From 2ce2bd5cf115e0b6f8b91878079adec61333b8a0 Mon Sep 17 00:00:00 2001
From: Eve <139727413+netrunnereve@users.noreply.github.com>
Date: Mon, 22 Sep 2025 17:06:53 +0000
Subject: [PATCH 1/4] run the x64 ci on regular machines

---
 .github/workflows/build.yml | 32 ++++++++++++++++++++++++++++----
 ci/run.sh                   | 22 +++++++++++-----------
 2 files changed, 39 insertions(+), 15 deletions(-)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index afdcdd07b05cc..675aa702a433b 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -1251,17 +1251,29 @@ jobs:
 # TODO: simplify the following workflows using a matrix
 # TODO: run lighter CI on PRs and the full CI only on master (if needed)
   ggml-ci-x64-cpu-low-perf:
-    runs-on: [self-hosted, Linux, X64, CPU, low-perf]
+    runs-on: ubuntu-22.04
 
     steps:
       - name: Clone
         id: checkout
         uses: actions/checkout@v4
 
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.16
+        with:
+          key: ggml-ci-x64-cpu-low-perf
+          evict-old-files: 1d
+
+      - name: Dependencies
+        id: depends
+        run: |
+          sudo apt-get update
+          sudo apt-get install build-essential libcurl4-openssl-dev
+
       - name: Test
         id: ggml-ci
         run: |
-          bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
+          LLAMA_ARG_THREADS=$(nproc) GG_BUILD_LOW_PERF=1 bash ./ci/run.sh ./tmp/results ./tmp/mnt
 
   ggml-ci-arm64-cpu-low-perf:
     runs-on: [self-hosted, Linux, ARM64, CPU, low-perf]
@@ -1277,17 +1289,29 @@ jobs:
           bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
 
   ggml-ci-x64-cpu-high-perf:
-    runs-on: [self-hosted, Linux, X64, CPU, high-perf]
+    runs-on: ubuntu-22.04
 
     steps:
       - name: Clone
         id: checkout
         uses: actions/checkout@v4
 
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.16
+        with:
+          key: ggml-ci-x64-cpu-high-perf
+          evict-old-files: 1d
+
+      - name: Dependencies
+        id: depends
+        run: |
+          sudo apt-get update
+          sudo apt-get install build-essential libcurl4-openssl-dev
+
       - name: Test
         id: ggml-ci
         run: |
-          bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
+          LLAMA_ARG_THREADS=$(nproc) bash ./ci/run.sh ./tmp/results ./tmp/mnt
 
   ggml-ci-arm64-cpu-high-perf:
     runs-on: [self-hosted, Linux, ARM64, CPU, high-perf]
diff --git a/ci/run.sh b/ci/run.sh
index cb90f7a7850fc..a86f663b85717 100755
--- a/ci/run.sh
+++ b/ci/run.sh
@@ -339,16 +339,16 @@ function gg_run_qwen3_0_6b {
 
     wiki_test="${path_wiki}/wiki.test.raw"
 
-    ./bin/llama-quantize ${model_bf16} ${model_q8_0} q8_0
-    ./bin/llama-quantize ${model_bf16} ${model_q4_0} q4_0
-    ./bin/llama-quantize ${model_bf16} ${model_q4_1} q4_1
-    ./bin/llama-quantize ${model_bf16} ${model_q5_0} q5_0
-    ./bin/llama-quantize ${model_bf16} ${model_q5_1} q5_1
-    ./bin/llama-quantize ${model_bf16} ${model_q2_k} q2_k
-    ./bin/llama-quantize ${model_bf16} ${model_q3_k} q3_k
-    ./bin/llama-quantize ${model_bf16} ${model_q4_k} q4_k
-    ./bin/llama-quantize ${model_bf16} ${model_q5_k} q5_k
-    ./bin/llama-quantize ${model_bf16} ${model_q6_k} q6_k
+    ./bin/llama-quantize ${model_bf16} ${model_q8_0} q8_0 $(nproc)
+    ./bin/llama-quantize ${model_bf16} ${model_q4_0} q4_0 $(nproc)
+    ./bin/llama-quantize ${model_bf16} ${model_q4_1} q4_1 $(nproc)
+    ./bin/llama-quantize ${model_bf16} ${model_q5_0} q5_0 $(nproc)
+    ./bin/llama-quantize ${model_bf16} ${model_q5_1} q5_1 $(nproc)
+    ./bin/llama-quantize ${model_bf16} ${model_q2_k} q2_k $(nproc)
+    ./bin/llama-quantize ${model_bf16} ${model_q3_k} q3_k $(nproc)
+    ./bin/llama-quantize ${model_bf16} ${model_q4_k} q4_k $(nproc)
+    ./bin/llama-quantize ${model_bf16} ${model_q5_k} q5_k $(nproc)
+    ./bin/llama-quantize ${model_bf16} ${model_q6_k} q6_k $(nproc)
 
     (time ./bin/llama-cli -no-cnv --model ${model_f16}  -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-f16.log
     (time ./bin/llama-cli -no-cnv --model ${model_bf16} -ngl 99 -c 1024 -s 1234 -n 64 --ignore-eos -p "I believe the meaning of life is" ) 2>&1 | tee -a $OUT/${ci}-tg-bf16.log
@@ -421,7 +421,7 @@ function gg_run_qwen3_0_6b {
 function gg_sum_qwen3_0_6b {
     gg_printf '### %s\n\n' "${ci}"
 
-    gg_printf 'Pythia 2.8B:\n'
+    gg_printf 'Qwen3 0.6B:\n'
     gg_printf '- status: %s\n' "$(cat $OUT/${ci}.exit)"
     gg_printf '- perplexity:\n%s\n' "$(cat $OUT/${ci}-ppl.log)"
     gg_printf '- imatrix:\n```\n%s\n```\n' "$(cat $OUT/${ci}-imatrix-sum.log)"

From 51d0b497d2057a581b7138fda13312591c5f51fb Mon Sep 17 00:00:00 2001
From: Eve <139727413+netrunnereve@users.noreply.github.com>
Date: Mon, 22 Sep 2025 21:18:13 +0000
Subject: [PATCH 2/4] set up the same thing for arm

fix test-quantize-perf just like #12306
---
 .github/workflows/build.yml  | 32 ++++++++++++++++++++++++++++----
 tests/test-quantize-perf.cpp | 11 +----------
 2 files changed, 29 insertions(+), 14 deletions(-)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 675aa702a433b..98c0f8d985e41 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -1276,17 +1276,29 @@ jobs:
           LLAMA_ARG_THREADS=$(nproc) GG_BUILD_LOW_PERF=1 bash ./ci/run.sh ./tmp/results ./tmp/mnt
 
   ggml-ci-arm64-cpu-low-perf:
-    runs-on: [self-hosted, Linux, ARM64, CPU, low-perf]
+    runs-on: ubuntu-22.04-arm
 
     steps:
       - name: Clone
         id: checkout
         uses: actions/checkout@v4
 
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.16
+        with:
+          key: ggml-ci-arm64-cpu-low-perf
+          evict-old-files: 1d
+
+      - name: Dependencies
+        id: depends
+        run: |
+          sudo apt-get update
+          sudo apt-get install build-essential libcurl4-openssl-dev
+
       - name: Test
         id: ggml-ci
         run: |
-          bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
+          LLAMA_ARG_THREADS=$(nproc) GG_BUILD_LOW_PERF=1 bash ./ci/run.sh ./tmp/results ./tmp/mnt
 
   ggml-ci-x64-cpu-high-perf:
     runs-on: ubuntu-22.04
@@ -1314,17 +1326,29 @@ jobs:
           LLAMA_ARG_THREADS=$(nproc) bash ./ci/run.sh ./tmp/results ./tmp/mnt
 
   ggml-ci-arm64-cpu-high-perf:
-    runs-on: [self-hosted, Linux, ARM64, CPU, high-perf]
+    runs-on: ubuntu-22.04-arm
 
     steps:
       - name: Clone
         id: checkout
         uses: actions/checkout@v4
 
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.16
+        with:
+          key: ggml-ci-arm64-cpu-high-perf
+          evict-old-files: 1d
+
+      - name: Dependencies
+        id: depends
+        run: |
+          sudo apt-get update
+          sudo apt-get install build-essential libcurl4-openssl-dev
+
       - name: Test
         id: ggml-ci
         run: |
-          GG_BUILD_NO_BF16=1 GG_BUILD_EXTRA_TESTS_0=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
+          LLAMA_ARG_THREADS=$(nproc) GG_BUILD_NO_BF16=1 GG_BUILD_EXTRA_TESTS_0=1 bash ./ci/run.sh ./tmp/results ./tmp/mnt
 
   ggml-ci-x64-nvidia-v100-cuda:
     runs-on: [self-hosted, Linux, X64, NVIDIA, V100]
diff --git a/tests/test-quantize-perf.cpp b/tests/test-quantize-perf.cpp
index 2882884938388..cac0782dee9a9 100644
--- a/tests/test-quantize-perf.cpp
+++ b/tests/test-quantize-perf.cpp
@@ -260,14 +260,7 @@ int main(int argc, char * argv[]) {
 
     int64_t iterations = params.iterations;
 
-
-    // Initialize GGML, ensures float conversion tables are initialized
-    struct ggml_init_params ggml_params = {
-        /* .mem_size   = */ 1*1024,
-        /* .mem_buffer = */ NULL,
-        /* .no_alloc   = */ true,
-    };
-    struct ggml_context * ctx = ggml_init(ggml_params);
+    ggml_cpu_init();
 
     for (int i = 0; i < GGML_TYPE_COUNT; i++) {
         ggml_type type = (ggml_type) i;
@@ -359,7 +352,5 @@ int main(int argc, char * argv[]) {
         }
     }
 
-    ggml_free(ctx);
-
     return 0;
 }

From a9a68096ee2be7c1a4be46c264bc3451689b0fcd Mon Sep 17 00:00:00 2001
From: Eve <139727413+netrunnereve@users.noreply.github.com>
Date: Tue, 23 Sep 2025 12:02:09 -0400
Subject: [PATCH 3/4] try to disable sve

---
 .github/workflows/build.yml | 2 +-
 ci/run.sh                   | 4 ++++
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 98c0f8d985e41..5c22fd33e2392 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -1348,7 +1348,7 @@ jobs:
       - name: Test
         id: ggml-ci
         run: |
-          LLAMA_ARG_THREADS=$(nproc) GG_BUILD_NO_BF16=1 GG_BUILD_EXTRA_TESTS_0=1 bash ./ci/run.sh ./tmp/results ./tmp/mnt
+          LLAMA_ARG_THREADS=$(nproc) GG_BUILD_NO_SVE=1 GG_BUILD_NO_BF16=1 GG_BUILD_EXTRA_TESTS_0=1 bash ./ci/run.sh ./tmp/results ./tmp/mnt
 
   ggml-ci-x64-nvidia-v100-cuda:
     runs-on: [self-hosted, Linux, X64, NVIDIA, V100]
diff --git a/ci/run.sh b/ci/run.sh
index a86f663b85717..7164b3f852169 100755
--- a/ci/run.sh
+++ b/ci/run.sh
@@ -103,6 +103,10 @@ if [ ! -z ${GG_BUILD_MUSA} ]; then
     MUSA_ARCH=${MUSA_ARCH:-21}
     CMAKE_EXTRA="${CMAKE_EXTRA} -DGGML_MUSA=ON -DMUSA_ARCHITECTURES=${MUSA_ARCH}"
 fi
+
+if [ ! -z ${GG_BUILD_NO_SVE} ]; then
+    CMAKE_EXTRA="${CMAKE_EXTRA} -DGGML_NATIVE=OFF -DGGML_CPU_ARM_ARCH=armv8.5-a+fp16+i8mm"
+fi
 ## helpers
 
 # download a file if it does not exist or if it is outdated

From c2df226f26934455599e00b8f70ca9c3240e8392 Mon Sep 17 00:00:00 2001
From: Eve <139727413+netrunnereve@users.noreply.github.com>
Date: Wed, 24 Sep 2025 11:28:11 -0400
Subject: [PATCH 4/4] add another sve run

---
 .github/workflows/build.yml | 25 +++++++++++++++++++++++++
 ci/run.sh                   |  1 +
 2 files changed, 26 insertions(+)

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 768d880a3865c..4f70232b1777f 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -1350,6 +1350,31 @@ jobs:
         run: |
           LLAMA_ARG_THREADS=$(nproc) GG_BUILD_NO_SVE=1 GG_BUILD_NO_BF16=1 GG_BUILD_EXTRA_TESTS_0=1 bash ./ci/run.sh ./tmp/results ./tmp/mnt
 
+  ggml-ci-arm64-cpu-high-perf-sve:
+    runs-on: ubuntu-22.04-arm
+
+    steps:
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v4
+
+      - name: ccache
+        uses: ggml-org/ccache-action@v1.2.16
+        with:
+          key: ggml-ci-arm64-cpu-high-perf-sve
+          evict-old-files: 1d
+
+      - name: Dependencies
+        id: depends
+        run: |
+          sudo apt-get update
+          sudo apt-get install build-essential libcurl4-openssl-dev
+
+      - name: Test
+        id: ggml-ci
+        run: |
+          LLAMA_ARG_THREADS=$(nproc) GG_BUILD_NO_BF16=1 GG_BUILD_EXTRA_TESTS_0=1 bash ./ci/run.sh ./tmp/results ./tmp/mnt
+
   ggml-ci-x64-nvidia-cuda:
     runs-on: [self-hosted, Linux, X64, NVIDIA]
 
diff --git a/ci/run.sh b/ci/run.sh
index 747b2cc401e4e..68cbfdf2f52aa 100755
--- a/ci/run.sh
+++ b/ci/run.sh
@@ -111,6 +111,7 @@ if [ ! -z ${GG_BUILD_MUSA} ]; then
 fi
 
 if [ ! -z ${GG_BUILD_NO_SVE} ]; then
+    # arm 9 and newer enables sve by default, adjust these flags depending on the cpu used
     CMAKE_EXTRA="${CMAKE_EXTRA} -DGGML_NATIVE=OFF -DGGML_CPU_ARM_ARCH=armv8.5-a+fp16+i8mm"
 fi
 ## helpers