From 5c84b256451e9873e6fac50fd18572217590e315 Mon Sep 17 00:00:00 2001
From: Oleksandr Kuvshynov <661042+okuvshynov@users.noreply.github.com>
Date: Sun, 19 Jul 2026 16:06:49 -0400
Subject: [PATCH] fix: old x86 Mac scalar -> vnni

Older macs still need -march, not -mcpu.

Before this fix, no ymm (or zmm, vnni, etc.).

Standalone repro:

```
% clang -O3  -mcpu=native glm.c -o /tmp/glm
clang: error: unsupported option '-mcpu=' for target 'x86_64-apple-darwin25.4.0'
% clang -O3  -mcpu=native glm.c -o /tmp/glm -lm
% otool -tv /tmp/glm | grep -c ymm
0
```

I'm not entirely sure why `-lm` would ignore `-mcpu=` rather than error, but either way we don't get vector instructions:

colibri itself:

```
% ulimit -l unlimited; OMP_NUM_THREADS=16 RAM_GB=700 PIN=auto PIN_GB=all PIN_FILL=1 ./coli run --ram 700 --model ~/projects/llms/glm5p2-i4 --ngen 256 "Explain the difference between concurrency and parallelism"
...
== GLM C engine (glm_moe_dsa), cache=8 experts/layer | experts@8-bit dense@8-bit | idot: scalar ==
```

After the fix, we'll get it:

```
% clang -O3  -march=native glm.c -o /tmp/glm -lm
% otool -tv /tmp/glm | grep -c ymm
2621
```

colibri run:
```
% ulimit -l unlimited; OMP_NUM_THREADS=16 RAM_GB=700 PIN=auto PIN_GB=all PIN_FILL=1 ./coli run --ram 700 --model ~/projects/llms/glm5p2-i4 --ngen 256 "Explain the difference between concurrency and parallelism"
...
== GLM C engine (glm_moe_dsa), cache=8 experts/layer | experts@8-bit dense@8-bit | idot: avx512-vnni ==
```
---
 c/Makefile | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/c/Makefile b/c/Makefile
index af17357..0341c0a 100644
--- a/c/Makefile
+++ b/c/Makefile
@@ -58,9 +58,14 @@ CFLAGS  = -O3 $(OMPC) -Wall -Wextra -Wno-unused-parameter -Wno-misleading-indent
 # Opt-in: ARCH=native appends -mcpu=native (arm64 clang uses -mcpu, not -march),
 # which unlocks the i8mm SMMLA int8/int4 dot kernels in colibri.c. ARCH unset ->
 # no -mcpu, default build byte-identical. Apple clang knows apple-m4 / native.
+# For older X86_64 Macs (for example, Mac Pro 2019) we need to use -march
 ifneq ($(ARCH),)
+ifneq (,$(X86_64))
+CFLAGS += -march=$(ARCH)
+else
 CFLAGS += -mcpu=$(ARCH)
 endif
+endif
 LDFLAGS = -lm $(OMPL)
 EXE     =
 else ifneq ($(IS_WIN),)
@@ -273,8 +278,9 @@ olmoe$(EXE): olmoe.c st.h json.h compat.h
 # Use a baseline that matches the compiler target. macOS already targets a
 # portable baseline when ARCH is empty; forcing the x86 value there breaks
 # Apple Silicon. Unknown targets use native rather than an invalid x86 flag.
+# Intel Macs need -march for vector instructions
 ifneq (,$(DARWIN))
-PORTABLE_ARCH =
+PORTABLE_ARCH = $(if $(X86_64),x86-64-v3,)
 else ifneq (,$(AARCH64))
 PORTABLE_ARCH = armv8-a
 else ifneq (,$(PPC64))