From 7689aded24f76cb02e7bb1c34ad5110a135a9e4b Mon Sep 17 00:00:00 2001
From: Shalini Salomi Bodapati <Shalini.Salomi.Bodapati@ibm.com>
Date: Thu, 11 Dec 2025 07:37:33 -0600
Subject: [PATCH] ggml-cpu: Enbale Matrix Math Accelerator for Power10

Adding -mcpu=power10 improves matrix multiplication performance when
running Ollama on PowerPC-based hardware. -mcpu=power10 needs to be
added in llamafile.go, so that powerpc optimized code(using Matrix
Multiply Assist) for llamafile_sgemm is enabled and is available in
ollama binary.

This changes adds -mcpu=power10 flag when built with build tag
ppc64le.power10 and this enables mma optimizations in ollama binary

-mcpu=power9 flag is added when built with build tag ppc64le.power9 and
this enables vsx optimizations in ollama binary.

When building on power10 machine use
go build --tags ppc64le.power10 .

When building on power9 machine use
go build --tags ppc64le.power9 .

Performance Impact:

Improved performance on Power10 Chips for Q4_0,Q8_0,FP32,BF16 Models.
Inference time with ollama run llama3:8b ( Q4_0 Model) ( ~ 30% less time
for a 50 word summarization of a prompt with 512 tokens.
with MMA enabled : 6.05 sec
without MMA (Base) : 8.45 sec

Signed-off-by: Shalini Salomi Bodapati <Shalini.Salomi.Bodapati@ibm.com>
---
 .../src/ggml-cpu/llamafile/llamafile_ppc64le_power10.go    | 7 +++++++
 .../src/ggml-cpu/llamafile/llamafile_ppc64le_power9.go     | 7 +++++++
 2 files changed, 14 insertions(+)
 create mode 100644 ml/backend/ggml/ggml/src/ggml-cpu/llamafile/llamafile_ppc64le_power10.go
 create mode 100644 ml/backend/ggml/ggml/src/ggml-cpu/llamafile/llamafile_ppc64le_power9.go

diff --git a/ml/backend/ggml/ggml/src/ggml-cpu/llamafile/llamafile_ppc64le_power10.go b/ml/backend/ggml/ggml/src/ggml-cpu/llamafile/llamafile_ppc64le_power10.go
new file mode 100644
index 000000000..0aaec68b2
--- /dev/null
+++ b/ml/backend/ggml/ggml/src/ggml-cpu/llamafile/llamafile_ppc64le_power10.go
@@ -0,0 +1,7 @@
+// +build ppc64le.power10
+
+package llamafile
+
+// #cgo CXXFLAGS: -std=c++17 -mcpu=power10
+// #cgo CPPFLAGS: -I${SRCDIR}/.. -I${SRCDIR}/../.. -I${SRCDIR}/../../../include
+import "C"
diff --git a/ml/backend/ggml/ggml/src/ggml-cpu/llamafile/llamafile_ppc64le_power9.go b/ml/backend/ggml/ggml/src/ggml-cpu/llamafile/llamafile_ppc64le_power9.go
new file mode 100644
index 000000000..6bdd69780
--- /dev/null
+++ b/ml/backend/ggml/ggml/src/ggml-cpu/llamafile/llamafile_ppc64le_power9.go
@@ -0,0 +1,7 @@
+// +build ppc64le.power9
+
+package llamafile
+
+// #cgo CXXFLAGS: -std=c++17 -mcpu=power9
+// #cgo CPPFLAGS: -I${SRCDIR}/.. -I${SRCDIR}/../.. -I${SRCDIR}/../../../include
+import "C"