(sc-is target signed-reg unsigned-reg descriptor-reg any-reg))
(inst mov target val))
;; Likewise if the value is small enough.
- ((typep val '(signed-byte 31))
+ ((typep val '(signed-byte 32))
(inst mov target val))
;; Otherwise go through the temporary register
(tmp-tn
(etypecase val
((integer 0 0)
(zeroize y))
- ((or (signed-byte 29) (unsigned-byte 29))
- (inst mov y (fixnumize val)))
(integer
- (move-immediate y (fixnumize val)))
+ (inst mov y (fixnumize val)))
(symbol
(load-symbol y val))
(character
;;; Arg is a fixnum or bignum, figure out which and load if necessary.
+#-#.(cl:if (cl:= sb!vm:n-fixnum-tag-bits 1) '(:and) '(:or))
(define-vop (move-to-word/integer)
- (:args (x :scs (descriptor-reg) :target eax))
+ (:args (x :scs (descriptor-reg) :target rax))
(:results (y :scs (signed-reg unsigned-reg)))
(:note "integer to untagged word coercion")
- (:temporary (:sc unsigned-reg :offset eax-offset
- :from (:argument 0) :to (:result 0) :target y) eax)
+ ;; I'm not convinced that increasing the demand for rAX is
+ ;; better than adding 1 byte to some instruction encodings.
+ ;; I'll leave it alone though.
+ (:temporary (:sc unsigned-reg :offset rax-offset
+ :from (:argument 0) :to (:result 0) :target y) rax)
(:generator 4
- (move eax x)
+ (move rax x)
(inst test al-tn fixnum-tag-mask)
(inst jmp :z FIXNUM)
- (loadw y eax bignum-digits-offset other-pointer-lowtag)
+ (loadw y rax bignum-digits-offset other-pointer-lowtag)
(inst jmp DONE)
FIXNUM
- (inst sar eax n-fixnum-tag-bits)
- (move y eax)
+ (inst sar rax n-fixnum-tag-bits)
+ (move y rax)
DONE))
+
+#+#.(cl:if (cl:= sb!vm:n-fixnum-tag-bits 1) '(:and) '(:or))
+(define-vop (move-to-word/integer)
+ (:args (x :scs (descriptor-reg) :target y))
+ (:results (y :scs (signed-reg unsigned-reg)))
+ (:note "integer to untagged word coercion")
+ (:temporary (:sc unsigned-reg) backup)
+ (:generator 4
+ (move y x)
+ (if (location= x y)
+ ;; It would be great if a principled way existed to advise GC of
+ ;; algebraic transforms such as 2*R being a conservative root.
+ ;; Until that is possible, emit straightforward code that uses
+ ;; a copy of the potential reference.
+ (move backup x)
+ (setf backup x))
+ (inst sar y 1) ; optimistically assume it's a fixnum
+ (inst jmp :nc DONE) ; no carry implies tag was 0
+ (loadw y backup bignum-digits-offset other-pointer-lowtag)
+ DONE))
+
(define-move-vop move-to-word/integer :move
(descriptor-reg) (signed-reg unsigned-reg))
-
;;; Result is a fixnum, so we can just shift. We need the result type
;;; restriction because of the control-stack ambiguity noted above.
(define-vop (move-from-word/fixnum)
(:generator 1
(cond ((and (sc-is x signed-reg unsigned-reg)
(not (location= x y)))
- ;; Uses 7 bytes, but faster on the Pentium
- (inst lea y (make-ea :qword :index x
- :scale (ash 1 n-fixnum-tag-bits))))
+ (if (= n-fixnum-tag-bits 1)
+ (inst lea y (make-ea :qword :base x :index x))
+ (inst lea y (make-ea :qword :index x
+ :scale (ash 1 n-fixnum-tag-bits)))))
(t
;; Uses: If x is a reg 2 + 3; if x = y uses only 3 bytes
(move y x)
;;; as the case may be. Fixnum case inline, bignum case in an assembly
;;; routine.
(define-vop (move-from-signed)
- (:args (x :scs (signed-reg unsigned-reg) :to :result))
- (:results (y :scs (any-reg descriptor-reg) :from :argument))
+ (:args (x :scs (signed-reg unsigned-reg) :to :result . #.(and (= 1 n-fixnum-tag-bits)
+ '(:target y))))
+ (:results (y :scs (any-reg descriptor-reg) . #.(and (> n-fixnum-tag-bits 1)
+ '(:from :argument))))
(:note "signed word to integer coercion")
;; Worst case cost to make sure people know they may be number consing.
(:generator 20
- (aver (not (location= x y)))
- (let ((done (gen-label)))
- (inst imul y x #.(ash 1 n-fixnum-tag-bits))
- (inst jmp :no done)
- (inst mov y x)
- (inst lea temp-reg-tn
- (make-ea :qword :disp
- (make-fixup (ecase (tn-offset y)
- (#.rax-offset 'alloc-signed-bignum-in-rax)
- (#.rcx-offset 'alloc-signed-bignum-in-rcx)
- (#.rdx-offset 'alloc-signed-bignum-in-rdx)
- (#.rbx-offset 'alloc-signed-bignum-in-rbx)
- (#.rsi-offset 'alloc-signed-bignum-in-rsi)
- (#.rdi-offset 'alloc-signed-bignum-in-rdi)
- (#.r8-offset 'alloc-signed-bignum-in-r8)
- (#.r9-offset 'alloc-signed-bignum-in-r9)
- (#.r10-offset 'alloc-signed-bignum-in-r10)
- (#.r12-offset 'alloc-signed-bignum-in-r12)
- (#.r13-offset 'alloc-signed-bignum-in-r13)
- (#.r14-offset 'alloc-signed-bignum-in-r14)
- (#.r15-offset 'alloc-signed-bignum-in-r15))
- :assembly-routine)))
- (inst call temp-reg-tn)
- (emit-label done))))
+ (cond ((= 1 n-fixnum-tag-bits)
+ (move y x)
+ (inst shl y 1)
+ (inst jmp :no DONE)
+ (if (location= y x)
+ (inst rcr y 1) ; we're about to cons a bignum. this RCR is noise
+ (inst mov y x)))
+ (t
+ (aver (not (location= x y)))
+ (inst imul y x #.(ash 1 n-fixnum-tag-bits))
+ (inst jmp :no DONE)
+ (inst mov y x)))
+ (inst lea temp-reg-tn
+ (make-ea :qword :disp
+ (make-fixup (ecase (tn-offset y)
+ (#.rax-offset 'alloc-signed-bignum-in-rax)
+ (#.rcx-offset 'alloc-signed-bignum-in-rcx)
+ (#.rdx-offset 'alloc-signed-bignum-in-rdx)
+ (#.rbx-offset 'alloc-signed-bignum-in-rbx)
+ (#.rsi-offset 'alloc-signed-bignum-in-rsi)
+ (#.rdi-offset 'alloc-signed-bignum-in-rdi)
+ (#.r8-offset 'alloc-signed-bignum-in-r8)
+ (#.r9-offset 'alloc-signed-bignum-in-r9)
+ (#.r10-offset 'alloc-signed-bignum-in-r10)
+ (#.r12-offset 'alloc-signed-bignum-in-r12)
+ (#.r13-offset 'alloc-signed-bignum-in-r13)
+ (#.r14-offset 'alloc-signed-bignum-in-r14)
+ (#.r15-offset 'alloc-signed-bignum-in-r15))
+ :assembly-routine)))
+ (inst call temp-reg-tn)
+ DONE))
(define-move-vop move-from-signed :move
(signed-reg) (descriptor-reg))
(:generator 20
(aver (not (location= x y)))
(let ((done (gen-label)))
- (inst mov y #.(ash lowtag-mask n-positive-fixnum-bits))
+ (inst mov y #.(ash (1- (ash 1 (1+ n-fixnum-tag-bits)))
+ n-positive-fixnum-bits))
;; The assembly routines test the sign flag from this one, so if
;; you change stuff here, make sure the sign flag doesn't get
;; overwritten before the CALL!
(inst test x y)
- ;; Faster but bigger then SHL Y 4. The cost of doing this
- ;; speculatively should be noise compared to bignum consing if
- ;; that is needed and saves one branch.
- (inst lea y (make-ea :qword :index x :scale (ash 1 n-fixnum-tag-bits)))
+ ;; Using LEA is faster but bigger than MOV+SHL; it also doesn't
+ ;; twiddle the sign flag. The cost of doing this speculatively
+ ;; should be noise compared to bignum consing if that is needed
+ ;; and saves one branch.
+ (if (= n-fixnum-tag-bits 1)
+ (inst lea y (make-ea :qword :base x :index x))
+ (inst lea y (make-ea :qword :index x
+ :scale (ash 1 n-fixnum-tag-bits))))
(inst jmp :z done)
(inst mov y x)
(inst lea temp-reg-tn
projects / sbcl.git / blobdiff