whispercpp 1.3.4 → 1.3.5

data/ext/sources/ggml/src/ggml-hexagon/htp/htp-ctx.h

@@ -0,0 +1,35 @@
+#ifndef HTP_CTX_H
+#define HTP_CTX_H
+
+#include "htp-dma.h"
+#include "worker-pool.h"
+
+#include <assert.h>
+#include <dspqueue.h>
+#include <stdatomic.h>
+#include <stdint.h>
+
+#define HTP_MAX_NTHREADS 10
+
+// Main context for htp DSP backend
+struct htp_context {
+    dspqueue_t            queue;
+    dma_queue *           dma[HTP_MAX_NTHREADS];
+    worker_pool_context_t worker_pool;
+    uint32_t              n_threads;
+
+    int thread_id;
+    int thread_prio;
+
+    uint8_t * vtcm_base;
+    size_t    vtcm_size;
+    uint32_t  vtcm_rctx;
+
+    atomic_bool vtcm_valid;
+    atomic_bool vtcm_inuse;
+    atomic_bool vtcm_needs_release;
+
+    uint32_t opmask;
+};
+
+#endif /* HTP_CTX_H */

data/ext/sources/ggml/src/ggml-hexagon/htp/htp-dma.c

@@ -0,0 +1,63 @@
+#include "htp-dma.h"
+
+#include <stdbool.h>
+#include <stdlib.h>
+#include <string.h>
+
+#pragma clang diagnostic ignored "-Wunused-function"
+
+static inline uint32_t pow2_ceil(uint32_t x) {
+    if (x <= 1) {
+        return 1;
+    }
+    int p = 2;
+    x--;
+    while (x >>= 1) {
+        p <<= 1;
+    }
+    return p;
+}
+
+dma_queue * dma_queue_create(size_t capacity) {
+    dma_queue * q = (dma_queue *) memalign(32, sizeof(dma_queue));
+    if (q == NULL) {
+        FARF(ERROR, "%s: failed to allocate DMA queue\n", __FUNCTION__);
+        return NULL;
+    }
+
+    capacity = pow2_ceil(capacity);
+
+    memset(q, 0, sizeof(dma_queue));
+    q->capacity = capacity;
+    q->idx_mask = capacity - 1;
+
+    q->desc = (hexagon_udma_descriptor_type1_t *) memalign(64, capacity * sizeof(hexagon_udma_descriptor_type1_t));
+    memset(q->desc, 0, capacity * sizeof(hexagon_udma_descriptor_type1_t));
+
+    q->dptr = (dma_ptr *) memalign(4, capacity * sizeof(dma_ptr));
+    memset(q->dptr, 0, capacity * sizeof(dma_ptr));
+
+    q->tail = &q->desc[capacity - 1];
+
+    if (!q->desc && !q->dptr) {
+        FARF(ERROR, "%s: failed to allocate DMA queue items\n", __FUNCTION__);
+        return NULL;
+    }
+
+    FARF(HIGH, "dma-queue: capacity %u\n", capacity);
+
+    return q;
+}
+
+void dma_queue_delete(dma_queue * q) {
+    if (!q) {
+        return;
+    }
+    free(q->desc);
+    free(q->dptr);
+    free(q);
+}
+
+void dma_queue_flush(dma_queue * q) {
+    while (dma_queue_pop(q).dst != NULL) ;
+}

data/ext/sources/ggml/src/ggml-hexagon/htp/htp-dma.h

@@ -0,0 +1,157 @@
+#ifndef HTP_DMA_H
+#define HTP_DMA_H
+
+#include <HAP_farf.h>
+#include <hexagon_protos.h>
+#include <hexagon_types.h>
+#include <stdbool.h>
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct {
+    void *dst;
+    const void *src;
+} dma_ptr;
+
+typedef struct {
+    hexagon_udma_descriptor_type1_t * desc;  // descriptor pointers
+    hexagon_udma_descriptor_type1_t * tail;  // tail pointer
+    dma_ptr                         * dptr;  // dst/src pointers
+    uint32_t                          push_idx;
+    uint32_t                          pop_idx;
+    uint32_t                          capacity;
+    uint32_t                          idx_mask;
+} dma_queue;
+
+dma_queue * dma_queue_create(size_t capacity);
+void        dma_queue_delete(dma_queue * q);
+void        dma_queue_flush(dma_queue * q);
+
+// TODO: technically we don't need these and could use Q6_dmstart/wait/etc instead
+// but those do not seem to always compiler properly.
+static inline void dmstart(void * next) {
+    asm volatile(" release(%0):at" : : "r"(next));
+    asm volatile(" dmstart(%0)" : : "r"(next));
+}
+
+static inline void dmlink(void * cur, void * next) {
+    asm volatile(" release(%0):at" : : "r"(next));
+    asm volatile(" dmlink(%0, %1)" : : "r"(cur), "r"(next));
+}
+
+static inline unsigned int dmpoll(void) {
+    unsigned int ret = 0;
+    asm volatile(" %0 = dmpoll" : "=r"(ret) : : "memory");
+    return ret;
+}
+
+static inline unsigned int dmwait(void) {
+    unsigned int ret = 0;
+    asm volatile(" %0 = dmwait" : "=r"(ret) : : "memory");
+    return ret;
+}
+
+static inline dma_ptr dma_make_ptr(void *dst, const void *src)
+{
+    dma_ptr p = { dst, src };
+    return p;
+}
+
+static inline bool dma_queue_push(dma_queue * q,
+                                  dma_ptr     dptr,
+                                  size_t      dst_row_size,
+                                  size_t      src_row_size,
+                                  size_t      width, // width in bytes. number of bytes to transfer per row
+                                  size_t      nrows) {
+    if (((q->push_idx + 1) & q->idx_mask) == q->pop_idx) {
+        FARF(ERROR, "dma-push: queue full\n");
+        return false;
+    }
+
+    hexagon_udma_descriptor_type1_t * desc = &q->desc[q->push_idx];
+
+    desc->next           = NULL;
+    desc->length         = 0;
+    desc->desctype       = HEXAGON_UDMA_DESC_DESCTYPE_TYPE1;
+    desc->dstbypass      = 1;
+    desc->srcbypass      = 1;
+#if __HVX_ARCH__ >= 73
+    desc->dstbypass      = 1;
+    desc->srcbypass      = 1;
+#else
+    desc->dstbypass      = 0;
+    desc->srcbypass      = 1;
+#endif
+    desc->order          = 0;
+    desc->dstate         = HEXAGON_UDMA_DESC_DSTATE_INCOMPLETE;
+    desc->src            = (void *) dptr.src;
+    desc->dst            = (void *) dptr.dst;
+    desc->allocation     = 0;
+    desc->padding        = 0;
+    desc->roiwidth       = width;
+    desc->roiheight      = nrows;
+    desc->srcstride      = src_row_size;
+    desc->dststride      = dst_row_size;
+    desc->srcwidthoffset = 0;
+    desc->dstwidthoffset = 0;
+
+    q->dptr[q->push_idx] = dptr;
+
+    dmlink(q->tail, desc);
+    q->tail = desc;
+
+    // FARF(ERROR, "dma-push: i %u len %u dst %p src %p\n", q->push_idx, len, dst, src);
+    q->push_idx = (q->push_idx + 1) & q->idx_mask;
+    return true;
+}
+
+static inline bool dma_queue_push_ddr_to_vtcm(dma_queue * q,
+                                              dma_ptr     dptr,
+                                              size_t      dst_row_size,
+                                              size_t      src_row_size,
+                                              size_t      nrows) {
+    return dma_queue_push(q, dptr, dst_row_size, src_row_size, src_row_size, nrows);
+}
+
+
+static inline bool dma_queue_push_vtcm_to_ddr(dma_queue * q,
+                                              dma_ptr     dptr,
+                                              size_t      dst_row_size,
+                                              size_t      src_row_size,
+                                              size_t      nrows) {
+    return dma_queue_push(q, dptr, dst_row_size, src_row_size, dst_row_size, nrows);
+}
+
+static inline dma_ptr dma_queue_pop(dma_queue * q) {
+    dma_ptr dptr  = { NULL };
+
+    if (q->push_idx == q->pop_idx) {
+        return dptr;
+    }
+
+    hexagon_udma_descriptor_type1_t * desc = &q->desc[q->pop_idx];
+
+    // Wait for desc to complete
+    while (1) {
+        dmpoll();
+        if (desc->dstate == HEXAGON_UDMA_DESC_DSTATE_COMPLETE) {
+            break;
+        }
+        // FARF(ERROR, "dma-pop: waiting for DMA : %u\n", q->pop_idx);
+    }
+
+    dptr = q->dptr[q->pop_idx];
+
+    // FARF(ERROR, "dma-pop: i %u dst %p\n", q->pop_idx, dst);
+    q->pop_idx = (q->pop_idx + 1) & q->idx_mask;
+    return dptr;
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+
+#endif /* HTP_DMA_H */

data/ext/sources/ggml/src/ggml-hexagon/htp/htp-msg.h

@@ -0,0 +1,165 @@
+#ifndef HTP_MSG_H
+#define HTP_MSG_H
+
+#include <assert.h>
+
+// ggml-common.h must be included prio to this header
+
+// Mask to enable various stages of the Ops.
+// Used for debugging and profiling.
+enum {
+    HTP_OPMASK_QUEUE    = (1 << 0),  // Enable Queueing (ie calls into the DSP)
+    HTP_OPMASK_QUANTIZE = (1 << 1),  // Enable Quantize
+    HTP_OPMASK_COMPUTE  = (1 << 2),  // Enable Compute
+};
+
+// Op flags
+enum {
+    HTP_OPFLAGS_SKIP_QUANTIZE = (1 << 0),  // Skip dynamic quantization (reuse quantized tensors)
+    HTP_OPFLAGS_SKIP_COMPUTE  = (1 << 1),  // Skip actual computation (used for profiling)
+    HTP_OPFLAGS_EARLY_WAKEUP  = (1 << 2)   // Send early wakeup notification
+};
+
+enum htp_status {
+    HTP_STATUS_OK             = 1,
+    HTP_STATUS_INTERNAL_ERR   = 2,
+    HTP_STATUS_NO_SUPPORT     = 3,
+    HTP_STATUS_INVAL_PARAMS   = 4,
+    HTP_STATUS_VTCM_TOO_SMALL = 5,
+};
+
+// The values must match the ggml_type.
+// Duplicated here because we can't include full ggml.h in the htp build.
+// We have some static_asserts in the cpp code to ensure things are in sync.
+enum htp_data_type {
+    HTP_TYPE_F32   = 0,
+    HTP_TYPE_F16   = 1,
+    HTP_TYPE_Q4_0  = 2,
+    HTP_TYPE_Q8_0  = 8,
+    HTP_TYPE_I32   = 26,
+    HTP_TYPE_I64   = 27,
+    HTP_TYPE_MXFP4 = 39,
+    HTP_TYPE_COUNT
+};
+
+// These values are manually translated over to HTP
+// !!!! DO NOT ALTER THE ORDER OF THE FIRST FOUR ENUMS !!!!
+enum htp_op {
+    HTP_OP_MUL            = 0,
+    HTP_OP_ADD            = 1,
+    HTP_OP_SUB            = 2,
+    HTP_OP_DIV            = 3,
+    HTP_OP_MUL_MAT        = 4,
+    HTP_OP_MUL_MAT_ID     = 5,
+    HTP_OP_RMS_NORM       = 6,
+    HTP_OP_UNARY_SILU     = 7,
+    HTP_OP_UNARY_GELU     = 8,
+    HTP_OP_GLU_SWIGLU     = 9,
+    HTP_OP_GLU_SWIGLU_OAI = 10,
+    HTP_OP_SOFTMAX        = 11,
+    HTP_OP_ADD_ID         = 12,
+    HTP_OP_ROPE           = 13,
+    HTP_OP_FLASH_ATTN_EXT = 14,
+    HTP_OP_SET_ROWS       = 15,
+    HTP_OP_SCALE          = 16,
+    HTP_OP_GET_ROWS       = 17,
+    INVALID
+};
+
+static inline size_t htp_type_block_size(uint32_t t) {
+    switch (t) {
+        case HTP_TYPE_F32:
+            return 1;
+        case HTP_TYPE_F16:
+            return 1;
+        case HTP_TYPE_Q4_0:
+            return QK4_0;
+        case HTP_TYPE_Q8_0:
+            return QK8_0;
+        case HTP_TYPE_MXFP4:
+            return QK_MXFP4;
+        default:
+            assert(0 && "unsupported HTP data type");
+    }
+    return 0;
+}
+
+static inline size_t htp_type_nbytes(uint32_t t) {
+    switch (t) {
+        case HTP_TYPE_F32:
+            return 4;
+        case HTP_TYPE_F16:
+            return 2;
+        case HTP_TYPE_Q4_0:
+            return sizeof(block_q4_0);
+        case HTP_TYPE_Q8_0:
+            return sizeof(block_q8_0);
+        case HTP_TYPE_MXFP4:
+            return sizeof(block_mxfp4);
+        default:
+            assert(0 && "unsupported HTP data type");
+    }
+    return 0;
+}
+
+static const char * htp_type_name(uint32_t t) {
+    switch (t) {
+        case HTP_TYPE_F32:
+            return "fp32";
+        case HTP_TYPE_F16:
+            return "fp16";
+        case HTP_TYPE_Q4_0:
+            return "q4_0";
+        case HTP_TYPE_Q8_0:
+            return "q8_0";
+        case HTP_TYPE_MXFP4:
+            return "mxfp4";
+    }
+    return 0;
+}
+
+// Internal types
+#define QK_Q4_0x4x2  256  // 4x Q4_0 blocks packed with next 4x Q4_0 blocks (size in bytes 128)
+#define QK_Q8_0x4x2  256  // 4x Q8_0 blocks concat with next 4x Q8_0 blocks
+#define QK_MXFP4x4x2 256  // 4x MXFP4 blocks concat with next 4x MXFP4 blocks
+
+#define HTP_MAX_DIMS 4
+
+struct htp_tensor {
+    uint32_t data;                // Buffer offset in the messages, and data pointer on the NSP
+    uint32_t type;                // Data type
+    uint32_t ne[HTP_MAX_DIMS];    // Number of elements
+    uint32_t nb[HTP_MAX_DIMS];    // Stride in bytes (see ggml.h ggml_tensor)
+};
+
+#define HTP_MAX_OP_PARAMS 64
+
+struct htp_general_req {
+    uint32_t op;  // GGML/HTP Op
+    int32_t  op_params[HTP_MAX_OP_PARAMS / sizeof(int32_t)];
+    // Params for the op, e.g. epsilon of RMS norm
+    uint32_t flags;          // Request flags
+
+    struct htp_tensor src0;  // Input0 tensor
+    struct htp_tensor src1;  // Input1 tensor
+    struct htp_tensor src2;  // Input2 tensor
+    struct htp_tensor src3;  // Input3 tensor
+    struct htp_tensor src4;  // Input4 tensor
+    struct htp_tensor dst;   // Output tensor
+
+    // should be multiple of 64 bytes (cacheline)
+};
+
+struct htp_general_rsp {
+    uint32_t op;           // GGML/HTP Op
+    uint32_t status;       // HTP_STATUS_...
+    uint32_t prof_usecs;   // Number of usec per request
+    uint32_t prof_cycles;  // Number of cycles per request
+    uint32_t prof_pkts;    // Number of instruction packets per request
+    uint8_t  unused[44];   // Pad to 64 bytes
+};
+
+#define HTP_MAX_MESSAGE_SIZE   sizeof(struct htp_general_req)
+#define HTP_MAX_PACKET_BUFFERS 8
+
+#endif /* HTP_MSG_H */

data/ext/sources/ggml/src/ggml-hexagon/htp/htp-ops.h

@@ -0,0 +1,92 @@
+#ifndef HTP_OPS_H
+#define HTP_OPS_H
+
+#include "htp-ctx.h"
+#include "htp-msg.h"
+#include "worker-pool.h"
+#include "ops-utils.h"
+
+#include <assert.h>
+#include <stdint.h>
+
+// ggml-common.h must be included prior to this header
+
+struct htp_spad {
+    uint8_t * data;
+    size_t    stride;
+    size_t    size;
+    size_t    size_per_thread;
+};
+
+struct htp_ops_context {
+    struct htp_context * ctx;
+
+    enum htp_op op;
+    int32_t     op_params[HTP_MAX_OP_PARAMS / sizeof(int32_t)];
+
+    struct htp_tensor src0;
+    struct htp_tensor src1;
+    struct htp_tensor src2;
+    struct htp_tensor src3;
+    struct htp_tensor src4;
+    struct htp_tensor dst;
+
+    struct htp_spad src0_spad;
+    struct htp_spad src1_spad;
+    struct htp_spad src2_spad;
+    struct htp_spad src3_spad;
+    struct htp_spad dst_spad;
+
+    worker_pool_context_t * wpool;      // worker pool
+    uint32_t                n_threads;  // num threads
+
+    uint32_t src0_nrows_per_thread;
+    uint32_t src1_nrows_per_thread;
+
+    struct fastdiv_values src0_div1;  // fastdiv values for ne1
+    struct fastdiv_values src0_div2;  // fastdiv values for ne2
+    struct fastdiv_values src0_div3;  // fastdiv values for ne3
+    struct fastdiv_values src0_div21; // fastdiv values for ne2 * ne1
+
+    struct fastdiv_values src1_div1;  // fastdiv values for ne1
+    struct fastdiv_values src1_div2;  // fastdiv values for ne2
+    struct fastdiv_values src1_div3;  // fastdiv values for ne3
+    struct fastdiv_values src1_div21; // fastdiv values for ne2 * ne1
+
+    struct fastdiv_values src3_div1;  // fastdiv values for ne1
+    struct fastdiv_values src3_div2;  // fastdiv values for ne2
+    struct fastdiv_values src3_div3;  // fastdiv values for ne3
+    struct fastdiv_values src3_div21; // fastdiv values for ne2 * ne1
+
+    struct fastdiv_values broadcast_rk2;
+    struct fastdiv_values broadcast_rk3;
+    struct fastdiv_values broadcast_rv2;
+    struct fastdiv_values broadcast_rv3;
+
+    struct fastdiv_values mm_div_ne12_ne1; // fastdiv values for ne12 * ne1
+    struct fastdiv_values mm_div_ne1;      // fastdiv values for ne1
+    struct fastdiv_values mm_div_r2;       // fastdiv values for ne12 / ne02
+    struct fastdiv_values mm_div_r3;       // fastdiv values for ne13 / ne03
+
+    struct fastdiv_values set_rows_div_ne12; // fastdiv values for ne12
+    struct fastdiv_values set_rows_div_ne11; // fastdiv values for ne11
+
+    struct fastdiv_values get_rows_div_ne10;      // fastdiv values for ne10
+    struct fastdiv_values get_rows_div_ne10_ne11; // fastdiv values for ne10 * ne11
+
+    uint32_t flags;
+};
+
+int op_matmul(struct htp_ops_context * octx);
+int op_matmul_id(struct htp_ops_context * octx);
+int op_binary(struct htp_ops_context * octx);
+int op_unary(struct htp_ops_context * octx);
+int op_activations(struct htp_ops_context * octx);
+int op_softmax(struct htp_ops_context * octx);
+int op_add_id(struct htp_ops_context * octx);
+int op_rope(struct htp_ops_context * octx);
+int op_flash_attn_ext(struct htp_ops_context * octx);
+int op_set_rows(struct htp_ops_context * octx);
+int op_get_rows(struct htp_ops_context * octx);
+
+#endif /* HTP_OPS_H */

data/ext/sources/ggml/src/ggml-hexagon/htp/htp_iface.idl

@@ -0,0 +1,16 @@
+// FastRPC IDL interface for GGML HTP
+
+#ifndef HTP_IDL
+#define HTP_IDL
+
+#include "AEEStdDef.idl"
+#include "remote.idl"
+
+interface htp_iface : remote_handle64 {
+    AEEResult start(in uint32 sess_id, in uint64 dsp_queue_id, in uint32 n_hvx);
+    AEEResult stop();
+    AEEResult enable_etm();
+    AEEResult disable_etm();
+};
+
+#endif /* HTP_IDL */

data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-exp.c

@@ -0,0 +1,94 @@
+#pragma clang diagnostic ignored "-Wunused-variable"
+#pragma clang diagnostic ignored "-Wunused-function"
+#pragma clang diagnostic ignored "-Wunused-but-set-variable"
+
+#include <hexagon_protos.h>
+#include <hexagon_types.h>
+#include <math.h>
+#include <string.h>
+
+#define GGML_COMMON_DECL_C
+#include "ggml-common.h"
+#include "htp-ctx.h"
+#include "htp-dma.h"
+#include "htp-msg.h"
+#include "htp-ops.h"
+#include "hvx-utils.h"
+#include "ops-utils.h"
+
+static inline HVX_Vector hvx_vec_exp_fp32_guard(HVX_Vector in_vec, HVX_Vector max_exp, HVX_Vector inf) {
+    const HVX_VectorPred pred0 = Q6_Q_vcmp_gt_VsfVsf(in_vec, max_exp);
+
+    HVX_Vector out = hvx_vec_exp_fp32(in_vec);
+
+    return Q6_V_vmux_QVV(pred0, inf, out);
+}
+
+void hvx_exp_f32(const uint8_t * restrict src, uint8_t * restrict dst, const int num_elems, bool negate) {
+    int left_over       = num_elems & (VLEN_FP32 - 1);
+    int num_elems_whole = num_elems - left_over;
+
+    int unaligned_addr = 0;
+    int unaligned_loop = 0;
+    if ((0 == htp_is_aligned((void *) src, VLEN)) || (0 == htp_is_aligned((void *) dst, VLEN))) {
+        FARF(HIGH, "hvx_exp_f32: unaligned address in hvx op, possibly slower execution\n");
+        unaligned_addr = 1;
+    }
+    // assert((0 == unaligned_addr) || (0 == num_elems_whole));
+    if ((1 == unaligned_addr) && (num_elems_whole != 0)) {
+        unaligned_loop = 1;
+        FARF(HIGH, "hvx_exp_f32: unaligned loop in hvx op, possibly slower execution\n");
+    }
+
+    HVX_Vector vec_out = Q6_V_vzero();
+
+    static const float kInf    = INFINITY;
+    static const float kMaxExp = 88.02f;  // log(INF)
+
+    const HVX_Vector max_exp = hvx_vec_splat_fp32(kMaxExp);
+    const HVX_Vector inf     = hvx_vec_splat_fp32(kInf);
+
+    if (0 == unaligned_loop) {
+        HVX_Vector * p_vec_in1 = (HVX_Vector *) src;
+        HVX_Vector * p_vec_out = (HVX_Vector *) dst;
+
+        #pragma unroll(4)
+        for (int i = 0; i < num_elems_whole; i += VLEN_FP32) {
+            if (true == negate) {
+                HVX_Vector neg_vec_in = hvx_vec_neg_fp32(*p_vec_in1++);
+                *p_vec_out++          = hvx_vec_exp_fp32_guard(neg_vec_in, max_exp, inf);
+            } else {
+                *p_vec_out++ = hvx_vec_exp_fp32_guard(*p_vec_in1++, max_exp, inf);
+            }
+        }
+    } else {
+        #pragma unroll(4)
+        for (int i = 0; i < num_elems_whole; i += VLEN_FP32) {
+            HVX_Vector in = *(HVX_UVector *) (src + i * SIZEOF_FP32);
+
+            if (true == negate) {
+                HVX_Vector neg_vec_in                    = hvx_vec_neg_fp32(in);
+                *(HVX_UVector *) (dst + i * SIZEOF_FP32) = hvx_vec_exp_fp32_guard(neg_vec_in, max_exp, inf);
+            } else {
+                *(HVX_UVector *) (dst + i * SIZEOF_FP32) = hvx_vec_exp_fp32_guard(in, max_exp, inf);
+            }
+        }
+    }
+
+    if (left_over > 0) {
+        const float * srcf = (float *) src + num_elems_whole;
+        float *       dstf = (float *) dst + num_elems_whole;
+
+        HVX_Vector in = *(HVX_UVector *) srcf;
+
+        if (true == negate) {
+            HVX_Vector neg_vec_in = hvx_vec_neg_fp32(in);
+
+            vec_out = hvx_vec_exp_fp32_guard(neg_vec_in, max_exp, inf);
+        } else {
+            vec_out = hvx_vec_exp_fp32_guard(in, max_exp, inf);
+        }
+
+        hvx_vec_store_u((void *) dstf, left_over * SIZEOF_FP32, vec_out);
+    }
+}

data/ext/sources/ggml/src/ggml-hexagon/htp/hvx-inverse.c

@@ -0,0 +1,72 @@
+#pragma clang diagnostic ignored "-Wunused-variable"
+#pragma clang diagnostic ignored "-Wunused-function"
+#pragma clang diagnostic ignored "-Wunused-but-set-variable"
+
+#include <hexagon_protos.h>
+#include <hexagon_types.h>
+#include <math.h>
+#include <string.h>
+
+#define GGML_COMMON_DECL_C
+#include "ggml-common.h"
+#include "htp-ctx.h"
+#include "htp-dma.h"
+#include "htp-msg.h"
+#include "htp-ops.h"
+#include "hvx-utils.h"
+#include "ops-utils.h"
+
+static inline HVX_Vector hvx_vec_inverse_fp32_guard(HVX_Vector v_sf, HVX_Vector nan_inf_mask) {
+    HVX_Vector out = hvx_vec_inverse_fp32(v_sf);
+
+    HVX_Vector           masked_out = Q6_V_vand_VV(out, nan_inf_mask);
+    const HVX_VectorPred pred       = Q6_Q_vcmp_eq_VwVw(nan_inf_mask, masked_out);
+
+    return Q6_V_vmux_QVV(pred, Q6_V_vzero(), out);
+}
+
+void hvx_inverse_f32(const uint8_t * restrict src, uint8_t * restrict dst, const int num_elems) {
+    int left_over       = num_elems & (VLEN_FP32 - 1);
+    int num_elems_whole = num_elems - left_over;
+
+    int unaligned_addr = 0;
+    int unaligned_loop = 0;
+    if ((0 == htp_is_aligned((void *) src, VLEN)) || (0 == htp_is_aligned((void *) dst, VLEN))) {
+        FARF(HIGH, "hvx_inverse_f32: unaligned address in hvx op, possibly slower execution\n");
+        unaligned_addr = 1;
+    }
+    // assert((0 == unaligned_addr) || (0 == num_elems_whole));
+    if ((1 == unaligned_addr) && (num_elems_whole != 0)) {
+        unaligned_loop = 1;
+        FARF(HIGH, "hvx_inverse_f32: unaligned loop in hvx op, possibly slower execution\n");
+    }
+
+    static const uint32_t kNanInfMask  = 0x7f800000;
+    const HVX_Vector      nan_inf_mask = Q6_V_vsplat_R(kNanInfMask);
+
+    if (0 == unaligned_loop) {
+        HVX_Vector * p_vec_in  = (HVX_Vector *) src;
+        HVX_Vector * p_vec_out = (HVX_Vector *) dst;
+
+        #pragma unroll(4)
+        for (int i = 0; i < num_elems_whole; i += VLEN_FP32) {
+            *p_vec_out++ = hvx_vec_inverse_fp32_guard(*p_vec_in++, nan_inf_mask);
+        }
+    } else {
+        #pragma unroll(4)
+        for (int i = 0; i < num_elems_whole; i += VLEN_FP32) {
+            HVX_Vector in                            = *(HVX_UVector *) (src + i * SIZEOF_FP32);
+            *(HVX_UVector *) (dst + i * SIZEOF_FP32) = hvx_vec_inverse_fp32_guard(in, nan_inf_mask);
+        }
+    }
+
+    if (left_over > 0) {
+        const float * srcf = (float *) src + num_elems_whole;
+        float *       dstf = (float *) dst + num_elems_whole;
+
+        HVX_Vector in  = *(HVX_UVector *) srcf;
+        HVX_Vector out = hvx_vec_inverse_fp32_guard(in, nan_inf_mask);
+
+        hvx_vec_store_u((void *) dstf, left_over * SIZEOF_FP32, out);
+    }
+}