From 9d4d14c9b00e8fc05708140f8339d1f74be07d18 Mon Sep 17 00:00:00 2001
From: Justine Tunney <jtunney@mozilla.com>
Date: Sun, 21 Apr 2024 11:56:15 -0700
Subject: [PATCH] Address review comments

---
 CMakeLists.txt | 16 +++++-----------
 sgemm.cpp      |  6 ++++--
 2 files changed, 9 insertions(+), 13 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index f134a153b..58a1805ba 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -43,17 +43,11 @@ else()
     set(LLAMA_METAL_DEFAULT OFF)
 endif()
 
-# TODO: fix this for Android CI
-#       https://github.com/ggerganov/llama.cpp/pull/6716#issuecomment-2061509191
-#if (CMAKE_SYSTEM_NAME MATCHES "ANDROID")
-#    set(LLAMA_LLAMAFILE_DEFAULT OFF)
-#else()
-#    set(LLAMA_LLAMAFILE_DEFAULT ON)
-#endif()
-
-# TODO: temporary disable until MoE is fixed
-#       https://github.com/ggerganov/llama.cpp/pull/6716
-set(LLAMA_LLAMAFILE_DEFAULT OFF)
+if (CMAKE_SYSTEM_NAME MATCHES "ANDROID")
+    set(LLAMA_LLAMAFILE_DEFAULT OFF)
+else()
+    set(LLAMA_LLAMAFILE_DEFAULT ON)
+endif()
 
 # general
 option(BUILD_SHARED_LIBS                "build shared libraries"                                OFF)
diff --git a/sgemm.cpp b/sgemm.cpp
index 80c3b19a6..799723b2f 100644
--- a/sgemm.cpp
+++ b/sgemm.cpp
@@ -512,8 +512,8 @@ class tinyBLAS_Q0_ARM {
         for (int job = start; job < end; ++job) {
             int ii = m0 + job / xtiles * RM;
             int jj = n0 + job % xtiles * RN;
-            D Cv[RN][RM] = {};
-            for (int l = 0; l < k; l += KN)
+            float32x4_t Cv[RN][RM] = {};
+            for (int l = 0; l < k; ++l)
                 for (int j = 0; j < RN; ++j)
                     for (int i = 0; i < RM; ++i)
                         Cv[j][i] = vmlaq_n_f32(Cv[j][i],
@@ -534,6 +534,7 @@ class tinyBLAS_Q0_ARM {
     inline int8x16_t load_lo(const block_q8_0 *b) {
         return vld1q_s8(b->qs);
     }
+
     inline int8x16_t load_hi(const block_q8_0 *b) {
         return vld1q_s8(b->qs + 16);
     }
@@ -543,6 +544,7 @@ class tinyBLAS_Q0_ARM {
                                                      vdupq_n_u8(0x0f))),
                         vdupq_n_s8(0x8));
     }
+
     inline int8x16_t load_hi(const block_q4_0 *b) {
         return vsubq_s8(vreinterpretq_s8_u8(vshrq_n_u8(vld1q_u8(b->qs), 4)),
                         vdupq_n_s8(0x8));