cpu::aarch64::quantized_matmul
 
-#endif // defined(__aarch64__) || defined(__ARM_NEON)
+#endif // defined(__aarch64__) && defined(__ARM_NEON)

torchao/experimental/kernels/cpu/aarch64/matmul/fp32_a_input_channelwise_8bit_b_1x16x4_f32_impl.h

@@ -6,7 +6,7 @@
 
 #pragma once
 
-#if defined(__aarch64__) || defined(__ARM_NEON)
+#if defined(__aarch64__) && defined(__ARM_NEON)
 
 #include <algorithm>
 #include <cassert>
@@ -278,4 +278,4 @@ void kernel(
 } // namespace fp32_a_input_channelwise_8bit_b_1x16x4_f32
 } // namespace torchao::kernels::cpu::aarch64::quantized_matmul
 
-#endif // defined(__aarch64__) || defined(__ARM_NEON)
+#endif // defined(__aarch64__) && defined(__ARM_NEON)

torchao/experimental/kernels/cpu/aarch64/matmul/matmul.h

@@ -10,7 +10,7 @@
 
 #pragma once
 
-#if defined(__aarch64__) || defined(__ARM_NEON)
+#if defined(__aarch64__) && defined(__ARM_NEON)
 
 #include <arm_neon.h>
 
@@ -92,4 +92,4 @@ void kernel(
 #include <torchao/experimental/kernels/cpu/aarch64/matmul/channelwise_8bit_a_channelwise_8bit_b_1x8x16_f32_neondot-impl.h>
 #include <torchao/experimental/kernels/cpu/aarch64/matmul/fp32_a_input_channelwise_8bit_b_1x16x4_f32_impl.h>
 
-#endif // defined(__aarch64__) || defined(__ARM_NEON)
+#endif // defined(__aarch64__) && defined(__ARM_NEON)

torchao/experimental/kernels/cpu/interface/quantized_matmul.h

@@ -11,11 +11,11 @@
 #include <torchao/experimental/kernels/cpu/fallback/matmul/channelwise_8bit_a_channelwise_8bit_b.h>
 #include <torchao/experimental/kernels/cpu/fallback/matmul/fp32_a_channelwise_8bit_b_fp32_c.h>
 
-#if defined(__aarch64__) || defined(__ARM_NEON)
+#if defined(__aarch64__) && defined(__ARM_NEON)
 #include <torchao/experimental/kernels/cpu/aarch64/matmul/channelwise_8bit_a_channelwise_8bit_b_1x16x16_f32_smlal-impl.h>
 #include <torchao/experimental/kernels/cpu/aarch64/matmul/channelwise_8bit_a_channelwise_8bit_b_1x8x16_f32_neondot-impl.h>
 #include <torchao/experimental/kernels/cpu/aarch64/matmul/fp32_a_input_channelwise_8bit_b_1x16x4_f32_impl.h>
-#endif // defined(__aarch64__) || defined(__ARM_NEON)
+#endif // defined(__aarch64__) && defined(__ARM_NEON)
 
 namespace torchao::kernels::cpu::quantized_matmul {
 
@@ -67,15 +67,15 @@ get_int8_a_int8_b_channelwise_qmatmul(
     bool b_transposed,
     int& a_stride_m,
     int& b_stride_n) {
-#if defined(__aarch64__) || defined(__ARM_NEON)
+#if defined(__aarch64__) && defined(__ARM_NEON)
   if (!a_transposed && b_transposed && n >= 8) {
     a_stride_m = k;
     b_stride_n = k;
     return aarch64::quantized_matmul::
         channelwise_8bit_a_channelwise_8bit_b_1x8x16_f32_neondot::
             kernel<true, true, false, true>;
   }
-#endif // defined(__aarch64__) || defined(__ARM_NEON)
+#endif // defined(__aarch64__) && defined(__ARM_NEON)
   assert(!a_transposed);
   if (b_transposed) {
     a_stride_m = k;
@@ -134,14 +134,14 @@ get_fp32_a_input_channelwise_8bit_b_f32_c_matmul(
     bool b_transposed,
     int& a_stride_m,
     int& b_stride_n) {
-#if defined(__aarch64__) || defined(__ARM_NEON)
+#if defined(__aarch64__) && defined(__ARM_NEON)
   if (!a_transposed && !b_transposed && n >= 16) {
     a_stride_m = k;
     b_stride_n = n;
     return aarch64::quantized_matmul::
         fp32_a_input_channelwise_8bit_b_1x16x4_f32::kernel<true, false, false>;
   }
-#endif // defined(__aarch64__) || defined(__ARM_NEON)
+#endif // defined(__aarch64__) && defined(__ARM_NEON)
   assert(!a_transposed);
   if (b_transposed) {
     a_stride_m = k;