fabi/Xenia-Canary
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

use much faster exp2/cos approximations in ffmpeg, large decrease in cpu usage on my machine on decoder thread

Browse Source 
properly byteswap r13 for spinlock
Add PPCOpcodeBits
stub out broken fpscr updating in ppc_hir_builder. it's just code that repeatedly does nothing right now.
add note about 0 opcode bytes being executed to ppc_frontend
Add assert to check that function end is greater than function start, can happen with malformed functions
Disable prefetch and cachecontrol by default, automatic hardware prefetchers already do the job for the most part
minor cleanup in simplification_pass, dont loop optimizations, let the pass manager do it for us
Add experimental "delay_via_maybeyield" cvar, which uses MaybeYield to "emulate" the db16cyc instruction
Add much faster/simpler way of directly calling guest functions, no longer have to do a byte by byte search through the generated code
Generate label string ids on the fly
Fix unused function warnings for prefetch on clang, fix many other clang warnings
Eliminated majority of CallNativeSafes by replacing them with naive generic code paths.
^ Vector rotate left, vector shift left, vector shift right, vector shift arithmetic right, and vector average are included
These naive paths are implemented small loops that stash the two inputs to the stack and load them in gprs from there, they are not particularly fast but should be an order of magnitude faster than callnativesafe
to a host function, which would involve a call, stashing all volatile registers, an indirect call, potentially setting up a stack frame for the arrays that the inputs get stashed to, the actual operations, a return, loading all volatile registers, a return, etc
Added the fast SHR_V128 path back in
Implement signed vector average byte, signed vector average word. previously we were emitting no code for them. signed vector average byte appears in many games
Fix bug with signed vector average 32, we were doing unsigned shift, turning negative values into big positive ones potentially
This commit is contained in:
chss95cs@gmail.com
2022-10-30 08:48:58 -07:00
parent d8b7b3ecec
commit 550d1d0a7c
26 changed files with 881 additions and 547 deletions
Download Patch File Download Diff File Expand all files Collapse all files

28
src/xenia/kernel/xboxkrnl/xboxkrnl_threading.cc
View File

@@ -9,7 +9,6 @@
#include <algorithm>
#include <vector>
#include "xenia/base/atomic.h"
#include "xenia/base/clock.h"
#include "xenia/base/logging.h"
@@ -964,7 +963,7 @@ uint32_t xeKeKfAcquireSpinLock(uint32_t* lock, uint64_t r13 = 1) {
PrefetchForCAS(lock);
assert_true(*lock != static_cast<uint32_t>(r13));
// Lock.
while (!xe::atomic_cas(0, static_cast<uint32_t>(r13), lock)) {
while (!xe::atomic_cas(0, xe::byte_swap(static_cast<uint32_t>(r13)), lock)) {
// Spin!
// TODO(benvanik): error on deadlock?
xe::threading::MaybeYield();
@@ -978,7 +977,7 @@ uint32_t xeKeKfAcquireSpinLock(uint32_t* lock, uint64_t r13 = 1) {
}
dword_result_t KfAcquireSpinLock_entry(lpdword_t lock_ptr,
ppc_context_t& ppc_context) {
const ppc_context_t& ppc_context) {
auto lock = reinterpret_cast<uint32_t*>(lock_ptr.host_address());
return xeKeKfAcquireSpinLock(lock, ppc_context->r[13]);
}
@@ -997,7 +996,7 @@ void xeKeKfReleaseSpinLock(uint32_t* lock, dword_t old_irql) {
}
void KfReleaseSpinLock_entry(lpdword_t lock_ptr, dword_t old_irql,
ppc_context_t& ppc_ctx) {
const ppc_context_t& ppc_ctx) {
auto lock = reinterpret_cast<uint32_t*>(lock_ptr.host_address());
assert_true(*lock_ptr == static_cast<uint32_t>(ppc_ctx->r[13]));
@@ -1014,14 +1013,14 @@ DECLARE_XBOXKRNL_EXPORT2(KfReleaseSpinLock, kThreading, kImplemented,
kHighFrequency);
// todo: this is not accurate
void KeAcquireSpinLockAtRaisedIrql_entry(lpdword_t lock_ptr,
ppc_context_t& ppc_ctx) {
const ppc_context_t& ppc_ctx) {
// Lock.
auto lock = reinterpret_cast<uint32_t*>(lock_ptr.host_address());
// must not be our own thread
assert_true(*lock_ptr != static_cast<uint32_t>(ppc_ctx->r[13]));
PrefetchForCAS(lock);
while (!xe::atomic_cas(0, static_cast<uint32_t>(ppc_ctx->r[13]), lock)) {
while (!xe::atomic_cas(0, xe::byte_swap(static_cast<uint32_t>(ppc_ctx->r[13])), lock)) {
#if XE_ARCH_AMD64 == 1
// todo: this is just a nop if they don't have SMT, which is not great
// either...
@@ -1036,12 +1035,12 @@ DECLARE_XBOXKRNL_EXPORT3(KeAcquireSpinLockAtRaisedIrql, kThreading,
kImplemented, kBlocking, kHighFrequency);
dword_result_t KeTryToAcquireSpinLockAtRaisedIrql_entry(
lpdword_t lock_ptr, ppc_context_t& ppc_ctx) {
lpdword_t lock_ptr, const ppc_context_t& ppc_ctx) {
// Lock.
auto lock = reinterpret_cast<uint32_t*>(lock_ptr.host_address());
assert_true(*lock_ptr != static_cast<uint32_t>(ppc_ctx->r[13]));
PrefetchForCAS(lock);
if (!xe::atomic_cas(0, static_cast<uint32_t>(ppc_ctx->r[13]), lock)) {
if (!xe::atomic_cas(0, xe::byte_swap(static_cast<uint32_t>(ppc_ctx->r[13])), lock)) {
return 0;
}
return 1;
@@ -1050,7 +1049,7 @@ DECLARE_XBOXKRNL_EXPORT4(KeTryToAcquireSpinLockAtRaisedIrql, kThreading,
kImplemented, kBlocking, kHighFrequency, kSketchy);
void KeReleaseSpinLockFromRaisedIrql_entry(lpdword_t lock_ptr,
ppc_context_t& ppc_ctx) {
const ppc_context_t& ppc_ctx) {
// Unlock.
assert_true(*lock_ptr == static_cast<uint32_t>(ppc_ctx->r[13]));
auto lock = reinterpret_cast<uint32_t*>(lock_ptr.host_address());
@@ -1283,7 +1282,8 @@ void ExInitializeReadWriteLock_entry(pointer_t<X_ERWLOCK> lock_ptr) {
}
DECLARE_XBOXKRNL_EXPORT1(ExInitializeReadWriteLock, kThreading, kImplemented);
void ExAcquireReadWriteLockExclusive_entry(pointer_t<X_ERWLOCK> lock_ptr, ppc_context_t& ppc_context) {
void ExAcquireReadWriteLockExclusive_entry(pointer_t<X_ERWLOCK> lock_ptr,
const ppc_context_t& ppc_context) {
auto old_irql = xeKeKfAcquireSpinLock(&lock_ptr->spin_lock, ppc_context->r[13]);
int32_t lock_count = ++lock_ptr->lock_count;
@@ -1301,7 +1301,7 @@ DECLARE_XBOXKRNL_EXPORT2(ExAcquireReadWriteLockExclusive, kThreading,
kImplemented, kBlocking);
dword_result_t ExTryToAcquireReadWriteLockExclusive_entry(
pointer_t<X_ERWLOCK> lock_ptr, ppc_context_t& ppc_context) {
pointer_t<X_ERWLOCK> lock_ptr, const ppc_context_t& ppc_context) {
auto old_irql =
xeKeKfAcquireSpinLock(&lock_ptr->spin_lock, ppc_context->r[13]);
@@ -1320,7 +1320,7 @@ DECLARE_XBOXKRNL_EXPORT1(ExTryToAcquireReadWriteLockExclusive, kThreading,
kImplemented);
void ExAcquireReadWriteLockShared_entry(pointer_t<X_ERWLOCK> lock_ptr,
ppc_context_t& ppc_context) {
const ppc_context_t& ppc_context) {
auto old_irql = xeKeKfAcquireSpinLock(&lock_ptr->spin_lock, ppc_context->r[13]);
int32_t lock_count = ++lock_ptr->lock_count;
@@ -1340,7 +1340,7 @@ DECLARE_XBOXKRNL_EXPORT2(ExAcquireReadWriteLockShared, kThreading, kImplemented,
kBlocking);
dword_result_t ExTryToAcquireReadWriteLockShared_entry(
pointer_t<X_ERWLOCK> lock_ptr, ppc_context_t& ppc_context) {
pointer_t<X_ERWLOCK> lock_ptr, const ppc_context_t& ppc_context) {
auto old_irql =
xeKeKfAcquireSpinLock(&lock_ptr->spin_lock, ppc_context->r[13]);
@@ -1361,7 +1361,7 @@ DECLARE_XBOXKRNL_EXPORT1(ExTryToAcquireReadWriteLockShared, kThreading,
kImplemented);
void ExReleaseReadWriteLock_entry(pointer_t<X_ERWLOCK> lock_ptr,
ppc_context_t& ppc_context) {
const ppc_context_t& ppc_context) {
auto old_irql =
xeKeKfAcquireSpinLock(&lock_ptr->spin_lock, ppc_context->r[13]);