vkd3d-shader/dxil: Implement the DXIL CAST instruction.

This commit is contained in:
Conor McCarthy 2023-11-07 13:11:51 +10:00 committed by Alexandre Julliard
parent bd77cbb33f
commit 90d178bf12
Notes: Alexandre Julliard 2023-11-10 00:09:32 +01:00
Approved-by: Giovanni Mascellani (@giomasce)
Approved-by: Henri Verbeet (@hverbeet)
Approved-by: Alexandre Julliard (@julliard)
Merge-Request: https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/436
9 changed files with 197 additions and 24 deletions

View File

@ -289,6 +289,23 @@ enum dx_intrinsic_opcode
DX_CBUFFER_LOAD_LEGACY = 59,
};
enum dxil_cast_code
{
CAST_TRUNC = 0,
CAST_ZEXT = 1,
CAST_SEXT = 2,
CAST_FPTOUI = 3,
CAST_FPTOSI = 4,
CAST_UITOFP = 5,
CAST_SITOFP = 6,
CAST_FPTRUNC = 7,
CAST_FPEXT = 8,
CAST_PTRTOINT = 9,
CAST_INTTOPTR = 10,
CAST_BITCAST = 11,
CAST_ADDRSPACECAST = 12,
};
struct sm6_pointer_info
{
const struct sm6_type *type;
@ -3248,6 +3265,159 @@ static void sm6_parser_emit_call(struct sm6_parser *sm6, const struct dxil_recor
fn_value->u.function.name, &operands[1], operand_count - 1, ins, dst);
}
static enum vkd3d_shader_opcode sm6_map_cast_op(uint64_t code, const struct sm6_type *from,
const struct sm6_type *to, struct sm6_parser *sm6)
{
enum vkd3d_shader_opcode op = VKD3DSIH_INVALID;
bool from_int, to_int, from_fp, to_fp;
bool is_valid = false;
from_int = sm6_type_is_integer(from);
to_int = sm6_type_is_integer(to);
from_fp = sm6_type_is_floating_point(from);
to_fp = sm6_type_is_floating_point(to);
/* NOTE: DXIL currently doesn't use vectors here. */
if ((!from_int && !from_fp) || (!to_int && !to_fp))
{
FIXME("Unhandled cast of type class %u to type class %u.\n", from->class, to->class);
vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND,
"Cast of type class %u to type class %u is not implemented.", from->class, to->class);
return VKD3DSIH_INVALID;
}
if (to->u.width == 8 || from->u.width == 8)
{
FIXME("Unhandled 8-bit value.\n");
vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND,
"Cast to/from an 8-bit type is not implemented.");
return VKD3DSIH_INVALID;
}
/* DXC emits minimum precision types as 16-bit. These must be emitted
* as 32-bit in VSIR, so all width extensions to 32 bits are no-ops. */
switch (code)
{
case CAST_TRUNC:
/* nop or min precision. TODO: native 16-bit */
if (to->u.width == from->u.width || (to->u.width == 16 && from->u.width == 32))
op = VKD3DSIH_NOP;
else
op = VKD3DSIH_UTOU;
is_valid = from_int && to_int && to->u.width <= from->u.width;
break;
case CAST_ZEXT:
case CAST_SEXT:
/* nop or min precision. TODO: native 16-bit */
if (to->u.width == from->u.width || (to->u.width == 32 && from->u.width == 16))
{
op = VKD3DSIH_NOP;
is_valid = from_int && to_int;
}
else if (to->u.width > from->u.width)
{
op = (code == CAST_ZEXT) ? VKD3DSIH_UTOU : VKD3DSIH_ITOI;
assert(from->u.width == 1 || to->u.width == 64);
is_valid = from_int && to_int;
}
break;
case CAST_FPTOUI:
op = VKD3DSIH_FTOU;
is_valid = from_fp && to_int && to->u.width > 1;
break;
case CAST_FPTOSI:
op = VKD3DSIH_FTOI;
is_valid = from_fp && to_int && to->u.width > 1;
break;
case CAST_UITOFP:
op = VKD3DSIH_UTOF;
is_valid = from_int && to_fp;
break;
case CAST_SITOFP:
op = VKD3DSIH_ITOF;
is_valid = from_int && to_fp;
break;
case CAST_FPTRUNC:
/* TODO: native 16-bit */
op = (from->u.width == 64) ? VKD3DSIH_DTOF : VKD3DSIH_NOP;
is_valid = from_fp && to_fp;
break;
case CAST_FPEXT:
/* TODO: native 16-bit */
op = (to->u.width == 64) ? VKD3DSIH_FTOD : VKD3DSIH_NOP;
is_valid = from_fp && to_fp;
break;
case CAST_BITCAST:
op = VKD3DSIH_MOV;
is_valid = to->u.width == from->u.width;
break;
default:
FIXME("Unhandled cast op %"PRIu64".\n", code);
vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND,
"Cast operation %"PRIu64" is unhandled.\n", code);
return VKD3DSIH_INVALID;
}
if (!is_valid)
{
FIXME("Invalid types %u and/or %u for op %"PRIu64".\n", from->class, to->class, code);
vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND,
"Cast operation %"PRIu64" from type class %u, width %u to type class %u, width %u is invalid.\n",
code, from->class, from->u.width, to->class, to->u.width);
return VKD3DSIH_INVALID;
}
return op;
}
static void sm6_parser_emit_cast(struct sm6_parser *sm6, const struct dxil_record *record,
struct vkd3d_shader_instruction *ins, struct sm6_value *dst)
{
struct vkd3d_shader_src_param *src_param;
enum vkd3d_shader_opcode handler_idx;
const struct sm6_value *value;
const struct sm6_type *type;
unsigned int i = 0;
if (!(value = sm6_parser_get_value_by_ref(sm6, record, NULL, &i)))
return;
if (!dxil_record_validate_operand_count(record, i + 2, i + 2, sm6))
return;
if (!(type = sm6_parser_get_type(sm6, record->operands[i++])))
return;
dst->type = type;
if (sm6_type_is_pointer(type))
{
*dst = *value;
dst->type = type;
ins->handler_idx = VKD3DSIH_NOP;
return;
}
if ((handler_idx = sm6_map_cast_op(record->operands[i], value->type, type, sm6)) == VKD3DSIH_INVALID)
return;
vsir_instruction_init(ins, &sm6->p.location, handler_idx);
if (handler_idx == VKD3DSIH_NOP)
{
dst->u.reg = value->u.reg;
return;
}
src_param = instruction_src_params_alloc(ins, 1, sm6);
src_param_init_from_value(src_param, value);
instruction_dst_param_init_ssa_scalar(ins, sm6);
/* bitcast */
if (handler_idx == VKD3DSIH_MOV)
src_param->reg.data_type = dst->u.reg.data_type;
}
static void sm6_parser_emit_extractval(struct sm6_parser *sm6, const struct dxil_record *record,
struct vkd3d_shader_instruction *ins, struct sm6_value *dst)
{
@ -3464,6 +3634,9 @@ static enum vkd3d_result sm6_parser_function_init(struct sm6_parser *sm6, const
case FUNC_CODE_INST_CALL:
sm6_parser_emit_call(sm6, record, code_block, ins, dst);
break;
case FUNC_CODE_INST_CAST:
sm6_parser_emit_cast(sm6, record, ins, dst);
break;
case FUNC_CODE_INST_EXTRACTVAL:
sm6_parser_emit_extractval(sm6, record, ins, dst);
break;

View File

@ -10,7 +10,7 @@ float4 main() : SV_TARGET
[test]
uniform 0 float4 5.0 16.0 0.0 0.0
todo(sm>=6) draw quad
draw quad
probe all rgba (21.0, -11.0, 80.0, 0.0)
[pixel shader]
@ -25,7 +25,7 @@ float4 main() : SV_TARGET
[test]
uniform 0 float4 5.0 16.0 0.0 0.0
todo(sm>=6) draw quad
draw quad
probe all rgba (5.0, 5.0, -5.0, 3.0)
[pixel shader]
@ -40,7 +40,7 @@ float4 main() : SV_TARGET
[test]
uniform 0 float4 42.0 5.0 0.0 0.0
todo(sm>=6) draw quad
draw quad
probe all rgba (8.0, -8.0, -8.0, 8.0)
[pixel shader]
@ -55,7 +55,7 @@ float4 main() : SV_TARGET
[test]
uniform 0 float4 42.0 5.0 0.0 0.0
todo(sm>=6) draw quad
draw quad
probe all rgba (2.0, -2.0, 2.0, -2.0)
[pixel shader]
@ -70,7 +70,7 @@ float4 main() : SV_TARGET
[test]
uniform 0 float4 45.0 5.0 0.0 0.0
todo(sm>=6) draw quad
draw quad
probe all rgba (9.0, -9.0, -9.0, 9.0)
[pixel shader]
@ -85,7 +85,7 @@ float4 main() : SV_TARGET
[test]
uniform 0 float4 45.0 5.0 0.0 0.0
todo(sm>=6) draw quad
draw quad
probe all rgba (0.0, 0.0, 0.0, 0.0)
[pixel shader]
@ -117,5 +117,5 @@ float4 main() : sv_target
[test]
uniform 0 float4 45.0 5.0 50.0 10.0
uniform 4 float4 3.0 8.0 2.0 5.0
todo(sm>=6) draw quad
draw quad
probe all rgba (9.0, 5.0, 1.0, 3.0)

View File

@ -20,7 +20,7 @@ float4 main() : sv_target
[test]
uniform 0 float4 123.0 -2.0 456 0.01
todo(sm>=6) draw quad
draw quad
probe (320,240) rgba (123.0, -2.0, 456.0, 0.01)
[pixel shader]

View File

@ -20,7 +20,7 @@ float4 main() : sv_target
[test]
uniform 0 uint4 123 0xc0000000 456 0x7fd69345
todo(sm>=6) draw quad
draw quad
probe (320,240) rgba (123.0, 3221225472.0, 456.0, 2144768896.0)
@ -37,7 +37,7 @@ float4 main() : sv_target
uniform 0 uint4 11 12 0 0
uniform 4 uint4 13 14 0 0
uniform 8 uint4 20 21 22 23
todo(sm>=6) draw quad
draw quad
probe (320,240) rgba (13.0, 21.0, 0.0, 0.0)

View File

@ -11,7 +11,7 @@ float4 main() : sv_target
[test]
uniform 0 float4 -0.5 6.5 7.5 3.4
todo(sm>=6) draw quad
draw quad
probe all rgba (1912.0, 1657.0, -127.0, 867.0) 1
[pixel shader]
@ -24,5 +24,5 @@ float4 main() : sv_target
[test]
uniform 0 float4 -0.5 6.5 7.5 3.4
todo(sm>=6) draw quad
draw quad
probe all rgba (-127.0, -127.0, -127.0, -127.0) 1

View File

@ -18,7 +18,7 @@ float4 main() : sv_target
[test]
uniform 0 float4 -1.9 -1.0 2.9 4.0
todo draw quad
todo(sm<6) draw quad
probe all rgba (-1.0, -1.0, 2.0, 4.0)
% As above, but cast "x" to float4 first.
@ -89,5 +89,5 @@ float4 main() : sv_target
[test]
uniform 0 int4 -2 0 1 -3000000
todo draw quad
todo(sm<6) draw quad
probe all rgba (-1.0, 0.0, 1.0, -3000000.0)

View File

@ -32,7 +32,7 @@ float4 main() : SV_TARGET
uniform 0 int4 2 3 4 0
uniform 4 int4 0 -10 10 1000000
uniform 8 int4 0 1 -1 1000000
todo(sm>=6) draw quad
draw quad
probe all rgba (2.0, -10.0, -2.0, 1e12)

View File

@ -12,17 +12,17 @@ uniform 0 float4 1.0 2.0 3.0 4.0
uniform 4 float4 5.0 6.0 7.0 8.0
uniform 8 float4 9.0 10.0 11.0 12.0
uniform 12 float4 0 0 0 0
todo(sm>=6) draw quad
todo(sm>=6) probe all rgba (1.0, 2.0, 3.0, 4.0)
draw quad
probe all rgba (1.0, 2.0, 3.0, 4.0)
uniform 12 float4 1 0 0 0
todo(sm>=6) draw quad
todo(sm>=6) probe all rgba (5.0, 6.0, 7.0, 8.0)
draw quad
probe all rgba (5.0, 6.0, 7.0, 8.0)
uniform 12 float4 0 1 0 0
todo(sm>=6) draw quad
todo(sm>=6) probe all rgba (5.0, 6.0, 7.0, 8.0)
draw quad
probe all rgba (5.0, 6.0, 7.0, 8.0)
uniform 12 float4 1 1 0 0
todo(sm>=6) draw quad
todo(sm>=6) probe all rgba (9.0, 10.0, 11.0, 12.0)
draw quad
probe all rgba (9.0, 10.0, 11.0, 12.0)
[pixel shader]

View File

@ -52,5 +52,5 @@ void main(float4 position : SV_Position, float2 t0 : TEXCOORD0,
}
[test]
todo(sm>=6) draw triangle strip 4
draw triangle strip 4
probe all rgba (10.0, 8.0, 7.0, 3.0)