March 2024 - Wine-commits - list.winehq.org

Conor McCarthy : vkd3d-shader/spirv: Introduce a data_type_is_floating_point() helper function.
by Alexandre Julliard 11 Mar '24

11 Mar '24

Module: vkd3d Branch: master Commit: 58123c2e10614b5fac0aa522a1ec4bb2b3ab73b8 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/58123c2e10614b5fac0aa522a1ec4… Author: Conor McCarthy <cmccarthy(a)codeweavers.com> Date: Wed Dec 13 15:25:57 2023 +1000 vkd3d-shader/spirv: Introduce a data_type_is_floating_point() helper function. --- libs/vkd3d-shader/spirv.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index b4782776..329b24ba 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -223,6 +223,11 @@ enum vkd3d_shader_input_sysval_semantic vkd3d_siv_from_sysval_indexed(enum vkd3d } } +static bool data_type_is_floating_point(enum vkd3d_data_type data_type) +{ + return data_type == VKD3D_DATA_FLOAT || data_type == VKD3D_DATA_DOUBLE; +} + #define VKD3D_SPIRV_VERSION 0x00010000 #define VKD3D_SPIRV_GENERATOR_ID 18 #define VKD3D_SPIRV_GENERATOR_VERSION 11 @@ -4104,7 +4109,7 @@ static uint32_t spirv_compiler_emit_abs(struct spirv_compiler *compiler, uint32_t type_id; type_id = spirv_compiler_get_type_id_for_reg(compiler, reg, write_mask); - if (reg->data_type == VKD3D_DATA_FLOAT || reg->data_type == VKD3D_DATA_DOUBLE) + if (data_type_is_floating_point(reg->data_type)) return vkd3d_spirv_build_op_glsl_std450_fabs(builder, type_id, val_id); FIXME("Unhandled data type %#x.\n", reg->data_type); @@ -4118,7 +4123,7 @@ static uint32_t spirv_compiler_emit_neg(struct spirv_compiler *compiler, uint32_t type_id; type_id = spirv_compiler_get_type_id_for_reg(compiler, reg, write_mask); - if (reg->data_type == VKD3D_DATA_FLOAT || reg->data_type == VKD3D_DATA_DOUBLE) + if (data_type_is_floating_point(reg->data_type)) return vkd3d_spirv_build_op_fnegate(builder, type_id, val_id); else if (data_type_is_integer(reg->data_type)) return vkd3d_spirv_build_op_snegate(builder, type_id, val_id); @@ -4302,7 +4307,7 @@ static uint32_t spirv_compiler_emit_sat(struct spirv_compiler *compiler, } type_id = spirv_compiler_get_type_id_for_reg(compiler, reg, write_mask); - if (reg->data_type == VKD3D_DATA_FLOAT || reg->data_type == VKD3D_DATA_DOUBLE) + if (data_type_is_floating_point(reg->data_type)) return vkd3d_spirv_build_op_glsl_std450_nclamp(builder, type_id, val_id, zero_id, one_id); FIXME("Unhandled data type %#x.\n", reg->data_type);

1 0

Conor McCarthy : vkd3d-shader/spirv: Introduce HALF and UINT16 types for minimum precision.
by Alexandre Julliard 11 Mar '24

11 Mar '24

Module: vkd3d Branch: master Commit: 066ea75945a570616ba072e834af170a09d19f37 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/066ea75945a570616ba072e834af1… Author: Conor McCarthy <cmccarthy(a)codeweavers.com> Date: Wed Dec 13 15:40:48 2023 +1000 vkd3d-shader/spirv: Introduce HALF and UINT16 types for minimum precision. Minimum precision types must always be implemented as 32-bit to match how reduced precision works in SPIR-V. --- libs/vkd3d-shader/d3d_asm.c | 2 + libs/vkd3d-shader/spirv.c | 84 ++++++++++++++++++++++++++++++-- libs/vkd3d-shader/vkd3d_shader_private.h | 8 ++- 3 files changed, 87 insertions(+), 7 deletions(-) diff --git a/libs/vkd3d-shader/d3d_asm.c b/libs/vkd3d-shader/d3d_asm.c index 2aabd5b1..7b30d260 100644 --- a/libs/vkd3d-shader/d3d_asm.c +++ b/libs/vkd3d-shader/d3d_asm.c @@ -649,6 +649,8 @@ static void shader_dump_data_type(struct vkd3d_d3d_asm_compiler *compiler, enum [VKD3D_DATA_UINT8 ] = "uint8", [VKD3D_DATA_UINT64 ] = "uint64", [VKD3D_DATA_BOOL ] = "bool", + [VKD3D_DATA_UINT16 ] = "uint16", + [VKD3D_DATA_HALF ] = "half", }; const char *name; diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index 329b24ba..56c89f76 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -225,7 +225,7 @@ enum vkd3d_shader_input_sysval_semantic vkd3d_siv_from_sysval_indexed(enum vkd3d static bool data_type_is_floating_point(enum vkd3d_data_type data_type) { - return data_type == VKD3D_DATA_FLOAT || data_type == VKD3D_DATA_DOUBLE; + return data_type == VKD3D_DATA_HALF || data_type == VKD3D_DATA_FLOAT || data_type == VKD3D_DATA_DOUBLE; } #define VKD3D_SPIRV_VERSION 0x00010000 @@ -1843,6 +1843,7 @@ static uint32_t vkd3d_spirv_get_type_id_for_data_type(struct vkd3d_spirv_builder { switch (data_type) { + case VKD3D_DATA_HALF: /* Minimum precision. TODO: native 16-bit */ case VKD3D_DATA_FLOAT: case VKD3D_DATA_SNORM: case VKD3D_DATA_UNORM: @@ -1850,6 +1851,7 @@ static uint32_t vkd3d_spirv_get_type_id_for_data_type(struct vkd3d_spirv_builder break; case VKD3D_DATA_INT: case VKD3D_DATA_UINT: + case VKD3D_DATA_UINT16: /* Minimum precision. TODO: native 16-bit */ return vkd3d_spirv_get_op_type_int(builder, 32, data_type == VKD3D_DATA_INT); break; case VKD3D_DATA_DOUBLE: @@ -3758,6 +3760,70 @@ static uint32_t spirv_compiler_emit_bool_to_double(struct spirv_compiler *compil return vkd3d_spirv_build_op_select(builder, type_id, val_id, true_id, false_id); } +/* Based on the implementation in the OpenGL Mathematics library. */ +static uint32_t half_to_float(uint16_t value) +{ + uint32_t s = (value & 0x8000u) << 16; + uint32_t e = (value >> 10) & 0x1fu; + uint32_t m = value & 0x3ffu; + + if (!e) + { + if (!m) + { + /* Plus or minus zero */ + return s; + } + else + { + /* Denormalized number -- renormalize it */ + + while (!(m & 0x400u)) + { + m <<= 1; + --e; + } + + ++e; + m &= ~0x400u; + } + } + else if (e == 31u) + { + /* Positive or negative infinity for zero 'm'. + * Nan for non-zero 'm' -- preserve sign and significand bits */ + return s | 0x7f800000u | (m << 13); + } + + /* Normalized number */ + e += 127u - 15u; + m <<= 13; + + /* Assemble s, e and m. */ + return s | (e << 23) | m; +} + +static uint32_t convert_raw_constant32(enum vkd3d_data_type data_type, unsigned int uint_value) +{ + int16_t i; + + /* TODO: native 16-bit support. */ + if (data_type != VKD3D_DATA_UINT16 && data_type != VKD3D_DATA_HALF) + return uint_value; + + if (data_type == VKD3D_DATA_HALF) + return half_to_float(uint_value); + + /* Values in DXIL have no signedness, so it is ambiguous whether 16-bit constants should or + * should not be sign-extended when 16-bit execution is not supported. The AMD RX 580 Windows + * driver has no 16-bit support, and sign-extends all 16-bit constant ints to 32 bits. These + * results differ from SM 5. The RX 6750 XT supports 16-bit execution, so constants are not + * extended, and results match SM 5. It seems best to replicate the sign-extension, and if + * execution is 16-bit, the values will be truncated. */ + i = uint_value; + return (int32_t)i; +} + static uint32_t spirv_compiler_emit_load_constant(struct spirv_compiler *compiler, const struct vkd3d_shader_register *reg, uint32_t swizzle, uint32_t write_mask) { @@ -3770,14 +3836,15 @@ static uint32_t spirv_compiler_emit_load_constant(struct spirv_compiler *compile if (reg->dimension == VSIR_DIMENSION_SCALAR) { for (i = 0; i < component_count; ++i) - values[i] = *reg->u.immconst_u32; + values[i] = convert_raw_constant32(reg->data_type, reg->u.immconst_u32[0]); } else { for (i = 0, j = 0; i < VKD3D_VEC4_SIZE; ++i) { if (write_mask & (VKD3DSP_WRITEMASK_0 << i)) - values[j++] = reg->u.immconst_u32[vsir_swizzle_get_component(swizzle, i)]; + values[j++] = convert_raw_constant32(reg->data_type, + reg->u.immconst_u32[vsir_swizzle_get_component(swizzle, i)]); } } @@ -3921,6 +3988,13 @@ static uint32_t spirv_compiler_emit_constant_array(struct spirv_compiler *compil switch (icb->data_type) { + case VKD3D_DATA_HALF: + case VKD3D_DATA_UINT16: + /* Scalar only. */ + for (i = 0; i < element_count; ++i) + elements[i] = vkd3d_spirv_get_op_constant(builder, elem_type_id, + convert_raw_constant32(icb->data_type, icb->data[i])); + break; case VKD3D_DATA_FLOAT: case VKD3D_DATA_INT: case VKD3D_DATA_UINT: @@ -6908,7 +6982,7 @@ static void spirv_compiler_emit_bool_cast(struct spirv_compiler *compiler, assert(src->reg.data_type == VKD3D_DATA_BOOL && dst->reg.data_type != VKD3D_DATA_BOOL); val_id = spirv_compiler_emit_load_src(compiler, src, dst->write_mask); - if (dst->reg.data_type == VKD3D_DATA_FLOAT) + if (dst->reg.data_type == VKD3D_DATA_HALF || dst->reg.data_type == VKD3D_DATA_FLOAT) { val_id = spirv_compiler_emit_bool_to_float(compiler, 1, val_id, instruction->handler_idx == VKD3DSIH_ITOF); } @@ -6917,7 +6991,7 @@ static void spirv_compiler_emit_bool_cast(struct spirv_compiler *compiler, /* ITOD is not supported. Frontends which emit bool casts must use ITOF for double. */ val_id = spirv_compiler_emit_bool_to_double(compiler, 1, val_id, instruction->handler_idx == VKD3DSIH_ITOF); } - else if (dst->reg.data_type == VKD3D_DATA_UINT) + else if (dst->reg.data_type == VKD3D_DATA_UINT16 || dst->reg.data_type == VKD3D_DATA_UINT) { val_id = spirv_compiler_emit_bool_to_int(compiler, 1, val_id, instruction->handler_idx == VKD3DSIH_ITOI); } diff --git a/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d-shader/vkd3d_shader_private.h index a666544c..d6f9d016 100644 --- a/libs/vkd3d-shader/vkd3d_shader_private.h +++ b/libs/vkd3d-shader/vkd3d_shader_private.h @@ -623,14 +623,16 @@ enum vkd3d_data_type VKD3D_DATA_UINT8, VKD3D_DATA_UINT64, VKD3D_DATA_BOOL, + VKD3D_DATA_UINT16, + VKD3D_DATA_HALF, VKD3D_DATA_COUNT, }; static inline bool data_type_is_integer(enum vkd3d_data_type data_type) { - return data_type == VKD3D_DATA_INT || data_type == VKD3D_DATA_UINT8 || data_type == VKD3D_DATA_UINT - || data_type == VKD3D_DATA_UINT64; + return data_type == VKD3D_DATA_INT || data_type == VKD3D_DATA_UINT8 || data_type == VKD3D_DATA_UINT16 + || data_type == VKD3D_DATA_UINT || data_type == VKD3D_DATA_UINT64; } static inline bool data_type_is_bool(enum vkd3d_data_type data_type) @@ -1531,10 +1533,12 @@ static inline enum vkd3d_shader_component_type vkd3d_component_type_from_data_ty { switch (data_type) { + case VKD3D_DATA_HALF: /* Minimum precision. TODO: native 16-bit */ case VKD3D_DATA_FLOAT: case VKD3D_DATA_UNORM: case VKD3D_DATA_SNORM: return VKD3D_SHADER_COMPONENT_FLOAT; + case VKD3D_DATA_UINT16: /* Minimum precision. TODO: native 16-bit */ case VKD3D_DATA_UINT: return VKD3D_SHADER_COMPONENT_UINT; case VKD3D_DATA_INT:

1 0

Conor McCarthy : vkd3d-shader/dxil: Support 16-bit types.
by Alexandre Julliard 11 Mar '24

11 Mar '24

Module: vkd3d Branch: master Commit: 4f0e5b9069c38bd7d3fde73e3e5131442c1d16d4 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/4f0e5b9069c38bd7d3fde73e3e513… Author: Conor McCarthy <cmccarthy(a)codeweavers.com> Date: Tue Jun 13 16:51:11 2023 +1000 vkd3d-shader/dxil: Support 16-bit types. --- libs/vkd3d-shader/dxil.c | 11 +++++++++-- tests/hlsl/minimum-precision.shader_test | 16 +++++++++++++--- 2 files changed, 22 insertions(+), 5 deletions(-) diff --git a/libs/vkd3d-shader/dxil.c b/libs/vkd3d-shader/dxil.c index dc810cb7..de51588b 100644 --- a/libs/vkd3d-shader/dxil.c +++ b/libs/vkd3d-shader/dxil.c @@ -2210,6 +2210,8 @@ static enum vkd3d_data_type vkd3d_data_type_from_sm6_type(const struct sm6_type return VKD3D_DATA_BOOL; case 8: return VKD3D_DATA_UINT8; + case 16: + return VKD3D_DATA_UINT16; case 32: return VKD3D_DATA_UINT; case 64: @@ -2223,6 +2225,8 @@ static enum vkd3d_data_type vkd3d_data_type_from_sm6_type(const struct sm6_type { switch (type->u.width) { + case 16: + return VKD3D_DATA_HALF; case 32: return VKD3D_DATA_FLOAT; case 64: @@ -2887,7 +2891,7 @@ static enum vkd3d_result sm6_parser_constants_init(struct sm6_parser *sm6, const } if (type->u.width == 16) - FIXME("Half float type is not supported yet.\n"); + dst->u.reg.u.immconst_u32[0] = record->operands[0]; else if (type->u.width == 32) dst->u.reg.u.immconst_f32[0] = bitcast_uint64_to_float(record->operands[0]); else if (type->u.width == 64) @@ -5130,7 +5134,10 @@ static enum vkd3d_shader_opcode sm6_map_cast_op(uint64_t code, const struct sm6_ break; case CAST_ZEXT: case CAST_SEXT: - /* nop or min precision. TODO: native 16-bit */ + /* nop or min precision. TODO: native 16-bit. + * Extension instructions could be emitted for min precision, but in Windows + * the AMD RX 580 simply drops such instructions, which makes sense as no + * assumptions should be made about any behaviour which depends on bit width. */ if (to->u.width == from->u.width || (to->u.width == 32 && from->u.width == 16)) { op = VKD3DSIH_NOP; diff --git a/tests/hlsl/minimum-precision.shader_test b/tests/hlsl/minimum-precision.shader_test index 773b2b39..f8d4eb25 100644 --- a/tests/hlsl/minimum-precision.shader_test +++ b/tests/hlsl/minimum-precision.shader_test @@ -30,16 +30,26 @@ uint4 main() : sv_target return uint4(u.x + i, u.y + j, 0, 0); } +[require] +shader model >= 4.0 +shader model < 6.0 + [test] uniform 0 uint4 0 0 0 0 draw quad probe all rgbaui (0x7fff, 0xffff, 0, 0) +[require] +shader model >= 6.0 + +[test] +uniform 0 uint4 0 0 0 0 +draw quad +probe all rgbaui (0x7fff, 0xffffffff, 0, 0) + % The code d3dcompiler_47 produces for this appears correct, but the result % is still zero in Windows. -[require] -shader model >= 6.0 [pixel shader] uniform min16uint4 u; @@ -55,7 +65,7 @@ uint4 main() : sv_target uniform 0 uint4 0 0 0 0 uniform 4 uint 2 draw quad -probe all rgbaui (0x7fff, 0xffff, 0, 0) +probe all rgbaui (0x7fff, 0xffffffff, 0, 0) uniform 0 uint4 0 0 0 0 uniform 4 uint 0 draw quad

1 0

Giovanni Mascellani : tests: Mark the HS punned array test as buggy on MoltenVK.
by Alexandre Julliard 11 Mar '24

11 Mar '24

Module: vkd3d Branch: master Commit: f6fe9a39dbf5accfaa4d3a3aca2b400362e0ed39 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/f6fe9a39dbf5accfaa4d3a3aca2b4… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Fri Mar 8 16:53:43 2024 +0100 tests: Mark the HS punned array test as buggy on MoltenVK. MoltenVK doesn't support the Component decoration on tessellation shaders. --- tests/d3d12.c | 1 + 1 file changed, 1 insertion(+) diff --git a/tests/d3d12.c b/tests/d3d12.c index 49b40a67..3e75fada 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -38139,6 +38139,7 @@ static void test_hull_shader_punned_array(void) transition_resource_state(command_list, context.render_target, D3D12_RESOURCE_STATE_RENDER_TARGET, D3D12_RESOURCE_STATE_COPY_SOURCE); + bug_if(is_mvk_device(context.device)) check_sub_resource_uint(context.render_target, 0, queue, command_list, 0xcc996633, 0); destroy_test_context(&context);

1 0

Giovanni Mascellani : tests: Pop the test context when leaving its scope.
by Alexandre Julliard 11 Mar '24

11 Mar '24

Module: vkd3d Branch: master Commit: cf816a255cb8249a20cdf0e21230977cfd867ceb URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/cf816a255cb8249a20cdf0e212309… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Fri Mar 8 17:01:53 2024 +0100 tests: Pop the test context when leaving its scope. --- tests/d3d12.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/tests/d3d12.c b/tests/d3d12.c index 3e75fada..ce907c96 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -32383,7 +32383,10 @@ static void test_resource_allocation_info(void) ok(infos1[i].SizeInBytes == sizes[i], "Got unexpected size %"PRIu64".\n", infos1[i].SizeInBytes); if (!i) + { + vkd3d_test_pop_context(); continue; + } ok(infos1[i].Offset - infos1[i - 1].Offset >= infos1[i - 1].SizeInBytes, "Got unexpected prev size %"PRIu64", prev offset %"PRIu64", offset %"PRIu64".\n",

1 0

Giovanni Mascellani : tests: Mark the coverage test as buggy on MoltenVK.
by Alexandre Julliard 11 Mar '24

11 Mar '24

Module: vkd3d Branch: master Commit: 0c63824d9ff723753ec0cc774af5e1778c414c71 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/0c63824d9ff723753ec0cc774af5e… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Fri Mar 8 23:14:58 2024 +0100 tests: Mark the coverage test as buggy on MoltenVK. Apparently MoltenVK doesn't set the SampleMask built-in to the fragment coverage mask. --- tests/d3d12.c | 1 + 1 file changed, 1 insertion(+) diff --git a/tests/d3d12.c b/tests/d3d12.c index ce907c96..7655c4e9 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -33454,6 +33454,7 @@ static void test_coverage(void) reset_command_list(command_list, context.allocator); transition_resource_state(command_list, texture, D3D12_RESOURCE_STATE_UNORDERED_ACCESS, D3D12_RESOURCE_STATE_COPY_SOURCE); + bug_if(is_mvk_device(context.device)) check_sub_resource_uint(texture, 0, queue, command_list, tests[i].sample_mask & 0xf, 0); reset_command_list(command_list, context.allocator);

1 0

Giovanni Mascellani : tests: Relax precision a bit for an interpolation test.
by Alexandre Julliard 11 Mar '24

11 Mar '24

Module: vkd3d Branch: master Commit: 5e43f9980935271d75325f49f97b0f6d6f12e59d URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/5e43f9980935271d75325f49f97b0… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Fri Mar 8 23:19:56 2024 +0100 tests: Relax precision a bit for an interpolation test. Otherwise the test fails on Intel, NVIDIA and llvmpipe. --- tests/hlsl/nointerpolation.shader_test | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/hlsl/nointerpolation.shader_test b/tests/hlsl/nointerpolation.shader_test index 94584ffe..d2b7d8a4 100644 --- a/tests/hlsl/nointerpolation.shader_test +++ b/tests/hlsl/nointerpolation.shader_test @@ -144,7 +144,7 @@ shader model >= 6.0 [test] draw triangle list 3 -probe (320,240) rgba (0.250390589, 0.499088615, 0.250520796, 1.0) 1 +probe (320,240) rgba (0.250390589, 0.499088615, 0.250520796, 1.0) 4 [require] shader model >= 4.0

1 0

Giovanni Mascellani : tests: Clear the UAV before testing.
by Alexandre Julliard 11 Mar '24

11 Mar '24

Module: vkd3d Branch: master Commit: dff22c9ed70811e4134c50d3d6d516b465e6a068 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/dff22c9ed70811e4134c50d3d6d51… Author: Giovanni Mascellani <gmascellani(a)codeweavers.com> Date: Fri Mar 8 23:38:21 2024 +0100 tests: Clear the UAV before testing. There is no guarantee that a resource is cleared upon creation. This fixes the test on NVIDIA and llvmpipe. --- tests/d3d12.c | 11 ----------- tests/d3d12_test_utils.h | 11 +++++++++++ tests/hlsl_d3d12.c | 7 +++++++ 3 files changed, 18 insertions(+), 11 deletions(-) diff --git a/tests/d3d12.c b/tests/d3d12.c index 7655c4e9..e3771b7f 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -128,17 +128,6 @@ static void check_alignment_(unsigned int line, uint64_t size, uint64_t alignmen size, aligned_size); } -static void uav_barrier(ID3D12GraphicsCommandList *list, ID3D12Resource *resource) -{ - D3D12_RESOURCE_BARRIER barrier; - - barrier.Type = D3D12_RESOURCE_BARRIER_TYPE_UAV; - barrier.Flags = D3D12_RESOURCE_BARRIER_FLAG_NONE; - barrier.UAV.pResource = resource; - - ID3D12GraphicsCommandList_ResourceBarrier(list, 1, &barrier); -} - static const DXGI_FORMAT depth_stencil_formats[] = { DXGI_FORMAT_R32G8X24_TYPELESS, diff --git a/tests/d3d12_test_utils.h b/tests/d3d12_test_utils.h index cf27b40f..4fe6fa71 100644 --- a/tests/d3d12_test_utils.h +++ b/tests/d3d12_test_utils.h @@ -269,6 +269,17 @@ static void transition_sub_resource_state(ID3D12GraphicsCommandList *list, ID3D1 ID3D12GraphicsCommandList_ResourceBarrier(list, 1, &barrier); } +static inline void uav_barrier(ID3D12GraphicsCommandList *list, ID3D12Resource *resource) +{ + D3D12_RESOURCE_BARRIER barrier; + + barrier.Type = D3D12_RESOURCE_BARRIER_TYPE_UAV; + barrier.Flags = D3D12_RESOURCE_BARRIER_FLAG_NONE; + barrier.UAV.pResource = resource; + + ID3D12GraphicsCommandList_ResourceBarrier(list, 1, &barrier); +} + #define create_command_queue(a, b, c) create_command_queue_(__LINE__, a, b, c) static inline ID3D12CommandQueue *create_command_queue_(unsigned int line, ID3D12Device *device, D3D12_COMMAND_LIST_TYPE type, int priority) diff --git a/tests/hlsl_d3d12.c b/tests/hlsl_d3d12.c index b2273ffb..4188d720 100644 --- a/tests/hlsl_d3d12.c +++ b/tests/hlsl_d3d12.c @@ -516,10 +516,17 @@ static void test_thread_id(void) for (i = 0; i < 3; ++i) { + const UINT clear_value[4] = {0}; + textures[i] = create_default_texture3d(device, 16, 8, 8, 1, DXGI_FORMAT_R32G32B32A32_UINT, D3D12_RESOURCE_FLAG_ALLOW_UNORDERED_ACCESS, D3D12_RESOURCE_STATE_UNORDERED_ACCESS); ID3D12Device_CreateUnorderedAccessView(device, textures[i], NULL, NULL, get_cpu_descriptor_handle(&context, heap, i)); + ID3D12GraphicsCommandList_ClearUnorderedAccessViewUint(command_list, + get_gpu_descriptor_handle(&context, heap, i), + get_cpu_descriptor_handle(&context, heap, i), + textures[i], clear_value, 0, NULL); + uav_barrier(command_list, textures[i]); } cs_code = compile_shader(cs_source, "cs_5_0");

1 0

Henri Verbeet : vkd3d-shader/glsl: Use location information from the current instruction.
by Alexandre Julliard 11 Mar '24

11 Mar '24

Module: vkd3d Branch: master Commit: 038764985ac8cb842c2be8948a15f25c3935feba URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/038764985ac8cb842c2be8948a15f… Author: Henri Verbeet <hverbeet(a)codeweavers.com> Date: Fri Nov 17 18:13:02 2023 +0100 vkd3d-shader/glsl: Use location information from the current instruction. --- libs/vkd3d-shader/glsl.c | 7 +++---- libs/vkd3d-shader/vkd3d_shader_main.c | 9 ++++----- libs/vkd3d-shader/vkd3d_shader_private.h | 2 +- 3 files changed, 8 insertions(+), 10 deletions(-) diff --git a/libs/vkd3d-shader/glsl.c b/libs/vkd3d-shader/glsl.c index c234caf8..e020bf3e 100644 --- a/libs/vkd3d-shader/glsl.c +++ b/libs/vkd3d-shader/glsl.c @@ -28,7 +28,7 @@ struct vkd3d_glsl_generator }; struct vkd3d_glsl_generator *vkd3d_glsl_generator_create(const struct vkd3d_shader_version *version, - struct vkd3d_shader_message_context *message_context, const struct vkd3d_shader_location *location) + struct vkd3d_shader_message_context *message_context) { struct vkd3d_glsl_generator *generator; @@ -38,7 +38,6 @@ struct vkd3d_glsl_generator *vkd3d_glsl_generator_create(const struct vkd3d_shad memset(generator, 0, sizeof(*generator)); generator->version = *version; vkd3d_string_buffer_init(&generator->buffer); - generator->location = *location; generator->message_context = message_context; return generator; } @@ -73,6 +72,8 @@ static void shader_glsl_ret(struct vkd3d_glsl_generator *generator, static void vkd3d_glsl_handle_instruction(struct vkd3d_glsl_generator *generator, const struct vkd3d_shader_instruction *instruction) { + generator->location = instruction->location; + switch (instruction->handler_idx) { case VKD3DSIH_DCL_INPUT: @@ -101,10 +102,8 @@ int vkd3d_glsl_generator_generate(struct vkd3d_glsl_generator *generator, vkd3d_string_buffer_printf(&generator->buffer, "#version 440\n\n"); vkd3d_string_buffer_printf(&generator->buffer, "void main()\n{\n"); - generator->location.column = 0; for (i = 0; i < program->instructions.count; ++i) { - generator->location.line = i + 1; vkd3d_glsl_handle_instruction(generator, &program->instructions.elements[i]); } diff --git a/libs/vkd3d-shader/vkd3d_shader_main.c b/libs/vkd3d-shader/vkd3d_shader_main.c index 42580f55..59a98d1b 100644 --- a/libs/vkd3d-shader/vkd3d_shader_main.c +++ b/libs/vkd3d-shader/vkd3d_shader_main.c @@ -1553,6 +1553,7 @@ static int vkd3d_shader_parser_compile(struct vkd3d_shader_parser *parser, struct vkd3d_shader_code *out, struct vkd3d_shader_message_context *message_context) { struct vkd3d_shader_scan_descriptor_info1 scan_descriptor_info; + struct vsir_program *program = &parser->program; struct vkd3d_glsl_generator *glsl_generator; struct vkd3d_shader_compile_info scan_info; int ret; @@ -1562,22 +1563,20 @@ static int vkd3d_shader_parser_compile(struct vkd3d_shader_parser *parser, switch (compile_info->target_type) { case VKD3D_SHADER_TARGET_D3D_ASM: - ret = vkd3d_dxbc_binary_to_text(&parser->program, &parser->shader_desc, - compile_info, out, VSIR_ASM_FLAG_NONE); + ret = vkd3d_dxbc_binary_to_text(program, &parser->shader_desc, compile_info, out, VSIR_ASM_FLAG_NONE); break; case VKD3D_SHADER_TARGET_GLSL: if ((ret = scan_with_parser(&scan_info, message_context, &scan_descriptor_info, parser)) < 0) return ret; - if (!(glsl_generator = vkd3d_glsl_generator_create(&parser->program.shader_version, - message_context, &parser->location))) + if (!(glsl_generator = vkd3d_glsl_generator_create(&program->shader_version, message_context))) { ERR("Failed to create GLSL generator.\n"); vkd3d_shader_free_scan_descriptor_info1(&scan_descriptor_info); return VKD3D_ERROR; } - ret = vkd3d_glsl_generator_generate(glsl_generator, &parser->program, out); + ret = vkd3d_glsl_generator_generate(glsl_generator, program, out); vkd3d_glsl_generator_destroy(glsl_generator); vkd3d_shader_free_scan_descriptor_info1(&scan_descriptor_info); break; diff --git a/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d-shader/vkd3d_shader_private.h index d6f9d016..ea52f151 100644 --- a/libs/vkd3d-shader/vkd3d_shader_private.h +++ b/libs/vkd3d-shader/vkd3d_shader_private.h @@ -1506,7 +1506,7 @@ int shader_parse_input_signature(const struct vkd3d_shader_code *dxbc, struct vkd3d_glsl_generator; struct vkd3d_glsl_generator *vkd3d_glsl_generator_create(const struct vkd3d_shader_version *version, - struct vkd3d_shader_message_context *message_context, const struct vkd3d_shader_location *location); + struct vkd3d_shader_message_context *message_context); int vkd3d_glsl_generator_generate(struct vkd3d_glsl_generator *generator, struct vsir_program *program, struct vkd3d_shader_code *out); void vkd3d_glsl_generator_destroy(struct vkd3d_glsl_generator *generator);

1 0

Evan Tang : vkd3d: Enable VK_EXT_fragment_shader_interlock if available.
by Alexandre Julliard 11 Mar '24

11 Mar '24

Module: vkd3d Branch: master Commit: 03a80cde2ea20ca54e8f8e5b1fabaf0200b1c757 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/03a80cde2ea20ca54e8f8e5b1faba… Author: Evan Tang <etang(a)codeweavers.com> Date: Fri Oct 13 16:23:21 2023 -0500 vkd3d: Enable VK_EXT_fragment_shader_interlock if available. --- libs/vkd3d/device.c | 29 +++++++++++++++++++++++-- libs/vkd3d/vkd3d_private.h | 3 ++- tests/hlsl/rasteriser-ordered-views.shader_test | 2 +- 3 files changed, 30 insertions(+), 4 deletions(-) diff --git a/libs/vkd3d/device.c b/libs/vkd3d/device.c index da9c1d96..7841a811 100644 --- a/libs/vkd3d/device.c +++ b/libs/vkd3d/device.c @@ -94,6 +94,7 @@ static const struct vkd3d_optional_extension_info optional_device_extensions[] = VK_EXTENSION(EXT_DEBUG_MARKER, EXT_debug_marker), VK_EXTENSION(EXT_DEPTH_CLIP_ENABLE, EXT_depth_clip_enable), VK_EXTENSION(EXT_DESCRIPTOR_INDEXING, EXT_descriptor_indexing), + VK_EXTENSION(EXT_FRAGMENT_SHADER_INTERLOCK, EXT_fragment_shader_interlock), VK_EXTENSION(EXT_MUTABLE_DESCRIPTOR_TYPE, EXT_mutable_descriptor_type), VK_EXTENSION(EXT_ROBUSTNESS_2, EXT_robustness2), VK_EXTENSION(EXT_SHADER_DEMOTE_TO_HELPER_INVOCATION, EXT_shader_demote_to_helper_invocation), @@ -789,6 +790,7 @@ struct vkd3d_physical_device_info VkPhysicalDeviceConditionalRenderingFeaturesEXT conditional_rendering_features; VkPhysicalDeviceDepthClipEnableFeaturesEXT depth_clip_features; VkPhysicalDeviceDescriptorIndexingFeaturesEXT descriptor_indexing_features; + VkPhysicalDeviceFragmentShaderInterlockFeaturesEXT fragment_shader_interlock_features; VkPhysicalDeviceRobustness2FeaturesEXT robustness2_features; VkPhysicalDeviceShaderDemoteToHelperInvocationFeaturesEXT demote_features; VkPhysicalDeviceTexelBufferAlignmentFeaturesEXT texel_buffer_alignment_features; @@ -808,6 +810,7 @@ static void vkd3d_physical_device_info_init(struct vkd3d_physical_device_info *i VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT *vertex_divisor_properties; VkPhysicalDeviceTexelBufferAlignmentPropertiesEXT *buffer_alignment_properties; VkPhysicalDeviceDescriptorIndexingFeaturesEXT *descriptor_indexing_features; + VkPhysicalDeviceFragmentShaderInterlockFeaturesEXT *fragment_shader_interlock_features; VkPhysicalDeviceRobustness2FeaturesEXT *robustness2_features; VkPhysicalDeviceVertexAttributeDivisorFeaturesEXT *vertex_divisor_features; VkPhysicalDeviceTexelBufferAlignmentFeaturesEXT *buffer_alignment_features; @@ -825,6 +828,7 @@ static void vkd3d_physical_device_info_init(struct vkd3d_physical_device_info *i conditional_rendering_features = &info->conditional_rendering_features; depth_clip_features = &info->depth_clip_features; descriptor_indexing_features = &info->descriptor_indexing_features; + fragment_shader_interlock_features = &info->fragment_shader_interlock_features; robustness2_features = &info->robustness2_features; descriptor_indexing_properties = &info->descriptor_indexing_properties; maintenance3_properties = &info->maintenance3_properties; @@ -846,6 +850,8 @@ static void vkd3d_physical_device_info_init(struct vkd3d_physical_device_info *i vk_prepend_struct(&info->features2, depth_clip_features); descriptor_indexing_features->sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DESCRIPTOR_INDEXING_FEATURES_EXT; vk_prepend_struct(&info->features2, descriptor_indexing_features); + fragment_shader_interlock_features->sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FRAGMENT_SHADER_INTERLOCK_FEATURES_EXT; + vk_prepend_struct(&info->features2, fragment_shader_interlock_features); robustness2_features->sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_ROBUSTNESS_2_FEATURES_EXT; vk_prepend_struct(&info->features2, robustness2_features); demote_features->sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_DEMOTE_TO_HELPER_INVOCATION_FEATURES_EXT; @@ -1158,6 +1164,7 @@ static void vkd3d_trace_physical_device_limits(const struct vkd3d_physical_devic static void vkd3d_trace_physical_device_features(const struct vkd3d_physical_device_info *info) { + const VkPhysicalDeviceFragmentShaderInterlockFeaturesEXT *fragment_shader_interlock_features; const VkPhysicalDeviceConditionalRenderingFeaturesEXT *conditional_rendering_features; const VkPhysicalDeviceShaderDemoteToHelperInvocationFeaturesEXT *demote_features; const VkPhysicalDeviceTexelBufferAlignmentFeaturesEXT *buffer_alignment_features; @@ -1279,6 +1286,15 @@ static void vkd3d_trace_physical_device_features(const struct vkd3d_physical_dev TRACE(" VkPhysicalDeviceDepthClipEnableFeaturesEXT:\n"); TRACE(" depthClipEnable: %#x.\n", depth_clip_features->depthClipEnable); + fragment_shader_interlock_features = &info->fragment_shader_interlock_features; + TRACE(" VkPhysicalDeviceFragmentShaderInterlockFeaturesEXT:\n"); + TRACE(" fragmentShaderSampleInterlock: %#x.\n.", + fragment_shader_interlock_features->fragmentShaderSampleInterlock); + TRACE(" fragmentShaderPixelInterlock: %#x\n.", + fragment_shader_interlock_features->fragmentShaderPixelInterlock); + TRACE(" fragmentShaderShadingRateInterlock: %#x\n.", + fragment_shader_interlock_features->fragmentShaderShadingRateInterlock); + demote_features = &info->demote_features; TRACE(" VkPhysicalDeviceShaderDemoteToHelperInvocationFeaturesEXT:\n"); TRACE(" shaderDemoteToHelperInvocation: %#x.\n", demote_features->shaderDemoteToHelperInvocation); @@ -1476,6 +1492,7 @@ static HRESULT vkd3d_init_device_caps(struct d3d12_device *device, uint32_t *device_extension_count, bool **user_extension_supported) { const struct vkd3d_vk_instance_procs *vk_procs = &device->vkd3d_instance->vk_procs; + VkPhysicalDeviceFragmentShaderInterlockFeaturesEXT *fragment_shader_interlock; const struct vkd3d_optional_device_extensions_info *optional_extensions; VkPhysicalDeviceDescriptorIndexingFeaturesEXT *descriptor_indexing; VkPhysicalDevice physical_device = device->vk_physical_device; @@ -1539,8 +1556,6 @@ static HRESULT vkd3d_init_device_caps(struct d3d12_device *device, device->feature_options.TypedUAVLoadAdditionalFormats = features->shaderStorageImageReadWithoutFormat && d3d12_device_supports_typed_uav_load_additional_formats(device); - /* GL_INTEL_fragment_shader_ordering, no Vulkan equivalent. */ - device->feature_options.ROVsSupported = FALSE; /* GL_INTEL_conservative_rasterization, no Vulkan equivalent. */ device->feature_options.ConservativeRasterizationTier = D3D12_CONSERVATIVE_RASTERIZATION_TIER_NOT_SUPPORTED; device->feature_options.MaxGPUVirtualAddressBitsPerResource = 40; /* FIXME */ @@ -1619,6 +1634,12 @@ static HRESULT vkd3d_init_device_caps(struct d3d12_device *device, *user_extension_supported, vulkan_info, "device", device->vkd3d_instance->config_flags & VKD3D_CONFIG_FLAG_VULKAN_DEBUG); + fragment_shader_interlock = &physical_device_info->fragment_shader_interlock_features; + if (!fragment_shader_interlock->fragmentShaderSampleInterlock + || !fragment_shader_interlock->fragmentShaderPixelInterlock) + vulkan_info->EXT_fragment_shader_interlock = false; + device->feature_options.ROVsSupported = vulkan_info->EXT_fragment_shader_interlock; + if (!physical_device_info->conditional_rendering_features.conditionalRendering) vulkan_info->EXT_conditional_rendering = false; if (!physical_device_info->depth_clip_features.depthClipEnable) @@ -1675,6 +1696,10 @@ static HRESULT vkd3d_init_device_caps(struct d3d12_device *device, vulkan_info->shader_extensions[vulkan_info->shader_extension_count++] = VKD3D_SHADER_SPIRV_EXTENSION_EXT_DESCRIPTOR_INDEXING; + if (vulkan_info->EXT_fragment_shader_interlock) + vulkan_info->shader_extensions[vulkan_info->shader_extension_count++] + = VKD3D_SHADER_SPIRV_EXTENSION_EXT_FRAGMENT_SHADER_INTERLOCK; + if (vulkan_info->EXT_shader_stencil_export) vulkan_info->shader_extensions[vulkan_info->shader_extension_count++] = VKD3D_SHADER_SPIRV_EXTENSION_EXT_STENCIL_EXPORT; diff --git a/libs/vkd3d/vkd3d_private.h b/libs/vkd3d/vkd3d_private.h index 4f82b211..aff2c440 100644 --- a/libs/vkd3d/vkd3d_private.h +++ b/libs/vkd3d/vkd3d_private.h @@ -55,7 +55,7 @@ #define VKD3D_MAX_COMPATIBLE_FORMAT_COUNT 6u #define VKD3D_MAX_QUEUE_FAMILY_COUNT 3u -#define VKD3D_MAX_SHADER_EXTENSIONS 4u +#define VKD3D_MAX_SHADER_EXTENSIONS 5u #define VKD3D_MAX_SHADER_STAGES 5u #define VKD3D_MAX_VK_SYNC_OBJECTS 4u #define VKD3D_MAX_DEVICE_BLOCKED_QUEUES 16u @@ -133,6 +133,7 @@ struct vkd3d_vulkan_info bool EXT_debug_marker; bool EXT_depth_clip_enable; bool EXT_descriptor_indexing; + bool EXT_fragment_shader_interlock; bool EXT_mutable_descriptor_type; bool EXT_robustness2; bool EXT_shader_demote_to_helper_invocation; diff --git a/tests/hlsl/rasteriser-ordered-views.shader_test b/tests/hlsl/rasteriser-ordered-views.shader_test index a7aa7827..3b8f32ab 100644 --- a/tests/hlsl/rasteriser-ordered-views.shader_test +++ b/tests/hlsl/rasteriser-ordered-views.shader_test @@ -64,6 +64,6 @@ shader model >= 5.0 rov [test] -draw triangle list 93 +todo(sm>=6) draw triangle list 93 probe all rui(0x6aaaaaaa) probe uav 1 all rui(0x2aaaaaaa)

1 0