// 576p60/720p60/1080i60 works, 1080p60 does not work (firmware limitation)
// Audio comes out as 8-channel 24-bit raw audio.
+#if (defined(__i386__) || defined(__x86_64__)) && defined(__GNUC__)
+#define HAS_MULTIVERSIONING 1
+#endif
+
#include <assert.h>
#include <errno.h>
#include <libusb.h>
+#include <unistd.h>
#include <netinet/in.h>
#include <sched.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
-#ifdef __SSE4_1__
+#if HAS_MULTIVERSIONING
#include <immintrin.h>
#endif
#include "bmusb.h"
using namespace std;
using namespace std::placeholders;
+#define USB_VENDOR_BLACKMAGIC 0x1edb
#define MIN_WIDTH 640
#define HEADER_SIZE 44
//#define HEADER_SIZE 0
#define FRAME_SIZE (8 << 20) // 8 MB.
#define USB_VIDEO_TRANSFER_SIZE (128 << 10) // 128 kB.
+card_connected_callback_t BMUSBCapture::card_connected_callback = nullptr;
+
namespace {
FILE *audiofp;
}
}
-#ifdef __SSE4_1__
-
#if 0
void avx2_dump(const char *name, __m256i n)
{
}
#endif
+#ifndef HAS_MULTIVERSIONING
+
+const uint8_t *add_to_frame_fastpath(FrameAllocator::Frame *current_frame, const uint8_t *start, const uint8_t *limit, const char sync_char)
+{
+ // No fast path possible unless we have multiversioning.
+ return start;
+}
+
+#else // defined(HAS_MULTIVERSIONING)
+
+const uint8_t *add_to_frame_fastpath_core(FrameAllocator::Frame *current_frame, const uint8_t *aligned_start, const uint8_t *limit, const char sync_char);
+
// Does a memcpy and memchr in one to reduce processing time.
// Note that the benefit is somewhat limited if your L3 cache is small,
// as you'll (unfortunately) spend most of the time loading the data
// up until the first instance of "sync_char" (usually a bit before, actually).
// This is fine, since 0x00 bytes shouldn't really show up in normal picture
// data, and what we really need this for is the 00 00 ff ff marker in video data.
+__attribute__((target("default")))
+const uint8_t *add_to_frame_fastpath(FrameAllocator::Frame *current_frame, const uint8_t *start, const uint8_t *limit, const char sync_char)
+{
+ // No fast path possible unless we have SSE 4.1 or higher.
+ return start;
+}
+
+__attribute__((target("sse4.1", "avx2")))
const uint8_t *add_to_frame_fastpath(FrameAllocator::Frame *current_frame, const uint8_t *start, const uint8_t *limit, const char sync_char)
{
if (current_frame->data == nullptr ||
assert(((limit - aligned_start) % 64) == 0);
}
-#if __AVX2__
+ return add_to_frame_fastpath_core(current_frame, aligned_start, limit, sync_char);
+}
+
+__attribute__((target("avx2")))
+const uint8_t *add_to_frame_fastpath_core(FrameAllocator::Frame *current_frame, const uint8_t *aligned_start, const uint8_t *limit, const char sync_char)
+{
const __m256i needle = _mm256_set1_epi8(sync_char);
const __restrict __m256i *in = (const __m256i *)aligned_start;
}
current_frame->len = (uint8_t *)out - current_frame->data;
}
-#else
+
+ //printf("managed to fastpath %ld/%ld bytes\n", (const uint8_t *)in - (const uint8_t *)aligned_start, orig_bytes);
+ return (const uint8_t *)in;
+}
+
+__attribute__((target("sse4.1")))
+const uint8_t *add_to_frame_fastpath_core(FrameAllocator::Frame *current_frame, const uint8_t *aligned_start, const uint8_t *limit, const char sync_char)
+{
const __m128i needle = _mm_set1_epi8(sync_char);
const __m128i *in = (const __m128i *)aligned_start;
}
current_frame->len = (uint8_t *)out - current_frame->data;
}
-#endif
//printf("managed to fastpath %ld/%ld bytes\n", (const uint8_t *)in - (const uint8_t *)aligned_start, orig_bytes);
-
return (const uint8_t *)in;
}
-#endif
+
+#endif // defined(HAS_MULTIVERSIONING)
void decode_packs(const libusb_transfer *xfr,
const char *sync_pattern,
const uint8_t *start = xfr->buffer + offset;
const uint8_t *limit = start + pack->actual_length;
while (start < limit) { // Usually runs only one iteration.
-#ifdef __SSE4_1__
start = add_to_frame_fastpath(current_frame, start, limit, sync_pattern[0]);
if (start == limit) break;
assert(start < limit);
-#endif
const unsigned char* start_next_frame = (const unsigned char *)memmem(start, limit - start, sync_pattern, sync_length);
if (start_next_frame == nullptr) {
void BMUSBCapture::cb_xfr(struct libusb_transfer *xfr)
{
- if (xfr->status != LIBUSB_TRANSFER_COMPLETED) {
- fprintf(stderr, "transfer status %d\n", xfr->status);
+ if (xfr->status != LIBUSB_TRANSFER_COMPLETED &&
+ xfr->status != LIBUSB_TRANSFER_NO_DEVICE) {
+ fprintf(stderr, "error: transfer status %d\n", xfr->status);
libusb_free_transfer(xfr);
exit(3);
}
assert(xfr->user_data != nullptr);
BMUSBCapture *usb = static_cast<BMUSBCapture *>(xfr->user_data);
+ if (xfr->status == LIBUSB_TRANSFER_NO_DEVICE) {
+ if (!usb->disconnected) {
+ fprintf(stderr, "Device went away, stopping transfers.\n");
+ usb->disconnected = true;
+ if (usb->card_disconnected_callback) {
+ usb->card_disconnected_callback();
+ }
+ }
+ // Don't reschedule the transfer; the loop will stop by itself.
+ return;
+ }
+
if (xfr->type == LIBUSB_TRANSFER_TYPE_ISOCHRONOUS) {
if (xfr->endpoint == 0x84) {
decode_packs(xfr, "DeckLinkAudioResyncT", 20, &usb->current_audio_frame, "audio", bind(&BMUSBCapture::start_new_audio_block, usb, _1));
}
}
+int BMUSBCapture::cb_hotplug(libusb_context *ctx, libusb_device *dev, libusb_hotplug_event event, void *user_data)
+{
+ if (card_connected_callback != nullptr) {
+ libusb_device_descriptor desc;
+ if (libusb_get_device_descriptor(dev, &desc) < 0) {
+ fprintf(stderr, "Error getting device descriptor for hotplugged device %p, killing hotplug\n", dev);
+ libusb_unref_device(dev);
+ return 1;
+ }
+
+ if ((desc.idVendor == USB_VENDOR_BLACKMAGIC && desc.idProduct == 0xbd3b) ||
+ (desc.idVendor == USB_VENDOR_BLACKMAGIC && desc.idProduct == 0xbd4f)) {
+ card_connected_callback(dev); // Callback takes ownership.
+ return 0;
+ }
+ }
+ libusb_unref_device(dev);
+ return 0;
+}
+
void BMUSBCapture::usb_thread_func()
{
sched_param param;
libusb_device *device;
};
+const char *get_product_name(uint16_t product)
+{
+ if (product == 0xbd3b) {
+ return "Intensity Shuttle";
+ } else if (product == 0xbd4f) {
+ return "UltraStudio SDI";
+ } else {
+ assert(false);
+ return nullptr;
+ }
+}
+
+string get_card_description(int id, uint8_t bus, uint8_t port, uint16_t product)
+{
+ const char *product_name = get_product_name(product);
+
+ char buf[256];
+ snprintf(buf, sizeof(buf), "USB card %d: Bus %03u Device %03u %s",
+ id, bus, port, product_name);
+ return buf;
+}
+
libusb_device_handle *open_card(int card_index, string *description)
-{
+{
libusb_device **devices;
ssize_t num_devices = libusb_get_device_list(nullptr, &devices);
if (num_devices == -1) {
uint8_t bus = libusb_get_bus_number(devices[i]);
uint8_t port = libusb_get_port_number(devices[i]);
- if (!(desc.idVendor == 0x1edb && desc.idProduct == 0xbd3b) &&
- !(desc.idVendor == 0x1edb && desc.idProduct == 0xbd4f)) {
+ if (!(desc.idVendor == USB_VENDOR_BLACKMAGIC && desc.idProduct == 0xbd3b) &&
+ !(desc.idVendor == USB_VENDOR_BLACKMAGIC && desc.idProduct == 0xbd4f)) {
libusb_unref_device(devices[i]);
continue;
}
});
for (size_t i = 0; i < found_cards.size(); ++i) {
- const char *product_name = nullptr;
- if (found_cards[i].product == 0xbd3b) {
- product_name = "Intensity Shuttle";
- } else if (found_cards[i].product == 0xbd4f) {
- product_name = "UltraStudio SDI";
- } else {
- assert(false);
- }
-
- char buf[256];
- snprintf(buf, sizeof(buf), "USB card %d: Bus %03u Device %03u %s",
- int(i), found_cards[i].bus, found_cards[i].port, product_name);
+ string tmp_description = get_card_description(i, found_cards[i].bus, found_cards[i].port, found_cards[i].product);
+ fprintf(stderr, "%s\n", tmp_description.c_str());
if (i == size_t(card_index)) {
- *description = buf;
+ *description = tmp_description;
}
- fprintf(stderr, "%s\n", buf);
}
if (size_t(card_index) >= found_cards.size()) {
return devh;
}
+libusb_device_handle *open_card(unsigned card_index, libusb_device *dev, string *description)
+{
+ uint8_t bus = libusb_get_bus_number(dev);
+ uint8_t port = libusb_get_port_number(dev);
+
+ libusb_device_descriptor desc;
+ if (libusb_get_device_descriptor(dev, &desc) < 0) {
+ fprintf(stderr, "Error getting device descriptor for device %p\n", dev);
+ exit(1);
+ }
+
+ *description = get_card_description(card_index, bus, port, desc.idProduct);
+
+ libusb_device_handle *devh;
+ int rc = libusb_open(dev, &devh);
+ if (rc < 0) {
+ fprintf(stderr, "Error opening card %p: %s\n", dev, libusb_error_name(rc));
+ exit(1);
+ }
+
+ return devh;
+}
+
void BMUSBCapture::configure_card()
{
if (video_frame_allocator == nullptr) {
exit(1);
}
- devh = open_card(card_index, &description);
+ if (dev == nullptr) {
+ devh = open_card(card_index, &description);
+ } else {
+ devh = open_card(card_index, dev, &description);
+ libusb_unref_device(dev);
+ }
if (!devh) {
fprintf(stderr, "Error finding USB device\n");
exit(1);
void BMUSBCapture::start_bm_thread()
{
+ // Devices leaving are discovered by seeing the isochronous packets
+ // coming back with errors, so only care about devices joining.
+ if (card_connected_callback != nullptr) {
+ if (libusb_hotplug_register_callback(
+ nullptr, LIBUSB_HOTPLUG_EVENT_DEVICE_ARRIVED, LIBUSB_HOTPLUG_NO_FLAGS,
+ USB_VENDOR_BLACKMAGIC, LIBUSB_HOTPLUG_MATCH_ANY, LIBUSB_HOTPLUG_MATCH_ANY,
+ &BMUSBCapture::cb_hotplug, nullptr, nullptr) < 0) {
+ fprintf(stderr, "libusb_hotplug_register_callback() failed\n");
+ exit(1);
+ }
+ }
+
should_quit = false;
usb_thread = thread(&BMUSBCapture::usb_thread_func);
}