ggml : fix row condition for i8mm kernels (llama/10561)

ggml-ci

diff --git a/ggml/src/ggml-cpu/ggml-cpu-quants.c b/ggml/src/ggml-cpu/ggml-cpu-quants.c
index f0e276b69879550b0d5028b3e4f1cd4f6ea3ba9a..11e8df253d5caca18bd4fd1815c6af16ecc5d601 100644 (file)
--- a/ggml/src/ggml-cpu/ggml-cpu-quants.c
+++ b/ggml/src/ggml-cpu/ggml-cpu-quants.c
@@ -1813,11 +1813,13 @@ void ggml_vec_dot_q4_0_q8_0(int n, float * restrict s, size_t bs, const void * r
             sumv0 = vmlaq_f32(sumv0,(vcvtq_f32_s32(vmmlaq_s32((vmmlaq_s32((vmmlaq_s32((vmmlaq_s32(vdupq_n_s32(0), l0, r0)),
                                                                                 l1, r1)), l2, r2)), l3, r3))), scale);
         }
-        float32x4_t sumv1 = vextq_f32(sumv0, sumv0, 2);
+
+        float32x4_t sumv1 = vextq_f32 (sumv0, sumv0, 2);
         float32x4_t sumv2 = vzip1q_f32(sumv0, sumv1);
 
-        vst1_f32(s,      vget_low_f32(sumv2));
+        vst1_f32(s,      vget_low_f32 (sumv2));
         vst1_f32(s + bs, vget_high_f32(sumv2));
+
         return;
     }
 #endif

diff --git a/ggml/src/ggml-cpu/ggml-cpu.c b/ggml/src/ggml-cpu/ggml-cpu.c
index fea867440424e869c55af8d98188f75f4a741ef6..1c88e5d81ab6ccc97208edf3d8d1c56e2db58311 100644 (file)
--- a/ggml/src/ggml-cpu/ggml-cpu.c
+++ b/ggml/src/ggml-cpu/ggml-cpu.c
@@ -7576,14 +7576,6 @@ UseGgmlGemm2:;
     // This is the size of the rest of the dimensions of the result
     const int64_t nr1 = ne1 * ne2 * ne3;
 
-    // dot kernels can handle 1 row and col at a time, but mmla kernels can process 2 rows and cols
-    int64_t num_rows_per_vec_dot = vec_dot_num_rows;
-    // TODO: currently the mmla kernels support only even numbered rows/cols.
-    // this check can be removed once they are extended to support odd numbered rows/cols too
-    if ((nr0 % 2 != 0) || (ne11 % 2 != 0)) {
-        num_rows_per_vec_dot = 1;
-    }
-
     // Now select a reasonable chunk size.
     int chunk_size = 16;
 
@@ -7646,6 +7638,15 @@ UseGgmlGemm2:;
         const int64_t ir1_start = dr1 * ith1;
         const int64_t ir1_end = MIN(ir1_start + dr1, nr1);
 
+        // dot kernels can handle 1 row and col at a time, but mmla kernels can process 2 rows and cols
+        int64_t num_rows_per_vec_dot = vec_dot_num_rows;
+
+        // TODO: currently the mmla kernels support only even numbered rows/cols.
+        // this check can be removed once they are extended to support odd numbered rows/cols too
+        if ((nr0 % 2 != 0) || (ne11 % 2 != 0) || ((ir0_end - ir0_start) % 2 != 0) || ((ir1_end - ir1_start) % 2 != 0)) {
+            num_rows_per_vec_dot = 1;
+        }
+
         ggml_compute_forward_mul_mat_one_chunk(params, dst, type, num_rows_per_vec_dot, ir0_start, ir0_end, ir1_start, ir1_end);
 
         if (nth >= nchunk0 * nchunk1) {