local-llm-rn 1.0.0

package/cpp/ggml/src/ggml-hexagon/htp/get-rows-ops.c

@@ -0,0 +1,111 @@
+#pragma clang diagnostic ignored "-Wunused-variable"
+#pragma clang diagnostic ignored "-Wunused-function"
+#pragma clang diagnostic ignored "-Wunused-but-set-variable"
+
+#include <HAP_farf.h>
+#include <HAP_perf.h>
+
+#include <math.h>
+#include <string.h>
+
+#define GGML_COMMON_DECL_C
+#include "ggml-common.h"
+#include "htp-ctx.h"
+#include "htp-msg.h"
+#include "htp-ops.h"
+#include "hvx-utils.h"
+
+struct get_rows_context {
+    struct htp_ops_context * octx;
+    uint32_t src1_nrows_per_thread;
+    struct fastdiv_values get_rows_div_ne10;
+    struct fastdiv_values get_rows_div_ne10_ne11;
+};
+
+#define get_rows_preamble \
+    const uint32_t ne00 = octx->src0.ne[0]; \
+    const uint32_t ne01 = octx->src0.ne[1]; \
+    const uint32_t ne02 = octx->src0.ne[2]; \
+    const uint32_t ne03 = octx->src0.ne[3]; \
+                                            \
+    const uint32_t ne10 = octx->src1.ne[0]; \
+    const uint32_t ne11 = octx->src1.ne[1]; \
+    const uint32_t ne12 = octx->src1.ne[2]; \
+                                            \
+    const uint32_t nb01 = octx->src0.nb[1]; \
+    const uint32_t nb02 = octx->src0.nb[2]; \
+    const uint32_t nb03 = octx->src0.nb[3]; \
+                                            \
+    const uint32_t nb10 = octx->src1.nb[0]; \
+    const uint32_t nb11 = octx->src1.nb[1]; \
+    const uint32_t nb12 = octx->src1.nb[2]; \
+                                            \
+    const uint32_t nb1 = octx->dst.nb[1];   \
+    const uint32_t nb2 = octx->dst.nb[2];   \
+    const uint32_t nb3 = octx->dst.nb[3];   \
+                                            \
+    const uint32_t nr = ne10 * ne11 * ne12;
+
+static void get_rows_thread_f32_f32(unsigned int nth, unsigned int ith, void *data) {
+    struct get_rows_context * grctx = (struct get_rows_context *)data;
+    struct htp_ops_context * octx = grctx->octx;
+    get_rows_preamble;
+
+    // parallelize by src1 elements (which correspond to dst rows)
+    const uint32_t dr  = grctx->src1_nrows_per_thread;
+    const uint32_t ir0 = dr * ith;
+    const uint32_t ir1 = (ir0 + dr < nr) ? (ir0 + dr) : nr;
+
+    const bool is_i32 = (octx->src1.type == HTP_TYPE_I32);
+
+    for (uint32_t i = ir0; i < ir1; ++i) {
+        const uint32_t i12 = fastdiv(i, &grctx->get_rows_div_ne10_ne11);
+        const uint32_t rem = i - i12 * ne11 * ne10;
+        const uint32_t i11 = fastdiv(rem, &grctx->get_rows_div_ne10);
+        const uint32_t i10 = rem - i11 * ne10;
+
+        const uintptr_t src1_addr = octx->src1.data + i10*nb10 + i11*nb11 + i12*nb12;
+
+        uint32_t i01 = is_i32 ? *(int32_t *)src1_addr : *(int64_t *)src1_addr;
+
+        if (i01 >= ne01) {
+            // invalid index, skip for now to avoid crash
+            continue;
+        }
+
+        const uintptr_t src0_ptr = octx->src0.data + i01*nb01 + i11*nb02 + i12*nb03;
+        const uintptr_t dst_ptr  = octx->dst.data  + i10*nb1  + i11*nb2  + i12*nb3;
+        hvx_copy_f32_uu((uint8_t *)dst_ptr, (const uint8_t *)src0_ptr, ne00);
+    }
+}
+
+int op_get_rows(struct htp_ops_context * octx) {
+    get_rows_preamble;
+
+    if (octx->src0.type != HTP_TYPE_F32) {
+        return HTP_STATUS_NO_SUPPORT;
+    }
+
+    if (octx->dst.type != HTP_TYPE_F32) {
+        return HTP_STATUS_NO_SUPPORT;
+    }
+
+    if (octx->src1.type != HTP_TYPE_I32 && octx->src1.type != HTP_TYPE_I64) {
+        return HTP_STATUS_NO_SUPPORT;
+    }
+
+    if (octx->flags & HTP_OPFLAGS_SKIP_COMPUTE) {
+        return HTP_STATUS_OK;
+    }
+
+    struct get_rows_context grctx;
+    grctx.octx = octx;
+    grctx.get_rows_div_ne10      = init_fastdiv_values(octx->src1.ne[0]);
+    grctx.get_rows_div_ne10_ne11 = init_fastdiv_values(octx->src1.ne[0] * octx->src1.ne[1]);
+
+    const uint32_t n_jobs = MIN(nr, octx->n_threads);
+    grctx.src1_nrows_per_thread = (nr + n_jobs - 1) / n_jobs;
+
+    worker_pool_run_func(octx->ctx->worker_pool, get_rows_thread_f32_f32, &grctx, n_jobs);
+    return HTP_STATUS_OK;
+}

package/cpp/ggml/src/ggml-hexagon/htp/hex-dma.c

@@ -0,0 +1,63 @@
+#include "hex-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) ;
+}

package/cpp/ggml/src/ggml-hexagon/htp/hex-dma.h

@@ -0,0 +1,182 @@
+#ifndef HTP_DMA_H
+#define HTP_DMA_H
+
+#include <HAP_farf.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 width %u nrows %d dst %p src %p\n", q->push_idx, width, nrows, dptr.dst, dptr.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 src %p\n", q->pop_idx, dptr.dst, dptr.src);
+    q->pop_idx = (q->pop_idx + 1) & q->idx_mask;
+    return dptr;
+}
+
+static inline dma_ptr dma_queue_pop_nowait(dma_queue * q) {
+    dma_ptr dptr  = { NULL };
+
+    if (q->push_idx == q->pop_idx) {
+        return dptr;
+    }
+
+    dptr = q->dptr[q->pop_idx];
+
+    // FARF(ERROR, "dma-pop-nowait: i %u dst %p src %p\n", q->pop_idx, dptr.dst, dptr.src);
+    q->pop_idx = (q->pop_idx + 1) & q->idx_mask;
+    return dptr;
+}
+
+static inline bool dma_queue_empty(dma_queue * q) {
+    return q->push_idx == q->pop_idx;
+}
+
+static inline uint32_t dma_queue_depth(dma_queue * q) {
+    return (q->push_idx - q->pop_idx) & q->idx_mask;
+}
+
+static inline uint32_t dma_queue_capacity(dma_queue * q) {
+    return q->capacity;
+}
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+
+#endif /* HTP_DMA_H */

package/cpp/ggml/src/ggml-hexagon/htp/hex-dump.h

@@ -0,0 +1,77 @@
+#ifndef HEX_DUMP_H
+#define HEX_DUMP_H
+
+#include <HAP_farf.h>
+
+static inline void hex_dump_int8_line(char * pref, const int8_t * x, int n) {
+    char str[1024], *p = str, *p_end = str + sizeof(str);
+    p += snprintf(p, p_end - p, "%s: ", pref);
+    for (int i = 0; i < n && p < p_end; i++) {
+        p += snprintf(p, p_end - p, "%d, ", x[i]);
+    }
+    FARF(HIGH, "%s\n", str);
+}
+
+static inline void hex_dump_uint8_line(char * pref, const uint8_t * x, uint32_t n) {
+    char str[1024], *p = str, *p_end = str + sizeof(str);
+    p += snprintf(p, p_end - p, "%s: ", pref);
+    for (int i = 0; i < n && p < p_end; i++) {
+        p += snprintf(p, p_end - p, "%d, ", x[i]);
+    }
+    FARF(HIGH, "%s\n", str);
+}
+
+static inline void hex_dump_int32_line(char * pref, const int32_t * x, uint32_t n) {
+    char str[1024], *p = str, *p_end = str + sizeof(str);
+    p += snprintf(p, p_end - p, "%s: ", pref);
+    for (int i = 0; i < n; i++) {
+        p += snprintf(p, p_end - p, "%d, ", (int) x[i]);
+    }
+    FARF(HIGH, "%s\n", str);
+}
+
+static inline void hex_dump_f16_line(char * pref, const __fp16 * x, uint32_t n) {
+    char str[1024], *p = str, *p_end = str + sizeof(str);
+    p += snprintf(p, p_end - p, "%s: ", pref);
+    for (int i = 0; i < n; i++) {
+        p += snprintf(p, p_end - p, "%.6f, ", (float) x[i]);
+    }
+    FARF(HIGH, "%s\n", str);
+}
+
+static inline void hex_dump_f32_line(char * pref, const float * x, uint32_t n) {
+    char str[1024], *p = str, *p_end = str + sizeof(str);
+    p += snprintf(p, p_end - p, "%s: ", pref);
+    for (int i = 0; i < n; i++) {
+        p += snprintf(p, p_end - p, "%.6f, ", x[i]);
+    }
+    FARF(HIGH, "%s\n", str);
+}
+
+static inline void hex_dump_f32(char * pref, const float * x, uint32_t n) {
+    uint32_t n0 = n / 16;
+    uint32_t n1 = n % 16;
+
+    uint32_t i = 0;
+    for (; i < n0; i++) {
+        hex_dump_f32_line(pref, x + (16 * i), 16);
+    }
+    if (n1) {
+        hex_dump_f32_line(pref, x + (16 * i), n1);
+    }
+}
+
+static inline void hex_dump_f16(char * pref, const __fp16 * x, uint32_t n) {
+    uint32_t n0 = n / 16;
+    uint32_t n1 = n % 16;
+
+    uint32_t i = 0;
+    for (; i < n0; i++) {
+        hex_dump_f16_line(pref, x + (16 * i), 16);
+    }
+    if (n1) {
+        hex_dump_f16_line(pref, x + (16 * i), n1);
+    }
+}
+
+#endif /* HEX_DUMP_H */

package/cpp/ggml/src/ggml-hexagon/htp/hex-fastdiv.h

@@ -0,0 +1,37 @@
+#ifndef HEX_FASTDIV_H
+#define HEX_FASTDIV_H
+
+// See https://gmplib.org/~tege/divcnst-pldi94.pdf figure 4.1.
+// Precompute mp (m' in the paper) and L such that division
+// can be computed using a multiply (high 32b of 64b result)
+// and a shift:
+//
+// n/d = (mulhi(n, mp) + n) >> L;
+struct fastdiv_values {
+    uint32_t mp;
+    uint32_t l;
+};
+
+static inline struct fastdiv_values init_fastdiv_values(uint32_t d) {
+    struct fastdiv_values result = { 0, 0 };
+    // compute L = ceil(log2(d));
+    while (result.l < 32 && ((uint32_t) 1 << result.l) < d) {
+        ++(result.l);
+    }
+
+    result.mp = (uint32_t) (((uint64_t) 1 << 32) * (((uint64_t) 1 << result.l) - d) / d + 1);
+    return result;
+}
+
+static inline uint32_t fastdiv(uint32_t n, const struct fastdiv_values * vals) {
+    // Compute high 32 bits of n * mp
+    const uint32_t hi = (uint32_t) (((uint64_t) n * vals->mp) >> 32);  // mulhi(n, mp)
+    // add n, apply bit shift
+    return (hi + n) >> vals->l;
+}
+
+static inline uint32_t fastmodulo(uint32_t n, uint32_t d, const struct fastdiv_values * vals) {
+    return n - fastdiv(n, vals) * d;
+}
+
+#endif /* HEX_FASTDIV_H */

package/cpp/ggml/src/ggml-hexagon/htp/hex-utils.h

@@ -0,0 +1,51 @@
+#ifndef HEX_UTILS_H
+#define HEX_UTILS_H
+
+#include <stdbool.h>
+#include <stdint.h>
+
+#include "hexagon_types.h"
+
+#include "hex-fastdiv.h"
+#include "hex-dump.h"
+
+#ifndef MAX
+#define MAX(a, b) ((a) > (b) ? (a) : (b))
+#endif
+
+#ifndef MIN
+#define MIN(a, b) ((a) < (b) ? (a) : (b))
+#endif
+
+static inline uint64_t hex_get_cycles() {
+    uint64_t cycles = 0;
+    asm volatile(" %0 = c15:14\n" : "=r"(cycles));
+    return cycles;
+}
+
+static inline uint64_t hex_get_pktcnt() {
+    uint64_t pktcnt;
+    asm volatile(" %0 = c19:18\n" : "=r"(pktcnt));
+    return pktcnt;
+}
+
+static inline int32_t hex_is_aligned(void * addr, uint32_t align) {
+    return ((size_t) addr & (align - 1)) == 0;
+}
+
+static inline int32_t hex_is_one_chunk(void * addr, uint32_t n, uint32_t chunk_size) {
+    uint32_t left_off  = (size_t) addr & (chunk_size - 1);
+    uint32_t right_off = left_off + n;
+    return right_off <= chunk_size;
+}
+
+static inline uint32_t hex_round_up(uint32_t n, uint32_t m) {
+    return m * ((n + m - 1) / m);
+}
+
+static inline void hex_l2fetch(const void * p, uint32_t width, uint32_t stride, uint32_t height) {
+    const uint64_t control = Q6_P_combine_RR(stride, Q6_R_combine_RlRl(width, height));
+    Q6_l2fetch_AP((void *) p, control);
+}
+
+#endif /* HEX_UTILS_H */

package/cpp/ggml/src/ggml-hexagon/htp/htp-ctx.h

@@ -0,0 +1,35 @@
+#ifndef HTP_CTX_H
+#define HTP_CTX_H
+
+#include "hex-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 */