Make it work after rebase

Mozilla-Ocho · May 10, 2024 · a3bca82 · a3bca82
1 parent 897be80
commit a3bca82
Show file tree

Hide file tree

Showing 2 changed files with 6 additions and 7 deletions.
diff --git a/llamafile/BUILD.mk b/llamafile/BUILD.mk
@@ -89,7 +89,7 @@ o/$(MODE)/llamafile:					\
 #
 
 o/$(MODE)/llamafile/sgemm.o: private CXXFLAGS += -Os
-o/$(MODE)/llamafile/iqk_mul_mat.o: private TARGET_ARCH += -Xx86_64-mtune=skylake -Xx86_64-mavx2 -Xx86_64-mfma
+o/$(MODE)/llamafile/iqk_mul_mat.o: private TARGET_ARCH += -Xx86_64-mtune=skylake -Xx86_64-mavx2 -Xx86_64-mfma -Xx86_64-mf16c
 o/$(MODE)/llamafile/tinyblas_cpu_sgemm_amd_avx.o: private TARGET_ARCH += -Xx86_64-mtune=sandybridge -Xx86_64-mf16c
 o/$(MODE)/llamafile/tinyblas_cpu_mixmul_amd_avx.o: private TARGET_ARCH += -Xx86_64-mtune=sandybridge -Xx86_64-mf16c
 o/$(MODE)/llamafile/tinyblas_cpu_sgemm_amd_fma.o: private TARGET_ARCH += -Xx86_64-mtune=bdver2 -Xx86_64-mf16c -Xx86_64-mfma

diff --git a/llamafile/tinyblas_cpu_sgemm.inc b/llamafile/tinyblas_cpu_sgemm.inc
@@ -320,12 +320,11 @@ bool llamafile_sgemm(long m, long n, long k, const void *A, long lda, const void
                 return true;
             }
         }
-    }
-#endif
-    if ((Btype == GGML_TYPE_Q8_0 || Btype == GGML_TYPE_Q8_1) && Ctype == GGML_TYPE_F32) {
-        assert (QK8_0 == QK8_1 == QK4_0 == QK4_1 == QK5_0 == QK5_1 == 32);
-        if (iqk_mul_mat(m, n, k*QK8_0, Atype, A, B, (float *)C, ldc, ith, nth)) {
-            return true;
+        if ((Btype == GGML_TYPE_Q8_0 || Btype == GGML_TYPE_Q8_1) && Ctype == GGML_TYPE_F32) {
+            assert (QK8_0 == QK8_1 == QK4_0 == QK4_1 == QK5_0 == QK5_1 == 32);
+            if (iqk_mul_mat(m, n, k * QK8_0, Atype, A, B, (float *)C, ldc, ith, nth)) {
+                return true;
+            }
         }
     }
 #endif