vkd3d-shader: Fix f32tof16 instruction.

Signed-off-by: Józef Kucia <jkucia@codeweavers.com>
Signed-off-by: Henri Verbeet <hverbeet@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>
This commit is contained in:
Józef Kucia 2018-10-18 14:59:44 +02:00 committed by Alexandre Julliard
parent dc9fdf5d31
commit 55abf3d6d4
2 changed files with 33 additions and 5 deletions

View File

@ -4760,12 +4760,13 @@ static void vkd3d_dxbc_compiler_emit_f16tof32(struct vkd3d_dxbc_compiler *compil
static void vkd3d_dxbc_compiler_emit_f32tof16(struct vkd3d_dxbc_compiler *compiler,
const struct vkd3d_shader_instruction *instruction)
{
uint32_t instr_set_id, type_id, scalar_type_id, src_id, zero_id, constituents[2], result_id;
uint32_t instr_set_id, type_id, scalar_type_id, src_id, zero_id, constituents[2];
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;
uint32_t components[VKD3D_VEC4_SIZE];
unsigned int i, j;
DWORD write_mask;
unsigned int i;
instr_set_id = vkd3d_spirv_get_glsl_std450_instr_set(builder);
type_id = vkd3d_spirv_get_type_id(builder, VKD3D_TYPE_FLOAT, 2);
@ -4773,7 +4774,7 @@ static void vkd3d_dxbc_compiler_emit_f32tof16(struct vkd3d_dxbc_compiler *compil
zero_id = vkd3d_dxbc_compiler_get_constant_float(compiler, 0.0f);
/* FIXME: Consider a single PackHalf2x16 intruction per 2 components. */
for (i = 0; i < VKD3D_VEC4_SIZE; ++i)
for (i = 0, j = 0; i < VKD3D_VEC4_SIZE; ++i)
{
if (!(write_mask = dst->write_mask & (VKD3DSP_WRITEMASK_0 << i)))
continue;
@ -4783,10 +4784,12 @@ static void vkd3d_dxbc_compiler_emit_f32tof16(struct vkd3d_dxbc_compiler *compil
constituents[1] = zero_id;
src_id = vkd3d_spirv_build_op_composite_construct(builder,
type_id, constituents, ARRAY_SIZE(constituents));
result_id = vkd3d_spirv_build_op_ext_inst(builder, scalar_type_id,
components[j++] = vkd3d_spirv_build_op_ext_inst(builder, scalar_type_id,
instr_set_id, GLSLstd450PackHalf2x16, &src_id, 1);
vkd3d_dxbc_compiler_emit_store_reg(compiler, &dst->reg, write_mask, result_id);
}
vkd3d_dxbc_compiler_emit_store_dst_components(compiler,
dst, vkd3d_component_type_from_data_type(dst->reg.data_type), components);
}
static void vkd3d_dxbc_compiler_emit_comparison_instruction(struct vkd3d_dxbc_compiler *compiler,

View File

@ -6608,6 +6608,29 @@ static void test_shader_instructions(void)
0x06000082, 0x001020f2, 0x00000000, 0x00208e46, 0x00000000, 0x00000000, 0x0100003e,
};
static const D3D12_SHADER_BYTECODE ps_f32tof16 = {ps_f32tof16_code, sizeof(ps_f32tof16_code)};
static const DWORD ps_f32tof16_2_code[] =
{
#if 0
ps_5_0
dcl_globalFlags refactoringAllowed
dcl_constantbuffer cb0[1], immediateIndexed
dcl_output o0.xyzw
dcl_temps 1
mov r0.xyzw, cb0[0].xyzw
f32tof16 r0.xyzw, r0.wzyx
mov o0.xyzw, r0.xyzw
ret
#endif
0x43425844, 0x607c82d2, 0x940cc7c2, 0xe9de23c6, 0x696beb90, 0x00000001, 0x000000ec, 0x00000003,
0x0000002c, 0x0000003c, 0x00000070, 0x4e475349, 0x00000008, 0x00000000, 0x00000008, 0x4e47534f,
0x0000002c, 0x00000001, 0x00000008, 0x00000020, 0x00000000, 0x00000000, 0x00000001, 0x00000000,
0x0000000f, 0x545f5653, 0x65677261, 0xabab0074, 0x58454853, 0x00000074, 0x00000050, 0x0000001d,
0x0100086a, 0x04000059, 0x00208e46, 0x00000000, 0x00000001, 0x03000065, 0x001020f2, 0x00000000,
0x02000068, 0x00000001, 0x06000036, 0x001000f2, 0x00000000, 0x00208e46, 0x00000000, 0x00000000,
0x05000082, 0x001000f2, 0x00000000, 0x001001b6, 0x00000000, 0x05000036, 0x001020f2, 0x00000000,
0x00100e46, 0x00000000, 0x0100003e,
};
static const D3D12_SHADER_BYTECODE ps_f32tof16_2 = {ps_f32tof16_2_code, sizeof(ps_f32tof16_2_code)};
static const DWORD ps_imad_code[] =
{
#if 0
@ -7355,6 +7378,8 @@ static void test_shader_instructions(void)
{&ps_f32tof16, {.f = {{0.0f, 1.0f, -1.0f, 666.0f}}}, {{0, 0x3c00, 0xbc00, 0x6134}}},
{&ps_f32tof16_2, {.f = {{0.0f, 1.0f, -1.0f, 666.0f}}}, {{0x6134, 0xbc00, 0x3c00, 0}}},
{&ps_imad, {{{0, 0, 0, 0}, {0, 0, 0, 0}, {0, 0, 0, 0}}}, {{ 0, 0, 0, 0}}},
{&ps_imad, {{{0, 0, 0, 0}, {0, 0, 0, 0}, {1, 2, 0, 0}}}, {{ 1, 2, 0, 0}}},
{&ps_imad, {{{2, 3, 4, 5}, {5, 5, 5, 5}, {0, 0, 0, 0}}}, {{10, 15, 20, 25}}},