source: branches/ia32/compiler

Revision Log Mode:


Legend:

Added
Modified
Copied or renamed
Diff Rev Age Author Log Message
(edit) @9802   11 years rme New function UNSIGNED-TO-SIGNED. Add missing vinsns misc-ref-c-u16, …
(edit) @9801   11 years rme x862-%natural+, x862-%natural-, x862-%natural-logxor, …
(edit) @9768   11 years rme Add .SPaset1 (replacing .SPunused_5), and use it in x862-%aset1. …
(edit) @9756   11 years rme Not ready for vinsn aset1 yet.
(edit) @9755   11 years rme Add missing vinsn misc-set-c-u8
(edit) @9754   11 years rme The TEST and XCHG instructions don't use a direction bit (which, in …
(edit) @9743   11 years rme Add forgotten vinsn %ref-symbol-value.
(edit) @9742   11 years rme Merge r8577 (fix natural-shift-right by constant)
(edit) @9693   11 years rme In fixnum->char, be careful to use addb only when we know that dest is …
(edit) @9692   11 years rme Don't bother with branch hint prefixes. On Pentium M and Core 2 …
(edit) @9683   11 years rme Undo change that used temp1 to pass first arg to ternary builtins (of …
(edit) @9678   11 years rme Explicity mention x8632::ra0 as a vinsn temporary in several places. …
(edit) @9677   11 years rme x862-natural-compare: use with-add'l-imm-reg x862-spread-lambda-list: …
(edit) @9646   11 years rme On x8632, we're using the PPC32-style DEFINE-FIXEDSIZE-OBJECT macro, …
(edit) @9587   11 years rme New vinsns fp-stack-to-single, fp-stack-to-double.
(edit) @9586   11 years rme x862-i386-ff-call: when a foreign function returns a floating point …
(edit) @9584   11 years rme unbox-u32: remove apparently extraneous code %unbox-u32: delete; we …
(edit) @9583   11 years rme x862-unboxed-reg-for-aset: can't assume that (unsigned-byte 32) fits …
(edit) @9581   11 years rme x862-get-immediate-xxx: rearrange the placement of with-add'l-imm-reg forms
(edit) @9580   11 years rme x862-fixnup-fwd-refs: update for x8632 x862-vref1: in the 32-bit …
(edit) @9572   11 years rme misc-set-c-u32, misc-set-c-s32: scale the index appropriately.
(edit) @9561   11 years rme .SPsetqsym
(edit) @9560   11 years rme Use function-to-function-vector archmacro (instead of calling …
(edit) @9545   11 years rme add %unbox-u32
(edit) @9544   11 years rme Add PSRAD instruction (two register operand form only).
(edit) @9462   11 years rme x862-get-bit: don't use with-add'l-imm-reg before using vinsn …
(edit) @9457   11 years rme x862-set-bit: don't use with-add'l-imm-register before emitting vinsn …
(edit) @9456   11 years rme Don't use with-add'l-imm-reg in x862-set-bit, since the vinsn …
(edit) @9455   11 years rme Add x87 FPU instruction FSTP. This will be used by the x8632 port to …
(edit) @9451   11 years rme Modify vinsn set-c-flag-if-variable-logbitp not to require an …
(edit) @9378   11 years rme Update create-x86-function so that it updates the self-reference table …
(edit) @9377   11 years rme Add x8632-xdisassemble.
(edit) @9376   11 years rme Correct vinsn temps in fix-fixnum-overflow-ool{-and-branch}. Add PROVIDE.
(edit) @9375   11 years rme Add (require "X8632-VINSNS").
(edit) @9374   11 years rme Define imm0.bh (ah), arg_z.bh (bh), etc. Define subtag-forward-marker.
(edit) @9302   11 years rme x862-vset1: remove some apparently unecessary with-add'l-imm-reg forms …
(edit) @9265   11 years rme set-{single,double}-c-arg, reload-{single,double}-c-arg: use vinsn …
(edit) @9264   11 years rme x862-vref1: don't need add'l imm reg in 16-bit case (at least) …
(edit) @9233   11 years rme unbox-u32: test for two digit and one digit bignums (not three and two …
(edit) @9191   11 years rme tail-funcall: fix typo affecting funcalling a symbol init-nclosure: …
(edit) @9190   11 years rme archmacros lfun-vector, lfun-vector-lfun: make these no-ops.
(edit) @9160   11 years rme %define-x8632-lap-function: tweaks to get :code-size to DTRT.
(edit) @9124   11 years rme Correction to save-frame-variable-arg-count.
(edit) @9123   11 years rme Don't try to use movapd instruction to clear memory on tstack, since …
(edit) @9122   11 years rme %define-x8632-lap-function: pay attention to *x86-lap-fixed-code-words*
(edit) @9043   11 years rme x862-lambda: Pass the right args to x862-seq-bind-var.
(edit) @9042   11 years rme sigh.
(edit) @9041   11 years rme Add back new load-next-method-context that I accidentally blew away.
(edit) @9039   11 years rme Keep next-method-context in a slot in the TCR.
(edit) @9000   11 years rme Add unboxed0 and unboxed1 slots to TCR. These are just scratch …
(edit) @8999   11 years rme compose-digit LAP macro for x8632
(edit) @8902   11 years rme extract-tag-fixnum: simplify jump-known-function: just jump through …
(edit) @8887   12 years rme x862-allocate-global-registers: always return (values 0 nil) on x8632 …
(edit) @8876   12 years rme x862-nlexit: add x8632 case to mark nargs in use before calling …
(edit) @8875   12 years rme unbind-interrupt-level-inline: figure out what the 16 was for.
(edit) @8858   12 years rme Add vinsns gets32 and getu32.
(edit) @8841   12 years rme x862-%immediate-store: changes for 32-bit target. (It's amazing that …
(edit) @8834   12 years rme port r8832 from ia32-0319 debug branch
(edit) @8829   12 years rme x862-labels: use *x862-temp1* instead of x8664::temp1
(edit) @8824   12 years rme set-c-arg-from-mm0: use MOVQ, not MOVD.
(edit) @8823   12 years rme Get Darwin/IA-32 syscalls partially working. (64-bit args/results …
(edit) @8821   12 years rme x862-make-list: respect *x862-target-num-arg-regs* when forming …
(edit) @8750   12 years rme Corrections to init-nclosure vinsn. Add vinsns: syscall, …
(edit) @8749   12 years rme X862-MAKE-CLOSURE, X862-LABELS: changes for closures on IA-32. Also …
(edit) @8688   12 years rme x862-push-register, x862-pop-register: track cstack correctly for …
(edit) @8687   12 years rme Correct bogosities in set-z-flag-if-cons. Use explicit vinsn …
(edit) @8665   12 years rme Fix bad typo.
(edit) @8654   12 years rme x862-do-lexical-setq: target appropriate x8632 regs for SPgvset …
(edit) @8632   12 years rme compiler-macro consp: punt if cons-tag and null-tag are the same.
(edit) @8631   12 years rme Several nargs-related changes. (nargs on x8632 is a 32-bit register; …
(edit) @8630   12 years rme Fix cvttss2si instruction definition.
(edit) @8629   12 years rme Numerous changes.
(edit) @8628   12 years rme set-z-flag-if-consp vinsn; updated bit vector vinsns.
(edit) @8627   12 years rme Checkpoint work in progress. Includes: * bit vector changes (don't …
(edit) @8426   12 years rme Use temp1 for nargs. Update %kernel-global and kernel-global for X8632.
(edit) @8368   12 years rme Treat nargs as a 32-bit register.
(edit) @8367   12 years rme Treat nargs as a 32-bit register.
(edit) @8357   12 years rme Make nargs 32 bits; define subtagmask; define default-node-regs-mask; …
(edit) @8349   12 years rme Don't open-code funcall. Fixes to %symbol->symptr, …
(edit) @8348   12 years rme More messing around with calling X862-THREE-TARGETED-REG-FORMS.
(edit) @8261   12 years rme When target-arch is x8632, target %temp0 instead of %arg_x in a few …
(edit) @8256   12 years rme Try again with default-optionals. Should probably dump the partial …
(edit) @8255   12 years rme Make tag for xcode-vector match ppc32.
(edit) @8254   12 years rme Don't bother trying to use byte registers in extract-lisptag and …
(edit) @8233   12 years rme Preserve nargs/eax in default-optionals. Correct emit-aligned-label …
(edit) @8232   12 years rme Repair brain damage in extract-typecode-fixnum.
(edit) @8231   12 years rme Preserve nargs in funcall/tail-funcall. Only compare to low byte of …
(edit) @8225   12 years rme Try to get unconfused about things tagged with fulltag-imm.
(edit) @8211   12 years rme Avoid word-size operands in trap-unless-list. The fullword code is …
(edit) @8161   12 years rme In OP-INDIRE, don't set REX prefix when in 32-bit mode.
(edit) @8076   12 years rme Several changes and additions for IA-32, including …
(edit) @8075   12 years rme Try to get misc-dfloat-offset right; add scratch-mxcsr field to TCR. …
(edit) @8074   12 years rme *x86-lap-entry-offset* confusion; other tweaks.
(edit) @8073   12 years rme Let ANDL opcode take a signed or unsigned immediate; add immediate …
(edit) @7964   12 years rme Correct SHLD and SHRD operands for 32 and 16 bit variants. Define a …
(edit) @7962   12 years rme Add save[0-3] fields to the TCR. These are a caller-saved spill area …
(edit) @7906   12 years rme Add cmovc/cmovnc/cmovz/cmovnz/jc/jnc opcodes (which are the same thing …
(edit) @7881   12 years rme To avoid confusion, treat defx86lapfunction as an x86-64 thing, and …
(edit) @7878   12 years rme WITH-SHORT-STACK-FLOATS macro
(edit) @7877   12 years rme A few more vinsns, including some single-float ones.
Note: See TracRevisionLog for help on using the revision log.