July 2015 - Wine-Devel - list.winehq.org

wine CPU usage and CreateWaitableTimer
by andrea July 9, 2015

July 9, 2015

Hi, I'm trying to understand why AppleWin uses 100% CPU when running in wine vs ~1% in Windows. my case is wine 1.7.43 vs Windows 7 https://appdb.winehq.org/objectManager.php?sClass=version&iId=29403 https://bugs.winehq.org/show_bug.cgi?id=34945 basically 97% time is spent in a WaitForSingleObject on a semaphore signalled by a DirectSound IReferenceClock via AdvisePeriodic. The timer period is set to 1 millisecond. I have tried an other Windows call to create the timer: CreateWaitableTimer The latter uses about 50% CPU, so it is much better, but still not very good. This is true even on a simple app that does nothing other than waiting for this timer all the time. Does anybody know 1) is this normal? should these waits be almost free like they are in Windows? 2) what is the best implemented timer in wine that can be used for a periodic event (~1ms)? Thanks

3 10

Re: [5/7] ntdll: Implement RtlDecompressFragment.
by Matteo Bruni July 9, 2015

July 9, 2015

Hi, a bunch of nitpicks which you can take or ignore: 2015-07-09 3:07 GMT+02:00 Sebastian Lackner <sebastian(a)fds-team.de>: > Based on a patch by Michael Müller. > > For https://bugs.winehq.org/show_bug.cgi?id=37449 > > --- > dlls/ntdll/ntdll.spec | 2 > dlls/ntdll/rtl.c | 214 ++++++++++++++++++++++++++++++++++-- > dlls/ntdll/tests/rtl.c | 3 > dlls/ntoskrnl.exe/ntoskrnl.exe.spec | 2 > 4 files changed, 210 insertions(+), 11 deletions(-) > > diff --git a/dlls/ntdll/ntdll.spec b/dlls/ntdll/ntdll.spec > index 75dc647..ca3561d 100644 > --- a/dlls/ntdll/ntdll.spec > +++ b/dlls/ntdll/ntdll.spec > @@ -513,7 +513,7 @@ > @ stdcall RtlDecodePointer(ptr) > # @ stub RtlDecodeSystemPointer > @ stdcall RtlDecompressBuffer(long ptr long ptr long ptr) > -@ stub RtlDecompressFragment > +@ stdcall RtlDecompressFragment(long ptr long ptr long long ptr ptr) > @ stub RtlDefaultNpAcl > @ stub RtlDelete > @ stdcall RtlDeleteAce(ptr long) > diff --git a/dlls/ntdll/rtl.c b/dlls/ntdll/rtl.c > index d9e448a..9a196d2 100644 > --- a/dlls/ntdll/rtl.c > +++ b/dlls/ntdll/rtl.c > @@ -1309,17 +1309,219 @@ NTSTATUS WINAPI RtlCompressBuffer(USHORT format, PUCHAR uncompressed, ULONG unco > } > } > > +/* decompress a single LZNT1 chunk */ > +static PUCHAR lznt1_decompress_chunk(UCHAR *dst, ULONG dst_size, UCHAR *src, ULONG src_size) PUCHAR -> UCHAR * > +{ > + UCHAR *src_cur = src, *src_end = src + src_size; > + UCHAR *dst_cur = dst, *dst_end = dst + dst_size; > + ULONG displacement_bits, length_bits; > + ULONG code_displacement, code_length; > + WORD flags, code; > + > + while (src_cur < src_end && dst_cur < dst_end) > + { > + flags = 0x8000 | *src_cur++; > + while ((flags & 0xFF00) && src_cur < src_end) It looks like we're mostly using lowercase in hex literals. > + { > + if (flags & 1) > + { > + /* backwards reference */ > + if (src_cur + sizeof(WORD) > src_end) > + return NULL; > + > + code = *(WORD *)src_cur; > + src_cur += sizeof(WORD); > + > + /* find length / displacement bits */ > + for (displacement_bits = 12; displacement_bits > 4; displacement_bits--) > + if ((1 << (displacement_bits - 1)) < dst_cur - dst) break; > + > + length_bits = 16 - displacement_bits; > + code_length = (code & ((1 << length_bits) - 1)) + 3; > + code_displacement = (code >> length_bits) + 1; > + > + if (dst_cur < dst + code_displacement) > + return NULL; > + > + /* copy bytes of chunk - we can't use memcpy() > + * since source and dest can be overlapping */ I would also mention that you can't use memmove() either because you actually want to copy any overlapping byte over and over again, as necessary. > + while (code_length--) > + { > + if (dst_cur >= dst_end) return dst_cur; > + *dst_cur = *(dst_cur - code_displacement); > + dst_cur++; > + } > + } > + else > + { > + /* uncompressed data */ > + if (dst_cur >= dst_end) return dst_cur; > + *dst_cur++ = *src_cur++; > + } > + flags >>= 1; > + } > + } > + > + return dst_cur; > +} > + > +/* decompress data encoded with LZNT1 */ > +static NTSTATUS lznt1_decompress(UCHAR *dst, ULONG dst_size, UCHAR *src, ULONG src_size, > + ULONG offset, ULONG *final_size, UCHAR *workspace) > +{ > + UCHAR *src_cur = src, *src_end = src + src_size; > + UCHAR *dst_cur = dst, *dst_end = dst + dst_size; > + ULONG chunk_size, block_size; > + WORD chunk_header; > + UCHAR *ptr; > + > + if (src_cur + sizeof(WCHAR) > src_end) > + return STATUS_BAD_COMPRESSION_BUFFER; > + > + /* skip over chunks which have a distance >= 0x1000 to the destination offset */ > + while (offset >= 0x1000 && src_cur + sizeof(WCHAR) <= src_end) > + { > + chunk_header = *(WORD *)src_cur; > + src_cur += sizeof(WCHAR); Why this mixing of WORD and WCHAR? They do match in practice but still... > + if (!chunk_header) goto out; > + chunk_size = (chunk_header & 0xFFF) + 1; > + > + if (src_cur + chunk_size > src_end) > + return STATUS_BAD_COMPRESSION_BUFFER; > + > + src_cur += chunk_size; > + offset -= 0x1000; > + } > + > + /* check if a chunk is included partially */ > + if (offset && src_cur + sizeof(WCHAR) <= src_end) > + { > + chunk_header = *(WORD *)src_cur; > + src_cur += sizeof(WCHAR); > + if (!chunk_header) goto out; > + chunk_size = (chunk_header & 0xFFF) + 1; > + > + if (src_cur + chunk_size > src_end) > + return STATUS_BAD_COMPRESSION_BUFFER; > + > + if (dst_cur >= dst_end) > + goto out; > + > + if (chunk_header & 0x8000) > + { > + /* compressed chunk */ > + if (!workspace) return STATUS_ACCESS_VIOLATION; > + ptr = lznt1_decompress_chunk(workspace, 0x1000, src_cur, chunk_size); I guess in theory you could avoid decompressing the whole chunk if you need less than the whole chunk to be decompressed. It's not strictly required though so if doing that makes the code too convoluted then ignore it. > + if (!ptr) return STATUS_BAD_COMPRESSION_BUFFER; > + if (ptr - workspace > offset) > + { > + block_size = min((ptr - workspace) - offset, dst_end - dst_cur); > + memcpy(dst_cur, workspace + offset, block_size); > + dst_cur += block_size; > + } > + } > + else > + { > + /* uncompressed chunk */ > + if (chunk_size > offset) > + { > + block_size = min(chunk_size - offset, dst_end - dst_cur); > + memcpy(dst_cur, src_cur + offset, block_size); > + dst_cur += block_size; > + } > + } > + > + src_cur += chunk_size; > + } > + > + /* remaining blocks */ > + while (src_cur + sizeof(WCHAR) <= src_end) > + { > + chunk_header = *(WORD *)src_cur; > + src_cur += sizeof(WCHAR); > + if (!chunk_header) goto out; > + chunk_size = (chunk_header & 0xFFF) + 1; > + > + if (src_cur + chunk_size > src_end) > + return STATUS_BAD_COMPRESSION_BUFFER; > + > + /* fill space with padding */ > + block_size = ((dst_cur - dst) + offset) & 0xFFF; > + if (block_size) > + { > + block_size = 0x1000 - block_size; > + if (dst_cur + block_size >= dst_end) > + goto out; > + memset(dst_cur, 0, block_size); > + dst_cur += block_size; > + } I'm probably dense but I don't follow this padding part. Don't you still need to memset to 0 the "remaining part" even if dst_end <= dst_cur + block_size (obviously stopping at dst_end) and then proceed to decompress the chunk? Unless you're not supposed to decompress the last chunk in that case? > + > + if (dst_cur >= dst_end) > + goto out; > + > + if (chunk_header & 0x8000) > + { > + /* compressed chunk */ > + dst_cur = lznt1_decompress_chunk(dst_cur, dst_end - dst_cur, src_cur, chunk_size); > + if (!dst_cur) return STATUS_BAD_COMPRESSION_BUFFER; > + } > + else > + { > + /* uncompressed chunk */ > + block_size = min(chunk_size, dst_end - dst_cur); > + memcpy(dst_cur, src_cur, block_size); > + dst_cur += block_size; > + } > + > + src_cur += chunk_size; > + } > + > +out: > + if (final_size) > + *final_size = dst_cur - dst; > + > + return STATUS_SUCCESS; > + > +} > + > +/****************************************************************************** > + * RtlDecompressFragment [NTDLL.@] > + */ > +NTSTATUS RtlDecompressFragment(USHORT format, PUCHAR uncompressed, ULONG uncompressed_size, > + PUCHAR compressed, ULONG compressed_size, ULONG offset, > + PULONG final_size, PVOID workspace) > +{ > + TRACE("0x%04x, %p, %u, %p, %u, %u, %p, %p\n", format, uncompressed, > + uncompressed_size, compressed, compressed_size, offset, final_size, workspace); > + > + switch (format & ~COMPRESSION_ENGINE_MAXIMUM) > + { > + case COMPRESSION_FORMAT_LZNT1: > + return lznt1_decompress(uncompressed, uncompressed_size, compressed, > + compressed_size, offset, final_size, workspace); > + > + case COMPRESSION_FORMAT_NONE: > + case COMPRESSION_FORMAT_DEFAULT: > + return STATUS_INVALID_PARAMETER; > + > + default: > + FIXME("format %u not implemented\n", format); > + return STATUS_UNSUPPORTED_COMPRESSION; > + } > +} > + > + > /****************************************************************************** > * RtlDecompressBuffer [NTDLL.@] > */ > -NTSTATUS WINAPI RtlDecompressBuffer(USHORT CompressionFormat, PUCHAR UncompressedBuffer, > - ULONG UncompressedBufferSize, PUCHAR CompressedBuffer, > - ULONG CompressedBufferSize, PULONG FinalUncompressedSize) > +NTSTATUS WINAPI RtlDecompressBuffer(USHORT format, PUCHAR uncompressed, ULONG uncompressed_size, > + PUCHAR compressed, ULONG compressed_size, PULONG final_size) > { > - FIXME("0x%04x, %p, %u, %p, %u, %p :stub\n", CompressionFormat, UncompressedBuffer, UncompressedBufferSize, > - CompressedBuffer, CompressedBufferSize, FinalUncompressedSize); > + TRACE("0x%04x, %p, %u, %p, %u, %p\n", format, uncompressed, > + uncompressed_size, compressed, compressed_size, final_size); > > - return STATUS_NOT_IMPLEMENTED; > + return RtlDecompressFragment(format, uncompressed, uncompressed_size, > + compressed, compressed_size, 0, final_size, NULL); > } > > /*********************************************************************** > diff --git a/dlls/ntdll/tests/rtl.c b/dlls/ntdll/tests/rtl.c > index c6a7023..99fd413 100644 > --- a/dlls/ntdll/tests/rtl.c > +++ b/dlls/ntdll/tests/rtl.c > @@ -1663,11 +1663,8 @@ static void test_RtlCompressBuffer(void) > memset(buf2, 0x11, sizeof(buf2)); > status = pRtlDecompressBuffer(COMPRESSION_FORMAT_LZNT1, buf2, sizeof(buf2), > buf1, buf_size, &final_size); > - todo_wine > ok(status == STATUS_SUCCESS, "got wrong status 0x%08x\n", status); > - todo_wine > ok(final_size == sizeof(test_buffer), "got wrong final_size %u\n", final_size); > - todo_wine > ok(!memcmp(buf2, test_buffer, sizeof(test_buffer)), "got wrong decoded data\n"); > ok(buf2[sizeof(test_buffer)] == 0x11, "too many bytes written\n"); > > diff --git a/dlls/ntoskrnl.exe/ntoskrnl.exe.spec b/dlls/ntoskrnl.exe/ntoskrnl.exe.spec > index 11d1c62..0bb7d04 100644 > --- a/dlls/ntoskrnl.exe/ntoskrnl.exe.spec > +++ b/dlls/ntoskrnl.exe/ntoskrnl.exe.spec > @@ -977,7 +977,7 @@ > @ stub RtlCustomCPToUnicodeN > @ stdcall RtlDecompressBuffer(long ptr long ptr long ptr) ntdll.RtlDecompressBuffer > @ stub RtlDecompressChunks > -@ stub RtlDecompressFragment > +@ stdcall RtlDecompressFragment(long ptr long ptr long long ptr ptr) ntdll.RtlDecompressFragment > @ stub RtlDelete > @ stdcall RtlDeleteAce(ptr long) ntdll.RtlDeleteAce > @ stdcall RtlDeleteAtomFromAtomTable(ptr long) ntdll.RtlDeleteAtomFromAtomTable > -- > 2.4.5 > >

2 1

Re: gdiplus: GdipConvertToEmfPlus parameter succ is optional
by Nikolay Sivov July 9, 2015

July 9, 2015

Please add a test for this.

1 0

Re: libwine: Add support for tcc
by Alexandre Julliard July 8, 2015

July 8, 2015

André Hentschel <nerv(a)dawncrow.de> writes: > @@ -67,7 +67,7 @@ void DECLSPEC_NORETURN wine_switch_to_stack( void (*func)(void *), void *arg, vo > * Switch to the specified stack to call the function and return. > */ > > -#if defined(__i386__) && defined(__GNUC__) > +#if defined(__i386__) && (defined(__GNUC__) || defined(__TINYC__)) Are you planning to do this everywhere __GNUC__ is used? If so, this needs a better approach. -- Alexandre Julliard julliard(a)winehq.org

2 1

Re: [7/7] msvcirt: Implement ios::unsetf
by Marvin July 8, 2015

July 8, 2015

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=15160 Your paranoid android. === build (build) === Patch failed to apply

1 0

Re: [6/7] msvcirt: Implement ios::setf
by Marvin July 8, 2015

July 8, 2015

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=15159 Your paranoid android. === build (build) === Patch failed to apply

1 0

Re: [5/7] msvcirt: Implement ios::flags
by Marvin July 8, 2015

July 8, 2015

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=15158 Your paranoid android. === build (build) === Patch failed to apply

1 0

Re: [4/7] msvcirt: Implement ios static locking (try 2)
by Marvin July 8, 2015

July 8, 2015

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=15157 Your paranoid android. === build (build) === Patch failed to apply

1 0

Re: [3/7] msvcirt: Implement ios stream locking (try 2)
by Marvin July 8, 2015

July 8, 2015

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=15156 Your paranoid android. === build (build) === Patch failed to apply

1 0

Re: [2/7] msvcirt: Implement ios object locking (try 2)
by Marvin July 8, 2015

July 8, 2015

Hi, While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=15155 Your paranoid android. === build (build) === Patch failed to apply

1 0