1 //===---------------------------------------------------------------------===//
2 // Random ideas for the X86 backend.
3 //===---------------------------------------------------------------------===//
5 Add a MUL2U and MUL2S nodes to represent a multiply that returns both the
6 Hi and Lo parts (combination of MUL and MULH[SU] into one node). Add this to
7 X86, & make the dag combiner produce it when needed. This will eliminate one
8 imul from the code generated for:
10 long long test(long long X, long long Y) { return X*Y; }
12 by using the EAX result from the mul. We should add a similar node for
17 long long test(int X, int Y) { return (long long)X*Y; }
19 ... which should only be one imul instruction.
21 This can be done with a custom expander, but it would be nice to move this to
24 //===---------------------------------------------------------------------===//
26 This should be one DIV/IDIV instruction, not a libcall:
28 unsigned test(unsigned long long X, unsigned Y) {
32 This can be done trivially with a custom legalizer. What about overflow
33 though? http://gcc.gnu.org/bugzilla/show_bug.cgi?id=14224
35 //===---------------------------------------------------------------------===//
37 Improvements to the multiply -> shift/add algorithm:
38 http://gcc.gnu.org/ml/gcc-patches/2004-08/msg01590.html
40 //===---------------------------------------------------------------------===//
42 Improve code like this (occurs fairly frequently, e.g. in LLVM):
43 long long foo(int x) { return 1LL << x; }
45 http://gcc.gnu.org/ml/gcc-patches/2004-09/msg01109.html
46 http://gcc.gnu.org/ml/gcc-patches/2004-09/msg01128.html
47 http://gcc.gnu.org/ml/gcc-patches/2004-09/msg01136.html
49 Another useful one would be ~0ULL >> X and ~0ULL << X.
51 One better solution for 1LL << x is:
60 But that requires good 8-bit subreg support.
62 64-bit shifts (in general) expand to really bad code. Instead of using
63 cmovs, we should expand to a conditional branch like GCC produces.
65 //===---------------------------------------------------------------------===//
68 _Bool f(_Bool a) { return a!=1; }
75 //===---------------------------------------------------------------------===//
79 1. Dynamic programming based approach when compile time if not an
81 2. Code duplication (addressing mode) during isel.
82 3. Other ideas from "Register-Sensitive Selection, Duplication, and
83 Sequencing of Instructions".
84 4. Scheduling for reduced register pressure. E.g. "Minimum Register
85 Instruction Sequence Problem: Revisiting Optimal Code Generation for DAGs"
86 and other related papers.
87 http://citeseer.ist.psu.edu/govindarajan01minimum.html
89 //===---------------------------------------------------------------------===//
91 Should we promote i16 to i32 to avoid partial register update stalls?
93 //===---------------------------------------------------------------------===//
95 Leave any_extend as pseudo instruction and hint to register
96 allocator. Delay codegen until post register allocation.
98 //===---------------------------------------------------------------------===//
100 Count leading zeros and count trailing zeros:
102 int clz(int X) { return __builtin_clz(X); }
103 int ctz(int X) { return __builtin_ctz(X); }
105 $ gcc t.c -S -o - -O3 -fomit-frame-pointer -masm=intel
107 bsr %eax, DWORD PTR [%esp+4]
111 bsf %eax, DWORD PTR [%esp+4]
114 however, check that these are defined for 0 and 32. Our intrinsics are, GCC's
117 Another example (use predsimplify to eliminate a select):
119 int foo (unsigned long j) {
121 return __builtin_ffs (j) - 1;
126 //===---------------------------------------------------------------------===//
128 Use push/pop instructions in prolog/epilog sequences instead of stores off
129 ESP (certain code size win, perf win on some [which?] processors).
130 Also, it appears icc use push for parameter passing. Need to investigate.
132 //===---------------------------------------------------------------------===//
134 Only use inc/neg/not instructions on processors where they are faster than
135 add/sub/xor. They are slower on the P4 due to only updating some processor
138 //===---------------------------------------------------------------------===//
140 The instruction selector sometimes misses folding a load into a compare. The
141 pattern is written as (cmp reg, (load p)). Because the compare isn't
142 commutative, it is not matched with the load on both sides. The dag combiner
143 should be made smart enough to cannonicalize the load into the RHS of a compare
144 when it can invert the result of the compare for free.
146 //===---------------------------------------------------------------------===//
148 How about intrinsics? An example is:
149 *res = _mm_mulhi_epu16(*A, _mm_mul_epu32(*B, *C));
152 pmuludq (%eax), %xmm0
157 The transformation probably requires a X86 specific pass or a DAG combiner
158 target specific hook.
160 //===---------------------------------------------------------------------===//
162 In many cases, LLVM generates code like this:
171 on some processors (which ones?), it is more efficient to do this:
180 Doing this correctly is tricky though, as the xor clobbers the flags.
182 //===---------------------------------------------------------------------===//
184 We should generate bts/btr/etc instructions on targets where they are cheap or
185 when codesize is important. e.g., for:
187 void setbit(int *target, int bit) {
188 *target |= (1 << bit);
190 void clearbit(int *target, int bit) {
191 *target &= ~(1 << bit);
194 //===---------------------------------------------------------------------===//
196 Instead of the following for memset char*, 1, 10:
198 movl $16843009, 4(%edx)
199 movl $16843009, (%edx)
202 It might be better to generate
209 when we can spare a register. It reduces code size.
211 //===---------------------------------------------------------------------===//
213 Evaluate what the best way to codegen sdiv X, (2^C) is. For X/8, we currently
230 GCC knows several different ways to codegen it, one of which is this:
240 which is probably slower, but it's interesting at least :)
242 //===---------------------------------------------------------------------===//
244 The first BB of this code:
248 %V = call bool %foo()
249 br bool %V, label %T, label %F
266 It would be better to emit "cmp %al, 1" than a xor and test.
268 //===---------------------------------------------------------------------===//
270 Enable X86InstrInfo::convertToThreeAddress().
272 //===---------------------------------------------------------------------===//
274 We are currently lowering large (1MB+) memmove/memcpy to rep/stosl and rep/movsl
275 We should leave these as libcalls for everything over a much lower threshold,
276 since libc is hand tuned for medium and large mem ops (avoiding RFO for large
277 stores, TLB preheating, etc)
279 //===---------------------------------------------------------------------===//
281 Optimize this into something reasonable:
282 x * copysign(1.0, y) * copysign(1.0, z)
284 //===---------------------------------------------------------------------===//
286 Optimize copysign(x, *y) to use an integer load from y.
288 //===---------------------------------------------------------------------===//
290 %X = weak global int 0
293 %N = cast int %N to uint
294 %tmp.24 = setgt int %N, 0
295 br bool %tmp.24, label %no_exit, label %return
298 %indvar = phi uint [ 0, %entry ], [ %indvar.next, %no_exit ]
299 %i.0.0 = cast uint %indvar to int
300 volatile store int %i.0.0, int* %X
301 %indvar.next = add uint %indvar, 1
302 %exitcond = seteq uint %indvar.next, %N
303 br bool %exitcond, label %return, label %no_exit
317 jl LBB_foo_4 # return
318 LBB_foo_1: # no_exit.preheader
321 movl L_X$non_lazy_ptr, %edx
325 jne LBB_foo_2 # no_exit
326 LBB_foo_3: # return.loopexit
330 We should hoist "movl L_X$non_lazy_ptr, %edx" out of the loop after
331 remateralization is implemented. This can be accomplished with 1) a target
332 dependent LICM pass or 2) makeing SelectDAG represent the whole function.
334 //===---------------------------------------------------------------------===//
336 The following tests perform worse with LSR:
338 lambda, siod, optimizer-eval, ackermann, hash2, nestedloop, strcat, and Treesor.
340 //===---------------------------------------------------------------------===//
342 Teach the coalescer to coalesce vregs of different register classes. e.g. FR32 /
345 //===---------------------------------------------------------------------===//
353 Obviously it would have been better for the first mov (or any op) to store
354 directly %esp[0] if there are no other uses.
356 //===---------------------------------------------------------------------===//
358 Adding to the list of cmp / test poor codegen issues:
360 int test(__m128 *A, __m128 *B) {
361 if (_mm_comige_ss(*A, *B))
381 Note the setae, movzbl, cmpl, cmove can be replaced with a single cmovae. There
382 are a number of issues. 1) We are introducing a setcc between the result of the
383 intrisic call and select. 2) The intrinsic is expected to produce a i32 value
384 so a any extend (which becomes a zero extend) is added.
386 We probably need some kind of target DAG combine hook to fix this.
388 //===---------------------------------------------------------------------===//
390 We generate significantly worse code for this than GCC:
391 http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21150
392 http://gcc.gnu.org/bugzilla/attachment.cgi?id=8701
394 There is also one case we do worse on PPC.
396 //===---------------------------------------------------------------------===//
398 If shorter, we should use things like:
403 The former can also be used when the two-addressy nature of the 'and' would
404 require a copy to be inserted (in X86InstrInfo::convertToThreeAddress).
406 //===---------------------------------------------------------------------===//
410 char foo(int x) { return x; }
418 SIGN_EXTEND_INREG can be implemented as (sext (trunc)) to take advantage of
421 //===---------------------------------------------------------------------===//
425 typedef struct pair { float A, B; } pair;
426 void pairtest(pair P, float *FP) {
430 We currently generate this code with llvmgcc4:
442 we should be able to generate:
450 The issue is that llvmgcc4 is forcing the struct to memory, then passing it as
451 integer chunks. It does this so that structs like {short,short} are passed in
452 a single 32-bit integer stack slot. We should handle the safe cases above much
453 nicer, while still handling the hard cases.
455 While true in general, in this specific case we could do better by promoting
456 load int + bitcast to float -> load fload. This basically needs alignment info,
457 the code is already implemented (but disabled) in dag combine).
459 //===---------------------------------------------------------------------===//
461 Another instruction selector deficiency:
464 %tmp = load int (int)** %foo
465 %tmp = tail call int %tmp( int 3 )
471 movl L_foo$non_lazy_ptr, %eax
477 The current isel scheme will not allow the load to be folded in the call since
478 the load's chain result is read by the callseq_start.
480 //===---------------------------------------------------------------------===//
482 Don't forget to find a way to squash noop truncates in the JIT environment.
484 //===---------------------------------------------------------------------===//
486 Implement anyext in the same manner as truncate that would allow them to be
489 //===---------------------------------------------------------------------===//
491 How about implementing truncate / anyext as a property of machine instruction
492 operand? i.e. Print as 32-bit super-class register / 16-bit sub-class register.
493 Do this for the cases where a truncate / anyext is guaranteed to be eliminated.
494 For IA32 that is truncate from 32 to 16 and anyext from 16 to 32.
496 //===---------------------------------------------------------------------===//
506 imull $3, 4(%esp), %eax
508 Perhaps this is what we really should generate is? Is imull three or four
509 cycles? Note: ICC generates this:
511 leal (%eax,%eax,2), %eax
513 The current instruction priority is based on pattern complexity. The former is
514 more "complex" because it folds a load so the latter will not be emitted.
516 Perhaps we should use AddedComplexity to give LEA32r a higher priority? We
517 should always try to match LEA first since the LEA matching code does some
518 estimate to determine whether the match is profitable.
520 However, if we care more about code size, then imull is better. It's two bytes
521 shorter than movl + leal.
523 //===---------------------------------------------------------------------===//
525 Implement CTTZ, CTLZ with bsf and bsr.
527 //===---------------------------------------------------------------------===//
529 It appears gcc place string data with linkonce linkage in
530 .section __TEXT,__const_coal,coalesced instead of
531 .section __DATA,__const_coal,coalesced.
532 Take a look at darwin.h, there are other Darwin assembler directives that we
535 //===---------------------------------------------------------------------===//
537 int %foo(int* %a, int %t) {
541 cond_true: ; preds = %cond_true, %entry
542 %x.0.0 = phi int [ 0, %entry ], [ %tmp9, %cond_true ]
543 %t_addr.0.0 = phi int [ %t, %entry ], [ %tmp7, %cond_true ]
544 %tmp2 = getelementptr int* %a, int %x.0.0
545 %tmp3 = load int* %tmp2 ; <int> [#uses=1]
546 %tmp5 = add int %t_addr.0.0, %x.0.0 ; <int> [#uses=1]
547 %tmp7 = add int %tmp5, %tmp3 ; <int> [#uses=2]
548 %tmp9 = add int %x.0.0, 1 ; <int> [#uses=2]
549 %tmp = setgt int %tmp9, 39 ; <bool> [#uses=1]
550 br bool %tmp, label %bb12, label %cond_true
552 bb12: ; preds = %cond_true
556 is pessimized by -loop-reduce and -indvars
558 //===---------------------------------------------------------------------===//
560 u32 to float conversion improvement:
562 float uint32_2_float( unsigned u ) {
563 float fl = (int) (u & 0xffff);
564 float fh = (int) (u >> 16);
569 00000000 subl $0x04,%esp
570 00000003 movl 0x08(%esp,1),%eax
571 00000007 movl %eax,%ecx
572 00000009 shrl $0x10,%ecx
573 0000000c cvtsi2ss %ecx,%xmm0
574 00000010 andl $0x0000ffff,%eax
575 00000015 cvtsi2ss %eax,%xmm1
576 00000019 mulss 0x00000078,%xmm0
577 00000021 addss %xmm1,%xmm0
578 00000025 movss %xmm0,(%esp,1)
579 0000002a flds (%esp,1)
580 0000002d addl $0x04,%esp
583 //===---------------------------------------------------------------------===//
585 When using fastcc abi, align stack slot of argument of type double on 8 byte
586 boundary to improve performance.
588 //===---------------------------------------------------------------------===//
592 int f(int a, int b) {
593 if (a == 4 || a == 6)
605 //===---------------------------------------------------------------------===//
607 GCC's ix86_expand_int_movcc function (in i386.c) has a ton of interesting
608 simplifications for integer "x cmp y ? a : b". For example, instead of:
611 void f(int X, int Y) {
637 //===---------------------------------------------------------------------===//
639 Currently we don't have elimination of redundant stack manipulations. Consider
644 call fastcc void %test1( )
645 call fastcc void %test2( sbyte* cast (void ()* %test1 to sbyte*) )
649 declare fastcc void %test1()
651 declare fastcc void %test2(sbyte*)
654 This currently compiles to:
664 The add\sub pair is really unneeded here.
666 //===---------------------------------------------------------------------===//
668 We currently compile sign_extend_inreg into two shifts:
671 return (long)(signed char)X;
688 //===---------------------------------------------------------------------===//
690 Consider the expansion of:
692 uint %test3(uint %X) {
693 %tmp1 = rem uint %X, 255
697 Currently it compiles to:
700 movl $2155905153, %ecx
706 This could be "reassociated" into:
708 movl $2155905153, %eax
712 to avoid the copy. In fact, the existing two-address stuff would do this
713 except that mul isn't a commutative 2-addr instruction. I guess this has
714 to be done at isel time based on the #uses to mul?
716 //===---------------------------------------------------------------------===//
718 Make sure the instruction which starts a loop does not cross a cacheline
719 boundary. This requires knowning the exact length of each machine instruction.
720 That is somewhat complicated, but doable. Example 256.bzip2:
722 In the new trace, the hot loop has an instruction which crosses a cacheline
723 boundary. In addition to potential cache misses, this can't help decoding as I
724 imagine there has to be some kind of complicated decoder reset and realignment
725 to grab the bytes from the next cacheline.
727 532 532 0x3cfc movb (1809(%esp, %esi), %bl <<<--- spans 2 64 byte lines
728 942 942 0x3d03 movl %dh, (1809(%esp, %esi)
729 937 937 0x3d0a incl %esi
730 3 3 0x3d0b cmpb %bl, %dl
731 27 27 0x3d0d jnz 0x000062db <main+11707>
733 //===---------------------------------------------------------------------===//
735 In c99 mode, the preprocessor doesn't like assembly comments like #TRUNCATE.
737 //===---------------------------------------------------------------------===//
739 This could be a single 16-bit load.
742 if ((p[0] == 1) & (p[1] == 2)) return 1;
746 //===---------------------------------------------------------------------===//
748 We should inline lrintf and probably other libc functions.
750 //===---------------------------------------------------------------------===//
752 Start using the flags more. For example, compile:
754 int add_zf(int *x, int y, int a, int b) {
778 int add_zf(int *x, int y, int a, int b) {
802 //===---------------------------------------------------------------------===//
806 int foo(double X) { return isnan(X); }
817 the pxor is not needed, we could compare the value against itself.
819 //===---------------------------------------------------------------------===//
821 These two functions have identical effects:
823 unsigned int f(unsigned int i, unsigned int n) {++i; if (i == n) ++i; return i;}
824 unsigned int f2(unsigned int i, unsigned int n) {++i; i += i == n; return i;}
826 We currently compile them to:
834 jne LBB1_2 #UnifiedReturnBlock
838 LBB1_2: #UnifiedReturnBlock
848 leal 1(%ecx,%eax), %eax
851 both of which are inferior to GCC's:
869 //===---------------------------------------------------------------------===//
877 is currently compiled to:
888 It would be better to produce:
897 This can be applied to any no-return function call that takes no arguments etc.
898 Alternatively, the stack save/restore logic could be shrink-wrapped, producing
909 Both are useful in different situations. Finally, it could be shrink-wrapped
910 and tail called, like this:
917 pop %eax # realign stack.
920 Though this probably isn't worth it.
922 //===---------------------------------------------------------------------===//