@agency-lang/whisper-local 0.0.1

package/vendor/whisper.cpp/ggml/src/ggml-cuda/convert.cuh

@@ -0,0 +1,26 @@
+#include "common.cuh"
+
+#define CUDA_DEQUANTIZE_BLOCK_SIZE 256
+
+template<typename T>
+using to_t_cuda_t = void (*)(const void * x, T * y, int64_t k, cudaStream_t stream);
+
+typedef to_t_cuda_t<float> to_fp32_cuda_t;
+typedef to_t_cuda_t<half> to_fp16_cuda_t;
+typedef to_t_cuda_t<nv_bfloat16> to_bf16_cuda_t;
+
+to_fp16_cuda_t ggml_get_to_fp16_cuda(ggml_type type);
+
+to_bf16_cuda_t ggml_get_to_bf16_cuda(ggml_type type);
+
+to_fp32_cuda_t ggml_get_to_fp32_cuda(ggml_type type);
+
+// TODO more general support for non-contiguous inputs
+
+template<typename T>
+using to_t_nc_cuda_t = void (*)(const void * x, T * y,
+    int64_t ne00, int64_t ne01, int64_t ne02, int64_t ne03,
+    int64_t s01, int64_t s02, int64_t s03, cudaStream_t stream);
+
+typedef to_t_nc_cuda_t<half> to_fp16_nc_cuda_t;
+to_fp16_nc_cuda_t ggml_get_to_fp16_nc_cuda(ggml_type type);

package/vendor/whisper.cpp/ggml/src/ggml-cuda/count-equal.cu

@@ -0,0 +1,64 @@
+#include "common.cuh"
+#include "count-equal.cuh"
+
+#include <cstdint>
+
+template <typename T>
+static __global__ void count_equal(const T * __restrict__ x, const T * __restrict__ y, int64_t * __restrict__ dst, const int64_t dk, const int64_t k) {
+    const int64_t i0 = (int64_t) blockIdx.x*dk;
+    const int64_t i1 = min(i0 + dk, k);
+
+    int nequal = 0;
+
+    for (int64_t i = i0 + threadIdx.x; i < i1; i += WARP_SIZE) {
+        const T xi = x[i];
+        const T yi = y[i];
+        nequal += xi == yi;
+    }
+
+    nequal = warp_reduce_sum(nequal);
+
+    if (threadIdx.x != 0) {
+        return;
+    }
+
+    atomicAdd((int *) dst, nequal);
+}
+
+void ggml_cuda_count_equal(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
+    const ggml_tensor * src0 = dst->src[0];
+    const ggml_tensor * src1 = dst->src[1];
+
+    GGML_ASSERT(src0->type == src1->type);
+    GGML_ASSERT( dst->type == GGML_TYPE_I64);
+
+    GGML_ASSERT(ggml_are_same_shape(src0, src1));
+    GGML_ASSERT(ggml_is_contiguous(src0));
+    GGML_ASSERT(ggml_is_contiguous(src1));
+    GGML_ASSERT(ggml_is_contiguous(dst));
+
+    int64_t * dst_d  = (int64_t *) dst->data;
+
+    cudaStream_t stream = ctx.stream();
+    const int nsm = ggml_cuda_info().devices[ggml_cuda_get_device()].nsm;
+
+    const int64_t ne = ggml_nelements(src0);
+    GGML_ASSERT(ne < (1 << 30) && "atomicAdd implementation only supports int");
+    const int64_t dne = GGML_PAD((ne + 4*nsm - 1) / (4*nsm), CUDA_COUNT_EQUAL_CHUNK_SIZE);
+
+    CUDA_CHECK(cudaMemsetAsync(dst_d, 0, ggml_nbytes(dst), stream));
+
+    const dim3 blocks_dim(WARP_SIZE, 1, 1);
+    const dim3 blocks_num(std::min((int64_t)4*nsm, (ne + CUDA_COUNT_EQUAL_CHUNK_SIZE - 1)/CUDA_COUNT_EQUAL_CHUNK_SIZE), 1, 1);
+
+    switch (src0->type) {
+        case GGML_TYPE_I32: {
+            const int * src0_d = (const int *) src0->data;
+            const int * src1_d = (const int *) src1->data;
+            count_equal<<<blocks_num, blocks_dim, 0, stream>>>(src0_d, src1_d, dst_d, dne, ne);
+        } break;
+        default:
+            GGML_ASSERT(false);
+            break;
+    }
+}

package/vendor/whisper.cpp/ggml/src/ggml-cuda/count-equal.cuh

@@ -0,0 +1,5 @@
+#include "common.cuh"
+
+#define CUDA_COUNT_EQUAL_CHUNK_SIZE 128
+
+void ggml_cuda_count_equal(ggml_backend_cuda_context & ctx, ggml_tensor * dst);

package/vendor/whisper.cpp/ggml/src/ggml-cuda/cp-async.cuh

@@ -0,0 +1,57 @@
+// Simplified API for asynchronous data loading.
+
+#include "common.cuh"
+
+
+static __device__ __forceinline__ unsigned int ggml_cuda_cvta_generic_to_shared(void * generic_ptr) {
+#ifdef CP_ASYNC_AVAILABLE
+    return __cvta_generic_to_shared(generic_ptr);
+#else
+    GGML_UNUSED(generic_ptr);
+    NO_DEVICE_CODE;
+    return 0;
+#endif // CP_ASYNC_AVAILABLE
+}
+
+// Copies data from global to shared memory, cg == cache global.
+// Both the src and dst pointers must be aligned to 16 bit.
+// Shared memory uses 32 bit addressing, the pointer is passed as unsigned int.
+// Generic pointers can be converted to 32 bit shared memory pointers using __cvta_generic_to_shared.
+// Only the 16 bit copy is exposed because 4 and 8 bit copies did not yield performance improvements.
+template <int preload>
+static __device__ __forceinline__ void cp_async_cg_16(const unsigned int dst, const void * src) {
+    static_assert(preload == 0 || preload == 64 || preload == 128 || preload == 256, "bad preload");
+#ifdef CP_ASYNC_AVAILABLE
+#if CUDART_VERSION >= 11040
+    if (preload == 256) {
+        asm volatile("cp.async.cg.shared.global.L2::256B [%0], [%1], 16;"
+            : : "r"(dst), "l"(src));
+    } else if (preload == 128) {
+        asm volatile("cp.async.cg.shared.global.L2::128B [%0], [%1], 16;"
+            : : "r"(dst), "l"(src));
+    } else if (preload == 64) {
+        asm volatile("cp.async.cg.shared.global.L2::64B [%0], [%1], 16;"
+            : : "r"(dst), "l"(src));
+    } else
+#endif // CUDART_VERSION >= 11040
+    {
+        asm volatile("cp.async.cg.shared.global [%0], [%1], 16;"
+            : : "r"(dst), "l"(src));
+    }
+#else
+    GGML_UNUSED(dst);
+    GGML_UNUSED(src);
+    NO_DEVICE_CODE;
+#endif // CP_ASYNC_AVAILABLE
+}
+
+// Makes each thread wait until its asynchronous data copies are done.
+// This does NOT provide any additional synchronization.
+// In particular, when copying data with multiple warps a call to __syncthreads will be needed.
+static __device__ __forceinline__ void cp_async_wait_all() {
+#ifdef CP_ASYNC_AVAILABLE
+    asm volatile("cp.async.wait_all;");
+#else
+    NO_DEVICE_CODE;
+#endif // CP_ASYNC_AVAILABLE
+}