Commit b41b5925 for whisper.cpp
commit b41b59253ede8d4c71a0069a3e35383fe4ce4fc8
Author: Neo Zhang <zhang.jianyu@outlook.com>
Date: Thu Nov 12 14:54:58 2026 +0800
sycl : fix AOT for more than 4GB (llama/31873)
* fix aot for >4g garbage output issue; add --aot in build script
* correct the aot flag
* support no-xmx build, update the code for no-xmx building, rename nossl to no-ssl
diff --git a/ggml/src/ggml-sycl/CMakeLists.txt b/ggml/src/ggml-sycl/CMakeLists.txt
index 35d36e58..8312a830 100644
--- a/ggml/src/ggml-sycl/CMakeLists.txt
+++ b/ggml/src/ggml-sycl/CMakeLists.txt
@@ -221,7 +221,7 @@ if (GGML_SYCL_TARGET STREQUAL "INTEL")
if (NOT GGML_SYCL_DEVICE_ARCH)
target_link_options(ggml-sycl PRIVATE -Xs -ze-intel-greater-than-4GB-buffer-required)
else()
- message(STATUS "Skipping -ze-intel-greater-than-4GB-buffer-required for spir64_gen AOT")
+ message(STATUS "spir64_gen AOT: -ze-intel-greater-than-4GB-buffer-required passed to ocloc")
endif()
# Link against Intel oneMKL
@@ -251,7 +251,7 @@ if (GGML_SYCL_DEVICE_ARCH)
target_compile_options(
ggml-sycl PRIVATE
-fsycl-targets=spir64_gen
- "SHELL:-Xsycl-target-backend=spir64_gen \"-device ${GGML_SYCL_DEVICE_ARCH}\""
+ "SHELL:-Xsycl-target-backend=spir64_gen \"-device ${GGML_SYCL_DEVICE_ARCH} -options -ze-intel-greater-than-4GB-buffer-required\""
)
# Pass through parallel job (process) count for parallelising the
@@ -266,7 +266,7 @@ if (GGML_SYCL_DEVICE_ARCH)
target_link_options(
ggml-sycl PRIVATE
-fsycl-targets=spir64_gen
- "SHELL:-Xsycl-target-backend=spir64_gen \"-device ${GGML_SYCL_DEVICE_ARCH}\""
+ "SHELL:-Xsycl-target-backend=spir64_gen \"-device ${GGML_SYCL_DEVICE_ARCH} -options -ze-intel-greater-than-4GB-buffer-required\""
-fsycl-max-parallel-link-jobs=${GGML_SYCL_MAX_PARALLEL_LINK_JOBS}
)
diff --git a/ggml/src/ggml-sycl/common.hpp b/ggml/src/ggml-sycl/common.hpp
index 99da9713..f46e7978 100644
--- a/ggml/src/ggml-sycl/common.hpp
+++ b/ggml/src/ggml-sycl/common.hpp
@@ -237,9 +237,9 @@ typedef sycl::float2 dfloat2;
#define MMVQ_MAX_BATCH_SIZE 8
-#if defined(__INTEL_LLVM_COMPILER) && __has_include(<sycl/ext/intel/esimd/xmx/dpas.hpp>)
+#if defined(GGML_SYCL_SUPPORT_XMX) && defined(__INTEL_LLVM_COMPILER) && __has_include(<sycl/ext/intel/esimd/xmx/dpas.hpp>)
# define GGML_SYCL_MMVQ_HAS_XMX
-#endif // __INTEL_LLVM_COMPILER
+#endif // GGML_SYCL_SUPPORT_XMX && __INTEL_LLVM_COMPILER
// default of the most columns the XMX mul_mat_vec_q handles (GGML_SYCL_XMX_MAX_COLS), wider batches use other kernels
#define GGML_SYCL_XMX_MAX_COLS_DEFAULT 80
diff --git a/ggml/src/ggml-sycl/esimd.hpp b/ggml/src/ggml-sycl/esimd.hpp
index 85cc93e2..568173e7 100644
--- a/ggml/src/ggml-sycl/esimd.hpp
+++ b/ggml/src/ggml-sycl/esimd.hpp
@@ -2,11 +2,14 @@
#define GGML_SYCL_ESIMD_HPP
#include <sycl/ext/intel/esimd.hpp>
-#include <sycl/ext/intel/esimd/xmx/dpas.hpp>
#include <sycl/ext/intel/experimental/grf_size_properties.hpp>
#include "common.hpp"
+#ifdef GGML_SYCL_MMVQ_HAS_XMX
+#include <sycl/ext/intel/esimd/xmx/dpas.hpp>
+#endif
+
namespace ggml_sycl_esimd {
constexpr int GGML_SYCL_DMMV_ESIMD_WG_SIZE = 4;
@@ -586,6 +589,7 @@ template <> struct esimd_reorder_q_traits<GGML_TYPE_Q6_K> {
}
};
+
//
// Multi-column mul_mat_vec for reordered K-quant weights on the XMX engines (int8 DPAS).
// A thread owns 16 weight rows (B operand, one dword is 4 weights of a row), the q8_1 activation
@@ -1024,6 +1028,7 @@ template <typename T, int NC, bool FUSED> constexpr int xmx_grf() {
return FUSED || NC >= T::nc_wide ? 256 : 128;
}
+#ifdef GGML_SYCL_MMVQ_HAS_XMX
// One DPAS of a block of RC columns: the int32 products of the weights B and the activations A, as float and scaled
// per weight row. Every quant type sums one or two of these per group of 32.
template <int RC, int NT>
@@ -1394,6 +1399,8 @@ static void xmx_mul_mat_ncols(const void * vx,
}
}
+#endif // GGML_SYCL_MMVQ_HAS_XMX
+
} // namespace ggml_sycl_esimd
#endif // GGML_SYCL_ESIMD_HPP