int_asm.h File Reference

Detailed Description

Integer arithmetic with a guaranteed instruction shape.

Public entry point for arithmetic where which instructions get emitted matters as much as the value computed – currently the fixed-point MP3 decoder's inner loops, where a recognised idiom against an unrecognised one measured 2 instructions against 40 on riscv32.

function	meaning	riscv32	armv7-m
mulshift32(x,y)	high half of a signed 32x32 product	mulh	smull
mul_shift_round32<S>(x,y)	(x*y + 2^(S-1)) >> S, truncating narrow	10 insns	–
wrap_add32(x,y)	32-bit add, defined wraparound	add	add
wrap_sub32(x,y)	32-bit subtract, defined wraparound	sub	sub

mulshift32 truncates rather than rounding, and discards nothing above 2^-32 of the exact product. It is emphatically not operand truncation; see platforms/int_asm.h for the measurement that separates them.

Pinned by ci/tests/test_int_asm_codegen.py, which fails if any of these stops compiling to a single arithmetic instruction on a cross-target.

Definition in file int_asm.h.

#include "platforms/int_asm.h"

Include dependency graph for int_asm.h:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Namespaces
namespace	fl
	Base definition for an LED controller.
namespace	fl::math

Functions
template<int Shift>
FASTLED_FORCE_INLINE i32	fl::math::mul_shift_round32 (i32 x, i32 y) FL_NO_EXCEPT
FASTLED_FORCE_INLINE i32	fl::math::mulshift32 (i32 x, i32 y) FL_NO_EXCEPT
FASTLED_FORCE_INLINE i32	fl::math::wrap_add32 (i32 x, i32 y) FL_NO_EXCEPT
FASTLED_FORCE_INLINE i32	fl::math::wrap_sub32 (i32 x, i32 y) FL_NO_EXCEPT