Distribute and run LLMs with a single file.
C++ cross-platformggufllama-cpplocal-ailocal-inferencelocal-llmopen-source-aisingle-file-executablespeech-to-text
Fix stack spill caused by recent BLAS refactoring
Justine Tunneyaf7aa5df673a
diff --git a/llamafile/BUILD.mk b/llamafile/BUILD.mkindex 8e136974..a7482051 100644--- a/llamafile/BUILD.mk+++ b/llamafile/BUILD.mk@@ -45,10 +45,10 @@ o/$(MODE)/llamafile: \o/$(MODE)/llamafile/sgemm.o: private CXXFLAGS += -Oso/$(MODE)/llamafile/tinyblas_cpu_amd_fma.o: private TARGET_ARCH += -Xx86_64-mtune=bdver2 -Xx86_64-mfma-o/$(MODE)/llamafile/tinyblas_cpu_amd_avx2.o: private TARGET_ARCH += -Xx86_64-mtune=skylake -Xx86_64-mavx2 -Xx86_64-mfma-o/$(MODE)/llamafile/tinyblas_cpu_amd_avxvnni.o: private TARGET_ARCH += -Xx86_64-mtune=alderlake -Xx86_64-mavx2 -Xx86_64-mfma -Xx86_64-mavxvnni-o/$(MODE)/llamafile/tinyblas_cpu_amd_avx512f.o: private TARGET_ARCH += -Xx86_64-mtune=cannonlake -Xx86_64-mavx512f-o/$(MODE)/llamafile/tinyblas_cpu_amd_zen4.o: private TARGET_ARCH += -Xx86_64-mtune=znver4 -Xx86_64-mavx512vl -Xx86_64-mavx512vnni -Xx86_64-mavx512bf16+o/$(MODE)/llamafile/tinyblas_cpu_amd_avx2.o: private TARGET_ARCH += -Xx86_64-mtune=skylake -Xx86_64-mfma -Xx86_64-mavx2 -Xx86_64-mfma+o/$(MODE)/llamafile/tinyblas_cpu_amd_avxvnni.o: private TARGET_ARCH += -Xx86_64-mtune=alderlake -Xx86_64-mfma -Xx86_64-mavx2 -Xx86_64-mavxvnni+o/$(MODE)/llamafile/tinyblas_cpu_amd_avx512f.o: private TARGET_ARCH += -Xx86_64-mtune=cannonlake -Xx86_64-mfma -Xx86_64-mavx2 -Xx86_64-mavx512f+o/$(MODE)/llamafile/tinyblas_cpu_amd_zen4.o: private TARGET_ARCH += -Xx86_64-mtune=znver4 -Xx86_64-mfma -Xx86_64-mavx2 -Xx86_64-mavx512f -Xx86_64-mavx512vl -Xx86_64-mavx512vnni -Xx86_64-mavx512bf16o/$(MODE)/llamafile/tinyblas_cpu_arm82.o: private TARGET_ARCH += -Xaarch64-march=armv8.2-a+dotprod+fp16################################################################################