From a9dbb511fb1fcf28ade60604f3c59f4ba4958318 Mon Sep 17 00:00:00 2001
From: Tracy Sharpe <42477615+tracysh@users.noreply.github.com>
Date: Mon, 5 Apr 2021 22:46:04 -0700
Subject: [PATCH] MLAS: fix qgemm bus error with Android + ARM32 (#7250)

---
 onnxruntime/core/mlas/lib/qgemm.cpp | 58 ++++++++++++++---------------
 1 file changed, 28 insertions(+), 30 deletions(-)

diff --git a/onnxruntime/core/mlas/lib/qgemm.cpp b/onnxruntime/core/mlas/lib/qgemm.cpp
index a6c83e5951..b64669564c 100644
--- a/onnxruntime/core/mlas/lib/qgemm.cpp
+++ b/onnxruntime/core/mlas/lib/qgemm.cpp
@@ -1553,23 +1553,23 @@ MlasGemmU8X8CopyPackA<MLAS_GEMM_U8X8_KERNEL_NEON>(
             k -= 16;
         }
 
-        uint32x4_t GatherVector = vmovq_n_u32(0);
-
         while (k >= 4) {
 
-            GatherVector = vld1q_lane_u32(reinterpret_cast<const uint32_t*>(a0), GatherVector, 0);
+            uint32_t v0 = *reinterpret_cast<const uint32_t*>(a0);
             a0 += 4;
-            GatherVector = vld1q_lane_u32(reinterpret_cast<const uint32_t*>(a1), GatherVector, 1);
+            uint32_t v1 = *reinterpret_cast<const uint32_t*>(a1);
             a1 += 4;
-            GatherVector = vld1q_lane_u32(reinterpret_cast<const uint32_t*>(a2), GatherVector, 2);
+            uint32_t v2 = *reinterpret_cast<const uint32_t*>(a2);
             a2 += 4;
-            GatherVector = vld1q_lane_u32(reinterpret_cast<const uint32_t*>(a3), GatherVector, 3);
+            uint32_t v3 = *reinterpret_cast<const uint32_t*>(a3);
             a3 += 4;
 
-            uint8x16_t PackedVector = vreinterpretq_u8_u32(GatherVector);
-            vst1q_u8(D, PackedVector);
+            *reinterpret_cast<uint32_t*>(&D[0]) = v0;
+            *reinterpret_cast<uint32_t*>(&D[4]) = v1;
+            *reinterpret_cast<uint32_t*>(&D[8]) = v2;
+            *reinterpret_cast<uint32_t*>(&D[12]) = v3;
 
-            RowSums = vpadalq_u16(RowSums, vpaddlq_u8(PackedVector));
+            RowSums = vpadalq_u16(RowSums, vpaddlq_u8(vld1q_u8(D)));
 
             D += 16;
             k -= 4;
@@ -1633,19 +1633,18 @@ MlasGemmU8X8CopyPackA<MLAS_GEMM_U8X8_KERNEL_NEON>(
 
         size_t k = CountK;
         uint32x2_t RowSums = vmov_n_u32(0);
-        uint32x2_t GatherVector = vmov_n_u32(0);
 
         while (k >= 4) {
 
-            GatherVector = vld1_lane_u32(reinterpret_cast<const uint32_t*>(a0), GatherVector, 0);
+            uint32_t v0 = *reinterpret_cast<const uint32_t*>(a0);
             a0 += 4;
-            GatherVector = vld1_lane_u32(reinterpret_cast<const uint32_t*>(a1), GatherVector, 1);
+            uint32_t v1 = *reinterpret_cast<const uint32_t*>(a1);
             a1 += 4;
 
-            uint8x8_t PackedVector = vreinterpret_u8_u32(GatherVector);
-            vst1_u8(D, PackedVector);
+            *reinterpret_cast<uint32_t*>(&D[0]) = v0;
+            *reinterpret_cast<uint32_t*>(&D[4]) = v1;
 
-            RowSums = vpadal_u16(RowSums, vpaddl_u8(PackedVector));
+            RowSums = vpadal_u16(RowSums, vpaddl_u8(vld1_u8(D)));
 
             D += 8;
             k -= 4;
@@ -2030,23 +2029,23 @@ MlasGemmU8X8CopyPackA<MLAS_GEMM_U8X8_KERNEL_UDOT>(
             k -= 16;
         }
 
-        uint32x4_t GatherVector = vmovq_n_u32(0);
-
         while (k >= 4) {
 
-            GatherVector = vld1q_lane_u32(reinterpret_cast<const uint32_t*>(a0), GatherVector, 0);
+            uint32_t v0 = *reinterpret_cast<const uint32_t*>(a0);
             a0 += 4;
-            GatherVector = vld1q_lane_u32(reinterpret_cast<const uint32_t*>(a1), GatherVector, 1);
+            uint32_t v1 = *reinterpret_cast<const uint32_t*>(a1);
             a1 += 4;
-            GatherVector = vld1q_lane_u32(reinterpret_cast<const uint32_t*>(a2), GatherVector, 2);
+            uint32_t v2 = *reinterpret_cast<const uint32_t*>(a2);
             a2 += 4;
-            GatherVector = vld1q_lane_u32(reinterpret_cast<const uint32_t*>(a3), GatherVector, 3);
+            uint32_t v3 = *reinterpret_cast<const uint32_t*>(a3);
             a3 += 4;
 
-            uint8x16_t PackedVector = vreinterpretq_u8_u32(GatherVector);
-            vst1q_u8(D, PackedVector);
+            *reinterpret_cast<uint32_t*>(&D[0]) = v0;
+            *reinterpret_cast<uint32_t*>(&D[4]) = v1;
+            *reinterpret_cast<uint32_t*>(&D[8]) = v2;
+            *reinterpret_cast<uint32_t*>(&D[12]) = v3;
 
-            RowSums = vpadalq_u16(RowSums, vpaddlq_u8(PackedVector));
+            RowSums = vpadalq_u16(RowSums, vpaddlq_u8(vld1q_u8(D)));
 
             D += 16;
             k -= 4;
@@ -2117,19 +2116,18 @@ MlasGemmU8X8CopyPackA<MLAS_GEMM_U8X8_KERNEL_UDOT>(
 
         size_t k = CountK;
         uint32x2_t RowSums = vmov_n_u32(0);
-        uint32x2_t GatherVector = vmov_n_u32(0);
 
         while (k >= 4) {
 
-            GatherVector = vld1_lane_u32(reinterpret_cast<const uint32_t*>(a0), GatherVector, 0);
+            uint32_t v0 = *reinterpret_cast<const uint32_t*>(a0);
             a0 += 4;
-            GatherVector = vld1_lane_u32(reinterpret_cast<const uint32_t*>(a1), GatherVector, 1);
+            uint32_t v1 = *reinterpret_cast<const uint32_t*>(a1);
             a1 += 4;
 
-            uint8x8_t PackedVector = vreinterpret_u8_u32(GatherVector);
-            vst1_u8(D, PackedVector);
+            *reinterpret_cast<uint32_t*>(&D[0]) = v0;
+            *reinterpret_cast<uint32_t*>(&D[4]) = v1;
 
-            RowSums = vpadal_u16(RowSums, vpaddl_u8(PackedVector));
+            RowSums = vpadal_u16(RowSums, vpaddl_u8(vld1_u8(D)));
 
             D += 8;
             k -= 4;