diff --git a/ggml/src/ggml-sycl/dpct/helper.hpp b/ggml/src/ggml-sycl/dpct/helper.hpp index ef4609e32..41839ae84 100644 --- a/ggml/src/ggml-sycl/dpct/helper.hpp +++ b/ggml/src/ggml-sycl/dpct/helper.hpp @@ -1834,6 +1834,20 @@ namespace dpct template inline auto dp4a(T1 a, T2 b, T3 c) { +#if defined(__SYCL_DEVICE_ONLY__) && defined(__NVPTX__) && \ + defined(__SYCL_CUDA_ARCH__) && __SYCL_CUDA_ARCH__ >= 610 + dot_product_acc_t res; + if constexpr (std::is_same_v, uint32_t>) { + asm volatile("dp4a.u32.u32 %0, %1, %2, %3;" + : "=r"(res) + : "r"(a), "r"(b), "r"(c)); + } else { + asm volatile("dp4a.s32.s32 %0, %1, %2, %3;" + : "=r"(res) + : "r"(a), "r"(b), "r"(c)); + } + return res; +#else dot_product_acc_t res = c; auto va = extract_and_sign_or_zero_extend4(a); auto vb = extract_and_sign_or_zero_extend4(b); @@ -1842,6 +1856,7 @@ namespace dpct res += va[2] * vb[2]; res += va[3] * vb[3]; return res; +#endif } struct sub_sat