vkd3d-shader/hlsl: Implement the frexp() intrinsic.

This commit is contained in:
Petrichor Park
2024-07-08 13:32:59 -05:00
committed by Henri Verbeet
parent ea6104cf5e
commit d964d55b4a
Notes: Henri Verbeet 2025-08-21 16:34:21 +02:00
Approved-by: Elizabeth Figura (@zfigura)
Approved-by: Henri Verbeet (@hverbeet)
Merge-Request: https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/1000
2 changed files with 65 additions and 17 deletions

View File

@@ -3983,6 +3983,53 @@ static bool intrinsic_frac(struct hlsl_ctx *ctx,
return !!add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_FRACT, arg, loc); return !!add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_FRACT, arg, loc);
} }
static bool intrinsic_frexp(struct hlsl_ctx *ctx,
const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
{
struct hlsl_type *type, *uint_dim_type, *int_dim_type, *bool_dim_type;
struct hlsl_ir_function_decl *func;
char *body;
static const char template[] =
"%s frexp(%s x, out %s exp)\n"
"{\n"
/* If x is zero, always return zero for exp and mantissa. */
" %s is_nonzero_mask = x != 0.0;\n"
" %s bits = asuint(x);\n"
/* Subtract 126, not 127, to increase the exponent */
" %s exp_int = asint((bits & 0x7f800000u) >> 23) - 126;\n"
/* Clear the given exponent and replace it with the bit pattern
* for 2^-1 */
" %s mantissa = asfloat((bits & 0x007fffffu) | 0x3f000000);\n"
" exp = is_nonzero_mask * %s(exp_int);\n"
" return is_nonzero_mask * mantissa;\n"
"}\n";
if (!elementwise_intrinsic_float_convert_args(ctx, params, loc))
return false;
type = params->args[0]->data_type;
if (type->e.numeric.type == HLSL_TYPE_DOUBLE)
{
hlsl_fixme(ctx, loc, "frexp() on doubles.");
return false;
}
type = hlsl_get_numeric_type(ctx, type->class, HLSL_TYPE_FLOAT, type->e.numeric.dimx, type->e.numeric.dimy);
uint_dim_type = hlsl_get_numeric_type(ctx, type->class, HLSL_TYPE_UINT, type->e.numeric.dimx, type->e.numeric.dimy);
int_dim_type = hlsl_get_numeric_type(ctx, type->class, HLSL_TYPE_INT, type->e.numeric.dimx, type->e.numeric.dimy);
bool_dim_type = hlsl_get_numeric_type(ctx, type->class, HLSL_TYPE_BOOL, type->e.numeric.dimx, type->e.numeric.dimy);
if (!(body = hlsl_sprintf_alloc(ctx, template, type->name, type->name, type->name,
bool_dim_type->name, uint_dim_type->name, int_dim_type->name, type->name, type->name)))
return false;
func = hlsl_compile_internal_function(ctx, "frexp", body);
vkd3d_free(body);
if (!func)
return false;
return !!add_user_call(ctx, func, params, false, loc);
}
static bool intrinsic_fwidth(struct hlsl_ctx *ctx, static bool intrinsic_fwidth(struct hlsl_ctx *ctx,
const struct parse_initializer *params, const struct vkd3d_shader_location *loc) const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
{ {
@@ -5283,6 +5330,7 @@ intrinsic_functions[] =
{"floor", 1, true, intrinsic_floor}, {"floor", 1, true, intrinsic_floor},
{"fmod", 2, true, intrinsic_fmod}, {"fmod", 2, true, intrinsic_fmod},
{"frac", 1, true, intrinsic_frac}, {"frac", 1, true, intrinsic_frac},
{"frexp", 2, true, intrinsic_frexp},
{"fwidth", 1, true, intrinsic_fwidth}, {"fwidth", 1, true, intrinsic_fwidth},
{"isinf", 1, true, intrinsic_isinf}, {"isinf", 1, true, intrinsic_isinf},
{"ldexp", 2, true, intrinsic_ldexp}, {"ldexp", 2, true, intrinsic_ldexp},

View File

@@ -1,4 +1,4 @@
[pixel shader todo(sm<6)] [pixel shader todo(sm<4)]
uniform float4 f; uniform float4 f;
float4 main() : sv_target float4 main() : sv_target
@@ -10,52 +10,52 @@ float4 main() : sv_target
[test] [test]
uniform 0 float4 3.1415927 0.0 0.0 0.0 uniform 0 float4 3.1415927 0.0 0.0 0.0
todo(sm<6 | msl) draw quad todo(sm<4 | msl & sm>=6) draw quad
probe (0, 0) f32(2.0, 0.785398185, 0.0, 0.0) probe (0, 0) f32(2.0, 0.785398185, 0.0, 0.0)
uniform 0 float4 -3.1415927 0.0 0.0 0.0 uniform 0 float4 -3.1415927 0.0 0.0 0.0
todo(sm<6 | msl) draw quad todo(sm<4 | msl & sm>=6) draw quad
if(sm<4) probe (0, 0) f32(2.0, -0.785398185, 0.0, 0.0) if(sm<4) probe (0, 0) f32(2.0, -0.785398185, 0.0, 0.0)
% Starting with shader model 4, negative inputs give positive mantissa. % Starting with shader model 4, negative inputs give positive mantissa.
if(sm>=4) probe (0, 0) f32(2.0, 0.785398185, 0.0, 0.0) if(sm>=4) probe (0, 0) f32(2.0, 0.785398185, 0.0, 0.0)
uniform 0 float4 7604.123 0.0 0.0 0.0 uniform 0 float4 7604.123 0.0 0.0 0.0
todo(sm<6 | msl) draw quad todo(sm<4 | msl & sm>=6) draw quad
probe (0, 0) f32(13.0, 0.92823765, 0.0, 0.0) probe (0, 0) f32(13.0, 0.92823765, 0.0, 0.0)
uniform 0 float4 0.00001234 0.0 0.0 0.0 uniform 0 float4 0.00001234 0.0 0.0 0.0
todo(sm<6 | msl) draw quad todo(sm<4 | msl & sm>=6) draw quad
probe (0, 0) f32(-16.0, 0.8087142, 0.0, 0.0) probe (0, 0) f32(-16.0, 0.8087142, 0.0, 0.0)
uniform 0 float4 0.0 0.0 0.0 0.0 uniform 0 float4 0.0 0.0 0.0 0.0
todo(sm<6 | msl) draw quad todo(sm<4 | msl & sm>=6) draw quad
probe (0, 0) f32(0.0, 0.0, 0.0, 0.0) probe (0, 0) f32(0.0, 0.0, 0.0, 0.0)
uniform 0 float4 -0.0 0.0 0.0 0.0 uniform 0 float4 -0.0 0.0 0.0 0.0
todo(sm<6 | msl) draw quad todo(sm<4 | msl & sm>=6) draw quad
probe (0, 0) f32(0.0, 0.0, 0.0, 0.0) probe (0, 0) f32(0.0, 0.0, 0.0, 0.0)
uniform 0 float4 INF 0.0 0.0 0.0 uniform 0 float4 INF 0.0 0.0 0.0
todo(sm<6 | msl) draw quad todo(sm<4 | msl & sm>=6) draw quad
if(sm<4) probe (0, 0) f32(-NAN, -NAN, 0, 0) if(sm<4) probe (0, 0) f32(-NAN, -NAN, 0, 0)
if(sm>=4) probe (0, 0) f32(129, 0.5, 0, 0) if(sm>=4) probe (0, 0) f32(129, 0.5, 0, 0)
uniform 0 float4 -INF 0.0 0.0 0.0 uniform 0 float4 -INF 0.0 0.0 0.0
todo(sm<6 | msl) draw quad todo(sm<4 | msl & sm>=6) draw quad
if(sm<4) probe (0, 0) f32(-NAN, -NAN, 0, 0) if(sm<4) probe (0, 0) f32(-NAN, -NAN, 0, 0)
if(sm>=4) probe (0, 0) f32(129, 0.5, 0, 0) if(sm>=4) probe (0, 0) f32(129, 0.5, 0, 0)
uniform 0 float4 NAN 0 0 0 uniform 0 float4 NAN 0 0 0
todo(sm<6 | msl) draw quad todo(sm<4 | msl & sm>=6) draw quad
if(sm<4) probe (0, 0) f32(NAN, NAN, 0, 0) if(sm<4) probe (0, 0) f32(NAN, NAN, 0, 0)
if(sm>=4) probe (0, 0) f32(129, 0.75, 0, 0) if(sm>=4) probe (0, 0) f32(129, 0.75, 0, 0)
% Subnormals. % Subnormals.
uniform 0 uint4 0x0007ffff 0 0 0 uniform 0 uint4 0x0007ffff 0 0 0
todo(sm<6 | msl) draw quad todo(sm<4 | msl & sm>=6) draw quad
probe (0, 0) f32(0, 0, 0, 0) probe (0, 0) f32(0, 0, 0, 0)
[pixel shader todo(sm<6)] [pixel shader todo(sm<4)]
float4 main() : sv_target float4 main() : sv_target
{ {
int arg = 7604; int arg = 7604;
@@ -66,10 +66,10 @@ float4 main() : sv_target
} }
[test] [test]
todo(sm<6) draw quad todo(sm<4) draw quad
probe (0, 0) f32(13.0, 0.0, 0.0, 0.0) probe (0, 0) f32(13.0, 0.0, 0.0, 0.0)
[pixel shader todo(sm<6)] [pixel shader todo(sm<4)]
float4 main() : sv_target float4 main() : sv_target
{ {
bool arg = true; bool arg = true;
@@ -80,11 +80,11 @@ float4 main() : sv_target
} }
[test] [test]
todo(sm<6) draw quad todo(sm<4) draw quad
if(sm<4) probe (0, 0) f32(0.0, 1.0, 0.0, 0.0) if(sm<4) probe (0, 0) f32(0.0, 1.0, 0.0, 0.0)
if(sm>=4) probe (0, 0) f32(1.0, 1.0, 0.0, 0.0) if(sm>=4) probe (0, 0) f32(1.0, 1.0, 0.0, 0.0)
[pixel shader todo(sm<6)] [pixel shader todo(sm<4)]
float4 main() : sv_target float4 main() : sv_target
{ {
half arg = 3.141; half arg = 3.141;
@@ -95,7 +95,7 @@ float4 main() : sv_target
} }
[test] [test]
todo(sm<6) draw quad todo(sm<4) draw quad
probe (0, 0) f32(2.0, 0.785250008, 0.0, 0.0) probe (0, 0) f32(2.0, 0.785250008, 0.0, 0.0)
[require] [require]