Cleanup float optimizations.
original commit: 1f631a219f31b34192f9eaa184303df357ce58bd
This commit is contained in:
Eric Dobson
parent
2d6cb5a4a4
commit
b9b8c9d67a
|
|
@ -1,6 +1,7 @@
|
|||
#lang racket/base
|
||||
|
||||
(require syntax/parse unstable/sequence racket/dict racket/flonum
|
||||
(require syntax/parse unstable/sequence racket/dict racket/flonum racket/promise
|
||||
syntax/parse/experimental/specialize
|
||||
(for-template racket/base racket/flonum racket/unsafe/ops racket/math)
|
||||
"../utils/utils.rkt"
|
||||
(utils tc-utils)
|
||||
|
|
@ -27,6 +28,15 @@
|
|||
(mk-float-tbl (list #'abs #'sin #'cos #'tan #'asin #'acos #'atan #'log #'exp
|
||||
#'sqrt #'round #'floor #'ceiling #'truncate)))
|
||||
|
||||
(define-literal-syntax-class -)
|
||||
(define-literal-syntax-class /)
|
||||
(define-literal-syntax-class sqr)
|
||||
(define-literal-syntax-class zero?)
|
||||
(define-literal-syntax-class add1)
|
||||
(define-literal-syntax-class sub1)
|
||||
(define-literal-syntax-class ->float^ (exact->inexact real->double-flonum))
|
||||
(define-literal-syntax-class ->single-float^ (exact->inexact real->single-flonum))
|
||||
|
||||
(define-syntax-class (float-op tbl)
|
||||
#:commit
|
||||
(pattern i:id
|
||||
|
|
@ -34,27 +44,13 @@
|
|||
#:with unsafe (begin (add-disappeared-use #'i)
|
||||
(dict-ref tbl #'i))))
|
||||
|
||||
(define-syntax-class float-expr
|
||||
#:commit
|
||||
(pattern e:expr
|
||||
#:when (subtypeof? #'e -Flonum)
|
||||
#:with opt ((optimize) #'e)))
|
||||
(define-syntax-class single-float-expr
|
||||
#:commit
|
||||
(pattern e:expr
|
||||
#:when (subtypeof? #'e -SingleFlonum)
|
||||
#:with opt ((optimize) #'e)))
|
||||
(define-syntax-class int-expr
|
||||
#:commit
|
||||
(pattern e:expr
|
||||
#:when (subtypeof? #'e -Integer)
|
||||
#:with opt ((optimize) #'e)))
|
||||
(define-syntax-class real-expr
|
||||
#:commit
|
||||
(pattern e:expr
|
||||
#:when (subtypeof? #'e -Real)
|
||||
#:with opt ((optimize) #'e)))
|
||||
|
||||
(define-syntax-class/specialize float-expr (subtyped-expr -Flonum))
|
||||
(define-syntax-class/specialize single-float-expr (subtyped-expr -SingleFlonum))
|
||||
(define-syntax-class/specialize int-expr (subtyped-expr -Integer))
|
||||
(define-syntax-class/specialize real-expr (subtyped-expr -Real))
|
||||
(define-syntax-class/specialize unary-float-op (float-op unary-float-ops))
|
||||
(define-syntax-class/specialize binary-float-op (float-op binary-float-ops))
|
||||
(define-syntax-class/specialize binary-float-comp (float-op binary-float-comps))
|
||||
|
||||
;; if the result of an operation is of type float, its non float arguments
|
||||
;; can be promoted, and we can use unsafe float operations
|
||||
|
|
@ -62,23 +58,22 @@
|
|||
;; can result in float (as opposed to real) results
|
||||
(define-syntax-class float-arg-expr
|
||||
#:commit
|
||||
#:attributes (opt)
|
||||
;; we can convert literals right away
|
||||
(pattern (quote n)
|
||||
#:when (and (real? (syntax->datum #'n))
|
||||
(exact? (syntax->datum #'n)))
|
||||
#:with opt
|
||||
(datum->syntax #'here (exact->inexact (syntax->datum #'n))))
|
||||
#:when (and (real? (syntax->datum #'n))
|
||||
(exact? (syntax->datum #'n)))
|
||||
#:with opt #`'#,(exact->inexact (syntax->datum #'n)))
|
||||
(pattern e:fixnum-expr
|
||||
#:with opt #'(unsafe-fx->fl e.opt))
|
||||
#:attr opt (delay #'(unsafe-fx->fl e.opt)))
|
||||
(pattern e:int-expr
|
||||
#:with opt #'(->fl e.opt))
|
||||
(pattern e:float-expr
|
||||
#:with opt #'e.opt)
|
||||
#:attr opt (delay #'(->fl e.opt)))
|
||||
(pattern :float-expr)
|
||||
;; reals within float expressions are not always valid to optimize because
|
||||
;; of the exact 0 problem, but since float-opt-expr checks whether the
|
||||
;; surrounding expressing is of type Float and not just Real, this is safe
|
||||
(pattern e:real-expr
|
||||
#:with opt #'(real->double-flonum e)))
|
||||
#:attr opt (delay #'(real->double-flonum e.opt))))
|
||||
|
||||
(define (log-float-real-missed-opt stx irritants)
|
||||
(log-missed-optimization
|
||||
|
|
@ -90,21 +85,28 @@
|
|||
" To fix, change the highlighted expression(s) to have Float type(s)."))
|
||||
stx irritants))
|
||||
|
||||
(define float-opt-msg "Float arithmetic specialization.")
|
||||
(define-syntax-rule (log-fl-opt opt-label)
|
||||
(log-opt opt-label "Float arithmetic specialization."))
|
||||
|
||||
(define (maybe-exact-rational? stx)
|
||||
(and (subtypeof? stx -Real)
|
||||
(not (subtypeof? stx -Flonum))
|
||||
(not (subtypeof? stx -Int))))
|
||||
|
||||
|
||||
(define-syntax-class float-opt-expr
|
||||
#:commit
|
||||
(pattern (#%plain-app (~var op (float-op unary-float-ops)) f:float-arg-expr)
|
||||
#:literal-sets (kernel-literals)
|
||||
(pattern (#%plain-app op:unary-float-op f:float-arg-expr)
|
||||
#:when (let* ([safe-to-opt? (subtypeof? this-syntax -Flonum)]
|
||||
[missed-optimization? (and (not safe-to-opt?)
|
||||
(in-real-layer? this-syntax))])
|
||||
(when missed-optimization?
|
||||
(log-float-real-missed-opt this-syntax (list #'f)))
|
||||
safe-to-opt?)
|
||||
#:with opt
|
||||
(begin (log-optimization "unary float" float-opt-msg this-syntax)
|
||||
#'(op.unsafe f.opt)))
|
||||
(pattern (#%plain-app (~var op (float-op binary-float-ops))
|
||||
#:do [(log-fl-opt "unary float")]
|
||||
#:with opt #'(op.unsafe f.opt))
|
||||
(pattern (#%plain-app op:binary-float-op
|
||||
;; for now, accept anything that can be coerced to float
|
||||
;; finer-grained checking is done below
|
||||
f1:float-arg-expr
|
||||
|
|
@ -183,110 +185,67 @@
|
|||
"This expression has a Float type, but the highlighted subexpression(s) use exact arithmetic. The extra precision of the exact arithmetic will be lost. Using Float types in these subexpression(s) may result in performance gains without significant precision loss."
|
||||
this-syntax extra-precision-subexprs)))
|
||||
safe-to-opt?)
|
||||
#:with opt
|
||||
(begin (log-optimization "binary float" float-opt-msg this-syntax)
|
||||
(n-ary->binary #'op.unsafe #'f1.opt #'f2.opt #'(fs.opt ...))))
|
||||
(pattern (#%plain-app (~var op (float-op binary-float-comps))
|
||||
f1:float-expr
|
||||
f2:float-expr)
|
||||
#:with opt
|
||||
(begin (log-optimization "binary float comp" float-opt-msg this-syntax)
|
||||
#'(op.unsafe f1.opt f2.opt)))
|
||||
(pattern (#%plain-app (~var op (float-op binary-float-comps))
|
||||
#:do [(log-fl-opt "binary float")]
|
||||
#:with opt (n-ary->binary #'op.unsafe #'f1.opt #'f2.opt #'(fs.opt ...)))
|
||||
(pattern (#%plain-app op:binary-float-comp f1:float-expr f2:float-expr)
|
||||
#:do [(log-fl-opt "binary float comp")]
|
||||
#:with opt #'(op.unsafe f1.opt f2.opt))
|
||||
(pattern (#%plain-app op:binary-float-comp
|
||||
f1:float-expr
|
||||
f2:float-expr
|
||||
fs:float-expr ...)
|
||||
#:with opt
|
||||
(begin (log-optimization "multi float comp" float-opt-msg this-syntax)
|
||||
(n-ary-comp->binary #'op.unsafe #'f1.opt #'f2.opt #'(fs.opt ...))))
|
||||
#:do [(log-fl-opt "multi float comp")]
|
||||
#:with opt (n-ary-comp->binary #'op.unsafe #'f1.opt #'f2.opt #'(fs.opt ...)))
|
||||
|
||||
(pattern (#%plain-app (~and op (~literal -)) f:float-expr)
|
||||
#:with opt
|
||||
(begin (log-optimization "unary float" float-opt-msg this-syntax)
|
||||
(add-disappeared-use #'op)
|
||||
#'(unsafe-fl* -1.0 f.opt)))
|
||||
(pattern (#%plain-app (~and op (~literal /)) f:float-expr)
|
||||
#:with opt
|
||||
(begin (log-optimization "unary float" float-opt-msg this-syntax)
|
||||
(add-disappeared-use #'op)
|
||||
#'(unsafe-fl/ 1.0 f.opt)))
|
||||
(pattern (#%plain-app (~and op (~literal sqr)) f:float-expr)
|
||||
#:with opt
|
||||
(begin (log-optimization "unary float" float-opt-msg this-syntax)
|
||||
(add-disappeared-use #'op)
|
||||
#'(let ([tmp f.opt]) (unsafe-fl* tmp tmp))))
|
||||
(pattern (#%plain-app op:-^ f:float-expr)
|
||||
#:do [(log-fl-opt "unary float")]
|
||||
#:with opt #'(unsafe-fl* -1.0 f.opt))
|
||||
(pattern (#%plain-app op:/^ f:float-expr)
|
||||
#:do [(log-fl-opt "unary float")]
|
||||
#:with opt #'(unsafe-fl/ 1.0 f.opt))
|
||||
(pattern (#%plain-app op:sqr^ f:float-expr)
|
||||
#:do [(log-fl-opt "unary float")]
|
||||
#:with opt #'(let ([tmp f.opt]) (unsafe-fl* tmp tmp)))
|
||||
|
||||
;; we can optimize exact->inexact if we know we're giving it an Integer
|
||||
(pattern (#%plain-app (~and op (~or (~literal exact->inexact)
|
||||
(~literal real->double-flonum)))
|
||||
n:int-expr)
|
||||
#:with opt
|
||||
(begin (log-optimization "int to float" float-opt-msg this-syntax)
|
||||
(add-disappeared-use #'op)
|
||||
#'(->fl n.opt)))
|
||||
(pattern (#%plain-app op:->float^ n:int-expr)
|
||||
#:do [(log-fl-opt "int to float")]
|
||||
#:with opt #'(->fl n.opt))
|
||||
;; we can get rid of it altogether if we're giving it a float
|
||||
(pattern (#%plain-app (~and op (~or (~literal exact->inexact)
|
||||
(~literal real->double-flonum)))
|
||||
f:float-expr)
|
||||
#:with opt
|
||||
(begin (log-optimization "float to float" float-opt-msg this-syntax)
|
||||
(add-disappeared-use #'op)
|
||||
#'f.opt))
|
||||
(pattern (#%plain-app op:->float^ f:float-expr)
|
||||
#:do [(log-fl-opt "float to float")]
|
||||
#:with opt #'f.opt)
|
||||
;; same for single-flonums
|
||||
(pattern (#%plain-app (~and op (~or (~literal exact->inexact)
|
||||
(~literal real->single-flonum)))
|
||||
f:single-float-expr)
|
||||
#:with opt
|
||||
(begin (log-optimization "single-float to single-float"
|
||||
float-opt-msg this-syntax)
|
||||
(add-disappeared-use #'op)
|
||||
#'f.opt))
|
||||
(pattern (#%plain-app op:->single-float^ f:single-float-expr)
|
||||
#:do [(log-fl-opt "single-float to single-float")]
|
||||
#:with opt #'f.opt)
|
||||
|
||||
(pattern (#%plain-app (~and op (~literal zero?)) f:float-expr)
|
||||
#:with opt
|
||||
(begin (log-optimization "float zero?" float-opt-msg this-syntax)
|
||||
(add-disappeared-use #'op)
|
||||
#'(unsafe-fl= f.opt 0.0)))
|
||||
(pattern (#%plain-app op:zero?^ f:float-expr)
|
||||
#:do [(log-fl-opt "float zero?")]
|
||||
#:with opt #'(unsafe-fl= f.opt 0.0))
|
||||
|
||||
(pattern (#%plain-app (~and op (~literal add1)) n:float-expr)
|
||||
#:with opt
|
||||
(begin (log-optimization "float add1" float-opt-msg this-syntax)
|
||||
(add-disappeared-use #'op)
|
||||
#'(unsafe-fl+ n.opt 1.0)))
|
||||
(pattern (#%plain-app (~and op (~literal sub1)) n:float-expr)
|
||||
#:with opt
|
||||
(begin (log-optimization "float sub1" float-opt-msg this-syntax)
|
||||
(add-disappeared-use #'op)
|
||||
#'(unsafe-fl- n.opt 1.0)))
|
||||
(pattern (#%plain-app op:add1^ n:float-expr)
|
||||
#:do [(log-fl-opt "float add1")]
|
||||
#:with opt #'(unsafe-fl+ n.opt 1.0))
|
||||
(pattern (#%plain-app op:sub1^ n:float-expr)
|
||||
#:do [(log-fl-opt "float sub1")]
|
||||
#:with opt #'(unsafe-fl- n.opt 1.0))
|
||||
|
||||
;; warn about (potentially) exact real arithmetic, in general
|
||||
;; Note: These patterns don't perform optimization. They only produce logging
|
||||
;; for consumption by Optimization Coach.
|
||||
(pattern (#%plain-app (~var op (float-op binary-float-ops))
|
||||
n ...)
|
||||
#:when (maybe-exact-rational? this-syntax)
|
||||
#:with opt
|
||||
(begin (log-optimization-info "exact real arith"
|
||||
this-syntax)
|
||||
this-syntax))
|
||||
(pattern (#%plain-app (~var op (float-op binary-float-comps))
|
||||
n ...)
|
||||
;; can't look at return type, since it's always bool
|
||||
#:when (andmap maybe-exact-rational? (syntax->list #'(n ...)))
|
||||
#:with opt
|
||||
(begin (log-optimization-info "exact real arith"
|
||||
this-syntax)
|
||||
this-syntax))
|
||||
(pattern (#%plain-app (~var op (float-op unary-float-ops))
|
||||
n ...)
|
||||
#:when (maybe-exact-rational? this-syntax)
|
||||
#:with opt
|
||||
(begin (log-optimization-info "exact real arith"
|
||||
this-syntax)
|
||||
this-syntax))
|
||||
(pattern (#%plain-app op:binary-float-op n:opt-expr ...)
|
||||
#:when (maybe-exact-rational? this-syntax)
|
||||
#:do [(log-opt-info "exact real arith")]
|
||||
#:with opt #'(op n.opt ...))
|
||||
(pattern (#%plain-app op:binary-float-comp n:opt-expr ...)
|
||||
;; can't look at return type, since it's always bool
|
||||
#:when (andmap maybe-exact-rational? (syntax->list #'(n ...)))
|
||||
#:do [(log-opt-info "exact real arith")]
|
||||
#:with opt #'(op n.opt ...))
|
||||
(pattern (#%plain-app op:unary-float-op n:opt-expr ...)
|
||||
#:when (maybe-exact-rational? this-syntax)
|
||||
#:do [(log-opt-info "exact real arith")]
|
||||
#:with opt #'(op n.opt ...))
|
||||
)
|
||||
|
||||
(define (maybe-exact-rational? stx)
|
||||
(and (subtypeof? stx -Real)
|
||||
(not (subtypeof? stx -Flonum))
|
||||
(not (subtypeof? stx -Int))))
|
||||
|
|
|
|||
|
|
@ -1,15 +1,13 @@
|
|||
#;#;
|
||||
#<<END
|
||||
TR missed opt: fixnum.rkt 37:0 (+ (ann z Fixnum) 234) -- out of fixnum range
|
||||
TR missed opt: fixnum.rkt 38:0 (* (ann x Index) (ann y Index)) -- out of fixnum range
|
||||
TR missed opt: fixnum.rkt 41:0 (+ (+ 300 301) (+ 301 302)) -- out of fixnum range
|
||||
TR opt: fixnum.rkt 34:10 (* x y) -- fixnum bounded expr
|
||||
TR opt: fixnum.rkt 41:15 (+ 301 302) -- fixnum bounded expr
|
||||
TR opt: fixnum.rkt 41:15 (+ 301 302) -- fixnum bounded expr
|
||||
TR opt: fixnum.rkt 41:3 (+ 300 301) -- fixnum bounded expr
|
||||
TR opt: fixnum.rkt 41:3 (+ 300 301) -- fixnum bounded expr
|
||||
TR opt: fixnum.rkt 42:17 (+ 301 302) -- fixnum bounded expr
|
||||
TR opt: fixnum.rkt 42:5 (+ 300 301) -- fixnum bounded expr
|
||||
TR missed opt: fixnum.rkt 35:0 (+ (ann z Fixnum) 234) -- out of fixnum range
|
||||
TR missed opt: fixnum.rkt 36:0 (* (ann x Index) (ann y Index)) -- out of fixnum range
|
||||
TR missed opt: fixnum.rkt 39:0 (+ (+ 300 301) (+ 301 302)) -- out of fixnum range
|
||||
TR opt: fixnum.rkt 32:10 (* x y) -- fixnum bounded expr
|
||||
TR opt: fixnum.rkt 39:15 (+ 301 302) -- fixnum bounded expr
|
||||
TR opt: fixnum.rkt 39:3 (+ 300 301) -- fixnum bounded expr
|
||||
TR opt: fixnum.rkt 40:17 (+ 301 302) -- fixnum bounded expr
|
||||
TR opt: fixnum.rkt 40:5 (+ 300 301) -- fixnum bounded expr
|
||||
END
|
||||
#<<END
|
||||
468
|
||||
|
|
|
|||
|
|
@ -1,24 +1,23 @@
|
|||
#;#;
|
||||
#<<END
|
||||
TR info: precision-loss.rkt 39:3 (* 3/4 2/3) -- exact real arith
|
||||
TR info: precision-loss.rkt 43:3 (- 3/4) -- exact real arith
|
||||
TR info: precision-loss.rkt 45:39 (+ 1/4 3/4) -- exact real arith
|
||||
TR info: precision-loss.rkt 45:39 (+ 1/4 3/4) -- exact real arith
|
||||
TR info: precision-loss.rkt 51:0 (* (* (r 3/4) 2/3) (car (list (* 2.0 (* (r 3/4) 2/3)))) 2.0) -- exact real arith
|
||||
TR info: precision-loss.rkt 51:3 (* (r 3/4) 2/3) -- exact real arith
|
||||
TR info: precision-loss.rkt 52:14 (* 2.0 (* (r 3/4) 2/3)) -- exact real arith
|
||||
TR info: precision-loss.rkt 52:21 (* (r 3/4) 2/3) -- exact real arith
|
||||
TR missed opt: precision-loss.rkt 39:0 (+ (* 3/4 2/3) 2.0) -- exact ops inside float expr -- caused by: 39:3 (* 3/4 2/3)
|
||||
TR missed opt: precision-loss.rkt 43:0 (+ (- 3/4) 2.0) -- exact ops inside float expr -- caused by: 43:3 (- 3/4)
|
||||
TR missed opt: precision-loss.rkt 45:0 (+ (vector-ref (quote #(2/3 1/2 3/4)) (assert (+ 1/4 3/4) exact-integer?)) 2.0) -- all args float-arg-expr, result not Float -- caused by: 45:3 (vector-ref (quote #(2/3 1/2 3/4)) (assert (+ 1/4 3/4) exact-integer?))
|
||||
TR missed opt: precision-loss.rkt 51:0 (* (* (r 3/4) 2/3) (car (list (* 2.0 (* (r 3/4) 2/3)))) 2.0) -- all args float-arg-expr, result not Float -- caused by: 51:3 (* (r 3/4) 2/3), 52:3 (car (list (* 2.0 (* (r 3/4) 2/3))))
|
||||
TR missed opt: precision-loss.rkt 51:3 (* (r 3/4) 2/3) -- all args float-arg-expr, result not Float -- caused by: 51:6 (r 3/4), 51:14 2/3
|
||||
TR missed opt: precision-loss.rkt 52:14 (* 2.0 (* (r 3/4) 2/3)) -- all args float-arg-expr, result not Float -- caused by: 52:21 (* (r 3/4) 2/3)
|
||||
TR missed opt: precision-loss.rkt 52:21 (* (r 3/4) 2/3) -- all args float-arg-expr, result not Float -- caused by: 52:24 (r 3/4), 52:32 2/3
|
||||
TR opt: precision-loss.rkt 39:0 (+ (* 3/4 2/3) 2.0) -- binary float
|
||||
TR opt: precision-loss.rkt 41:0 (+ 3/4 2.0) -- binary float
|
||||
TR opt: precision-loss.rkt 43:0 (+ (- 3/4) 2.0) -- binary float
|
||||
TR opt: precision-loss.rkt 52:3 (car (list (* 2.0 (* (r 3/4) 2/3)))) -- pair
|
||||
TR info: precision-loss.rkt 38:3 (* 3/4 2/3) -- exact real arith
|
||||
TR info: precision-loss.rkt 42:3 (- 3/4) -- exact real arith
|
||||
TR info: precision-loss.rkt 44:39 (+ 1/4 3/4) -- exact real arith
|
||||
TR info: precision-loss.rkt 50:0 (* (* (r 3/4) 2/3) (car (list (* 2.0 (* (r 3/4) 2/3)))) 2.0) -- exact real arith
|
||||
TR info: precision-loss.rkt 50:3 (* (r 3/4) 2/3) -- exact real arith
|
||||
TR info: precision-loss.rkt 51:14 (* 2.0 (* (r 3/4) 2/3)) -- exact real arith
|
||||
TR info: precision-loss.rkt 51:21 (* (r 3/4) 2/3) -- exact real arith
|
||||
TR missed opt: precision-loss.rkt 38:0 (+ (* 3/4 2/3) 2.0) -- exact ops inside float expr -- caused by: 38:3 (* 3/4 2/3)
|
||||
TR missed opt: precision-loss.rkt 42:0 (+ (- 3/4) 2.0) -- exact ops inside float expr -- caused by: 42:3 (- 3/4)
|
||||
TR missed opt: precision-loss.rkt 44:0 (+ (vector-ref (quote #(2/3 1/2 3/4)) (assert (+ 1/4 3/4) exact-integer?)) 2.0) -- all args float-arg-expr, result not Float -- caused by: 44:3 (vector-ref (quote #(2/3 1/2 3/4)) (assert (+ 1/4 3/4) exact-integer?))
|
||||
TR missed opt: precision-loss.rkt 50:0 (* (* (r 3/4) 2/3) (car (list (* 2.0 (* (r 3/4) 2/3)))) 2.0) -- all args float-arg-expr, result not Float -- caused by: 50:3 (* (r 3/4) 2/3), 51:3 (car (list (* 2.0 (* (r 3/4) 2/3))))
|
||||
TR missed opt: precision-loss.rkt 50:3 (* (r 3/4) 2/3) -- all args float-arg-expr, result not Float -- caused by: 50:6 (r 3/4), 50:14 2/3
|
||||
TR missed opt: precision-loss.rkt 51:14 (* 2.0 (* (r 3/4) 2/3)) -- all args float-arg-expr, result not Float -- caused by: 51:21 (* (r 3/4) 2/3)
|
||||
TR missed opt: precision-loss.rkt 51:21 (* (r 3/4) 2/3) -- all args float-arg-expr, result not Float -- caused by: 51:24 (r 3/4), 51:32 2/3
|
||||
TR opt: precision-loss.rkt 38:0 (+ (* 3/4 2/3) 2.0) -- binary float
|
||||
TR opt: precision-loss.rkt 40:0 (+ 3/4 2.0) -- binary float
|
||||
TR opt: precision-loss.rkt 42:0 (+ (- 3/4) 2.0) -- binary float
|
||||
TR opt: precision-loss.rkt 51:3 (car (list (* 2.0 (* (r 3/4) 2/3)))) -- pair
|
||||
END
|
||||
#<<END
|
||||
2.5
|
||||
|
|
|
|||
|
|
@ -1,11 +1,10 @@
|
|||
#;#;
|
||||
#<<END
|
||||
TR opt: float-complex-fixnum.rkt 18:0 (+ (modulo 2 1) 1.0+2.0i 3.0+6.0i) -- unboxed binary float complex
|
||||
TR opt: float-complex-fixnum.rkt 18:16 1.0+2.0i -- unboxed literal
|
||||
TR opt: float-complex-fixnum.rkt 18:25 3.0+6.0i -- unboxed literal
|
||||
TR opt: float-complex-fixnum.rkt 18:3 (modulo 2 1) -- binary nonzero fixnum
|
||||
TR opt: float-complex-fixnum.rkt 18:3 (modulo 2 1) -- binary nonzero fixnum
|
||||
TR opt: float-complex-fixnum.rkt 18:3 (modulo 2 1) -- float-arg-expr in complex ops
|
||||
TR opt: float-complex-fixnum.rkt 17:0 (+ (modulo 2 1) 1.0+2.0i 3.0+6.0i) -- unboxed binary float complex
|
||||
TR opt: float-complex-fixnum.rkt 17:16 1.0+2.0i -- unboxed literal
|
||||
TR opt: float-complex-fixnum.rkt 17:25 3.0+6.0i -- unboxed literal
|
||||
TR opt: float-complex-fixnum.rkt 17:3 (modulo 2 1) -- binary nonzero fixnum
|
||||
TR opt: float-complex-fixnum.rkt 17:3 (modulo 2 1) -- float-arg-expr in complex ops
|
||||
END
|
||||
#<<END
|
||||
4.0+8.0i
|
||||
|
|
|
|||
|
|
@ -1,16 +1,12 @@
|
|||
#;#;
|
||||
#<<END
|
||||
TR info: float-complex-sin.rkt 25:13 (sin (* t 6.28)) -- exact real arith
|
||||
TR info: float-complex-sin.rkt 25:13 (sin (* t 6.28)) -- exact real arith
|
||||
TR info: float-complex-sin.rkt 25:18 (* t 6.28) -- exact real arith
|
||||
TR info: float-complex-sin.rkt 25:18 (* t 6.28) -- exact real arith
|
||||
TR missed opt: float-complex-sin.rkt 25:13 (sin (* t 6.28)) -- all args float-arg-expr, result not Float -- caused by: 25:18 (* t 6.28)
|
||||
TR missed opt: float-complex-sin.rkt 25:13 (sin (* t 6.28)) -- all args float-arg-expr, result not Float -- caused by: 25:18 (* t 6.28)
|
||||
TR missed opt: float-complex-sin.rkt 25:18 (* t 6.28) -- all args float-arg-expr, result not Float -- caused by: 25:21 t
|
||||
TR missed opt: float-complex-sin.rkt 25:18 (* t 6.28) -- all args float-arg-expr, result not Float -- caused by: 25:21 t
|
||||
TR opt: float-complex-sin.rkt 25:10 (+ (sin (* t 6.28)) 0.0+0.0i) -- unboxed binary float complex
|
||||
TR opt: float-complex-sin.rkt 25:13 (sin (* t 6.28)) -- float-arg-expr in complex ops
|
||||
TR opt: float-complex-sin.rkt 25:30 0.0+0.0i -- unboxed literal
|
||||
TR info: float-complex-sin.rkt 21:13 (sin (* t 6.28)) -- exact real arith
|
||||
TR info: float-complex-sin.rkt 21:18 (* t 6.28) -- exact real arith
|
||||
TR missed opt: float-complex-sin.rkt 21:13 (sin (* t 6.28)) -- all args float-arg-expr, result not Float -- caused by: 21:18 (* t 6.28)
|
||||
TR missed opt: float-complex-sin.rkt 21:18 (* t 6.28) -- all args float-arg-expr, result not Float -- caused by: 21:21 t
|
||||
TR opt: float-complex-sin.rkt 21:10 (+ (sin (* t 6.28)) 0.0+0.0i) -- unboxed binary float complex
|
||||
TR opt: float-complex-sin.rkt 21:13 (sin (* t 6.28)) -- float-arg-expr in complex ops
|
||||
TR opt: float-complex-sin.rkt 21:30 0.0+0.0i -- unboxed literal
|
||||
END
|
||||
#<<END
|
||||
-0.0031853017931379904+0.0i
|
||||
|
|
|
|||
Loading…
Reference in New Issue
Block a user