[Mesa-dev] [PATCH 02/50] glsl: Add "built-in" functions to do neg(fp64) (v2)

[Roland Scheidegger]

Am 13.03.2018 um 05:24 schrieb Dave Airlie:
> From: Elie Tournier <tournier.elie at gmail.com>
> 
> v2: use mix.
> 
> Signed-off-by: Elie Tournier <elie.tournier at collabora.com>
> ---
>  src/compiler/glsl/builtin_float64.h     | 51 +++++++++++++++++++++++++++++++++
>  src/compiler/glsl/builtin_functions.cpp |  4 +++
>  src/compiler/glsl/builtin_functions.h   |  3 ++
>  src/compiler/glsl/float64.glsl          | 24 ++++++++++++++++
>  src/compiler/glsl/glcpp/glcpp-parse.y   |  1 +
>  5 files changed, 83 insertions(+)
> 
> diff --git a/src/compiler/glsl/builtin_float64.h b/src/compiler/glsl/builtin_float64.h
> index 7b57231..2898fc9 100644
> --- a/src/compiler/glsl/builtin_float64.h
> +++ b/src/compiler/glsl/builtin_float64.h
> @@ -17,3 +17,54 @@ fabs64(void *mem_ctx, builtin_available_predicate avail)
>     sig->replace_parameters(&sig_parameters);
>     return sig;
>  }
> +ir_function_signature *
> +is_nan(void *mem_ctx, builtin_available_predicate avail)
> +{
> +   ir_function_signature *const sig =
> +      new(mem_ctx) ir_function_signature(glsl_type::bool_type, avail);
> +   ir_factory body(&sig->body, mem_ctx);
> +   sig->is_defined = true;
> +
> +   exec_list sig_parameters;
> +
> +   ir_variable *const r000C = new(mem_ctx) ir_variable(glsl_type::uvec2_type, "a", ir_var_function_in);
> +   sig_parameters.push_tail(r000C);
> +   ir_expression *const r000D = lshift(swizzle_y(r000C), body.constant(int(1)));
> +   ir_expression *const r000E = gequal(r000D, body.constant(4292870144u));
> +   ir_expression *const r000F = nequal(swizzle_x(r000C), body.constant(0u));
> +   ir_expression *const r0010 = bit_and(swizzle_y(r000C), body.constant(1048575u));
> +   ir_expression *const r0011 = nequal(r0010, body.constant(0u));
> +   ir_expression *const r0012 = logic_or(r000F, r0011);
> +   ir_expression *const r0013 = logic_and(r000E, r0012);
> +   body.emit(ret(r0013));
> +
> +   sig->replace_parameters(&sig_parameters);
> +   return sig;
> +}
> +ir_function_signature *
> +fneg64(void *mem_ctx, builtin_available_predicate avail)
> +{
> +   ir_function_signature *const sig =
> +      new(mem_ctx) ir_function_signature(glsl_type::uvec2_type, avail);
> +   ir_factory body(&sig->body, mem_ctx);
> +   sig->is_defined = true;
> +
> +   exec_list sig_parameters;
> +
> +   ir_variable *const r0014 = new(mem_ctx) ir_variable(glsl_type::uvec2_type, "a", ir_var_function_in);
> +   sig_parameters.push_tail(r0014);
> +   ir_expression *const r0015 = lshift(swizzle_y(r0014), body.constant(int(1)));
> +   ir_expression *const r0016 = gequal(r0015, body.constant(4292870144u));
> +   ir_expression *const r0017 = nequal(swizzle_x(r0014), body.constant(0u));
> +   ir_expression *const r0018 = bit_and(swizzle_y(r0014), body.constant(1048575u));
> +   ir_expression *const r0019 = nequal(r0018, body.constant(0u));
> +   ir_expression *const r001A = logic_or(r0017, r0019);
> +   ir_expression *const r001B = logic_and(r0016, r001A);
> +   ir_expression *const r001C = bit_xor(swizzle_y(r0014), body.constant(2147483648u));
> +   body.emit(assign(r0014, expr(ir_triop_csel, r001B, swizzle_y(r0014), r001C), 0x02));
> +
> +   body.emit(ret(r0014));
> +
> +   sig->replace_parameters(&sig_parameters);
> +   return sig;
> +}
> diff --git a/src/compiler/glsl/builtin_functions.cpp b/src/compiler/glsl/builtin_functions.cpp
> index 133a896..9d88a31 100644
> --- a/src/compiler/glsl/builtin_functions.cpp
> +++ b/src/compiler/glsl/builtin_functions.cpp
> @@ -3346,6 +3346,10 @@ builtin_builder::create_builtins()
>                  generate_ir::fabs64(mem_ctx, integer_functions_supported),
>                  NULL);
>  
> +   add_function("__builtin_fneg64",
> +                generate_ir::fneg64(mem_ctx, integer_functions_supported),
> +                NULL);
> +
>  #undef F
>  #undef FI
>  #undef FIUD_VEC
> diff --git a/src/compiler/glsl/builtin_functions.h b/src/compiler/glsl/builtin_functions.h
> index deaf640..adec424 100644
> --- a/src/compiler/glsl/builtin_functions.h
> +++ b/src/compiler/glsl/builtin_functions.h
> @@ -70,6 +70,9 @@ udivmod64(void *mem_ctx, builtin_available_predicate avail);
>  ir_function_signature *
>  fabs64(void *mem_ctx, builtin_available_predicate avail);
>  
> +ir_function_signature *
> +fneg64(void *mem_ctx, builtin_available_predicate avail);
> +
>  }
>  
>  #endif /* BULITIN_FUNCTIONS_H */
> diff --git a/src/compiler/glsl/float64.glsl b/src/compiler/glsl/float64.glsl
> index d798d7e..fedf8b7 100644
> --- a/src/compiler/glsl/float64.glsl
> +++ b/src/compiler/glsl/float64.glsl
> @@ -6,6 +6,7 @@
>  
>  #version 130
>  #extension GL_ARB_shader_bit_encoding : enable
> +#extension GL_EXT_shader_integer_mix : enable
>  
>  /* Software IEEE floating-point rounding mode.
>   * GLSL spec section "4.7.1 Range and Precision":
> @@ -27,3 +28,26 @@ fabs64(uvec2 a)
>     a.y &= 0x7FFFFFFFu;
>     return a;
>  }
> +
> +/* Returns 1 if the double-precision floating-point value `a' is a NaN;
> + * otherwise returns 0.
> + */
> +bool
> +is_nan(uvec2 a)
> +{
> +   return (0xFFE00000u <= (a.y<<1)) &&
> +      ((a.x != 0u) || ((a.y & 0x000FFFFFu) != 0u));
> +}
> +
> +/* Negate value of a Float64 :
> + * Toggle the sign bit
> + */
> +uvec2
> +fneg64(uvec2 a)
> +{
> +   uint t = a.y;
> +
> +   t ^= (1u << 31);
> +   a.y = mix(t, a.y, is_nan(a));

Why do you care about NaNs here?
If anything I would be more concerned about +0 getting converted into -0
and vice versa (not that I think it's really a problem).
But I don't see why the sign of the NaN is a problem - it's still a NaN.

Roland



> +   return a;
> +}
> diff --git a/src/compiler/glsl/glcpp/glcpp-parse.y b/src/compiler/glsl/glcpp/glcpp-parse.y
> index 4e7affa..b9506d8 100644
> --- a/src/compiler/glsl/glcpp/glcpp-parse.y
> +++ b/src/compiler/glsl/glcpp/glcpp-parse.y
> @@ -2369,6 +2369,7 @@ _glcpp_parser_handle_version_declaration(glcpp_parser_t *parser, intmax_t versio
>           add_builtin_define(parser, "__have_builtin_builtin_idiv64", 1);
>           add_builtin_define(parser, "__have_builtin_builtin_imod64", 1);
>           add_builtin_define(parser, "__have_builtin_builtin_fabs64", 1);
> +         add_builtin_define(parser, "__have_builtin_builtin_fneg64", 1);
>        }
>     }
>  
>