wine/vkd3d
0
0
Fork 0
mirror of https://gitlab.winehq.org/wine/vkd3d.git synced 2024-11-21 16:46:41 -08:00
Code Issues Packages Projects Releases Wiki Activity

vkd3d-shader/hlsl: Store SM4 RCP in the vsir program.

Browse Source
This commit is contained in:
Francisco Casas 2024-11-05 19:13:25 -03:00 committed by Henri Verbeet
parent 3c4889add3
commit 2e3f4a01bf
Notes: Henri Verbeet 2024-11-06 23:02:23 +01:00 
Approved-by: Elizabeth Figura (@zfigura)
Approved-by: Henri Verbeet (@hverbeet)
Merge-Request: https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/1251
2 changed files with 45 additions and 39 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

43
libs/vkd3d-shader/hlsl_codegen.c
View File

@ -8018,6 +8018,33 @@ static void sm4_generate_vsir_expr_with_two_destinations(struct hlsl_ctx *ctx, s
vsir_src_from_hlsl_node(&ins->src[i], ctx, expr->operands[i].node, dst_param->write_mask); vsir_src_from_hlsl_node(&ins->src[i], ctx, expr->operands[i].node, dst_param->write_mask);
} }
static void sm4_generate_vsir_rcp_using_div(struct hlsl_ctx *ctx,
struct vsir_program *program, const struct hlsl_ir_expr *expr)
{
struct hlsl_ir_node *operand = expr->operands[0].node;
const struct hlsl_ir_node *instr = &expr->node;
struct vkd3d_shader_dst_param *dst_param;
struct hlsl_constant_value value = {0};
struct vkd3d_shader_instruction *ins;
VKD3D_ASSERT(type_is_float(expr->node.data_type));
if (!(ins = generate_vsir_add_program_instruction(ctx, program, &instr->loc, VKD3DSIH_DIV, 1, 2)))
return;
dst_param = &ins->dst[0];
vsir_dst_from_hlsl_node(dst_param, ctx, instr);
value.u[0].f = 1.0f;
value.u[1].f = 1.0f;
value.u[2].f = 1.0f;
value.u[3].f = 1.0f;
vsir_src_from_hlsl_constant_value(&ins->src[0], ctx, &value,
VKD3D_DATA_FLOAT, instr->data_type->dimx, dst_param->write_mask);
vsir_src_from_hlsl_node(&ins->src[1], ctx, operand, dst_param->write_mask);
}
static bool sm4_generate_vsir_instr_expr(struct hlsl_ctx *ctx, static bool sm4_generate_vsir_instr_expr(struct hlsl_ctx *ctx,
struct vsir_program *program, struct hlsl_ir_expr *expr, const char *dst_type_name) struct vsir_program *program, struct hlsl_ir_expr *expr, const char *dst_type_name)
{ {
@ -8141,6 +8168,22 @@ static bool sm4_generate_vsir_instr_expr(struct hlsl_ctx *ctx,
return false; return false;
} }
case HLSL_OP1_RCP:
switch (dst_type->e.numeric.type)
{
case HLSL_TYPE_FLOAT:
/* SM5 comes with a RCP opcode */
if (hlsl_version_ge(ctx, 5, 0))
generate_vsir_instr_expr_single_instr_op(ctx, program, expr, VKD3DSIH_RCP, 0, 0, true);
else
sm4_generate_vsir_rcp_using_div(ctx, program, expr);
return true;
default:
hlsl_fixme(ctx, &expr->node.loc, "SM4 %s rcp expression.", dst_type_name);
return false;
}
case HLSL_OP1_REINTERPRET: case HLSL_OP1_REINTERPRET:
generate_vsir_instr_expr_single_instr_op(ctx, program, expr, VKD3DSIH_MOV, 0, 0, true); generate_vsir_instr_expr_single_instr_op(ctx, program, expr, VKD3DSIH_MOV, 0, 0, true);
return true; return true;

41
libs/vkd3d-shader/tpf.c
View File

@ -5331,7 +5331,6 @@ static bool type_is_float(const struct hlsl_type *type)
static void write_sm4_expr(const struct tpf_compiler *tpf, const struct hlsl_ir_expr *expr) static void write_sm4_expr(const struct tpf_compiler *tpf, const struct hlsl_ir_expr *expr)
{ {
const struct vkd3d_shader_version *version = &tpf->program->shader_version;
const struct hlsl_ir_node *arg1 = expr->operands[0].node; const struct hlsl_ir_node *arg1 = expr->operands[0].node;
const struct hlsl_ir_node *arg2 = expr->operands[1].node; const struct hlsl_ir_node *arg2 = expr->operands[1].node;
const struct hlsl_type *dst_type = expr->node.data_type; const struct hlsl_type *dst_type = expr->node.data_type;
@ -5344,44 +5343,6 @@ static void write_sm4_expr(const struct tpf_compiler *tpf, const struct hlsl_ir_
switch (expr->op) switch (expr->op)
{ {
case HLSL_OP1_RCP:
switch (dst_type->e.numeric.type)
{
case HLSL_TYPE_FLOAT:
/* SM5 comes with a RCP opcode */
if (vkd3d_shader_ver_ge(version, 5, 0))
{
write_sm4_unary_op(tpf, VKD3D_SM5_OP_RCP, &expr->node, arg1, 0);
}
else
{
/* For SM4, implement as DIV dst, 1.0, src */
struct sm4_instruction instr;
struct hlsl_constant_value one;
VKD3D_ASSERT(type_is_float(dst_type));
memset(&instr, 0, sizeof(instr));
instr.opcode = VKD3D_SM4_OP_DIV;
sm4_dst_from_node(&instr.dsts[0], &expr->node);
instr.dst_count = 1;
for (unsigned int i = 0; i < 4; i++)
one.u[i].f = 1.0f;
sm4_src_from_constant_value(&instr.srcs[0], &one, dst_type->dimx, instr.dsts[0].write_mask);
sm4_src_from_node(tpf, &instr.srcs[1], arg1, instr.dsts[0].write_mask);
instr.src_count = 2;
write_sm4_instruction(tpf, &instr);
}
break;
default:
hlsl_fixme(tpf->ctx, &expr->node.loc, "SM4 %s rcp expression.", dst_type_string->buffer);
}
break;
case HLSL_OP1_SAT: case HLSL_OP1_SAT:
VKD3D_ASSERT(type_is_float(dst_type)); VKD3D_ASSERT(type_is_float(dst_type));
write_sm4_unary_op(tpf, VKD3D_SM4_OP_MOV write_sm4_unary_op(tpf, VKD3D_SM4_OP_MOV
@ -5920,6 +5881,7 @@ static void tpf_handle_instruction(struct tpf_compiler *tpf, const struct vkd3d_
case VKD3DSIH_ADD: case VKD3DSIH_ADD:
case VKD3DSIH_AND: case VKD3DSIH_AND:
case VKD3DSIH_DIV:
case VKD3DSIH_DSX: case VKD3DSIH_DSX:
case VKD3DSIH_DSX_COARSE: case VKD3DSIH_DSX_COARSE:
case VKD3DSIH_DSX_FINE: case VKD3DSIH_DSX_FINE:
@ -5956,6 +5918,7 @@ static void tpf_handle_instruction(struct tpf_compiler *tpf, const struct vkd3d_
case VKD3DSIH_NEU: case VKD3DSIH_NEU:
case VKD3DSIH_NOT: case VKD3DSIH_NOT:
case VKD3DSIH_OR: case VKD3DSIH_OR:
case VKD3DSIH_RCP:
case VKD3DSIH_ROUND_NE: case VKD3DSIH_ROUND_NE:
case VKD3DSIH_ROUND_NI: case VKD3DSIH_ROUND_NI:
case VKD3DSIH_ROUND_PI: case VKD3DSIH_ROUND_PI: