May 2024 - Wine-Commits - list.winehq.org

Alexandre Julliard : ntdll: Update PROCESSOR_ARCHITECTURE variable correctly for ARM platforms.
by Alexandre Julliard May 14, 2024

May 14, 2024

Module: wine Branch: master Commit: 9dca2a707269dce7723e5a72204220560d3641d7 URL: https://gitlab.winehq.org/wine/wine/-/commit/9dca2a707269dce7723e5a72204220… Author: Alexandre Julliard <julliard(a)winehq.org> Date: Tue May 14 12:28:51 2024 +0200 ntdll: Update PROCESSOR_ARCHITECTURE variable correctly for ARM platforms. --- dlls/ntdll/env.c | 30 ++++++++++++++++++------------ 1 file changed, 18 insertions(+), 12 deletions(-) diff --git a/dlls/ntdll/env.c b/dlls/ntdll/env.c index 04e4ba94563..14f00ab1538 100644 --- a/dlls/ntdll/env.c +++ b/dlls/ntdll/env.c @@ -46,6 +46,18 @@ static inline SIZE_T get_env_length( const WCHAR *env ) return end + 1 - env; } +static UNICODE_STRING *get_machine_name( USHORT machine, UNICODE_STRING *str ) +{ + switch (machine) + { + case IMAGE_FILE_MACHINE_I386: RtlInitUnicodeString( str, L"x86" ); break; + case IMAGE_FILE_MACHINE_AMD64: RtlInitUnicodeString( str, L"AMD64" ); break; + case IMAGE_FILE_MACHINE_ARMNT: RtlInitUnicodeString( str, L"ARM" ); break; + case IMAGE_FILE_MACHINE_ARM64: RtlInitUnicodeString( str, L"ARM64" ); break; + default: RtlInitUnicodeString( str, L"Unknown" ); break; + } + return str; +} /*********************************************************************** * set_wow64_environment @@ -59,23 +71,17 @@ static void set_wow64_environment( WCHAR **env ) UNICODE_STRING arch6432_strW = RTL_CONSTANT_STRING( L"PROCESSOR_ARCHITEW6432" ); UNICODE_STRING valW = { 0, sizeof(buf), buf }; UNICODE_STRING nameW; + USHORT machine = RtlImageNtHeader( NtCurrentTeb()->Peb->ImageBaseAddress )->FileHeader.Machine; /* set the PROCESSOR_ARCHITECTURE variable */ - if (!RtlQueryEnvironmentVariable_U( *env, &arch6432_strW, &valW )) - { - if (is_win64) - { - RtlSetEnvironmentVariable( env, &arch_strW, &valW ); - RtlSetEnvironmentVariable( env, &arch6432_strW, NULL ); - } - } - else if (NtCurrentTeb64() && !RtlQueryEnvironmentVariable_U( *env, &arch_strW, &valW )) + RtlSetEnvironmentVariable( env, &arch_strW, get_machine_name( machine, &nameW )); + if (NtCurrentTeb()->WowTebOffset) { - RtlSetEnvironmentVariable( env, &arch6432_strW, &valW ); - RtlInitUnicodeString( &nameW, L"x86" ); - RtlSetEnvironmentVariable( env, &arch_strW, &nameW ); + RtlWow64GetProcessMachines( GetCurrentProcess(), NULL, &machine ); + RtlSetEnvironmentVariable( env, &arch6432_strW, get_machine_name( machine, &nameW )); } + else RtlSetEnvironmentVariable( env, &arch6432_strW, NULL ); /* set the ProgramFiles variables */

1 0

Alexandre Julliard : ntdll: Report the correct processor details on ARM platforms.
by Alexandre Julliard May 14, 2024

May 14, 2024

Module: wine Branch: master Commit: cd5eb0f143fa13795374530c603231cff40cbb8a URL: https://gitlab.winehq.org/wine/wine/-/commit/cd5eb0f143fa13795374530c603231… Author: Alexandre Julliard <julliard(a)winehq.org> Date: Tue May 14 10:45:37 2024 +0200 ntdll: Report the correct processor details on ARM platforms. --- dlls/ntdll/tests/info.c | 17 ++++---- dlls/ntdll/unix/system.c | 108 +++++++++++++++-------------------------------- 2 files changed, 42 insertions(+), 83 deletions(-) diff --git a/dlls/ntdll/tests/info.c b/dlls/ntdll/tests/info.c index 88086a83f59..8e8abbfb467 100644 --- a/dlls/ntdll/tests/info.c +++ b/dlls/ntdll/tests/info.c @@ -358,16 +358,15 @@ static void test_query_cpu(void) if (winetest_debug > 1) trace("Processor FeatureSet : %08lx\n", sci.ProcessorFeatureBits); ok( sci.ProcessorFeatureBits != 0, "Expected some features for this processor, got %08lx\n", sci.ProcessorFeatureBits); - - ok( sci.ProcessorLevel == sci2.ProcessorLevel, "ProcessorLevel differs %x / %x\n", - sci.ProcessorLevel, sci2.ProcessorLevel ); - ok( sci.ProcessorRevision == sci2.ProcessorRevision, "ProcessorRevision differs %x / %x\n", - sci.ProcessorRevision, sci2.ProcessorRevision ); - ok( sci.MaximumProcessors == sci2.MaximumProcessors, "MaximumProcessors differs %x / %x\n", - sci.MaximumProcessors, sci2.MaximumProcessors ); - ok( sci.ProcessorFeatureBits == sci2.ProcessorFeatureBits, "ProcessorFeatureBits differs %lx / %lx\n", - sci.ProcessorFeatureBits, sci2.ProcessorFeatureBits ); } + ok( sci.ProcessorLevel == sci2.ProcessorLevel, "ProcessorLevel differs %x / %x\n", + sci.ProcessorLevel, sci2.ProcessorLevel ); + ok( sci.ProcessorRevision == sci2.ProcessorRevision, "ProcessorRevision differs %x / %x\n", + sci.ProcessorRevision, sci2.ProcessorRevision ); + ok( sci.MaximumProcessors == sci2.MaximumProcessors, "MaximumProcessors differs %x / %x\n", + sci.MaximumProcessors, sci2.MaximumProcessors ); + ok( sci.ProcessorFeatureBits == sci2.ProcessorFeatureBits, "ProcessorFeatureBits differs %lx / %lx\n", + sci.ProcessorFeatureBits, sci2.ProcessorFeatureBits ); memset(&sci3, 0xcc, sizeof(sci3)); status = pNtQuerySystemInformation(SystemEmulationProcessorInformation, &sci3, sizeof(sci3), &len); diff --git a/dlls/ntdll/unix/system.c b/dlls/ntdll/unix/system.c index dcfc2ebd980..fc9dfa90355 100644 --- a/dlls/ntdll/unix/system.c +++ b/dlls/ntdll/unix/system.c @@ -538,71 +538,12 @@ static void get_cpuinfo( SYSTEM_CPU_INFORMATION *info ) info->ProcessorFeatureBits = cpu_features.ProcessorFeatureBits = features; } -#elif defined(__arm__) - -static inline void get_cpuinfo( SYSTEM_CPU_INFORMATION *info ) -{ - ULONGLONG features = 0; -#ifdef linux - char line[512]; - char *s, *value; - FILE *f = fopen("/proc/cpuinfo", "r"); - if (f) - { - while (fgets( line, sizeof(line), f )) - { - /* NOTE: the ':' is the only character we can rely on */ - if (!(value = strchr(line,':'))) continue; - /* terminate the valuename */ - s = value - 1; - while ((s >= line) && (*s == ' ' || *s == '\t')) s--; - s[1] = 0; - /* and strip leading spaces from value */ - value += 1; - while (*value == ' ' || *value == '\t') value++; - if ((s = strchr( value,'\n' ))) *s = 0; - if (!strcmp( line, "CPU architecture" )) - { - info->ProcessorLevel = atoi(value); - continue; - } - if (!strcmp( line, "CPU revision" )) - { - info->ProcessorRevision = atoi(value); - continue; - } - if (!strcmp( line, "Features" )) - { - if (strstr(value, "crc32")) features |= CPU_FEATURE_ARM_V8_CRC32; - if (strstr(value, "aes")) features |= CPU_FEATURE_ARM_V8_CRYPTO; - continue; - } - } - fclose( f ); - } -#elif defined(__FreeBSD__) - size_t valsize; - char buf[8]; - int value; - - valsize = sizeof(buf); - if (!sysctlbyname("hw.machine_arch", &buf, &valsize, NULL, 0) && sscanf(buf, "armv%i", &value) == 1) - info->ProcessorLevel = value; - - valsize = sizeof(value); - if (!sysctlbyname("hw.floatingpoint", &value, &valsize, NULL, 0)) features |= CPU_FEATURE_ARM_VFP_32; -#else - FIXME("CPU Feature detection not implemented.\n"); -#endif - info->ProcessorArchitecture = PROCESSOR_ARCHITECTURE_ARM; - info->ProcessorFeatureBits = cpu_features.ProcessorFeatureBits = features; -} - -#elif defined(__aarch64__) +#elif defined(__arm__) || defined(__aarch64__) static void get_cpuinfo( SYSTEM_CPU_INFORMATION *info ) { ULONGLONG features = 0; + unsigned int implementer = 0x41, part = 0, variant = 0, revision = 0; #ifdef linux char line[512]; char *s, *value; @@ -621,24 +562,23 @@ static void get_cpuinfo( SYSTEM_CPU_INFORMATION *info ) value += 1; while (*value == ' ' || *value == '\t') value++; if ((s = strchr( value,'\n' ))) *s = 0; - if (!strcmp( line, "CPU architecture" )) - { - info->ProcessorLevel = atoi(value); - continue; - } - if (!strcmp( line, "CPU revision" )) - { - info->ProcessorRevision = atoi(value); - continue; - } - if (!strcmp( line, "Features" )) + if (!strcmp( line, "CPU implementer" )) implementer = strtoul( value, NULL, 0); + else if (!strcmp( line, "CPU part" )) part = strtoul( value, NULL, 0); + else if (!strcmp( line, "CPU variant" )) variant = strtoul( value, NULL, 0); + else if (!strcmp( line, "CPU revision" )) revision = strtoul( value, NULL, 0); + else if (!strcmp( line, "Features" )) { +#ifdef __arm__ + if (strstr(value, "vfpv3")) features |= CPU_FEATURE_ARM_VFP_32; + if (strstr(value, "neon")) features |= CPU_FEATURE_ARM_NEON; +#else if (strstr(value, "crc32")) features |= CPU_FEATURE_ARM_V8_CRC32; if (strstr(value, "aes")) features |= CPU_FEATURE_ARM_V8_CRYPTO; if (strstr(value, "atomics")) features |= CPU_FEATURE_ARM_V81_ATOMIC; if (strstr(value, "asimddp")) features |= CPU_FEATURE_ARM_V82_DP; if (strstr(value, "jscvt")) features |= CPU_FEATURE_ARM_V83_JSCVT; if (strstr(value, "lrcpc")) features |= CPU_FEATURE_ARM_V83_LRCPC; +#endif continue; } } @@ -647,9 +587,29 @@ static void get_cpuinfo( SYSTEM_CPU_INFORMATION *info ) #else FIXME("CPU Feature detection not implemented.\n"); #endif - info->ProcessorLevel = max(info->ProcessorLevel, 8); +#ifdef __arm__ + info->ProcessorArchitecture = PROCESSOR_ARCHITECTURE_ARM; +#else info->ProcessorArchitecture = PROCESSOR_ARCHITECTURE_ARM64; +#endif info->ProcessorFeatureBits = cpu_features.ProcessorFeatureBits = features; + info->ProcessorLevel = part; + info->ProcessorRevision = (variant << 8) | revision; + cpu_id = (implementer << 24) | (variant << 20) | (0x0f << 16) | (part << 4) | revision; + switch (implementer) + { + case 0x41: strcpy( cpu_vendor, "ARM" ); break; + case 0x42: strcpy( cpu_vendor, "Broadcom" ); break; + case 0x43: strcpy( cpu_vendor, "Cavium" ); break; + case 0x44: strcpy( cpu_vendor, "DEC" ); break; + case 0x4e: strcpy( cpu_vendor, "Nvidia" ); break; + case 0x50: strcpy( cpu_vendor, "APM" ); break; + case 0x51: strcpy( cpu_vendor, "Qualcomm" ); break; + case 0x53: strcpy( cpu_vendor, "Samsung" ); break; + case 0x56: strcpy( cpu_vendor, "Marvell" ); break; + case 0x66: strcpy( cpu_vendor, "Faraday" ); break; + case 0x69: strcpy( cpu_vendor, "Intel" ); break; + } } #endif /* End architecture specific feature detection for CPUs */ @@ -1348,7 +1308,7 @@ void init_cpu_info(void) num = 1; FIXME("Detecting the number of processors is not supported.\n"); #endif - peb->NumberOfProcessors = num; + peb->NumberOfProcessors = cpu_info.MaximumProcessors = num; get_cpuinfo( &cpu_info ); TRACE( "<- CPU arch %d, level %d, rev %d, features 0x%x\n", (int)cpu_info.ProcessorArchitecture, (int)cpu_info.ProcessorLevel,

1 0

Alexandre Julliard : configure: Restore warning for missing PE compiler.
by Alexandre Julliard May 14, 2024

May 14, 2024

Module: wine Branch: master Commit: 4f68f67cb5d6c88a062220044f15b9cce8b2c418 URL: https://gitlab.winehq.org/wine/wine/-/commit/4f68f67cb5d6c88a062220044f15b9… Author: Alexandre Julliard <julliard(a)winehq.org> Date: Tue May 14 09:46:58 2024 +0200 configure: Restore warning for missing PE compiler. It was dropped in cda2886fd3018c9e8c12791238db481b528f6a65. Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=56665 --- configure | 11 +++++++++++ configure.ac | 3 +++ 2 files changed, 14 insertions(+) diff --git a/configure b/configure index 0468aea1422..2d1564bfea8 100755 --- a/configure +++ b/configure @@ -12027,6 +12027,17 @@ case $HOST_ARCH in arm|aarch64) test "x$PE_ARCHS" != x || as_fn_error $? "PE cross-compilation is required for $HOST_ARCH, please install clang/llvm-dlltool/lld, or llvm-mingw." "$LINENO" 5 DLLEXT="" ;; + *) + if test "x$PE_ARCHS" = "x" +then : + case "x$with_mingw" in + x) as_fn_append wine_notices "|Suitable PE cross-compiler not found, PE files won't be built." ;; + xno) ;; + *) as_fn_error $? "Suitable PE cross-compiler not found, PE files won't be built. +This is an error since --with-mingw was requested." "$LINENO" 5 ;; +esac + +fi ;; esac diff --git a/configure.ac b/configure.ac index 31c2fa2588e..71398ef9b29 100644 --- a/configure.ac +++ b/configure.ac @@ -1016,6 +1016,9 @@ case $HOST_ARCH in arm|aarch64) test "x$PE_ARCHS" != x || AC_MSG_ERROR([PE cross-compilation is required for $HOST_ARCH, please install clang/llvm-dlltool/lld, or llvm-mingw.]) DLLEXT="" ;; + *) + WINE_NOTICE_WITH(mingw,[test "x$PE_ARCHS" = "x"], + [Suitable PE cross-compiler not found, PE files won't be built.]) ;; esac dnl **** External libraries ****

1 0

Conor McCarthy : vkd3d-shader/spirv: Emit a trace instead of a warning message for ignored alignment.
by Alexandre Julliard May 13, 2024

May 13, 2024

Module: vkd3d Branch: master Commit: f47ac703ab2f6fd5d0ee2db1723485f10af9634a URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/f47ac703ab2f6fd5d0ee2db172348… Author: Conor McCarthy <cmccarthy(a)codeweavers.com> Date: Wed Apr 24 00:32:13 2024 +1000 vkd3d-shader/spirv: Emit a trace instead of a warning message for ignored alignment. --- libs/vkd3d-shader/spirv.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index af2be2e0..c1079166 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -3617,8 +3617,9 @@ static void spirv_compiler_emit_dereference_register(struct spirv_compiler *comp indexes[index_count++] = spirv_compiler_emit_register_addressing(compiler, &reg->idx[0]); } + /* Alignment is supported only in the Kernel execution model and is an optimisation only. */ if (reg->alignment) - WARN("Ignoring alignment %u.\n", reg->alignment); + TRACE("Ignoring alignment %u.\n", reg->alignment); if (index_count) { @@ -5904,8 +5905,9 @@ static void spirv_compiler_emit_dcl_indexable_temp(struct spirv_compiler *compil vsir_register_init(&reg, VKD3DSPR_IDXTEMP, VKD3D_DATA_FLOAT, 1); reg.idx[0].offset = temp->register_idx; + /* Alignment is supported only in the Kernel execution model and is an optimisation only. */ if (temp->alignment) - WARN("Ignoring alignment %u.\n", temp->alignment); + TRACE("Ignoring alignment %u.\n", temp->alignment); function_location = spirv_compiler_get_current_function_location(compiler); vkd3d_spirv_begin_function_stream_insertion(builder, function_location);

1 0

Conor McCarthy : vkd3d-shader/ir: Allow all DCL instructions to occur outside hull shader phases.
by Alexandre Julliard May 13, 2024

May 13, 2024

Module: vkd3d Branch: master Commit: 98f73ca2e5353b858f3249716dbb7ed957d8e3e1 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/98f73ca2e5353b858f3249716dbb7… Author: Conor McCarthy <cmccarthy(a)codeweavers.com> Date: Wed Apr 24 12:28:45 2024 +1000 vkd3d-shader/ir: Allow all DCL instructions to occur outside hull shader phases. --- libs/vkd3d-shader/ir.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) diff --git a/libs/vkd3d-shader/ir.c b/libs/vkd3d-shader/ir.c index c6ecbdd9..2c78447e 100644 --- a/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d-shader/ir.c @@ -5748,11 +5748,6 @@ static void vsir_validate_instruction(struct validation_context *ctx) instruction->declaration.max_tessellation_factor); return; - /* The DXIL parser can generate these outside phases, but this is not an issue. */ - case VKD3DSIH_DCL_INPUT: - case VKD3DSIH_DCL_OUTPUT: - return; - case VKD3DSIH_DCL_INPUT_PRIMITIVE: if (instruction->declaration.primitive_type.type == VKD3D_PT_UNDEFINED || instruction->declaration.primitive_type.type >= VKD3D_PT_COUNT) @@ -5810,7 +5805,9 @@ static void vsir_validate_instruction(struct validation_context *ctx) break; } - if (version->type == VKD3D_SHADER_TYPE_HULL && ctx->phase == VKD3DSIH_INVALID) + /* Only DCL instructions may occur outside hull shader phases. */ + if (!vsir_instruction_is_dcl(instruction) && version->type == VKD3D_SHADER_TYPE_HULL + && ctx->phase == VKD3DSIH_INVALID) validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_HANDLER, "Instruction %#x appear before any phase instruction in a hull shader.", instruction->handler_idx);

1 0

Conor McCarthy : vkd3d-shader/dxil: Delete bool not-equal-to-false comparison instructions.
by Alexandre Julliard May 13, 2024

May 13, 2024

Module: vkd3d Branch: master Commit: def87539de122956d9464664a61ef32f3b506bb0 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/def87539de122956d9464664a61ef… Author: Conor McCarthy <cmccarthy(a)codeweavers.com> Date: Wed Apr 24 12:23:27 2024 +1000 vkd3d-shader/dxil: Delete bool not-equal-to-false comparison instructions. --- libs/vkd3d-shader/dxil.c | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/libs/vkd3d-shader/dxil.c b/libs/vkd3d-shader/dxil.c index ab165910..22fdcf9f 100644 --- a/libs/vkd3d-shader/dxil.c +++ b/libs/vkd3d-shader/dxil.c @@ -6726,6 +6726,15 @@ static void sm6_parser_emit_cmp2(struct sm6_parser *sm6, const struct dxil_recor code = record->operands[i++]; + /* dxcompiler occasionally emits bool not-equal-to-false, which is a no-op. Bool comparisons + * do not otherwise occur, so deleting these avoids the need for backend support. */ + if (sm6_type_is_bool(type_a) && code == ICMP_NE && sm6_value_is_constant_zero(b)) + { + ins->handler_idx = VKD3DSIH_NOP; + *dst = *a; + return; + } + if ((!is_int && !is_fp) || is_int != (code >= ICMP_EQ)) { FIXME("Invalid operation %"PRIu64" on type class %u.\n", code, type_a->class);

1 0

Conor McCarthy : vkd3d-shader/spirv: Do not warn if no global flags are unhandled.
by Alexandre Julliard May 13, 2024

May 13, 2024

Module: vkd3d Branch: master Commit: e340855122b2eb0af03d6712b9d6a331b4a05c3d URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/e340855122b2eb0af03d6712b9d6a… Author: Conor McCarthy <cmccarthy(a)codeweavers.com> Date: Thu Nov 30 17:02:15 2023 +1000 vkd3d-shader/spirv: Do not warn if no global flags are unhandled. --- 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 a78af601..af2be2e0 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -5842,7 +5842,7 @@ static void spirv_compiler_emit_dcl_global_flags(struct spirv_compiler *compiler if (flags & ~(VKD3DSGF_REFACTORING_ALLOWED | VKD3DSGF_ENABLE_RAW_AND_STRUCTURED_BUFFERS)) FIXME("Unhandled global flags %#"PRIx64".\n", (uint64_t)flags); - else + else if (flags) WARN("Unhandled global flags %#"PRIx64".\n", (uint64_t)flags); }

1 0

Francisco Casas : vkd3d-shader/hlsl: Track bind count according to usage for uniforms.
by Alexandre Julliard May 13, 2024

May 13, 2024

Module: vkd3d Branch: master Commit: 573d5113440d992d4838e70bcffb75a62a6ddc4a URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/573d5113440d992d4838e70bcffb7… Author: Francisco Casas <fcasas(a)codeweavers.com> Date: Tue May 7 02:45:14 2024 -0400 vkd3d-shader/hlsl: Track bind count according to usage for uniforms. Where bind count means the number of registers required to include all dereferences to the variable within the shader as in hlsl_ir_var.bind_count[]. --- libs/vkd3d-shader/hlsl.h | 7 +-- libs/vkd3d-shader/hlsl_codegen.c | 101 ++++++++++++++++++++++++++++++--------- 2 files changed, 82 insertions(+), 26 deletions(-) diff --git a/libs/vkd3d-shader/hlsl.h b/libs/vkd3d-shader/hlsl.h index 209d59e1..0210cf9d 100644 --- a/libs/vkd3d-shader/hlsl.h +++ b/libs/vkd3d-shader/hlsl.h @@ -447,9 +447,10 @@ struct hlsl_ir_var enum hlsl_sampler_dim sampler_dim; struct vkd3d_shader_location first_sampler_dim_loc; } *objects_usage[HLSL_REGSET_LAST_OBJECT + 1]; - /* Minimum number of binds required to include all object components actually used in the shader. - * It may be less than the allocation size, e.g. for texture arrays. */ - unsigned int bind_count[HLSL_REGSET_LAST_OBJECT + 1]; + /* Minimum number of binds required to include all components actually used in the shader. + * It may be less than the allocation size, e.g. for texture arrays. + * The bind_count for HLSL_REGSET_NUMERIC is only used in uniforms for now. */ + unsigned int bind_count[HLSL_REGSET_LAST + 1]; /* Whether the shader performs dereferences with non-constant offsets in the variable. */ bool indexable; diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index a05b8e3f..25bd5d9c 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -4252,6 +4252,8 @@ static void register_deref_usage(struct hlsl_ctx *ctx, struct hlsl_deref *deref) { struct hlsl_ir_var *var = deref->var; enum hlsl_regset regset = hlsl_deref_get_regset(ctx, deref); + uint32_t required_bind_count; + struct hlsl_type *type; unsigned int index; if (!hlsl_regset_index_from_deref(ctx, deref, regset, &index)) @@ -4262,12 +4264,39 @@ static void register_deref_usage(struct hlsl_ctx *ctx, struct hlsl_deref *deref) var->objects_usage[regset][index].used = true; var->bind_count[regset] = max(var->bind_count[regset], index + 1); } + else if (regset == HLSL_REGSET_NUMERIC) + { + type = hlsl_deref_get_type(ctx, deref); + + hlsl_regset_index_from_deref(ctx, deref, regset, &index); + required_bind_count = align(index + type->reg_size[regset], 4) / 4; + var->bind_count[regset] = max(var->bind_count[regset], required_bind_count); + } + else + { + vkd3d_unreachable(); + } } -static bool track_object_components_usage(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context) +static bool track_components_usage(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context) { switch (instr->type) { + case HLSL_IR_LOAD: + { + struct hlsl_ir_load *load = hlsl_ir_load(instr); + + if (!load->src.var->is_uniform) + return false; + + /* These will are handled by validate_static_object_references(). */ + if (hlsl_deref_get_regset(ctx, &load->src) != HLSL_REGSET_NUMERIC) + return false; + + register_deref_usage(ctx, &load->src); + break; + } + case HLSL_IR_RESOURCE_LOAD: register_deref_usage(ctx, &hlsl_ir_resource_load(instr)->resource); if (hlsl_ir_resource_load(instr)->sampler.var) @@ -5106,14 +5135,15 @@ bool hlsl_component_index_range_from_deref(struct hlsl_ctx *ctx, const struct hl return true; } +/* Retrieves true if the index is constant, and false otherwise. In the latter case, the maximum + * possible index is retrieved, assuming there is not out-of-bounds access. */ bool hlsl_regset_index_from_deref(struct hlsl_ctx *ctx, const struct hlsl_deref *deref, enum hlsl_regset regset, unsigned int *index) { struct hlsl_type *type = deref->var->data_type; + bool index_is_constant = true; unsigned int i; - assert(regset <= HLSL_REGSET_LAST_OBJECT); - *index = 0; for (i = 0; i < deref->path_len; ++i) @@ -5122,37 +5152,62 @@ bool hlsl_regset_index_from_deref(struct hlsl_ctx *ctx, const struct hlsl_deref unsigned int idx = 0; assert(path_node); - if (path_node->type != HLSL_IR_CONSTANT) - return false; + if (path_node->type == HLSL_IR_CONSTANT) + { + /* We should always have generated a cast to UINT. */ + assert(path_node->data_type->class == HLSL_CLASS_SCALAR + && path_node->data_type->e.numeric.type == HLSL_TYPE_UINT); - /* We should always have generated a cast to UINT. */ - assert(path_node->data_type->class == HLSL_CLASS_SCALAR - && path_node->data_type->e.numeric.type == HLSL_TYPE_UINT); + idx = hlsl_ir_constant(path_node)->value.u[0].u; - idx = hlsl_ir_constant(path_node)->value.u[0].u; + switch (type->class) + { + case HLSL_CLASS_ARRAY: + if (idx >= type->e.array.elements_count) + return false; - switch (type->class) + *index += idx * type->e.array.type->reg_size[regset]; + break; + + case HLSL_CLASS_STRUCT: + *index += type->e.record.fields[idx].reg_offset[regset]; + break; + + case HLSL_CLASS_MATRIX: + *index += 4 * idx; + break; + + default: + vkd3d_unreachable(); + } + } + else { - case HLSL_CLASS_ARRAY: - if (idx >= type->e.array.elements_count) - return false; + index_is_constant = false; - *index += idx * type->e.array.type->reg_size[regset]; - break; + switch (type->class) + { + case HLSL_CLASS_ARRAY: + idx = type->e.array.elements_count - 1; + *index += idx * type->e.array.type->reg_size[regset]; + break; - case HLSL_CLASS_STRUCT: - *index += type->e.record.fields[idx].reg_offset[regset]; - break; + case HLSL_CLASS_MATRIX: + idx = hlsl_type_major_size(type) - 1; + *index += idx * 4; + break; - default: - vkd3d_unreachable(); + default: + vkd3d_unreachable(); + } } type = hlsl_get_element_type_from_path_index(ctx, type, path_node); } - assert(type->reg_size[regset] == 1); - return true; + assert(!(regset <= HLSL_REGSET_LAST_OBJECT) || (type->reg_size[regset] == 1)); + assert(!(regset == HLSL_REGSET_NUMERIC) || type->reg_size[regset] <= 4); + return index_is_constant; } bool hlsl_offset_from_deref(struct hlsl_ctx *ctx, const struct hlsl_deref *deref, unsigned int *offset) @@ -5449,7 +5504,7 @@ int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry compute_liveness(ctx, entry_func); while (hlsl_transform_ir(ctx, dce, body, NULL)); - hlsl_transform_ir(ctx, track_object_components_usage, body, NULL); + hlsl_transform_ir(ctx, track_components_usage, body, NULL); sort_synthetic_separated_samplers_first(ctx); if (profile->major_version < 4)

1 0

Francisco Casas : vkd3d-shader/hlsl: Improve tracking of used components running DCE before.
by Alexandre Julliard May 13, 2024

May 13, 2024

Module: vkd3d Branch: master Commit: e0a801e796c8c524890795dbebf804df3318e382 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/e0a801e796c8c524890795dbebf80… Author: Francisco Casas <fcasas(a)codeweavers.com> Date: Mon May 6 23:37:31 2024 -0400 vkd3d-shader/hlsl: Improve tracking of used components running DCE before. track_object_components_usage() had to be improved to also register derefs on resource stores. It was not doing it because it assumed that for every resource store there was a resource load already, which was true, before calling DCE. --- libs/vkd3d-shader/hlsl_codegen.c | 51 +++++++++++++++++++++++----------------- tests/hlsl/ternary.shader_test | 2 +- 2 files changed, 31 insertions(+), 22 deletions(-) diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index 4af55293..a05b8e3f 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -4248,34 +4248,38 @@ static bool track_object_components_sampler_dim(struct hlsl_ctx *ctx, struct hls return false; } -static bool track_object_components_usage(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context) +static void register_deref_usage(struct hlsl_ctx *ctx, struct hlsl_deref *deref) { - struct hlsl_ir_resource_load *load; - struct hlsl_ir_var *var; - enum hlsl_regset regset; + struct hlsl_ir_var *var = deref->var; + enum hlsl_regset regset = hlsl_deref_get_regset(ctx, deref); unsigned int index; - if (instr->type != HLSL_IR_RESOURCE_LOAD) - return false; - - load = hlsl_ir_resource_load(instr); - var = load->resource.var; - - regset = hlsl_deref_get_regset(ctx, &load->resource); + if (!hlsl_regset_index_from_deref(ctx, deref, regset, &index)) + return; - if (!hlsl_regset_index_from_deref(ctx, &load->resource, regset, &index)) - return false; + if (regset <= HLSL_REGSET_LAST_OBJECT) + { + var->objects_usage[regset][index].used = true; + var->bind_count[regset] = max(var->bind_count[regset], index + 1); + } +} - var->objects_usage[regset][index].used = true; - var->bind_count[regset] = max(var->bind_count[regset], index + 1); - if (load->sampler.var) +static bool track_object_components_usage(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context) +{ + switch (instr->type) { - var = load->sampler.var; - if (!hlsl_regset_index_from_deref(ctx, &load->sampler, HLSL_REGSET_SAMPLERS, &index)) - return false; + case HLSL_IR_RESOURCE_LOAD: + register_deref_usage(ctx, &hlsl_ir_resource_load(instr)->resource); + if (hlsl_ir_resource_load(instr)->sampler.var) + register_deref_usage(ctx, &hlsl_ir_resource_load(instr)->sampler); + break; + + case HLSL_IR_RESOURCE_STORE: + register_deref_usage(ctx, &hlsl_ir_resource_store(instr)->resource); + break; - var->objects_usage[HLSL_REGSET_SAMPLERS][index].used = true; - var->bind_count[HLSL_REGSET_SAMPLERS] = max(var->bind_count[HLSL_REGSET_SAMPLERS], index + 1); + default: + break; } return false; @@ -5440,6 +5444,11 @@ int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry hlsl_transform_ir(ctx, track_object_components_sampler_dim, body, NULL); if (profile->major_version >= 4) hlsl_transform_ir(ctx, lower_combined_samples, body, NULL); + + do + compute_liveness(ctx, entry_func); + while (hlsl_transform_ir(ctx, dce, body, NULL)); + hlsl_transform_ir(ctx, track_object_components_usage, body, NULL); sort_synthetic_separated_samplers_first(ctx); diff --git a/tests/hlsl/ternary.shader_test b/tests/hlsl/ternary.shader_test index 0ac78914..bd68dbc3 100644 --- a/tests/hlsl/ternary.shader_test +++ b/tests/hlsl/ternary.shader_test @@ -318,7 +318,7 @@ probe all rgba (2.0, 3.0, 3.0, 2.0) % Objects can be used, but their types have to be identical. -[pixel shader todo(sm<4)] +[pixel shader] Texture2D t; float4 main() : sv_target

1 0

Francisco Casas : vkd3d-shader/hlsl: Allocate SM1 numeric uniforms in decreasing bind count.
by Alexandre Julliard May 13, 2024

May 13, 2024

Module: vkd3d Branch: master Commit: 28d267b7c0238e4430a60859406fbdb5f9cf960e URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/28d267b7c0238e4430a60859406fb… Author: Francisco Casas <fcasas(a)codeweavers.com> Date: Fri May 3 02:51:44 2024 -0400 vkd3d-shader/hlsl: Allocate SM1 numeric uniforms in decreasing bind count. --- libs/vkd3d-shader/hlsl_codegen.c | 36 +++++++++++++++++++++++++++++ tests/hlsl/sm1-const-allocation.shader_test | 8 +++---- 2 files changed, 40 insertions(+), 4 deletions(-) diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index de778000..26179042 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -4553,11 +4553,47 @@ static void allocate_const_registers_recurse(struct hlsl_ctx *ctx, } } +static void sort_uniform_by_numeric_bind_count(struct list *sorted, struct hlsl_ir_var *to_sort) +{ + struct hlsl_ir_var *var; + + list_remove(&to_sort->extern_entry); + + LIST_FOR_EACH_ENTRY(var, sorted, struct hlsl_ir_var, extern_entry) + { + uint32_t to_sort_size = to_sort->bind_count[HLSL_REGSET_NUMERIC]; + uint32_t var_size = var->bind_count[HLSL_REGSET_NUMERIC]; + + if (to_sort_size > var_size) + { + list_add_before(&var->extern_entry, &to_sort->extern_entry); + return; + } + } + + list_add_tail(sorted, &to_sort->extern_entry); +} + +static void sort_uniforms_by_numeric_bind_count(struct hlsl_ctx *ctx) +{ + struct list sorted = LIST_INIT(sorted); + struct hlsl_ir_var *var, *next; + + LIST_FOR_EACH_ENTRY_SAFE(var, next, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) + { + if (var->is_uniform) + sort_uniform_by_numeric_bind_count(&sorted, var); + } + list_move_tail(&ctx->extern_vars, &sorted); +} + static void allocate_const_registers(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry_func) { struct register_allocator allocator = {0}; struct hlsl_ir_var *var; + sort_uniforms_by_numeric_bind_count(ctx); + LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) { unsigned int reg_size = var->data_type->reg_size[HLSL_REGSET_NUMERIC]; diff --git a/tests/hlsl/sm1-const-allocation.shader_test b/tests/hlsl/sm1-const-allocation.shader_test index e8a60255..3b14c42a 100644 --- a/tests/hlsl/sm1-const-allocation.shader_test +++ b/tests/hlsl/sm1-const-allocation.shader_test @@ -50,7 +50,7 @@ uniform 0 float4 1 2 3 4 uniform 4 float4 11 12 13 14 uniform 8 float4 21 22 23 24 draw quad -todo probe all rgba (21, 22, 23, 11) +probe all rgba (21, 22, 23, 11) [pixel shader] @@ -76,7 +76,7 @@ uniform 8 float4 21 22 23 24 uniform 12 float4 31 32 33 34 uniform 16 float4 41 42 43 44 draw quad -todo probe all rgba (41, 21, 0, 0) +probe all rgba (41, 21, 0, 0) [pixel shader] @@ -224,7 +224,7 @@ uniform 20 float4 51 52 53 54 uniform 24 float4 61 62 63 64 uniform 28 float4 71 72 73 74 draw quad -todo probe all rgba (74, 31, 63, 0) +probe all rgba (74, 31, 63, 0) [pixel shader] @@ -253,7 +253,7 @@ uniform 0 float4 1 2 3 4 uniform 4 float4 11 12 13 14 uniform 8 float4 21 22 23 24 draw quad -todo probe all rgba (21, 22, 1, 11) +probe all rgba (21, 22, 1, 11) [pixel shader]

1 0