Added a cudaMemcpy2DAsync fast path to ggml_cuda_cpy (llama/25057)

* [CUDA] Added a cudaMemcpy2DAsync fast path to ggml_cuda_cpy

Add a CUDA ggml_cpy fast path for same-type, same-shape strided copies that are just 2D pitched block copies.
When tensors are not fully contiguous but each row is contiguous, it now uses cudaMemcpy2DAsync instead of the slow element-wise scalar copy kernel.

This fixes the GDN recurrent snapshot update with -np 4, where rollback slots are separated by cache stride gaps.

* Add new tests that execute the new optimized strided copy path

* Return unsupported for strided copy in OpenVINO, as new tests are failing
This commit is contained in:
Gaurav Garg
2026-06-27 17:46:21 +05:30
committed by Georgi Gerganov
parent a4ac1c0369
commit 868367609f
2 changed files with 49 additions and 0 deletions

View File

@@ -386,6 +386,46 @@ static void ggml_cpy_f32_iq4_nl_cuda(
(cx, cdst, ne, ne00, ne01, ne02, nb00, nb01, nb02, nb03, ne10, ne11, ne12, nb10, nb11, nb12, nb13);
}
// check if a same-type copy reduces to a 2D strided copy (height rows of width
// contiguous bytes), so it can use cudaMemcpy2DAsync instead of the scalar kernel
static bool ggml_cuda_cpy_as_memcpy_2d(const ggml_tensor * src0, const ggml_tensor * src1,
size_t & width, size_t & height, size_t & spitch, size_t & dpitch) {
// require matching shape: a reshaped copy maps elements by flat order, which the
// prefix walk below does not handle
if (src0->type != src1->type || !ggml_are_same_shape(src0, src1)) {
return false;
}
// grow the contiguous prefix block shared by both tensors
size_t block_nb = ggml_element_size(src0);
int d = 0;
for (; d < GGML_MAX_DIMS; ++d) {
if (src0->nb[d] != block_nb || src1->nb[d] != block_nb) {
break;
}
block_nb *= src0->ne[d];
}
// d == 0: nothing contiguous; d == GGML_MAX_DIMS: fully contiguous (handled by memcpy)
if (d == 0 || d == GGML_MAX_DIMS) {
return false;
}
// dim d carries the rows; everything above it must be a single element
for (int i = d + 1; i < GGML_MAX_DIMS; ++i) {
if (src0->ne[i] != 1) {
return false;
}
}
width = block_nb;
height = src0->ne[d];
spitch = src0->nb[d];
dpitch = src1->nb[d];
return spitch >= width && dpitch >= width;
}
void ggml_cuda_cpy(ggml_backend_cuda_context & ctx, const ggml_tensor * src0, ggml_tensor * src1) {
const int64_t ne = ggml_nelements(src0);
GGML_ASSERT(ne == ggml_nelements(src1));
@@ -421,6 +461,8 @@ void ggml_cuda_cpy(ggml_backend_cuda_context & ctx, const ggml_tensor * src0, gg
const bool can_be_transposed = nb01 == (int64_t)ggml_element_size(src0) &&
src0->ne[3] == 1 && nb02 == ne00 * ne01 * (int64_t)ggml_element_size(src0);
size_t mc_width = 0, mc_height = 0, mc_spitch = 0, mc_dpitch = 0;
if (src0->type == src1->type && contiguous_srcs) {
GGML_ASSERT(ggml_nbytes(src0) == ggml_nbytes(src1));
#if defined(GGML_USE_MUSA) && defined(GGML_MUSA_MUDNN_COPY)
@@ -431,6 +473,9 @@ void ggml_cuda_cpy(ggml_backend_cuda_context & ctx, const ggml_tensor * src0, gg
{
CUDA_CHECK(cudaMemcpyAsync(src1_ddc, src0_ddc, ggml_nbytes(src0), cudaMemcpyDeviceToDevice, main_stream));
}
} else if (ggml_cuda_cpy_as_memcpy_2d(src0, src1, mc_width, mc_height, mc_spitch, mc_dpitch)) {
CUDA_CHECK(cudaMemcpy2DAsync(src1_ddc, mc_dpitch, src0_ddc, mc_spitch,
mc_width, mc_height, cudaMemcpyDeviceToDevice, main_stream));
} else if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_F32) {
if (can_be_transposed) {
ggml_cpy_scalar_cuda<float, float, true>

View File

@@ -1053,6 +1053,10 @@ static bool is_op_unsupported_case(const ggml_tensor * op) {
(op->ne[0] == 2 && op->ne[1] == 4 && op->ne[2] == 3 && op->ne[3] == 2)) {
return true;
}
// CPY into a strided view of a larger buffer (recurrent-state snapshots) not supported
if (op->view_src && ggml_nbytes(op) != ggml_nbytes(op->view_src)) {
return true;
}
break;
}
case GGML_OP_MUL_MAT: {