Optimize RWKV6 Operator Naming and Implement Multi-core CPU/ SYCL Acceleration (#10133)

* rwkv6: rename to wkv6 * rwkv6: support avx2 avx512 armv8 armv9 * rwkv6: update cuda file name * rwkv6: rename params * wkv on sycl * sycl: add some ops * sycl: Enhance OP support judgment * wkv6: drop armv9 and tranfer to GGML style ggml-ci * sync : ggml * update the function to use appropriate types * fix define error * Update ggml/src/ggml-cpu.c * add appropriate asserts * move element-wise functions outside * put the declaration outside the loop * rewrite to be more inline with the common pattern for distributing threads * use recommended way GGML_TENSOR_LOCALS --------- Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: Diego Devesa <slarengh@gmail.com> Co-authored-by: Plamen Minev <pacominev@gmail.com> Co-authored-by: Yuri Khrustalev <ykhrustalev@users.noreply.github.com> Co-authored-by: Meng, Hengyu <airdldl@163.com>
2024-11-07 18:19:10 +11:00 · 2024-11-07 18:19:10 +11:00 · 3bcd40b3c5
commit 3bcd40b3c5
parent 5c333e0140
22 changed files with 1977 additions and 1027 deletions
--- a/ggml/src/ggml-sycl/outprod.cpp
+++ b/ggml/src/ggml-sycl/outprod.cpp
@ -0,0 +1,55 @@
+#include <sycl/sycl.hpp>
+#include "outprod.hpp"
+
+
+void ggml_sycl_op_out_prod(ggml_backend_sycl_context& ctx, const ggml_tensor* src0,
+    const ggml_tensor* src1, ggml_tensor* dst) {
+
+
+    GGML_ASSERT(src0->type == GGML_TYPE_F32);
+    GGML_ASSERT(src1->type == GGML_TYPE_F32);
+    GGML_ASSERT(dst->type == GGML_TYPE_F32);
+    GGML_ASSERT(ggml_is_contiguous(src0));
+    GGML_ASSERT(ggml_is_contiguous(dst));
+
+    GGML_TENSOR_BINARY_OP_LOCALS
+
+    // Get SYCL queue
+    dpct::queue_ptr stream = ctx.stream();
+
+    // Dimension checks
+    GGML_ASSERT(ne01 == ne11);  // Inner dimensions must match
+    GGML_ASSERT(ne0 == ne00);   // Output rows match src0 rows
+    GGML_ASSERT(ne1 == ne10);   // Output cols match src1 cols
+
+    // Get data pointers
+    const float* src0_d = (const float*)src0->data;
+    const float* src1_d = (const float*)src1->data;
+    float* dst_d = (float*)dst->data;
+
+    // GEMM parameters
+    const float alpha = 1.0f;
+    const float beta = 0.0f;
+
+    // Handle transposition of src1
+    const bool src1_T = ggml_is_transposed(src1);
+    const oneapi::mkl::transpose src1_op =
+        src1_T ? oneapi::mkl::transpose::nontrans : oneapi::mkl::transpose::trans;
+    const int64_t ldb = (src1_T ? nb10 : nb11) / sizeof(float);
+
+    try {
+        // Perform matrix multiplication using oneMKL GEMM
+        oneapi::mkl::blas::gemm(*stream,
+            oneapi::mkl::transpose::nontrans, src1_op,
+            ne0, ne1, ne01,
+            alpha,
+            src0_d, ne00,
+            src1_d, ldb,
+            beta,
+            dst_d, ne0);
+    }
+    catch (sycl::exception const& exc) {
+        std::cerr << exc.what() << std::endl;
+        GGML_ASSERT(false);
+    }
+}