whispercpp 1.3.4 → 1.3.6

data/ext/sources/ggml/src/ggml-cuda/vecdotq.cuh

@@ -94,6 +94,15 @@ static __device__ __forceinline__ int2 get_int_from_table_16(const int & q4, con
 #endif
 }
 
+static __device__ __forceinline__ uint32_t unpack_ksigns(const uint8_t v) {
+    // v is a 7 bit int, with the 8th sign being encodable as popcnt
+    // with xor we can "correct" the bit instead of having to mask
+    const uint32_t p = __popc(v) & 1;
+    const uint32_t s = v ^ p << 7;
+    // broadcast over uint to allow for 0x08040201 / 0x80402010 as selectors
+    return s * 0x01010101;
+}
+
 // VDR = vec dot ratio, how many contiguous integers each thread processes when the vec dot kernel is called
 // MMVQ = mul_mat_vec_q, MMQ = mul_mat_q
 
@@ -905,22 +914,22 @@ static __device__ __forceinline__ float vec_dot_iq2_xxs_q8_1(
     int sumi = 0;
 #pragma unroll
     for (int k0 = 0; k0 < 8; k0 += 2) {
-        const int * grid_pos = (const int *) (iq2xxs_grid + aux8[k0/2]);
-        const int signs_packed = ksigns_iq2xs[(aux32 >> (7*k0/2)) & 0x7F];
+        const uint2 grid_pos = ((const uint2*)iq2xxs_grid)[aux8[k0/2]];
+        const uint32_t signs = unpack_ksigns(aux32 >> (7 * k0 / 2));
 
-        const int signs0 = __vcmpne4(((signs_packed & 0x03) << 7) | ((signs_packed & 0x0C) << 21), 0x00000000);
-        const int grid0 = __vsub4(grid_pos[0] ^ signs0, signs0);
+        const int signs0 = __vcmpne4(signs & 0x08040201, 0);
+        const int grid0 = __vsub4(grid_pos.x ^ signs0, signs0);
         const int u0 = get_int_b4(bq8_1[iqs/2].qs, k0 + 0);
         sumi = ggml_cuda_dp4a(grid0, u0, sumi);
 
-        const int signs1 = __vcmpne4(((signs_packed & 0x30) << 3) | ((signs_packed & 0xC0) << 17), 0x00000000);
-        const int grid1 = __vsub4(grid_pos[1] ^ signs1, signs1);
+        const int signs1 = __vcmpne4(signs & 0x80402010, 0);
+        const int grid1 = __vsub4(grid_pos.y ^ signs1, signs1);
         const int u1 = get_int_b4(bq8_1[iqs/2].qs, k0 + 1);
         sumi = ggml_cuda_dp4a(grid1, u1, sumi);
     }
 
-    const int ls = aux32 >> 28;
-    sumi = (ls*sumi + sumi/2)/4;
+    const int ls = aux32 >> 27 | 1; // (scale * 2 + 1)
+    sumi = sumi * ls / 8;           // (sumi * scale + sumi / 2) / 4
     const float d = __half2float(bq2->d) * __low2float(bq8_1[iqs/2].ds);
     return d * sumi;
 }
@@ -942,13 +951,15 @@ static __device__ __forceinline__ float vec_dot_iq2_xs_q8_1(
     int sumi1 = 0;
 #pragma unroll
     for (int l0 = 0; l0 < 8; l0 += 2) {
-        const uint32_t * grid_pos = (const uint32_t *)(iq2xs_grid + (q2[l0/2] & 0x000001FF));
-        const uint32_t * signs    = (const uint32_t *)(ksigns64   + (q2[l0/2] >> 9));
-
-        const int grid_l = __vsub4(grid_pos[0] ^ signs[0], signs[0]);
-        const int grid_h = __vsub4(grid_pos[1] ^ signs[1], signs[1]);
+        const uint2 grid_pos = ((const uint2*)iq2xs_grid)[q2[l0/2] & 0x1FF];
+        const uint32_t signs = unpack_ksigns(q2[l0/2] >> 9);
 
+        const int signs0 = __vcmpne4(signs & 0x08040201, 0);
+        const int grid_l = __vsub4(grid_pos.x ^ signs0, signs0);
         const int u0 = get_int_b4(bq8_1[iqs/2].qs, l0 + 0);
+
+        const int signs1 = __vcmpne4(signs & 0x80402010, 0);
+        const int grid_h = __vsub4(grid_pos.y ^ signs1, signs1);
         const int u1 = get_int_b4(bq8_1[iqs/2].qs, l0 + 1);
 
         if (l0 < 4) {
@@ -1028,13 +1039,16 @@ static __device__ __forceinline__ float vec_dot_iq3_xxs_q8_1(
 #pragma unroll
     for (int l0 = 0; l0 < 8; l0 += 2) {
         const int2 grid_pos = make_int2(iq3xxs_grid[q3[l0 + 0]], iq3xxs_grid[q3[l0 + 1]]);
+        const uint32_t signs = unpack_ksigns(aux32 >> (7*l0/2));
 
-        const int * signs = (const int *)(ksigns64 + ((aux32 >> (7*l0/2)) & 0x7F));
-
-        const int grid_l = __vsub4(grid_pos.x ^ signs[0], signs[0]);
-        const int grid_h = __vsub4(grid_pos.y ^ signs[1], signs[1]);
+        const int signs0 = __vcmpne4(signs & 0x08040201, 0);
+        const int grid_l = __vsub4(grid_pos.x ^ signs0, signs0);
 
         const int u0 = get_int_b4(bq8_1[iqs/2].qs, l0 + 0);
+
+        const int signs1 = __vcmpne4(signs & 0x80402010, 0);
+        const int grid_h = __vsub4(grid_pos.y ^ signs1, signs1);
+
         const int u1 = get_int_b4(bq8_1[iqs/2].qs, l0 + 1);
 
         sumi = ggml_cuda_dp4a(grid_l, u0, sumi);

data/ext/sources/ggml/src/ggml-cuda/vendors/cuda.h

@@ -10,6 +10,10 @@
 #include <cuda_fp8.h>
 #endif // CUDART_VERSION >= 12050
 
+#if CUDART_VERSION >= 12080
+#include <cuda_fp4.h>
+#endif // CUDART_VERSION >= 12080
+
 #if CUDART_VERSION < 11020
 #define CU_DEVICE_ATTRIBUTE_VIRTUAL_MEMORY_MANAGEMENT_SUPPORTED CU_DEVICE_ATTRIBUTE_VIRTUAL_ADDRESS_MANAGEMENT_SUPPORTED
 #define CUBLAS_TF32_TENSOR_OP_MATH CUBLAS_TENSOR_OP_MATH

data/ext/sources/ggml/src/ggml-cuda/vendors/hip.h

@@ -6,6 +6,10 @@
 #include <hip/hip_fp16.h>
 #include <hip/hip_bf16.h>
 
+#if defined(GGML_HIP_ROCWMMA_FATTN)
+#include <rocwmma/rocwmma-version.hpp>
+#endif // defined(GGML_HIP_ROCWMMA_FATTN)
+
 #define CUBLAS_GEMM_DEFAULT HIPBLAS_GEMM_DEFAULT
 #define CUBLAS_GEMM_DEFAULT_TENSOR_OP HIPBLAS_GEMM_DEFAULT
 #define CUBLAS_OP_N HIPBLAS_OP_N
@@ -15,6 +19,9 @@
 #define CUDA_R_16F  HIPBLAS_R_16F
 #define CUDA_R_16BF HIPBLAS_R_16B
 #define CUDA_R_32F  HIPBLAS_R_32F
+#define CUBLAS_SIDE_RIGHT HIPBLAS_SIDE_RIGHT
+#define CUBLAS_FILL_MODE_UPPER HIPBLAS_FILL_MODE_UPPER
+#define CUBLAS_DIAG_NON_UNIT HIPBLAS_DIAG_NON_UNIT
 #define CU_DEVICE_ATTRIBUTE_VIRTUAL_MEMORY_MANAGEMENT_SUPPORTED hipDeviceAttributeVirtualMemoryManagementSupported
 #define CU_MEM_ALLOC_GRANULARITY_RECOMMENDED hipMemAllocationGranularityRecommended
 #define CU_MEM_ALLOCATION_TYPE_PINNED hipMemAllocationTypePinned
@@ -26,6 +33,7 @@
 #define __shfl_xor_sync(mask, var, laneMask, width) __shfl_xor(var, laneMask, width)
 #define __all_sync(mask, var) __all(var)
 #define __any_sync(mask, var) __any(var)
+#define cublasStrsmBatched hipblasStrsmBatched
 #define cublasCreate hipblasCreate
 #define cublasDestroy hipblasDestroy
 #define cublasGemmEx hipblasGemmEx
@@ -37,9 +45,11 @@
 #define cublasSgemm hipblasSgemm
 #define cublasStatus_t hipblasStatus_t
 #define cublasOperation_t hipblasOperation_t
+#define cudaDevAttrCooperativeLaunch hipDeviceAttributeCooperativeLaunch
 #define cudaDeviceCanAccessPeer hipDeviceCanAccessPeer
 #define cudaDeviceDisablePeerAccess hipDeviceDisablePeerAccess
 #define cudaDeviceEnablePeerAccess hipDeviceEnablePeerAccess
+#define cudaDeviceGetAttribute hipDeviceGetAttribute
 #define cudaDeviceProp hipDeviceProp_t
 #define cudaDeviceSynchronize hipDeviceSynchronize
 #define cudaError_t hipError_t
@@ -62,6 +72,7 @@
 #define cudaHostRegisterPortable hipHostRegisterPortable
 #define cudaHostRegisterReadOnly hipHostRegisterReadOnly
 #define cudaHostUnregister hipHostUnregister
+#define cudaLaunchCooperativeKernel hipLaunchCooperativeKernel
 #define cudaLaunchHostFunc hipLaunchHostFunc
 #define cudaMalloc hipMalloc
 #define cudaMallocHost(ptr, size) hipHostMalloc(ptr, size, hipHostMallocDefault)
@@ -101,7 +112,7 @@
 #define cudaStreamNonBlocking hipStreamNonBlocking
 #define cudaStreamPerThread hipStreamPerThread
 #define cudaStreamSynchronize hipStreamSynchronize
-#define cudaStreamWaitEvent(stream, event, flags) hipStreamWaitEvent(stream, event, flags)
+#define cudaStreamWaitEvent hipStreamWaitEvent
 #define cudaGraphExec_t hipGraphExec_t
 #define cudaGraphNode_t hipGraphNode_t
 #define cudaKernelNodeParams hipKernelNodeParams
@@ -127,6 +138,8 @@
 #define cudaStream_t hipStream_t
 #define cudaSuccess hipSuccess
 #define cudaOccupancyMaxActiveBlocksPerMultiprocessor hipOccupancyMaxActiveBlocksPerMultiprocessor
+#define cudaFuncSetAttribute hipFuncSetAttribute
+#define cudaFuncAttributeMaxDynamicSharedMemorySize hipFuncAttributeMaxDynamicSharedMemorySize
 #define __trap() do { abort(); __builtin_unreachable(); } while(0)
 #define CUBLAS_STATUS_SUCCESS HIPBLAS_STATUS_SUCCESS
 #define CUBLAS_STATUS_NOT_INITIALIZED HIPBLAS_STATUS_NOT_INITIALIZED
@@ -194,6 +207,14 @@
 #define RDNA3
 #endif // defined(__GFX11__)
 
+#if defined(__gfx1150__) || defined(__gfx1151__)
+#define RDNA3_5
+#endif // defined(__gfx1150__) || defined(__gfx1151__)
+
+#if defined(RDNA3) && !defined(RDNA3_5)
+#define RDNA3_0
+#endif // defined(RDNA3) && !defined(RDNA3_5)
+
 #if defined(__gfx1030__) || defined(__gfx1031__) || defined(__gfx1032__) || defined(__gfx1033__) || \
     defined(__gfx1034__) || defined(__gfx1035__) || defined(__gfx1036__) || defined(__gfx1037__)
 #define RDNA2

data/ext/sources/ggml/src/ggml-cuda/vendors/musa.h

@@ -12,11 +12,16 @@
 #define CUBLAS_GEMM_DEFAULT_TENSOR_OP MUBLAS_GEMM_DEFAULT
 #define CUBLAS_OP_N MUBLAS_OP_N
 #define CUBLAS_OP_T MUBLAS_OP_T
+#define CUBLAS_DEFAULT_MATH MUBLAS_DEFAULT_MATH
+#define CUBLAS_SIDE_RIGHT MUBLAS_SIDE_RIGHT
+#define CUBLAS_FILL_MODE_UPPER MUBLAS_FILL_MODE_UPPER
+#define CUBLAS_DIAG_NON_UNIT MUBLAS_DIAG_NON_UNIT
 #define CUBLAS_STATUS_SUCCESS MUBLAS_STATUS_SUCCESS
 #define CUBLAS_TF32_TENSOR_OP_MATH MUBLAS_TENSOR_OP_MATH
 #define CUDA_R_16F  MUSA_R_16F
 #define CUDA_R_16BF MUSA_R_16BF
 #define CUDA_R_32F  MUSA_R_32F
+#define cublasStrsmBatched mublasStrsmBatched
 #define cublasComputeType_t cudaDataType_t
 #define cublasCreate mublasCreate
 #define cublasDestroy mublasDestroy
@@ -56,6 +61,7 @@
 #define cudaHostRegisterPortable musaHostRegisterPortable
 #define cudaHostRegisterReadOnly musaHostRegisterReadOnly
 #define cudaHostUnregister musaHostUnregister
+#define cudaLaunchCooperativeKernel musaLaunchCooperativeKernel
 #define cudaLaunchHostFunc musaLaunchHostFunc
 #define cudaMalloc musaMalloc
 #define cudaMallocHost musaMallocHost

data/ext/sources/ggml/src/ggml-hexagon/CMakeLists.txt

@@ -0,0 +1,117 @@
+file(TO_CMAKE_PATH "${HEXAGON_SDK_ROOT}"   HEXAGON_SDK_ROOT)
+file(TO_CMAKE_PATH "${HEXAGON_TOOLS_ROOT}" HEXAGON_TOOLS_ROOT)
+
+if (NOT IS_DIRECTORY "${HEXAGON_SDK_ROOT}")
+    message(FATAL_ERROR "Make sure HEXAGON_SDK_ROOT point to the correct Hexagon SDK installation.")
+endif()
+
+if (NOT IS_DIRECTORY "${HEXAGON_TOOLS_ROOT}")
+    message("Try to read HEXAGON_TOOLS_ROOT from hexagon_sdk.json")
+    file(READ "${HEXAGON_SDK_ROOT}/hexagon_sdk.json" HEXAGON_SDK_CONFIG_PATH)
+    string(JSON HEXAGON_TOOLS_PATH GET ${HEXAGON_SDK_CONFIG_PATH} "root" "tools" "info" 0 "path")
+    message("Found HEXAGON_TOOLS_PATH: ${HEXAGON_TOOLS_PATH}")
+    set(HEXAGON_TOOLS_ROOT "${HEXAGON_SDK_ROOT}/${HEXAGON_TOOLS_PATH}")
+    file(TO_CMAKE_PATH "${HEXAGON_TOOLS_ROOT}" HEXAGON_TOOLS_ROOT)
+    if (NOT IS_DIRECTORY "${HEXAGON_TOOLS_ROOT}")
+        message(FATAL_ERROR "Make sure HEXAGON_TOOLS_ROOT point to the correct Hexagon SDK installation.")
+    endif()
+endif()
+
+message(STATUS "hexagon: using ${HEXAGON_SDK_ROOT} and ${HEXAGON_TOOLS_ROOT} for building libggml-htp skels")
+
+include(${HEXAGON_SDK_ROOT}/build/cmake/hexagon_fun.cmake)
+include(ExternalProject)
+
+option(GGML_HEXAGON_HTP_DEBUG "ggml-hexagon: enable HTP debug output" OFF)
+set(GGML_HEXAGON_HTP_CERT  "$ENV{HEXAGON_HTP_CERT}" CACHE PATH "ggml-hexagon: enable HTP library signing using certificate")
+set(GGML_HEXAGON_FP32_QUANTIZE_GROUP_SIZE 128 CACHE STRING "ggml-hexagon: quantize group size (32, 64, or 128)")
+
+add_library(htp_iface OBJECT
+    ${CMAKE_CURRENT_BINARY_DIR}/htp_iface_stub.c)
+
+set_target_properties(htp_iface PROPERTIES POSITION_INDEPENDENT_CODE ON)
+target_include_directories(htp_iface PUBLIC
+    ${HEXAGON_SDK_ROOT}/incs
+    ${HEXAGON_SDK_ROOT}/incs/stddef
+    ${HEXAGON_SDK_ROOT}/utils/examples
+    ${CMAKE_CURRENT_SOURCE_DIR}/htp
+    ${CMAKE_CURRENT_BINARY_DIR})
+
+build_idl(htp/htp_iface.idl htp_iface)
+
+if (CMAKE_SYSTEM_NAME MATCHES Android)
+    target_link_options(htp_iface PUBLIC -llog -ldl)
+elseif (CMAKE_SYSTEM_NAME MATCHES Windows)
+    target_precompile_headers(htp_iface PUBLIC <sal.h>)
+else()
+    target_link_options(htp_iface PUBLIC -ldl)
+endif()
+
+set(TARGET_NAME ggml-hexagon)
+ggml_add_backend_library(${TARGET_NAME}
+    ggml-hexagon.cpp
+    htp-drv.cpp
+    htp-drv.h
+    libdl.h
+    ../../include/ggml-hexagon.h)
+
+target_link_libraries(${TARGET_NAME} PRIVATE htp_iface)
+target_include_directories(${TARGET_NAME} PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/htp ${CMAKE_CURRENT_BINARY_DIR})
+
+# Build HTP skels
+set(HTP_SKELS)
+function(build_htp_skel V)
+    ExternalProject_Add(htp-${V}
+        SOURCE_DIR ${CMAKE_CURRENT_SOURCE_DIR}/htp BUILD_ALWAYS ON
+        BUILD_BYPRODUCTS ${CMAKE_CURRENT_BINARY_DIR}/libggml-htp-${V}.so
+        CMAKE_ARGS
+            -DCMAKE_BUILD_TYPE=Release
+            -DCMAKE_TOOLCHAIN_FILE=${CMAKE_CURRENT_SOURCE_DIR}/htp/cmake-toolchain.cmake
+            -DCMAKE_INSTALL_LIBDIR=${CMAKE_CURRENT_BINARY_DIR}
+            -DHEXAGON_SDK_ROOT=${HEXAGON_SDK_ROOT}
+            -DHEXAGON_TOOLS_ROOT=${HEXAGON_TOOLS_ROOT}
+            -DHEXAGON_HTP_DEBUG=${GGML_HEXAGON_HTP_DEBUG}
+            -DGGML_HEXAGON_FP32_QUANTIZE_GROUP_SIZE=${GGML_HEXAGON_FP32_QUANTIZE_GROUP_SIZE}
+            -DDSP_VERSION=${V}
+            -DPREBUILT_LIB_DIR="toolv19_${V}")
+    list(APPEND HTP_SKELS ${CMAKE_CURRENT_BINARY_DIR}/libggml-htp-${V}.so)
+    set(HTP_SKELS ${HTP_SKELS} PARENT_SCOPE)
+endfunction()
+
+build_htp_skel(v68)
+build_htp_skel(v69)
+build_htp_skel(v73)
+build_htp_skel(v75)
+build_htp_skel(v79)
+build_htp_skel(v81)
+
+# Install Hexagon skels required at runtime
+install(FILES ${HTP_SKELS} TYPE LIB)
+
+if (CMAKE_SYSTEM_NAME MATCHES Windows AND GGML_HEXAGON_HTP_CERT)
+    file(TO_CMAKE_PATH "$ENV{WINDOWS_SDK_BIN}/arm64"      WINSDK_BIN0_ARM64)
+    file(TO_CMAKE_PATH "$ENV{WINDOWS_SDK_BIN}/x86"        WINSDK_BIN0_X86)
+    file(TO_CMAKE_PATH "$ENV{WindowsSdkVerBinPath}/arm64" WINSDK_BIN1_ARM64)
+    file(TO_CMAKE_PATH "$ENV{WindowsSdkVerBinPath}/x86"   WINSDK_BIN1_X86)
+
+    set(WINSDK_PATHS ${WINSDK_BIN0_ARM64} ${WINSDK_BIN0_X86} ${WINSDK_BIN1_ARM64} ${WINSDK_BIN1_X86})
+
+    find_program(INF2CAT  NAMES inf2cat.exe  PATHS ${WINSDK_PATHS} REQUIRED)
+    find_program(SIGNTOOL NAMES signtool.exe PATHS ${WINSDK_PATHS} REQUIRED)
+
+    message(STATUS "hexagon: using ${GGML_HEXAGON_HTP_CERT} to sign libggml-htp skels")
+
+    set(LIBGGML_HTP_CAT ${CMAKE_CURRENT_BINARY_DIR}/libggml-htp.cat)
+    add_custom_target(libggml-htp-cat
+        BYPRODUCTS ${LIBGGML_HTP_CAT}
+        DEPENDS libggml-htp.inf ${HTP_SKELS}
+        COMMAND ${CMAKE_COMMAND} -E copy ${CMAKE_CURRENT_SOURCE_DIR}/libggml-htp.inf ${CMAKE_CURRENT_BINARY_DIR}
+        COMMAND ${INF2CAT} /driver:${CMAKE_CURRENT_BINARY_DIR} /os:10_25H2_ARM64
+        COMMAND ${SIGNTOOL} sign /fd sha256 /f ${GGML_HEXAGON_HTP_CERT} ${LIBGGML_HTP_CAT}
+        COMMENT "generating and signing libggml-htp.cat file"
+        VERBATIM
+    )
+
+    add_dependencies(${TARGET_NAME} libggml-htp-cat)
+    install(FILES ${LIBGGML_HTP_CAT} TYPE LIB)
+endif()