February 2023 - Wine-commits

Rémi Bernon : ntdll: Implement HeapCompatibilityInformation.
by Alexandre Julliard 22 Feb '23

22 Feb '23

Module: wine Branch: master Commit: 7314029c907a5760fcf7502d37784e079078adf5 URL: https://gitlab.winehq.org/wine/wine/-/commit/7314029c907a5760fcf7502d37784e… Author: Rémi Bernon <rbernon(a)codeweavers.com> Date: Thu Dec 1 11:18:27 2022 +0100 ntdll: Implement HeapCompatibilityInformation. --- dlls/kernel32/tests/heap.c | 34 +++++++++++++++++++-------- dlls/ntdll/heap.c | 58 +++++++++++++++++++++++++++++++++++++++------- 2 files changed, 73 insertions(+), 19 deletions(-) diff --git a/dlls/kernel32/tests/heap.c b/dlls/kernel32/tests/heap.c index a81cb1b7a15..307b11771b3 100644 --- a/dlls/kernel32/tests/heap.c +++ b/dlls/kernel32/tests/heap.c @@ -827,11 +827,8 @@ static void test_HeapCreate(void) size = 0; SetLastError( 0xdeadbeef ); ret = pHeapQueryInformation( 0, HeapCompatibilityInformation, &compat_info, sizeof(compat_info), &size ); - todo_wine ok( !ret, "HeapQueryInformation succeeded\n" ); - todo_wine ok( GetLastError() == ERROR_NOACCESS, "got error %lu\n", GetLastError() ); - todo_wine ok( size == 0, "got size %Iu\n", size ); size = 0; @@ -871,7 +868,6 @@ static void test_HeapCreate(void) ok( ret, "HeapSetInformation failed, error %lu\n", GetLastError() ); ret = pHeapQueryInformation( heap, HeapCompatibilityInformation, &compat_info, sizeof(compat_info), &size ); ok( ret, "HeapQueryInformation failed, error %lu\n", GetLastError() ); - todo_wine ok( compat_info == 2, "got HeapCompatibilityInformation %lu\n", compat_info ); /* cannot be undone */ @@ -879,25 +875,44 @@ static void test_HeapCreate(void) compat_info = 0; SetLastError( 0xdeadbeef ); ret = pHeapSetInformation( heap, HeapCompatibilityInformation, &compat_info, sizeof(compat_info) ); - todo_wine ok( !ret, "HeapSetInformation succeeded\n" ); - todo_wine ok( GetLastError() == ERROR_GEN_FAILURE, "got error %lu\n", GetLastError() ); compat_info = 1; SetLastError( 0xdeadbeef ); ret = pHeapSetInformation( heap, HeapCompatibilityInformation, &compat_info, sizeof(compat_info) ); - todo_wine ok( !ret, "HeapSetInformation succeeded\n" ); - todo_wine ok( GetLastError() == ERROR_GEN_FAILURE, "got error %lu\n", GetLastError() ); ret = pHeapQueryInformation( heap, HeapCompatibilityInformation, &compat_info, sizeof(compat_info), &size ); ok( ret, "HeapQueryInformation failed, error %lu\n", GetLastError() ); - todo_wine ok( compat_info == 2, "got HeapCompatibilityInformation %lu\n", compat_info ); ret = HeapDestroy( heap ); ok( ret, "HeapDestroy failed, error %lu\n", GetLastError() ); + + /* cannot set LFH with HEAP_NO_SERIALIZE */ + + heap = HeapCreate( HEAP_NO_SERIALIZE, 0, 0 ); + ok( !!heap, "HeapCreate failed, error %lu\n", GetLastError() ); + ok( !((ULONG_PTR)heap & 0xffff), "wrong heap alignment\n" ); + + ret = pHeapQueryInformation( heap, HeapCompatibilityInformation, &compat_info, sizeof(compat_info), &size ); + ok( ret, "HeapQueryInformation failed, error %lu\n", GetLastError() ); + ok( compat_info == 0, "got HeapCompatibilityInformation %lu\n", compat_info ); + + compat_info = 2; + SetLastError( 0xdeadbeef ); + ret = pHeapSetInformation( heap, HeapCompatibilityInformation, &compat_info, sizeof(compat_info) ); + ok( !ret, "HeapSetInformation succeeded\n" ); + ok( GetLastError() == ERROR_INVALID_PARAMETER, "got error %lu\n", GetLastError() ); + ret = pHeapQueryInformation( heap, HeapCompatibilityInformation, &compat_info, sizeof(compat_info), &size ); + ok( ret, "HeapQueryInformation failed, error %lu\n", GetLastError() ); + ok( compat_info == 0, "got HeapCompatibilityInformation %lu\n", compat_info ); + + ret = HeapDestroy( heap ); + ok( ret, "HeapDestroy failed, error %lu\n", GetLastError() ); + + /* some allocation pattern automatically enables LFH */ heap = HeapCreate( 0, 0, 0 ); @@ -931,7 +946,6 @@ static void test_HeapCreate(void) ok( ret, "HeapSetInformation failed, error %lu\n", GetLastError() ); ret = pHeapQueryInformation( heap, HeapCompatibilityInformation, &compat_info, sizeof(compat_info), &size ); ok( ret, "HeapQueryInformation failed, error %lu\n", GetLastError() ); - todo_wine ok( compat_info == 2, "got HeapCompatibilityInformation %lu\n", compat_info ); for (i = 0; i < 0x11; i++) ptrs[i] = pHeapAlloc( heap, 0, 24 + 2 * sizeof(void *) ); diff --git a/dlls/ntdll/heap.c b/dlls/ntdll/heap.c index a2410957a44..32808754ddb 100644 --- a/dlls/ntdll/heap.c +++ b/dlls/ntdll/heap.c @@ -39,6 +39,13 @@ WINE_DEFAULT_DEBUG_CHANNEL(heap); +/* HeapCompatibilityInformation values */ + +#define HEAP_STD 0 +#define HEAP_LAL 1 +#define HEAP_LFH 2 + + /* undocumented RtlWalkHeap structure */ struct rtl_heap_entry @@ -194,6 +201,7 @@ struct heap DWORD force_flags; /* 0044/0074 */ /* end of the Windows 10 compatible struct layout */ + LONG compat_info; /* HeapCompatibilityInformation / heap frontend type */ struct list entry; /* Entry in process heap list */ struct list subheap_list; /* Sub-heap list */ struct list large_list; /* Large blocks list */ @@ -1378,6 +1386,7 @@ HANDLE WINAPI RtlCreateHeap( ULONG flags, void *addr, SIZE_T total_size, SIZE_T heap->ffeeffee = 0xffeeffee; heap->auto_flags = (flags & HEAP_GROWABLE); heap->flags = (flags & ~HEAP_SHARED); + heap->compat_info = HEAP_STD; heap->magic = HEAP_MAGIC; heap->grow_size = max( HEAP_DEF_SIZE, total_size ); heap->min_size = commit_size; @@ -2039,21 +2048,24 @@ ULONG WINAPI RtlGetProcessHeaps( ULONG count, HANDLE *heaps ) * RtlQueryHeapInformation (NTDLL.@) */ NTSTATUS WINAPI RtlQueryHeapInformation( HANDLE handle, HEAP_INFORMATION_CLASS info_class, - void *info, SIZE_T size_in, PSIZE_T size_out ) + void *info, SIZE_T size_in, SIZE_T *size_out ) { + struct heap *heap; + ULONG flags; + + TRACE( "handle %p, info_class %u, info %p, size_in %Iu, size_out %p.\n", handle, info_class, info, size_in, size_out ); + switch (info_class) { case HeapCompatibilityInformation: + if (!(heap = unsafe_heap_from_handle( handle, 0, &flags ))) return STATUS_ACCESS_VIOLATION; if (size_out) *size_out = sizeof(ULONG); - - if (size_in < sizeof(ULONG)) - return STATUS_BUFFER_TOO_SMALL; - - *(ULONG *)info = 0; /* standard heap */ + if (size_in < sizeof(ULONG)) return STATUS_BUFFER_TOO_SMALL; + *(ULONG *)info = ReadNoFence( &heap->compat_info ); return STATUS_SUCCESS; default: - FIXME("Unknown heap information class %u\n", info_class); + FIXME( "HEAP_INFORMATION_CLASS %u not implemented!\n", info_class ); return STATUS_INVALID_INFO_CLASS; } } @@ -2063,8 +2075,36 @@ NTSTATUS WINAPI RtlQueryHeapInformation( HANDLE handle, HEAP_INFORMATION_CLASS i */ NTSTATUS WINAPI RtlSetHeapInformation( HANDLE handle, HEAP_INFORMATION_CLASS info_class, void *info, SIZE_T size ) { - FIXME( "handle %p, info_class %d, info %p, size %Id stub!\n", handle, info_class, info, size ); - return STATUS_SUCCESS; + struct heap *heap; + ULONG flags; + + TRACE( "handle %p, info_class %u, info %p, size %Iu.\n", handle, info_class, info, size ); + + switch (info_class) + { + case HeapCompatibilityInformation: + { + ULONG compat_info; + + if (size < sizeof(ULONG)) return STATUS_BUFFER_TOO_SMALL; + if (!(heap = unsafe_heap_from_handle( handle, 0, &flags ))) return STATUS_INVALID_HANDLE; + if (heap->flags & HEAP_NO_SERIALIZE) return STATUS_INVALID_PARAMETER; + + compat_info = *(ULONG *)info; + if (compat_info != HEAP_STD && compat_info != HEAP_LFH) + { + FIXME( "HeapCompatibilityInformation %lu not implemented!\n", compat_info ); + return STATUS_UNSUCCESSFUL; + } + if (InterlockedCompareExchange( &heap->compat_info, compat_info, HEAP_STD ) != HEAP_STD) + return STATUS_UNSUCCESSFUL; + return STATUS_SUCCESS; + } + + default: + FIXME( "HEAP_INFORMATION_CLASS %u not implemented!\n", info_class ); + return STATUS_SUCCESS; + } } /***********************************************************************

1 0

Rémi Bernon : ntdll: Increase heap block tail_size capacity to 16 bits.
by Alexandre Julliard 22 Feb '23

22 Feb '23

Module: wine Branch: master Commit: 0f6ebd0152f1217e7d9271cff9afb2e35b130995 URL: https://gitlab.winehq.org/wine/wine/-/commit/0f6ebd0152f1217e7d9271cff9afb2… Author: Rémi Bernon <rbernon(a)codeweavers.com> Date: Thu Feb 2 23:27:17 2023 +0100 ntdll: Increase heap block tail_size capacity to 16 bits. We need this for larger LFH block bin steps. Use smaller block types to get more space, they are somewhat redundant with the block flags. --- dlls/ntdll/heap.c | 24 ++++++++++++++---------- 1 file changed, 14 insertions(+), 10 deletions(-) diff --git a/dlls/ntdll/heap.c b/dlls/ntdll/heap.c index 32808754ddb..f17064972a1 100644 --- a/dlls/ntdll/heap.c +++ b/dlls/ntdll/heap.c @@ -86,11 +86,14 @@ struct rtl_heap_entry struct block { - WORD block_size; /* block size in multiple of BLOCK_ALIGN */ + /* block size in multiple of BLOCK_ALIGN */ + WORD block_size; + /* unused size (used block) / high size bits (free block) */ + WORD tail_size; + /* offset to region base */ + WORD base_offset; + BYTE block_type; BYTE block_flags; - BYTE tail_size; /* unused size (used block) / high size bits (free block) */ - WORD base_offset; /* offset to the region base in multiple of REGION_ALIGN */ - WORD block_type; }; C_ASSERT( sizeof(struct block) == 8 ); @@ -131,10 +134,10 @@ typedef struct C_ASSERT( sizeof(ARENA_LARGE) == offsetof(ARENA_LARGE, block) + sizeof(struct block) ); C_ASSERT( sizeof(ARENA_LARGE) == 4 * BLOCK_ALIGN ); -#define BLOCK_TYPE_USED 0x5355 -#define BLOCK_TYPE_DEAD 0xdead -#define BLOCK_TYPE_FREE 0x5246 -#define BLOCK_TYPE_LARGE 0x614c +#define BLOCK_TYPE_USED 'u' +#define BLOCK_TYPE_DEAD 'D' +#define BLOCK_TYPE_FREE 'F' +#define BLOCK_TYPE_LARGE 'L' #define BLOCK_FILL_USED 0x55 #define BLOCK_FILL_TAIL 0xab @@ -157,8 +160,9 @@ C_ASSERT( sizeof(struct entry) <= HEAP_MIN_BLOCK_SIZE ); #define HEAP_MAX_BLOCK_REGION_SIZE (FIELD_MAX( struct block, base_offset ) * REGION_ALIGN) C_ASSERT( HEAP_MAX_USED_BLOCK_SIZE == 512 * 1024 * (sizeof(void *) / 4) - BLOCK_ALIGN ); -C_ASSERT( HEAP_MAX_FREE_BLOCK_SIZE == 128 * 1024 * 1024 * (sizeof(void *) / 4) - BLOCK_ALIGN ); -C_ASSERT( HEAP_MAX_BLOCK_REGION_SIZE >= HEAP_MAX_FREE_BLOCK_SIZE ); +C_ASSERT( HEAP_MAX_FREE_BLOCK_SIZE >= 128 * 1024 * 1024 * (sizeof(void *) / 4) - BLOCK_ALIGN ); +C_ASSERT( HEAP_MAX_BLOCK_REGION_SIZE >= 128 * 1024 * 1024 * (sizeof(void *) / 4) - BLOCK_ALIGN ); +C_ASSERT( HEAP_MAX_FREE_BLOCK_SIZE >= HEAP_MAX_BLOCK_REGION_SIZE ); /* minimum size to start allocating large blocks */ #define HEAP_MIN_LARGE_BLOCK_SIZE (HEAP_MAX_USED_BLOCK_SIZE - 0x1000)

1 0

Rémi Bernon : ntdll: Count allocations and automatically enable LFH.
by Alexandre Julliard 22 Feb '23

22 Feb '23

Module: wine Branch: master Commit: 632b08aa5bfa71c666f0b2560503c12e2e080b3e URL: https://gitlab.winehq.org/wine/wine/-/commit/632b08aa5bfa71c666f0b2560503c1… Author: Rémi Bernon <rbernon(a)codeweavers.com> Date: Tue Nov 15 00:11:12 2022 +0100 ntdll: Count allocations and automatically enable LFH. --- dlls/kernel32/tests/heap.c | 2 - dlls/ntdll/heap.c | 138 ++++++++++++++++++++++++++++++++++++++++++++- 2 files changed, 136 insertions(+), 4 deletions(-)

1 0

Rémi Bernon : ntdll: Split heap_resize_block into heap_resize_(block|large) helpers.
by Alexandre Julliard 22 Feb '23

22 Feb '23

Module: wine Branch: master Commit: b4e5aa8dfc7261504f734ce34b092eec32673362 URL: https://gitlab.winehq.org/wine/wine/-/commit/b4e5aa8dfc7261504f734ce34b092e… Author: Rémi Bernon <rbernon(a)codeweavers.com> Date: Mon Feb 6 17:54:32 2023 +0100 ntdll: Split heap_resize_block into heap_resize_(block|large) helpers. --- dlls/ntdll/heap.c | 92 ++++++++++++++++++++++++++++++------------------------- 1 file changed, 51 insertions(+), 41 deletions(-) diff --git a/dlls/ntdll/heap.c b/dlls/ntdll/heap.c index 4b7e379da74..197217c90a7 100644 --- a/dlls/ntdll/heap.c +++ b/dlls/ntdll/heap.c @@ -1767,57 +1767,47 @@ BOOLEAN WINAPI DECLSPEC_HOTPATCH RtlFreeHeap( HANDLE handle, ULONG flags, void * return !status; } -static NTSTATUS heap_resize_block( struct heap *heap, ULONG flags, struct block *block, SIZE_T block_size, +static NTSTATUS heap_resize_large( struct heap *heap, ULONG flags, struct block *block, SIZE_T block_size, SIZE_T size, SIZE_T *old_size, void **ret ) { - SUBHEAP *subheap = block_get_subheap( heap, block ); - SIZE_T old_bin, old_block_size; - struct entry *entry; - struct block *next; + ARENA_LARGE *large = CONTAINING_RECORD( block, ARENA_LARGE, block ); + SIZE_T old_block_size = large->block_size; + *old_size = large->data_size; - if (block_get_flags( block ) & BLOCK_FLAG_LARGE) - { - ARENA_LARGE *large = CONTAINING_RECORD( block, ARENA_LARGE, block ); - old_block_size = large->block_size; - *old_size = large->data_size; - - if (block_size < HEAP_MIN_LARGE_BLOCK_SIZE / 4) return STATUS_NO_MEMORY; /* shrinking large block to small block */ - if (old_block_size < block_size) return STATUS_NO_MEMORY; - - /* FIXME: we could remap zero-pages instead */ - valgrind_notify_resize( block + 1, *old_size, size ); - initialize_block( block, *old_size, size, flags ); + if (block_size < HEAP_MIN_LARGE_BLOCK_SIZE / 4) return STATUS_NO_MEMORY; /* shrinking large block to small block */ + if (old_block_size < block_size) return STATUS_NO_MEMORY; - large->data_size = size; - valgrind_make_noaccess( (char *)block + sizeof(*block) + large->data_size, - old_block_size - sizeof(*block) - large->data_size ); - - *ret = block + 1; - return STATUS_SUCCESS; - } + /* FIXME: we could remap zero-pages instead */ + valgrind_notify_resize( block + 1, *old_size, size ); + initialize_block( block, *old_size, size, flags ); - old_block_size = block_get_size( block ); - *old_size = old_block_size - block_get_overhead( block ); - old_bin = BLOCK_SIZE_BIN( old_block_size ); + large->data_size = size; + valgrind_make_noaccess( (char *)block + sizeof(*block) + large->data_size, + old_block_size - sizeof(*block) - large->data_size ); - if (block_size >= HEAP_MIN_LARGE_BLOCK_SIZE) return STATUS_NO_MEMORY; /* growing small block to large block */ + *ret = block + 1; + return STATUS_SUCCESS; +} - heap_lock( heap, flags ); +static NTSTATUS heap_resize_block( struct heap *heap, ULONG flags, struct block *block, SIZE_T block_size, + SIZE_T size, SIZE_T old_block_size, SIZE_T *old_size, void **ret ) +{ + SUBHEAP *subheap = block_get_subheap( heap, block ); + struct block *next; if (block_size > old_block_size) { - if (!(next = next_block( subheap, block )) || !(block_get_flags( next ) & BLOCK_FLAG_FREE) || - block_size > old_block_size + block_get_size( next ) || !subheap_commit( heap, subheap, block, block_size )) - { - heap_unlock( heap, flags ); - return STATUS_NO_MEMORY; - } + /* need to grow block, make sure it's followed by large enough free block */ + if (!(next = next_block( subheap, block ))) return STATUS_NO_MEMORY; + if (!(block_get_flags( next ) & BLOCK_FLAG_FREE)) return STATUS_NO_MEMORY; + if (block_size > old_block_size + block_get_size( next )) return STATUS_NO_MEMORY; + if (!subheap_commit( heap, subheap, block, block_size )) return STATUS_NO_MEMORY; } if ((next = next_block( subheap, block )) && (block_get_flags( next ) & BLOCK_FLAG_FREE)) { /* merge with next block if it is free */ - entry = (struct entry *)next; + struct entry *entry = (struct entry *)next; list_remove( &entry->entry ); old_block_size += block_get_size( next ); } @@ -1836,9 +1826,30 @@ static NTSTATUS heap_resize_block( struct heap *heap, ULONG flags, struct block if ((next = next_block( subheap, block ))) block_set_flags( next, BLOCK_FLAG_PREV_FREE, 0 ); + *ret = block + 1; + return STATUS_SUCCESS; +} + +static NTSTATUS heap_resize_in_place( struct heap *heap, ULONG flags, struct block *block, SIZE_T block_size, + SIZE_T size, SIZE_T *old_size, void **ret ) +{ + SIZE_T old_bin, old_block_size; + NTSTATUS status; + + if (block_get_flags( block ) & BLOCK_FLAG_LARGE) + return heap_resize_large( heap, flags, block, block_size, size, old_size, ret ); + + old_block_size = block_get_size( block ); + *old_size = old_block_size - block_get_overhead( block ); + old_bin = BLOCK_SIZE_BIN( old_block_size ); + + if (block_size >= HEAP_MIN_LARGE_BLOCK_SIZE) return STATUS_NO_MEMORY; /* growing small block to large block */ + + heap_lock( heap, flags ); + status = heap_resize_block( heap, flags, block, block_size, size, old_block_size, old_size, ret ); heap_unlock( heap, flags ); - if (heap->bins) + if (!status && heap->bins) { SIZE_T new_bin = BLOCK_SIZE_BIN( block_size ); InterlockedIncrement( &heap->bins[old_bin].count_freed ); @@ -1846,8 +1857,7 @@ static NTSTATUS heap_resize_block( struct heap *heap, ULONG flags, struct block if (!ReadNoFence( &heap->bins[new_bin].enabled )) bin_try_enable( heap, &heap->bins[new_bin] ); } - *ret = block + 1; - return STATUS_SUCCESS; + return status; } /*********************************************************************** @@ -1870,8 +1880,8 @@ void *WINAPI RtlReAllocateHeap( HANDLE handle, ULONG flags, void *ptr, SIZE_T si status = STATUS_NO_MEMORY; else if (!(block = unsafe_block_from_ptr( heap, heap_flags, ptr ))) status = STATUS_INVALID_PARAMETER; - else if ((status = heap_resize_block( heap, heap_flags, block, block_size, size, - &old_size, &ret ))) + else if ((status = heap_resize_in_place( heap, heap_flags, block, block_size, size, + &old_size, &ret ))) { if (flags & HEAP_REALLOC_IN_PLACE_ONLY) status = STATUS_NO_MEMORY;

1 0

Rémi Bernon : ntdll: Implement Low Fragmentation Heap frontend.
by Alexandre Julliard 22 Feb '23

22 Feb '23

Module: wine Branch: master Commit: 27665f35e4da13bac1e4dd8948a65f484c9dadfa URL: https://gitlab.winehq.org/wine/wine/-/commit/27665f35e4da13bac1e4dd8948a65f… Author: Rémi Bernon <rbernon(a)codeweavers.com> Date: Sat Feb 11 09:43:30 2023 +0100 ntdll: Implement Low Fragmentation Heap frontend. This implements the reduced fragmentation from the heap frontend, by carving smaller blocks out of larger allocated blocks. The super block and each sub-block are all flagged with BLOCK_FLAG_LFH. The super-block (struct group) uses a standard struct block header, as well as a list entry to be linked in free list, and a free bit map to track free sub-blocks. Sub-blocks reference their super block through the base_offset, instead of the subheap, using the block size as radix. --- dlls/kernel32/tests/heap.c | 20 ++-- dlls/ntdll/heap.c | 276 +++++++++++++++++++++++++++++++++++++++++++-- 2 files changed, 277 insertions(+), 19 deletions(-)

1 0

Rémi Bernon : ntdll: Use atomics and lock-free list for bin groups.
by Alexandre Julliard 22 Feb '23

22 Feb '23

Module: wine Branch: master Commit: 889743154e63766de17bd82489cac5940e81000f URL: https://gitlab.winehq.org/wine/wine/-/commit/889743154e63766de17bd82489cac5… Author: Rémi Bernon <rbernon(a)codeweavers.com> Date: Fri Feb 10 08:28:06 2023 +0100 ntdll: Use atomics and lock-free list for bin groups. --- dlls/kernel32/tests/heap.c | 1 - dlls/ntdll/heap.c | 50 +++++++++++++++++++--------------------------- 2 files changed, 21 insertions(+), 30 deletions(-) diff --git a/dlls/kernel32/tests/heap.c b/dlls/kernel32/tests/heap.c index 8c06e9d5f53..b0b56132393 100644 --- a/dlls/kernel32/tests/heap.c +++ b/dlls/kernel32/tests/heap.c @@ -1246,7 +1246,6 @@ static void test_HeapCreate(void) thread_params.flags = 0; SetEvent( thread_params.start_event ); res = WaitForSingleObject( thread_params.ready_event, 100 ); - todo_wine ok( !res, "WaitForSingleObject returned %#lx, error %lu\n", res, GetLastError() ); ret = HeapUnlock( heap ); ok( ret, "HeapUnlock failed, error %lu\n", GetLastError() ); diff --git a/dlls/ntdll/heap.c b/dlls/ntdll/heap.c index 1ea2c24a656..b2d08a1de7d 100644 --- a/dlls/ntdll/heap.c +++ b/dlls/ntdll/heap.c @@ -264,7 +264,7 @@ struct bin LONG enabled; /* list of groups with free blocks */ - struct list groups; + SLIST_HEADER groups; }; struct heap @@ -1535,7 +1535,7 @@ HANDLE WINAPI RtlCreateHeap( ULONG flags, void *addr, SIZE_T total_size, SIZE_T NtAllocateVirtualMemory( NtCurrentProcess(), (void *)&heap->bins, 0, &size, MEM_COMMIT, PAGE_READWRITE ); for (i = 0; i < BLOCK_SIZE_BIN_COUNT; ++i) - list_init( &heap->bins[i].groups ); + RtlInitializeSListHead( &heap->bins[i].groups ); } /* link it into the per-process heap list */ @@ -1690,7 +1690,7 @@ static NTSTATUS heap_allocate_block( struct heap *heap, ULONG flags, SIZE_T bloc /* header for every LFH block group */ struct DECLSPEC_ALIGN(BLOCK_ALIGN) group { - struct list entry; + SLIST_ENTRY entry; /* one bit for each free block and the highest bit for GROUP_FLAG_FREE */ LONG free_bits; /* first block of a group, required for alignment */ @@ -1699,7 +1699,6 @@ struct DECLSPEC_ALIGN(BLOCK_ALIGN) group #define GROUP_BLOCK_COUNT (sizeof(((struct group *)0)->free_bits) * 8 - 1) #define GROUP_FLAG_FREE (1u << GROUP_BLOCK_COUNT) -#define GROUP_FREE_BITS_MASK (GROUP_FLAG_FREE - 1) static inline UINT block_get_group_index( const struct block *block ) { @@ -1728,10 +1727,10 @@ static inline struct block *group_get_block( struct group *group, SIZE_T block_s /* lookup a free block using the group free_bits, the current thread must own the group */ static inline struct block *group_find_free_block( struct group *group, SIZE_T block_size ) { - ULONG i, free_bits = group->free_bits; + ULONG i, free_bits = ReadNoFence( &group->free_bits ); /* free_bits will never be 0 as the group is unlinked when it's fully used */ BitScanForward( &i, free_bits ); - group->free_bits &= ~(1 << i); + InterlockedAnd( &group->free_bits, ~(1 << i) ); return group_get_block( group, block_size, i ); } @@ -1745,11 +1744,15 @@ static struct group *group_allocate( struct heap *heap, ULONG flags, SIZE_T bloc group_size = offsetof( struct group, first_block ) + GROUP_BLOCK_COUNT * block_size; group_block_size = heap_get_block_size( heap, flags, group_size ); + heap_lock( heap, flags ); + if (group_block_size >= HEAP_MIN_LARGE_BLOCK_SIZE) status = heap_allocate_large( heap, flags & ~HEAP_ZERO_MEMORY, group_block_size, group_size, (void **)&group ); else status = heap_allocate_block( heap, flags & ~HEAP_ZERO_MEMORY, group_block_size, group_size, (void **)&group ); + heap_unlock( heap, flags ); + if (status) return NULL; block_set_flags( (struct block *)group - 1, 0, BLOCK_FLAG_LFH ); @@ -1775,6 +1778,8 @@ static NTSTATUS group_release( struct heap *heap, ULONG flags, struct bin *bin, struct block *block = (struct block *)group - 1; NTSTATUS status; + heap_lock( heap, flags ); + block_set_flags( block, BLOCK_FLAG_LFH, 0 ); if (block_get_flags( block ) & BLOCK_FLAG_LARGE) @@ -1782,21 +1787,18 @@ static NTSTATUS group_release( struct heap *heap, ULONG flags, struct bin *bin, else status = heap_free_block( heap, flags, block ); + heap_unlock( heap, flags ); + return status; } /* acquire a group from the bin, thread takes ownership of a shared group or allocates a new one */ static struct group *heap_acquire_bin_group( struct heap *heap, ULONG flags, SIZE_T block_size, struct bin *bin ) { - struct group *group; - struct list *entry; + SLIST_ENTRY *entry; - if ((entry = list_head( &bin->groups ))) - { - group = LIST_ENTRY( entry, struct group, entry ); - list_remove( &group->entry ); - return group; - } + if ((entry = RtlInterlockedPopEntrySList( &bin->groups ))) + return CONTAINING_RECORD( entry, struct group, entry ); return group_allocate( heap, flags, block_size ); } @@ -1805,9 +1807,9 @@ static struct group *heap_acquire_bin_group( struct heap *heap, ULONG flags, SIZ static NTSTATUS heap_release_bin_group( struct heap *heap, ULONG flags, struct bin *bin, struct group *group ) { /* try re-using the block group instead of releasing it */ - if (list_empty( &bin->groups )) + if (!RtlQueryDepthSList( &bin->groups )) { - list_add_tail( &bin->groups, &group->entry ); + RtlInterlockedPushEntrySList( &bin->groups, &group->entry ); return STATUS_SUCCESS; } @@ -1826,12 +1828,10 @@ static struct block *find_free_bin_block( struct heap *heap, ULONG flags, SIZE_T block = group_find_free_block( group, block_size ); /* serialize with heap_free_block_lfh: atomically set GROUP_FLAG_FREE when the free bits are all 0. */ - if (!group->free_bits) - group->free_bits = GROUP_FLAG_FREE; - else + if (ReadNoFence( &group->free_bits ) || InterlockedCompareExchange( &group->free_bits, GROUP_FLAG_FREE, 0 )) { /* if GROUP_FLAG_FREE isn't set, thread is responsible for putting it back into group list. */ - list_add_tail( &bin->groups, &group->entry ); + RtlInterlockedPushEntrySList( &bin->groups, &group->entry ); } return block; @@ -1851,7 +1851,6 @@ static NTSTATUS heap_allocate_block_lfh( struct heap *heap, ULONG flags, SIZE_T block_size = BLOCK_BIN_SIZE( BLOCK_SIZE_BIN( block_size ) ); - heap_lock( heap, flags ); if ((block = find_free_bin_block( heap, flags, block_size, bin ))) { block_set_type( block, BLOCK_TYPE_USED ); @@ -1861,7 +1860,6 @@ static NTSTATUS heap_allocate_block_lfh( struct heap *heap, ULONG flags, SIZE_T mark_block_tail( block, flags ); *ret = block + 1; } - heap_unlock( heap, flags ); return block ? STATUS_SUCCESS : STATUS_NO_MEMORY; } @@ -1884,20 +1882,14 @@ static NTSTATUS heap_free_block_lfh( struct heap *heap, ULONG flags, struct bloc block_set_flags( block, ~BLOCK_FLAG_LFH, BLOCK_FLAG_FREE ); mark_block_free( block + 1, (char *)block + block_size - (char *)(block + 1), flags ); - heap_lock( heap, flags ); - - group->free_bits |= (1 << i); - /* if this was the last used block in a group and GROUP_FLAG_FREE was set */ - if (group->free_bits == (GROUP_FREE_BITS_MASK | GROUP_FLAG_FREE)) + if (InterlockedOr( &group->free_bits, 1 << i ) == ~(1 << i)) { /* thread now owns the group, and can release it to its bin */ group->free_bits = ~GROUP_FLAG_FREE; status = heap_release_bin_group( heap, flags, bin, group ); } - heap_unlock( heap, flags ); - return status; }

1 0

Rémi Bernon : ntdll: Add a heap thread affinity and per-affinity bin group cache.
by Alexandre Julliard 22 Feb '23

22 Feb '23

Module: wine Branch: master Commit: 40b7c3e89a95d6ccb190b234d4ad13b3a8304495 URL: https://gitlab.winehq.org/wine/wine/-/commit/40b7c3e89a95d6ccb190b234d4ad13… Author: Rémi Bernon <rbernon(a)codeweavers.com> Date: Fri Feb 10 08:10:01 2023 +0100 ntdll: Add a heap thread affinity and per-affinity bin group cache. --- dlls/ntdll/heap.c | 90 ++++++++++++++++++++++++++++++++++++++++++++++--- dlls/ntdll/loader.c | 2 ++ dlls/ntdll/ntdll_misc.h | 1 + 3 files changed, 89 insertions(+), 4 deletions(-) diff --git a/dlls/ntdll/heap.c b/dlls/ntdll/heap.c index b2d08a1de7d..afe38eddaa8 100644 --- a/dlls/ntdll/heap.c +++ b/dlls/ntdll/heap.c @@ -255,6 +255,9 @@ C_ASSERT( BLOCK_BIN_SIZE( 0x80 ) == ~(SIZE_T)0 ); /* difference between block classes and all possible validation overhead must fit into block tail_size */ C_ASSERT( BIN_SIZE_STEP_7 + 3 * BLOCK_ALIGN <= FIELD_MAX( struct block, tail_size ) ); +static BYTE affinity_mapping[] = {20,6,31,15,14,29,27,4,18,24,26,13,0,9,2,30,17,7,23,25,10,19,12,3,22,21,5,16,1,28,11,8}; +static LONG next_thread_affinity; + /* a bin, tracking heap blocks of a certain size */ struct bin { @@ -265,8 +268,20 @@ struct bin /* list of groups with free blocks */ SLIST_HEADER groups; + + /* array of affinity reserved groups, interleaved with other bins to keep + * all pointers of the same affinity and different bin grouped together, + * and pointers of the same bin and different affinity away from each other, + * hopefully in separate cache lines. + */ + struct group **affinity_group_base; }; +static inline struct group **bin_get_affinity_group( struct bin *bin, BYTE affinity ) +{ + return bin->affinity_group_base + affinity * BLOCK_SIZE_BIN_COUNT; +} + struct heap { /* win32/win64 */ DWORD_PTR unknown1[2]; /* 0000/0000 */ @@ -1531,11 +1546,16 @@ HANDLE WINAPI RtlCreateHeap( ULONG flags, void *addr, SIZE_T total_size, SIZE_T if (heap->flags & HEAP_GROWABLE) { - SIZE_T size = sizeof(struct bin) * BLOCK_SIZE_BIN_COUNT; + SIZE_T size = (sizeof(struct bin) + sizeof(struct group *) * ARRAY_SIZE(affinity_mapping)) * BLOCK_SIZE_BIN_COUNT; NtAllocateVirtualMemory( NtCurrentProcess(), (void *)&heap->bins, 0, &size, MEM_COMMIT, PAGE_READWRITE ); - for (i = 0; i < BLOCK_SIZE_BIN_COUNT; ++i) + + for (i = 0; heap->bins && i < BLOCK_SIZE_BIN_COUNT; ++i) + { RtlInitializeSListHead( &heap->bins[i].groups ); + /* offset affinity_group_base to interleave the bin affinity group pointers */ + heap->bins[i].affinity_group_base = (struct group **)(heap->bins + BLOCK_SIZE_BIN_COUNT) + i; + } } /* link it into the per-process heap list */ @@ -1693,6 +1713,8 @@ struct DECLSPEC_ALIGN(BLOCK_ALIGN) group SLIST_ENTRY entry; /* one bit for each free block and the highest bit for GROUP_FLAG_FREE */ LONG free_bits; + /* affinity of the thread which last allocated from this group */ + LONG affinity; /* first block of a group, required for alignment */ struct block first_block; }; @@ -1792,11 +1814,30 @@ static NTSTATUS group_release( struct heap *heap, ULONG flags, struct bin *bin, return status; } +static inline ULONG heap_current_thread_affinity(void) +{ + ULONG affinity; + + if (!(affinity = NtCurrentTeb()->HeapVirtualAffinity)) + { + affinity = InterlockedIncrement( &next_thread_affinity ); + affinity = affinity_mapping[affinity % ARRAY_SIZE(affinity_mapping)]; + NtCurrentTeb()->HeapVirtualAffinity = affinity; + } + + return affinity; +} + /* acquire a group from the bin, thread takes ownership of a shared group or allocates a new one */ static struct group *heap_acquire_bin_group( struct heap *heap, ULONG flags, SIZE_T block_size, struct bin *bin ) { + ULONG affinity = NtCurrentTeb()->HeapVirtualAffinity; + struct group *group; SLIST_ENTRY *entry; + if ((group = InterlockedExchangePointer( (void *)bin_get_affinity_group( bin, affinity ), NULL ))) + return group; + if ((entry = RtlInterlockedPopEntrySList( &bin->groups ))) return CONTAINING_RECORD( entry, struct group, entry ); @@ -1806,8 +1847,16 @@ static struct group *heap_acquire_bin_group( struct heap *heap, ULONG flags, SIZ /* release a thread owned and fully freed group to the bin shared group, or free its memory */ static NTSTATUS heap_release_bin_group( struct heap *heap, ULONG flags, struct bin *bin, struct group *group ) { + ULONG affinity = group->affinity; + + /* using InterlockedExchangePointer here would possibly return a group that has used blocks, + * we prefer keeping our fully freed group instead for reduced memory consumption. + */ + if (!InterlockedCompareExchangePointer( (void *)bin_get_affinity_group( bin, affinity ), group, NULL )) + return STATUS_SUCCESS; + /* try re-using the block group instead of releasing it */ - if (!RtlQueryDepthSList( &bin->groups )) + if (RtlQueryDepthSList( &bin->groups ) <= ARRAY_SIZE(affinity_mapping)) { RtlInterlockedPushEntrySList( &bin->groups, &group->entry ); return STATUS_SUCCESS; @@ -1818,6 +1867,7 @@ static NTSTATUS heap_release_bin_group( struct heap *heap, ULONG flags, struct b static struct block *find_free_bin_block( struct heap *heap, ULONG flags, SIZE_T block_size, struct bin *bin ) { + ULONG affinity = heap_current_thread_affinity(); struct block *block; struct group *group; @@ -1825,13 +1875,16 @@ static struct block *find_free_bin_block( struct heap *heap, ULONG flags, SIZE_T * some other thread might still set the free bits if they are freeing blocks. */ if (!(group = heap_acquire_bin_group( heap, flags, block_size, bin ))) return NULL; + group->affinity = affinity; + block = group_find_free_block( group, block_size ); /* serialize with heap_free_block_lfh: atomically set GROUP_FLAG_FREE when the free bits are all 0. */ if (ReadNoFence( &group->free_bits ) || InterlockedCompareExchange( &group->free_bits, GROUP_FLAG_FREE, 0 )) { /* if GROUP_FLAG_FREE isn't set, thread is responsible for putting it back into group list. */ - RtlInterlockedPushEntrySList( &bin->groups, &group->entry ); + if ((group = InterlockedExchangePointer( (void *)bin_get_affinity_group( bin, affinity ), group ))) + RtlInterlockedPushEntrySList( &bin->groups, &group->entry ); } return block; @@ -1922,6 +1975,35 @@ static void bin_try_enable( struct heap *heap, struct bin *bin ) WriteRelease( &bin->enabled, TRUE ); } +static void heap_thread_detach_bin_groups( struct heap *heap ) +{ + ULONG i, affinity = NtCurrentTeb()->HeapVirtualAffinity; + + if (!heap->bins) return; + + for (i = 0; i < BLOCK_SIZE_BIN_COUNT; ++i) + { + struct bin *bin = heap->bins + i; + struct group *group; + if (!(group = InterlockedExchangePointer( (void *)bin_get_affinity_group( bin, affinity ), NULL ))) continue; + RtlInterlockedPushEntrySList( &bin->groups, &group->entry ); + } +} + +void heap_thread_detach(void) +{ + struct heap *heap; + + RtlEnterCriticalSection( &process_heap->cs ); + + LIST_FOR_EACH_ENTRY( heap, &process_heap->entry, struct heap, entry ) + heap_thread_detach_bin_groups( heap ); + + heap_thread_detach_bin_groups( process_heap ); + + RtlLeaveCriticalSection( &process_heap->cs ); +} + /*********************************************************************** * RtlAllocateHeap (NTDLL.@) */ diff --git a/dlls/ntdll/loader.c b/dlls/ntdll/loader.c index f254f1beefd..f533db060d0 100644 --- a/dlls/ntdll/loader.c +++ b/dlls/ntdll/loader.c @@ -3778,6 +3778,8 @@ void WINAPI LdrShutdownThread(void) /* don't call DbgUiGetThreadDebugObject as some apps hook it and terminate if called */ if (NtCurrentTeb()->DbgSsReserved[1]) NtClose( NtCurrentTeb()->DbgSsReserved[1] ); RtlFreeThreadActivationContextStack(); + + heap_thread_detach(); } diff --git a/dlls/ntdll/ntdll_misc.h b/dlls/ntdll/ntdll_misc.h index d1a7790991b..f6b77b79cde 100644 --- a/dlls/ntdll/ntdll_misc.h +++ b/dlls/ntdll/ntdll_misc.h @@ -127,5 +127,6 @@ static inline void ascii_to_unicode( WCHAR *dst, const char *src, size_t len ) /* FLS data */ extern TEB_FLS_DATA *fls_alloc_data(void) DECLSPEC_HIDDEN; +extern void heap_thread_detach(void) DECLSPEC_HIDDEN; #endif

1 0

Francisco Casas : vkd3d-shader/hlsl: Respect object reservations even if the object is unused.
by Alexandre Julliard 22 Feb '23

22 Feb '23

Module: vkd3d Branch: master Commit: 75359e6dbdaeb91124f7903756c800c72caeab8f URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/75359e6dbdaeb91124f7903756c80… Author: Francisco Casas <fcasas(a)codeweavers.com> Date: Tue Jan 31 12:45:54 2023 -0300 vkd3d-shader/hlsl: Respect object reservations even if the object is unused. --- libs/vkd3d-shader/hlsl_codegen.c | 2 +- tests/register-reservations.shader_test | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index d5534d8c..00cb3711 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -2908,7 +2908,7 @@ static const struct hlsl_ir_var *get_reserved_object(struct hlsl_ctx *ctx, enum LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, const struct hlsl_ir_var, extern_entry) { - if (var->last_read && var->reg_reservation.type == get_regset_name(regset) + if (var->reg_reservation.type == get_regset_name(regset) && var->reg_reservation.index == index) return var; } diff --git a/tests/register-reservations.shader_test b/tests/register-reservations.shader_test index 04de717b..f0287d00 100644 --- a/tests/register-reservations.shader_test +++ b/tests/register-reservations.shader_test @@ -34,7 +34,7 @@ float4 main() : sv_target [test] draw quad -todo probe all rgba (1.0, 1.0, 1.0, 99.0) +probe all rgba (1.0, 1.0, 1.0, 99.0) [pixel shader]

1 0

Francisco Casas : vkd3d-shader/hlsl: Allocate register reservations in a separate pass.
by Alexandre Julliard 22 Feb '23

22 Feb '23

Module: vkd3d Branch: master Commit: f8338ef08953686298bb31315a19e6044325bb49 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/f8338ef08953686298bb31315a19e… Author: Francisco Casas <fcasas(a)codeweavers.com> Date: Thu Nov 24 19:36:20 2022 -0300 vkd3d-shader/hlsl: Allocate register reservations in a separate pass. This refactoring is required for improving the allocation strategy so it works with multiple-register variables. --- libs/vkd3d-shader/hlsl_codegen.c | 77 +++++++++++++++++++++++++++------------- 1 file changed, 53 insertions(+), 24 deletions(-) diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index 00cb3711..65bd9d4c 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -2250,6 +2250,40 @@ static char get_regset_name(enum hlsl_regset regset) vkd3d_unreachable(); } +static void allocate_register_reservations(struct hlsl_ctx *ctx) +{ + struct hlsl_ir_var *var; + + LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) + { + enum hlsl_regset regset; + + if (!hlsl_type_is_resource(var->data_type)) + continue; + regset = hlsl_type_get_regset(var->data_type); + + if (var->reg_reservation.type) + { + if (var->reg_reservation.type != get_regset_name(regset)) + { + struct vkd3d_string_buffer *type_string; + + type_string = hlsl_type_to_string(ctx, var->data_type); + hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "Object of type '%s' must be bound to register type '%c'.", + type_string->buffer, get_regset_name(regset)); + hlsl_release_string_buffer(ctx, type_string); + } + else + { + var->regs[regset].allocated = true; + var->regs[regset].id = var->reg_reservation.index; + TRACE("Allocated reserved %s to %c%u.\n", var->name, var->reg_reservation.type, var->reg_reservation.index); + } + } + } +} + /* Compute the earliest and latest liveness for each variable. In the case that * a variable is accessed inside of a loop, we promote its liveness to extend * to at least the range of the entire loop. Note that we don't need to do this @@ -2901,15 +2935,17 @@ static void allocate_buffers(struct hlsl_ctx *ctx) } } -static const struct hlsl_ir_var *get_reserved_object(struct hlsl_ctx *ctx, enum hlsl_regset regset, +static const struct hlsl_ir_var *get_allocated_object(struct hlsl_ctx *ctx, enum hlsl_regset regset, uint32_t index) { const struct hlsl_ir_var *var; LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, const struct hlsl_ir_var, extern_entry) { - if (var->reg_reservation.type == get_regset_name(regset) - && var->reg_reservation.index == index) + if (!var->regs[regset].allocated) + continue; + + if (index == var->regs[regset].id) return var; } return NULL; @@ -2936,37 +2972,39 @@ static void allocate_objects(struct hlsl_ctx *ctx, enum hlsl_regset regset) LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) { - if (!var->last_read || var->data_type->reg_size[regset] == 0) + if (!var->last_read || !var->data_type->reg_size[regset]) continue; - if (var->reg_reservation.type == regset_name) + if (var->regs[regset].allocated) { - const struct hlsl_ir_var *reserved_object = get_reserved_object(ctx, regset, - var->reg_reservation.index); + const struct hlsl_ir_var *reserved_object; + unsigned int index = var->regs[regset].id; - if (var->reg_reservation.index < min_index) + reserved_object = get_allocated_object(ctx, regset, index); + + if (var->regs[regset].id < min_index) { assert(regset == HLSL_REGSET_UAVS); hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_OVERLAPPING_RESERVATIONS, "UAV index (%u) must be higher than the maximum render target index (%u).", - var->reg_reservation.index, min_index - 1); + var->regs[regset].id, min_index - 1); } else if (reserved_object && reserved_object != var) { hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_OVERLAPPING_RESERVATIONS, - "Multiple objects bound to %c%u.", regset_name, var->reg_reservation.index); + "Multiple objects bound to %c%u.", regset_name, index); hlsl_note(ctx, &reserved_object->loc, VKD3D_SHADER_LOG_ERROR, - "Object '%s' is already bound to %c%u.", reserved_object->name, regset_name, - var->reg_reservation.index); + "Object '%s' is already bound to %c%u.", reserved_object->name, + regset_name, index); } var->regs[regset].id = var->reg_reservation.index; var->regs[regset].allocated = true; TRACE("Allocated reserved %s to %c%u.\n", var->name, regset_name, var->regs[regset].id); } - else if (!var->reg_reservation.type) + else { - while (get_reserved_object(ctx, regset, index)) + while (get_allocated_object(ctx, regset, index)) ++index; var->regs[regset].id = index; @@ -2974,16 +3012,6 @@ static void allocate_objects(struct hlsl_ctx *ctx, enum hlsl_regset regset) TRACE("Allocated object to %c%u.\n", regset_name, index); ++index; } - else - { - struct vkd3d_string_buffer *type_string; - - type_string = hlsl_type_to_string(ctx, var->data_type); - hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, - "Object of type '%s' must be bound to register type '%c'.", - type_string->buffer, regset_name); - hlsl_release_string_buffer(ctx, type_string); - } } } @@ -3299,6 +3327,7 @@ int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry if (TRACE_ON()) rb_for_each_entry(&ctx->functions, dump_function, ctx); + allocate_register_reservations(ctx); allocate_temp_registers(ctx, entry_func); if (profile->major_version < 4) {

1 0

Francisco Casas : vkd3d-shader/hlsl: Avoid segfault on missing sampler.
by Alexandre Julliard 22 Feb '23

22 Feb '23

Module: vkd3d Branch: master Commit: 9fdff25a3b815860894c7f07166a69a5d013b6d2 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/9fdff25a3b815860894c7f07166a6… Author: Francisco Casas <fcasas(a)codeweavers.com> Date: Thu Sep 29 13:32:07 2022 -0300 vkd3d-shader/hlsl: Avoid segfault on missing sampler. --- libs/vkd3d-shader/hlsl_sm4.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/libs/vkd3d-shader/hlsl_sm4.c b/libs/vkd3d-shader/hlsl_sm4.c index 10bb1e3d..d69b8fea 100644 --- a/libs/vkd3d-shader/hlsl_sm4.c +++ b/libs/vkd3d-shader/hlsl_sm4.c @@ -2265,7 +2265,10 @@ static void write_sm4_resource_load(struct hlsl_ctx *ctx, case HLSL_RESOURCE_SAMPLE: if (!load->sampler.var) + { hlsl_fixme(ctx, &load->node.loc, "SM4 combined sample expression."); + return; + } write_sm4_sample(ctx, buffer, resource_type, &load->node, &load->resource, &load->sampler, coords, texel_offset); break;

1 0