ggml : require contiguous src for ROLL on CUDA and Metal (llama/25928)

ggml_roll only asserts nb[0] == ggml_type_size, so a permuted src is a
valid input, but the CUDA and Metal roll kernels index by ne alone and
never read the nb strides. A non-contiguous src therefore produced
silently wrong results. Neither backend declared a contiguity
requirement in supports_op, so the scheduler did not fall back to the
CPU implementation, which does handle strides correctly.

Add the requirement to both backends, matching the existing
GGML_OP_ROPE guard, and add a permuted test_roll case.

diff --git a/ggml/src/ggml-cuda/ggml-cuda.cu b/ggml/src/ggml-cuda/ggml-cuda.cu
index dec61932407097b7c82994b5e47bbab00cb6f60f..05e8d7f7368471a0119f6f691191fce3d238464f 100644 (file)
--- a/ggml/src/ggml-cuda/ggml-cuda.cu
+++ b/ggml/src/ggml-cuda/ggml-cuda.cu
@@ -5185,7 +5185,7 @@ static bool ggml_backend_cuda_device_supports_op(ggml_backend_dev_t dev, const g
             return max_bias == 0.0f;
         }
         case GGML_OP_ROLL:
-            if(op->src[0]->type == GGML_TYPE_F32) {
+            if(op->src[0]->type == GGML_TYPE_F32 && ggml_is_contiguous(op->src[0])) {
                 return true;
             }
             return false;

diff --git a/ggml/src/ggml-metal/ggml-metal-device.m b/ggml/src/ggml-metal/ggml-metal-device.m
index 2dc6eb8fdbc01c6736ab323f0472d5a659f6ae07..df85ab02cf54e328341353f1d9138deef0d60bd4 100644 (file)
--- a/ggml/src/ggml-metal/ggml-metal-device.m
+++ b/ggml/src/ggml-metal/ggml-metal-device.m
@@ -1268,8 +1268,9 @@ bool ggml_metal_device_supports_op(ggml_metal_device_t dev, const struct ggml_te
         case GGML_OP_ARGSORT:
         case GGML_OP_TOP_K:
         case GGML_OP_ARANGE:
-        case GGML_OP_ROLL:
             return true;
+        case GGML_OP_ROLL:
+            return ggml_is_contiguous(op->src[0]);
         case GGML_OP_FLASH_ATTN_EXT:
             // for new head sizes, add checks here
             if (op->src[0]->ne[0] != 32 &&