August 2018 - Wine-Devel - list.winehq.org

[PATCH v2] ntoskrnl: Emulate sti/cli instructions on x86_64
by Fabian Maurer Aug. 2, 2018

Aug. 2, 2018

Thanks to Anastasius Focht for the clear bug report. Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=45521 Signed-off-by: Fabian Maurer <dark.shadow4(a)web.de> --- dlls/ntoskrnl.exe/instr.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/dlls/ntoskrnl.exe/instr.c b/dlls/ntoskrnl.exe/instr.c index b2dac58fc4..67cdd327ee 100644 --- a/dlls/ntoskrnl.exe/instr.c +++ b/dlls/ntoskrnl.exe/instr.c @@ -815,6 +815,11 @@ static DWORD emulate_instruction( EXCEPTION_RECORD *rec, CONTEXT *context ) } break; /* Unable to emulate it */ } + + case 0xfa: /* cli */ + case 0xfb: /* sti */ + context->Rip += prefixlen + 1; + return ExceptionContinueExecution; } return ExceptionContinueSearch; /* Unable to emulate it */ } -- 2.18.0

1 0

kernel32: Add support for MOVEFILE_WRITE_THROUGH to MoveFile.
by Dmitry Timoshkov Aug. 2, 2018

Aug. 2, 2018

Signed-off-by: Dmitry Timoshkov <dmitry(a)baikal.ru> --- dlls/kernel32/path.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) diff --git a/dlls/kernel32/path.c b/dlls/kernel32/path.c index a389743171..6d2a9029b9 100644 --- a/dlls/kernel32/path.c +++ b/dlls/kernel32/path.c @@ -1283,6 +1283,7 @@ BOOL WINAPI MoveFileWithProgressW( LPCWSTR source, LPCWSTR dest, NTSTATUS status; HANDLE source_handle = 0, dest_handle; ANSI_STRING source_unix, dest_unix; + DWORD options; TRACE("(%s,%s,%p,%p,%04x)\n", debugstr_w(source), debugstr_w(dest), fnProgress, param, flag ); @@ -1293,9 +1294,6 @@ BOOL WINAPI MoveFileWithProgressW( LPCWSTR source, LPCWSTR dest, if (!dest) return DeleteFileW( source ); - if (flag & MOVEFILE_WRITE_THROUGH) - FIXME("MOVEFILE_WRITE_THROUGH unimplemented\n"); - /* check if we are allowed to rename the source */ if (!RtlDosPathNameToNtPathName_U( source, &nt_name, NULL, NULL )) @@ -1336,8 +1334,11 @@ BOOL WINAPI MoveFileWithProgressW( LPCWSTR source, LPCWSTR dest, SetLastError( ERROR_PATH_NOT_FOUND ); goto error; } - status = NtOpenFile( &dest_handle, GENERIC_READ | GENERIC_WRITE | SYNCHRONIZE, &attr, &io, 0, - FILE_NON_DIRECTORY_FILE | FILE_SYNCHRONOUS_IO_NONALERT ); + + options = FILE_NON_DIRECTORY_FILE | FILE_SYNCHRONOUS_IO_NONALERT; + if (flag & MOVEFILE_WRITE_THROUGH) + options |= FILE_WRITE_THROUGH; + status = NtOpenFile( &dest_handle, GENERIC_READ | GENERIC_WRITE | SYNCHRONIZE, &attr, &io, 0, options ); if (status == STATUS_SUCCESS) /* destination exists */ { NtClose( dest_handle ); -- 2.17.1

1 0

server: Add support for FILE_WRITE_THROUGH to create_file() request.
by Dmitry Timoshkov Aug. 2, 2018

Aug. 2, 2018

Signed-off-by: Dmitry Timoshkov <dmitry(a)baikal.ru> --- server/file.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/server/file.c b/server/file.c index 446621a997..9299530a4f 100644 --- a/server/file.c +++ b/server/file.c @@ -224,6 +224,11 @@ static struct object *create_file( struct fd *root, const char *nameptr, data_si default: set_error( STATUS_INVALID_PARAMETER ); goto done; } +#ifdef O_DIRECT + if (options & FILE_WRITE_THROUGH) + flags |= O_DIRECT; +#endif + if (sd) { const SID *owner = sd_get_owner( sd ); -- 2.17.1

1 0

[PATCH] libwine: Use getsegmentdata(3) on Mac OS to find the end of the __TEXT segment.
by Chip Davis Aug. 1, 2018

Aug. 1, 2018

Don't assume it ends with the fake PE header. This assumption doesn't hold on Mac OS: the __data section where it was placed is located after several other sections, all in the __DATA segment. Unfortunately, this causes Wine, when DEP/NX is turned off, to override the page protections for the start of the __DATA segment, removing write permission from them, leading to a crash when winemac.drv attempted to use an Objective-C class for the first time. Also, be sure to include the zero-fill (i.e. BSS) sections in the total size of the .data section. This should fix some tests that use large uninitialized arrays. Signed-off-by: Chip Davis <cdavis(a)codeweavers.com> --- Notes: Try 2: Don't make data_start the same as code_end. Try 3: Reduce duplication a bit. libs/wine/loader.c | 19 ++++++++++++++++--- 1 file changed, 16 insertions(+), 3 deletions(-) diff --git a/libs/wine/loader.c b/libs/wine/loader.c index c07042a583e6..0af3b8e5204d 100644 --- a/libs/wine/loader.c +++ b/libs/wine/loader.c @@ -49,6 +49,7 @@ #undef LoadResource #undef GetCurrentThread #include <pthread.h> +#include <mach-o/getsect.h> #else extern char **environ; #endif @@ -387,11 +388,15 @@ static void *map_dll( const IMAGE_NT_HEADERS *nt_descr ) IMAGE_NT_HEADERS *nt; IMAGE_SECTION_HEADER *sec; BYTE *addr; - DWORD code_start, data_start, data_end; + DWORD code_start, code_end, data_start, data_end; const size_t page_size = sysconf( _SC_PAGESIZE ); const size_t page_mask = page_size - 1; int delta, nb_sections = 2; /* code + data */ unsigned int i; +#ifdef __APPLE__ + Dl_info dli; + unsigned long data_size; +#endif size_t size = (sizeof(IMAGE_DOS_HEADER) + sizeof(IMAGE_NT_HEADERS) @@ -425,7 +430,15 @@ static void *map_dll( const IMAGE_NT_HEADERS *nt_descr ) delta = (const BYTE *)nt_descr - addr; code_start = page_size; data_start = delta & ~page_mask; +#ifdef __APPLE__ + /* Need the mach_header, not the PE header, to give to getsegmentdata(3) */ + dladdr(addr, &dli); + code_end = getsegmentdata(dli.dli_fbase, "__DATA", &data_size) - addr; + data_end = (code_end + data_size + page_mask) & ~page_mask; +#else + code_end = data_start; data_end = (nt->OptionalHeader.SizeOfImage + delta + page_mask) & ~page_mask; +#endif fixup_rva_ptrs( &nt->OptionalHeader.AddressOfEntryPoint, addr, 1 ); @@ -434,7 +447,7 @@ static void *map_dll( const IMAGE_NT_HEADERS *nt_descr ) #ifndef _WIN64 nt->OptionalHeader.BaseOfData = data_start; #endif - nt->OptionalHeader.SizeOfCode = data_start - code_start; + nt->OptionalHeader.SizeOfCode = code_end - code_start; nt->OptionalHeader.SizeOfInitializedData = data_end - data_start; nt->OptionalHeader.SizeOfUninitializedData = 0; nt->OptionalHeader.SizeOfImage = data_end; @@ -443,7 +456,7 @@ static void *map_dll( const IMAGE_NT_HEADERS *nt_descr ) /* Build the code section */ memcpy( sec->Name, ".text", sizeof(".text") ); - sec->SizeOfRawData = data_start - code_start; + sec->SizeOfRawData = code_end - code_start; sec->Misc.VirtualSize = sec->SizeOfRawData; sec->VirtualAddress = code_start; sec->PointerToRawData = code_start; -- 2.18.0

2 1

[PATCH] libwine: Use getsegmentdata(3) on Mac OS to find the end of the __TEXT segment.
by Chip Davis Aug. 1, 2018

Aug. 1, 2018

Don't assume it ends with the fake PE header. This assumption doesn't hold on Mac OS: the __data section where it was placed is located after several other sections, all in the __DATA segment. Unfortunately, this causes Wine, when DEP/NX is turned off, to override the page protections for the start of the __DATA segment, removing write permission from them, leading to a crash when winemac.drv attempted to use an Objective-C class for the first time. Also, be sure to include the zero-fill (i.e. BSS) sections in the total size of the .data section. This should fix some tests that use large uninitialized arrays. Signed-off-by: Chip Davis <cdavis(a)codeweavers.com> --- Notes: Try 2: Don't make data_start the same as code_end. libs/wine/loader.c | 19 ++++++++++++++++--- 1 file changed, 16 insertions(+), 3 deletions(-) diff --git a/libs/wine/loader.c b/libs/wine/loader.c index c07042a583e6..41ffb02fc43c 100644 --- a/libs/wine/loader.c +++ b/libs/wine/loader.c @@ -49,6 +49,7 @@ #undef LoadResource #undef GetCurrentThread #include <pthread.h> +#include <mach-o/getsect.h> #else extern char **environ; #endif @@ -387,11 +388,15 @@ static void *map_dll( const IMAGE_NT_HEADERS *nt_descr ) IMAGE_NT_HEADERS *nt; IMAGE_SECTION_HEADER *sec; BYTE *addr; - DWORD code_start, data_start, data_end; + DWORD code_start, code_end, data_start, data_end; const size_t page_size = sysconf( _SC_PAGESIZE ); const size_t page_mask = page_size - 1; int delta, nb_sections = 2; /* code + data */ unsigned int i; +#ifdef __APPLE__ + Dl_info dli; + unsigned long data_size; +#endif size_t size = (sizeof(IMAGE_DOS_HEADER) + sizeof(IMAGE_NT_HEADERS) @@ -425,7 +430,15 @@ static void *map_dll( const IMAGE_NT_HEADERS *nt_descr ) delta = (const BYTE *)nt_descr - addr; code_start = page_size; data_start = delta & ~page_mask; +#ifdef __APPLE__ + /* Need the mach_header, not the PE header, to give to getsegmentdata(3) */ + dladdr(addr, &dli); + code_end = getsegmentdata(dli.dli_fbase, "__DATA", &data_size) - addr; + data_end = (code_end + data_size + page_mask) & ~page_mask; +#else + code_end = delta & ~page_mask; data_end = (nt->OptionalHeader.SizeOfImage + delta + page_mask) & ~page_mask; +#endif fixup_rva_ptrs( &nt->OptionalHeader.AddressOfEntryPoint, addr, 1 ); @@ -434,7 +447,7 @@ static void *map_dll( const IMAGE_NT_HEADERS *nt_descr ) #ifndef _WIN64 nt->OptionalHeader.BaseOfData = data_start; #endif - nt->OptionalHeader.SizeOfCode = data_start - code_start; + nt->OptionalHeader.SizeOfCode = code_end - code_start; nt->OptionalHeader.SizeOfInitializedData = data_end - data_start; nt->OptionalHeader.SizeOfUninitializedData = 0; nt->OptionalHeader.SizeOfImage = data_end; @@ -443,7 +456,7 @@ static void *map_dll( const IMAGE_NT_HEADERS *nt_descr ) /* Build the code section */ memcpy( sec->Name, ".text", sizeof(".text") ); - sec->SizeOfRawData = data_start - code_start; + sec->SizeOfRawData = code_end - code_start; sec->Misc.VirtualSize = sec->SizeOfRawData; sec->VirtualAddress = code_start; sec->PointerToRawData = code_start; -- 2.18.0

2 1

[PATCH vkd3d 13/14] libs/vkd3d-shader: Zero-initialize unused shader output components.
by Józef Kucia Aug. 1, 2018

Aug. 1, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d-shader/spirv.c | 67 ++++++++++++++++++++++++++++++++++------------- 1 file changed, 49 insertions(+), 18 deletions(-) diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index f3344c4763ba..a5dae19ef7cb 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -1863,7 +1863,7 @@ struct vkd3d_dxbc_compiler bool after_declarations_section; const struct vkd3d_shader_signature *input_signature; const struct vkd3d_shader_signature *output_signature; - struct + struct vkd3d_shader_output_info { uint32_t id; enum vkd3d_component_type component_type; @@ -2129,11 +2129,17 @@ static uint32_t vkd3d_dxbc_compiler_get_constant_float(struct vkd3d_dxbc_compile return vkd3d_dxbc_compiler_get_constant(compiler, VKD3D_TYPE_FLOAT, 1, (uint32_t *)&value); } +static uint32_t vkd3d_dxbc_compiler_get_constant_vector(struct vkd3d_dxbc_compiler *compiler, + enum vkd3d_component_type component_type, unsigned int component_count, uint32_t value) +{ + const uint32_t values[] = {value, value, value, value}; + return vkd3d_dxbc_compiler_get_constant(compiler, component_type, component_count, values); +} + static uint32_t vkd3d_dxbc_compiler_get_constant_uint_vector(struct vkd3d_dxbc_compiler *compiler, uint32_t value, unsigned int component_count) { - const uint32_t values[] = {value, value, value, value}; - return vkd3d_dxbc_compiler_get_constant(compiler, VKD3D_TYPE_UINT, component_count, values); + return vkd3d_dxbc_compiler_get_constant_vector(compiler, VKD3D_TYPE_UINT, component_count, value); } static uint32_t vkd3d_dxbc_compiler_get_constant_float_vector(struct vkd3d_dxbc_compiler *compiler, @@ -6259,14 +6265,50 @@ void vkd3d_dxbc_compiler_handle_instruction(struct vkd3d_dxbc_compiler *compiler } } +static void vkd3d_dxbc_compiler_emit_store_shader_output(struct vkd3d_dxbc_compiler *compiler, + const struct vkd3d_shader_signature_element *output, + const struct vkd3d_shader_output_info *output_info, uint32_t val_id) +{ + struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; + uint32_t write_mask, use_mask, uninit_mask, swizzle; + unsigned int component_count; + uint32_t type_id, zero_id; + + write_mask = output->mask & 0xff; + use_mask = (output->mask >> 8) & 0xff; + swizzle = get_shader_output_swizzle(compiler, output->register_index); + val_id = vkd3d_dxbc_compiler_emit_swizzle(compiler, + val_id, VKD3D_TYPE_FLOAT, swizzle, write_mask); + + component_count = vkd3d_write_mask_component_count(write_mask); + + if (output_info->component_type != VKD3D_TYPE_FLOAT) + { + type_id = vkd3d_spirv_get_type_id(builder, output_info->component_type, component_count); + val_id = vkd3d_spirv_build_op_bitcast(builder, type_id, val_id); + } + + uninit_mask = write_mask & use_mask; + if (uninit_mask) + { + /* Set values to 0 for not initialized shader output components. */ + zero_id = vkd3d_dxbc_compiler_get_constant_vector(compiler, + output_info->component_type, VKD3D_VEC4_SIZE, 0); + val_id = vkd3d_dxbc_compiler_emit_vector_shuffle(compiler, + val_id, zero_id, uninit_mask, output_info->component_type, component_count); + } + + vkd3d_spirv_build_op_store(builder, output_info->id, val_id, SpvMemoryAccessMaskNone); +} + static void vkd3d_dxbc_compiler_emit_output_setup_function(struct vkd3d_dxbc_compiler *compiler) { - uint32_t void_id, type_id, ptr_type_id, function_type_id, function_id, val_id; uint32_t param_type_id[MAX_REG_OUTPUT + 1], param_id[MAX_REG_OUTPUT + 1] = {}; const struct vkd3d_shader_signature *signature = compiler->output_signature; + uint32_t void_id, type_id, ptr_type_id, function_type_id, function_id; struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; - DWORD write_mask, swizzle, variable_idx; unsigned int i, count; + DWORD variable_idx; function_id = compiler->output_setup_function_id; @@ -6309,19 +6351,8 @@ static void vkd3d_dxbc_compiler_emit_output_setup_function(struct vkd3d_dxbc_com if (!param_id[variable_idx]) continue; - write_mask = signature->elements[i].mask & 0xff; - swizzle = get_shader_output_swizzle(compiler, signature->elements[i].register_index); - val_id = vkd3d_dxbc_compiler_emit_swizzle(compiler, - param_id[variable_idx], VKD3D_TYPE_FLOAT, swizzle, write_mask); - - if (compiler->output_info[i].component_type != VKD3D_TYPE_FLOAT) - { - type_id = vkd3d_spirv_get_type_id(builder, compiler->output_info[i].component_type, - vkd3d_write_mask_component_count(write_mask)); - val_id = vkd3d_spirv_build_op_bitcast(builder, type_id, val_id); - } - - vkd3d_spirv_build_op_store(builder, compiler->output_info[i].id, val_id, SpvMemoryAccessMaskNone); + vkd3d_dxbc_compiler_emit_store_shader_output(compiler, + &signature->elements[i], &compiler->output_info[i], param_id[variable_idx]); } vkd3d_spirv_build_op_return(&compiler->spirv_builder); -- 2.16.4

2 1

[PATCH vkd3d 10/14] libs/vkd3d-shader: Fix vkd3d_find_signature_element_for_reg().
by Józef Kucia Aug. 1, 2018

Aug. 1, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> The exact match of write masks between a shader signature and dcl_input/dcl_output instructions isn't required. Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d-shader/spirv.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index 26052f2e2068..0549fb9a85bc 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -3016,7 +3016,7 @@ static const struct vkd3d_shader_signature_element *vkd3d_find_signature_element for (signature_idx = 0; signature_idx < signature->element_count; ++signature_idx) { if (signature->elements[signature_idx].register_index == reg_idx - && (signature->elements[signature_idx].mask & 0xff) == write_mask) + && (signature->elements[signature_idx].mask & write_mask) == write_mask) { if (signature_element_index) *signature_element_index = signature_idx; -- 2.16.4

2 1

[PATCH vkd3d 07/14] libs/vkd3d-shader: Translate partial derivative instructions.
by Józef Kucia Aug. 1, 2018

Aug. 1, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d-shader/spirv.c | 65 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 65 insertions(+) diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index 21d6a1a33288..797ea1c4018e 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -4954,6 +4954,63 @@ static void vkd3d_dxbc_compiler_emit_control_flow_instruction(struct vkd3d_dxbc_ } } +static void vkd3d_dxbc_compiler_emit_deriv_instruction(struct vkd3d_dxbc_compiler *compiler, + const struct vkd3d_shader_instruction *instruction) +{ + struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; + const struct vkd3d_shader_dst_param *dst = instruction->dst; + const struct vkd3d_shader_src_param *src = instruction->src; + const struct instruction_info *info; + uint32_t type_id, src_id, val_id; + unsigned int component_count; + unsigned int i; + + static const struct instruction_info + { + enum VKD3D_SHADER_INSTRUCTION_HANDLER handler_idx; + SpvOp op; + bool needs_derivative_control; + } + deriv_instructions[] = + { + {VKD3DSIH_DSX, SpvOpDPdx}, + {VKD3DSIH_DSX_COARSE, SpvOpDPdxCoarse, true}, + {VKD3DSIH_DSX_FINE, SpvOpDPdxFine, true}, + {VKD3DSIH_DSY, SpvOpDPdy}, + {VKD3DSIH_DSY_COARSE, SpvOpDPdyCoarse, true}, + {VKD3DSIH_DSY_FINE, SpvOpDPdyFine, true}, + }; + + info = NULL; + for (i = 0; i < ARRAY_SIZE(deriv_instructions); ++i) + { + if (deriv_instructions[i].handler_idx == instruction->handler_idx) + { + info = &deriv_instructions[i]; + break; + } + } + if (!info) + { + ERR("Unexpected instruction %#x.\n", instruction->handler_idx); + return; + } + + if (info->needs_derivative_control) + vkd3d_spirv_enable_capability(builder, SpvCapabilityDerivativeControl); + + assert(instruction->dst_count == 1); + assert(instruction->src_count == 1); + + component_count = vkd3d_write_mask_component_count(dst->write_mask); + type_id = vkd3d_spirv_get_type_id(builder, + vkd3d_component_type_from_data_type(dst->reg.data_type), component_count); + + src_id = vkd3d_dxbc_compiler_emit_load_src(compiler, src, dst->write_mask); + val_id = vkd3d_spirv_build_op_tr1(builder, &builder->function_stream, info->op, type_id, src_id); + vkd3d_dxbc_compiler_emit_store_dst(compiler, dst, val_id); +} + struct vkd3d_shader_image { uint32_t id; @@ -6060,6 +6117,14 @@ void vkd3d_dxbc_compiler_handle_instruction(struct vkd3d_dxbc_compiler *compiler case VKD3DSIH_TEXKILL: vkd3d_dxbc_compiler_emit_control_flow_instruction(compiler, instruction); break; + case VKD3DSIH_DSX: + case VKD3DSIH_DSX_COARSE: + case VKD3DSIH_DSX_FINE: + case VKD3DSIH_DSY: + case VKD3DSIH_DSY_COARSE: + case VKD3DSIH_DSY_FINE: + vkd3d_dxbc_compiler_emit_deriv_instruction(compiler, instruction); + break; case VKD3DSIH_LD: vkd3d_dxbc_compiler_emit_ld(compiler, instruction); break; -- 2.16.4

2 1

[PATCH vkd3d 14/14] tests: Add test for incompletely initialized pixel shader outputs.
by Józef Kucia Aug. 1, 2018

Aug. 1, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- tests/d3d12.c | 78 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 78 insertions(+) diff --git a/tests/d3d12.c b/tests/d3d12.c index dfc3da7bdec0..a0185951999d 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -7512,6 +7512,83 @@ static void test_shader_interstage_interface(void) destroy_test_context(&context); } +static void test_incompletely_initialized_shader_outputs(void) +{ + static const float white[] = {1.0f, 1.0f, 1.0f, 1.0f}; + ID3D12GraphicsCommandList *command_list; + struct test_context_desc desc; + struct test_context context; + ID3D12CommandQueue *queue; + unsigned int i; + + static const DWORD ps_code[] = + { +#if 0 + float3 color; + + void main(out float4 o : SV_Target) + { + o.xyz = color; + } +#endif + 0x43425844, 0xb5d86526, 0xc6c2455a, 0x8d02fe46, 0x5932818b, 0x00000001, 0x000000bc, 0x00000003, + 0x0000002c, 0x0000003c, 0x00000070, 0x4e475349, 0x00000008, 0x00000000, 0x00000008, 0x4e47534f, + 0x0000002c, 0x00000001, 0x00000008, 0x00000020, 0x00000000, 0x00000000, 0x00000003, 0x00000000, + 0x0000080f, 0x545f5653, 0x65677261, 0xabab0074, 0x58454853, 0x00000044, 0x00000050, 0x00000011, + 0x0100086a, 0x04000059, 0x00208e46, 0x00000000, 0x00000001, 0x03000065, 0x00102072, 0x00000000, + 0x06000036, 0x00102072, 0x00000000, 0x00208246, 0x00000000, 0x00000000, 0x0100003e, + }; + static const D3D12_SHADER_BYTECODE ps = {ps_code, sizeof(ps_code)}; + static struct + { + struct vec4 input; + struct vec4 output; + } + tests[] = + { + {{0.0f, 0.0f, 0.0f, 0.0f}, {0.0f, 0.0f, 0.0f, 0.0f}}, + {{1.0f, 1.0f, 1.0f, 1.0f}, {1.0f, 1.0f, 1.0f, 0.0f}}, + {{1.0f, 2.0f, 3.0f, 4.0f}, {1.0f, 2.0f, 3.0f, 0.0f}}, + }; + + memset(&desc, 0, sizeof(desc)); + desc.rt_format = DXGI_FORMAT_R32G32B32A32_FLOAT; + desc.no_root_signature = true; + if (!init_test_context(&context, &desc)) + return; + command_list = context.list; + queue = context.queue; + + context.root_signature = create_32bit_constants_root_signature(context.device, + 0, 4, D3D12_SHADER_VISIBILITY_PIXEL); + context.pipeline_state = create_pipeline_state(context.device, + context.root_signature, desc.rt_format, NULL, &ps, NULL); + + for (i = 0; i < ARRAY_SIZE(tests); ++i) + { + ID3D12GraphicsCommandList_ClearRenderTargetView(command_list, context.rtv, white, 0, NULL); + + ID3D12GraphicsCommandList_OMSetRenderTargets(command_list, 1, &context.rtv, FALSE, NULL); + ID3D12GraphicsCommandList_SetGraphicsRootSignature(command_list, context.root_signature); + ID3D12GraphicsCommandList_SetGraphicsRoot32BitConstants(command_list, 0, 4, &tests[i].input, 0); + ID3D12GraphicsCommandList_SetPipelineState(command_list, context.pipeline_state); + ID3D12GraphicsCommandList_IASetPrimitiveTopology(command_list, D3D_PRIMITIVE_TOPOLOGY_TRIANGLESTRIP); + ID3D12GraphicsCommandList_RSSetViewports(command_list, 1, &context.viewport); + ID3D12GraphicsCommandList_RSSetScissorRects(command_list, 1, &context.scissor_rect); + ID3D12GraphicsCommandList_DrawInstanced(command_list, 4, 1, 0, 0); + + transition_resource_state(command_list, context.render_target, + D3D12_RESOURCE_STATE_RENDER_TARGET, D3D12_RESOURCE_STATE_COPY_SOURCE); + check_sub_resource_vec4(context.render_target, 0, queue, command_list, &tests[i].output, 0); + + reset_command_list(command_list, context.allocator); + transition_resource_state(command_list, context.render_target, + D3D12_RESOURCE_STATE_COPY_SOURCE, D3D12_RESOURCE_STATE_RENDER_TARGET); + } + + destroy_test_context(&context); +} + static void check_descriptor_range_(unsigned int line, const D3D12_DESCRIPTOR_RANGE *range, const D3D12_DESCRIPTOR_RANGE *expected_range) { @@ -17992,6 +18069,7 @@ START_TEST(d3d12) run_test(test_shader_instructions); run_test(test_discard_instruction); run_test(test_shader_interstage_interface); + run_test(test_incompletely_initialized_shader_outputs); run_test(test_root_signature_byte_code); run_test(test_cs_constant_buffer); run_test(test_constant_buffer_relative_addressing); -- 2.16.4

2 1

[PATCH vkd3d 12/14] libs/vkd3d-shader: Factor out vkd3d_dxbc_compiler_emit_vector_shuffle().
by Józef Kucia Aug. 1, 2018

Aug. 1, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d-shader/spirv.c | 43 ++++++++++++++++++++++++++++--------------- 1 file changed, 28 insertions(+), 15 deletions(-) diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index 142993a42c44..f3344c4763ba 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -2475,6 +2475,30 @@ static uint32_t vkd3d_dxbc_compiler_emit_swizzle(struct vkd3d_dxbc_compiler *com val_id, VKD3DSP_WRITEMASK_ALL, component_type, swizzle, write_mask); } +static uint32_t vkd3d_dxbc_compiler_emit_vector_shuffle(struct vkd3d_dxbc_compiler *compiler, + uint32_t vector1_id, uint32_t vector2_id, uint32_t write_mask, + enum vkd3d_component_type component_type, unsigned int component_count) +{ + struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; + uint32_t components[VKD3D_VEC4_SIZE]; + unsigned int i, component_idx; + uint32_t type_id; + + assert(component_count <= ARRAY_SIZE(components)); + + for (i = 0, component_idx = 0; i < component_count; ++i) + { + if (write_mask & (VKD3DSP_WRITEMASK_0 << i)) + components[i] = VKD3D_VEC4_SIZE + component_idx++; + else + components[i] = i; + } + + type_id = vkd3d_spirv_get_type_id(builder, component_type, component_count); + return vkd3d_spirv_build_op_vector_shuffle(builder, + type_id, vector1_id, vector2_id, components, component_count); +} + static uint32_t vkd3d_dxbc_compiler_emit_load_constant(struct vkd3d_dxbc_compiler *compiler, const struct vkd3d_shader_register *reg, DWORD swizzle, DWORD write_mask) { @@ -2700,9 +2724,8 @@ static void vkd3d_dxbc_compiler_emit_store_reg(struct vkd3d_dxbc_compiler *compi const struct vkd3d_shader_register *reg, DWORD write_mask, uint32_t val_id) { struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; - unsigned int i, component_idx, component_count; - uint32_t components[VKD3D_VEC4_SIZE]; - uint32_t reg_id; + uint32_t reg_id, type_id, reg_val_id; + unsigned int component_count; assert(reg->type != VKD3DSPR_IMMCONST); assert(write_mask); @@ -2722,21 +2745,11 @@ static void vkd3d_dxbc_compiler_emit_store_reg(struct vkd3d_dxbc_compiler *compi if (component_count != VKD3D_VEC4_SIZE) { - uint32_t type_id, reg_val_id; - type_id = vkd3d_spirv_get_type_id(builder, VKD3D_TYPE_FLOAT, VKD3D_VEC4_SIZE); reg_val_id = vkd3d_spirv_build_op_load(builder, type_id, reg_id, SpvMemoryAccessMaskNone); - for (i = 0, component_idx = 0; i < ARRAY_SIZE(components); ++i) - { - if (write_mask & (VKD3DSP_WRITEMASK_0 << i)) - components[i] = VKD3D_VEC4_SIZE + component_idx++; - else - components[i] = i; - } - - val_id = vkd3d_spirv_build_op_vector_shuffle(builder, - type_id, reg_val_id, val_id, components, ARRAY_SIZE(components)); + val_id = vkd3d_dxbc_compiler_emit_vector_shuffle(compiler, + reg_val_id, val_id, write_mask, VKD3D_TYPE_FLOAT, VKD3D_VEC4_SIZE); } vkd3d_spirv_build_op_store(builder, reg_id, val_id, SpvMemoryAccessMaskNone); -- 2.16.4

2 1