Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Prefetch MFC list elements #5345

Merged
merged 6 commits into from Jul 23, 2019
Merged

Prefetch MFC list elements #5345

Changes from 5 commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
81431f4
Prefetch mfc list elemets to protect from overwriting
elad335 Nov 15, 2018
3f254bd
cleanup
elad335 Jul 20, 2019
5317cda
Merge branch 'master' into spu2
elad335 Jul 20, 2019
6c5a57f
Alice: the union madness fixed [to squash merge]
elad335 Jul 23, 2019
0de64d4
Merge branch 'master' into spu2
elad335 Jul 23, 2019
b52a851
happy optimization
elad335 Jul 23, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
94 changes: 70 additions & 24 deletions rpcs3/Emu/Cell/SPUThread.cpp
View file
Open in desktop
Expand Up @@ -1500,55 +1500,101 @@ bool spu_thread::do_dma_check(const spu_mfc_cmd& args)

bool spu_thread::do_list_transfer(spu_mfc_cmd& args)
{
struct list_element
// Amount of elements to fetch in one go
constexpr u32 fetch_size = 6;

struct alignas(8) list_element
{
be_t<u16> sb; // Stall-and-Notify bit (0x8000)
be_t<u16> ts; // List Transfer Size
be_t<u32> ea; // External Address Low
} item{};
};

while (args.size)
union
{
if (UNLIKELY(item.sb & 0x8000))
{
ch_stall_mask |= utils::rol32(1, args.tag);
list_element items[fetch_size];
alignas(v128) char bufitems[sizeof(items)];
};

if (!ch_stall_stat.get_count())
spu_mfc_cmd transfer;
transfer.eah = 0;
transfer.tag = args.tag;
transfer.cmd = MFC(args.cmd & ~MFC_LIST_MASK);

args.lsa &= 0x3fff0;

u32 index = fetch_size;

// Assume called with size greater than 0
while (true)
{
// Check if fetching is needed
if (index == fetch_size)
{
const auto src = _ptr<void>(args.eal & 0x3fff8);
if ((uptr)src % alignof(v128))
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can union be replaced with simple alignas(16) list_element items[fetch_size];, and the whole alignment check with two pathes with single memcpy?

{
ch_event_stat |= SPU_EVENT_SN;
// Unaligned
((u64*)+bufitems)[0] = ((u64*)src)[0];
((u64*)+bufitems)[1] = ((u64*)src)[1];
((u64*)+bufitems)[2] = ((u64*)src)[2];
((u64*)+bufitems)[3] = ((u64*)src)[3];
((u64*)+bufitems)[4] = ((u64*)src)[4];
((u64*)+bufitems)[5] = ((u64*)src)[5];
}
else
{
((v128*)+bufitems)[0] = ((v128*)src)[0];
((v128*)+bufitems)[1] = ((v128*)src)[1];
((v128*)+bufitems)[2] = ((v128*)src)[2];
}

ch_stall_stat.set_value(utils::rol32(1, args.tag) | ch_stall_stat.get_value());

args.tag |= 0x80; // Set stalled status
return false;
// Reset to elements array head
index = 0;
}

args.lsa &= 0x3fff0;
item = _ref<list_element>(args.eal & 0x3fff8);
const u32 size = items[index].ts & 0x7fff;
const u32 addr = items[index].ea;

const u32 size = item.ts & 0x7fff;
const u32 addr = item.ea;

LOG_TRACE(SPU, "LIST: addr=0x%x, size=0x%x, lsa=0x%05x, sb=0x%x", addr, size, args.lsa | (addr & 0xf), item.sb);
LOG_TRACE(SPU, "LIST: addr=0x%x, size=0x%x, lsa=0x%05x, sb=0x%x", addr, size, args.lsa | (addr & 0xf), items[index].sb);

if (size)
{
spu_mfc_cmd transfer;
transfer.eal = addr;
transfer.eah = 0;
transfer.lsa = args.lsa | (addr & 0xf);
transfer.tag = args.tag;
transfer.cmd = MFC(args.cmd & ~MFC_LIST_MASK);
transfer.size = size;

do_dma_transfer(transfer);
const u32 add_size = std::max<u32>(size, 16);
args.lsa += add_size;
}

args.eal += 8;
args.size -= 8;

if (!args.size)
{
// No more elements
break;
}

args.eal += 8;

if (UNLIKELY(items[index].sb & 0x8000))
{
ch_stall_mask |= utils::rol32(1, args.tag);

if (!ch_stall_stat.get_count())
{
ch_event_stat |= SPU_EVENT_SN;
}

ch_stall_stat.set_value(utils::rol32(1, args.tag) | ch_stall_stat.get_value());

args.tag |= 0x80; // Set stalled status
return false;
}

index++;
}

return true;
Expand Down Expand Up @@ -2040,7 +2086,7 @@ bool spu_thread::process_mfc_cmd()

if (LIKELY(do_dma_check(cmd)))
{
if (LIKELY(do_list_transfer(cmd)))
if (LIKELY(!cmd.size || do_list_transfer(cmd)))
{
return true;
}
Expand Down