RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-01 08:23:21 +01:00
Code Issues Projects Releases Wiki Activity
c922b4f2a3
llvm-mirror/test/CodeGen/X86/tail-opts.ll

433 lines
12 KiB
LLVM
Raw Normal View History

Don't enable the post-RA scheduler on x86 except at -O3. In its current form, it is too expensive in compile time. llvm-svn: 90781
2009-12-07 20:04:31 +01:00
; RUN: llc < %s -march=x86-64 -mtriple=x86_64-unknown-linux-gnu -asm-verbose=false -post-RA-scheduler=true | FileCheck %s
Add support for tail duplication to BranchFolding, and extend tail merging support to handle more cases. - Recognize several cases where tail merging is beneficial even when the tail size is smaller than the generic threshold. - Make use of MachineInstrDesc::isBarrier to help detect non-fallthrough blocks. - Check for and avoid disrupting fall-through edges in more cases. llvm-svn: 86871
2009-11-11 20:48:59 +01:00
declare void @bar(i32)
declare void @car(i32)
declare void @dar(i32)
declare void @ear(i32)
declare void @far(i32)
declare i1 @qux()
@GHJK = global i32 0
@HABC = global i32 0
; BranchFolding should tail-merge the stores since they all precede
; direct branches to the same place.
; CHECK: tail_merge_me:
; CHECK-NOT: GHJK
; CHECK: movl $0, GHJK(%rip)
; CHECK-NEXT: movl $1, HABC(%rip)
; CHECK-NOT: GHJK
define void @tail_merge_me() nounwind {
entry:
%a = call i1 @qux()
br i1 %a, label %A, label %next
next:
%b = call i1 @qux()
br i1 %b, label %B, label %C
A:
call void @bar(i32 0)
store i32 0, i32* @GHJK
br label %M
B:
call void @car(i32 1)
store i32 0, i32* @GHJK
br label %M
C:
call void @dar(i32 2)
store i32 0, i32* @GHJK
br label %M
M:
store i32 1, i32* @HABC
%c = call i1 @qux()
br i1 %c, label %return, label %altret
return:
call void @ear(i32 1000)
ret void
altret:
call void @far(i32 1001)
ret void
}
Delete useless trailing semicolons. llvm-svn: 92740
2010-01-05 18:55:26 +01:00
declare i8* @choose(i8*, i8*)
Add support for tail duplication to BranchFolding, and extend tail merging support to handle more cases. - Recognize several cases where tail merging is beneficial even when the tail size is smaller than the generic threshold. - Make use of MachineInstrDesc::isBarrier to help detect non-fallthrough blocks. - Check for and avoid disrupting fall-through edges in more cases. llvm-svn: 86871
2009-11-11 20:48:59 +01:00
; BranchFolding should tail-duplicate the indirect jump to avoid
; redundant branching.
; CHECK: tail_duplicate_me:
; CHECK: movl $0, GHJK(%rip)
Teach machine sink to 1) Do forward copy propagation. This makes it easier to estimate the cost of the instruction being sunk. 2) Break critical edges on demand, including cases where the value is used by PHI nodes. Critical edge splitting is not yet enabled by default. llvm-svn: 114227
2010-09-18 00:28:18 +02:00
; CHECK-NEXT: jmpq *%r
Add support for tail duplication to BranchFolding, and extend tail merging support to handle more cases. - Recognize several cases where tail merging is beneficial even when the tail size is smaller than the generic threshold. - Make use of MachineInstrDesc::isBarrier to help detect non-fallthrough blocks. - Check for and avoid disrupting fall-through edges in more cases. llvm-svn: 86871
2009-11-11 20:48:59 +01:00
; CHECK: movl $0, GHJK(%rip)
Teach machine sink to 1) Do forward copy propagation. This makes it easier to estimate the cost of the instruction being sunk. 2) Break critical edges on demand, including cases where the value is used by PHI nodes. Critical edge splitting is not yet enabled by default. llvm-svn: 114227
2010-09-18 00:28:18 +02:00
; CHECK-NEXT: jmpq *%r
Add support for tail duplication to BranchFolding, and extend tail merging support to handle more cases. - Recognize several cases where tail merging is beneficial even when the tail size is smaller than the generic threshold. - Make use of MachineInstrDesc::isBarrier to help detect non-fallthrough blocks. - Check for and avoid disrupting fall-through edges in more cases. llvm-svn: 86871
2009-11-11 20:48:59 +01:00
; CHECK: movl $0, GHJK(%rip)
Teach machine sink to 1) Do forward copy propagation. This makes it easier to estimate the cost of the instruction being sunk. 2) Break critical edges on demand, including cases where the value is used by PHI nodes. Critical edge splitting is not yet enabled by default. llvm-svn: 114227
2010-09-18 00:28:18 +02:00
; CHECK-NEXT: jmpq *%r
Add support for tail duplication to BranchFolding, and extend tail merging support to handle more cases. - Recognize several cases where tail merging is beneficial even when the tail size is smaller than the generic threshold. - Make use of MachineInstrDesc::isBarrier to help detect non-fallthrough blocks. - Check for and avoid disrupting fall-through edges in more cases. llvm-svn: 86871
2009-11-11 20:48:59 +01:00
define void @tail_duplicate_me() nounwind {
entry:
%a = call i1 @qux()
%c = call i8* @choose(i8* blockaddress(@tail_duplicate_me, %return),
i8* blockaddress(@tail_duplicate_me, %altret))
br i1 %a, label %A, label %next
next:
%b = call i1 @qux()
br i1 %b, label %B, label %C
A:
call void @bar(i32 0)
store i32 0, i32* @GHJK
br label %M
B:
call void @car(i32 1)
store i32 0, i32* @GHJK
br label %M
C:
call void @dar(i32 2)
store i32 0, i32* @GHJK
br label %M
M:
indirectbr i8* %c, [label %return, label %altret]
return:
call void @ear(i32 1000)
ret void
altret:
call void @far(i32 1001)
ret void
}
; BranchFolding shouldn't try to merge the tails of two blocks
; with only a branch in common, regardless of the fallthrough situation.
; CHECK: dont_merge_oddly:
; CHECK-NOT: ret
In the pre-RA scheduler, maintain cmp+br proximity. This is done by pushing physical register definitions close to their use, which happens to handle flag definitions if they're not glued to the branch. This seems to be generally a good thing though, so I didn't need to add a target hook yet. The primary motivation is to generate code closer to what people expect and rule out missed opportunity from enabling macro-op fusion. As a side benefit, we get several 2-5% gains on x86 benchmarks. There is one regression: SingleSource/Benchmarks/Shootout/lists slows down be -10%. But this is an independent scheduler bug that will be tracked separately. See rdar://problem/9283108. Incidentally, pre-RA scheduling is only half the solution. Fixing the later passes is tracked by: <rdar://problem/8932804> [pre-RA-sched] on x86, attempt to schedule CMP/TEST adjacent with condition jump Fixes: <rdar://problem/9262453> Scheduler unnecessary break of cmp/jump fusion llvm-svn: 129508
2011-04-14 07:15:06 +02:00
; CHECK: ucomiss %xmm{{[0-2]}}, %xmm{{[0-2]}}
Start function numbering at 0. llvm-svn: 101638
2010-04-17 18:29:15 +02:00
; CHECK-NEXT: jbe .LBB2_3
In the pre-RA scheduler, maintain cmp+br proximity. This is done by pushing physical register definitions close to their use, which happens to handle flag definitions if they're not glued to the branch. This seems to be generally a good thing though, so I didn't need to add a target hook yet. The primary motivation is to generate code closer to what people expect and rule out missed opportunity from enabling macro-op fusion. As a side benefit, we get several 2-5% gains on x86 benchmarks. There is one regression: SingleSource/Benchmarks/Shootout/lists slows down be -10%. But this is an independent scheduler bug that will be tracked separately. See rdar://problem/9283108. Incidentally, pre-RA scheduling is only half the solution. Fixing the later passes is tracked by: <rdar://problem/8932804> [pre-RA-sched] on x86, attempt to schedule CMP/TEST adjacent with condition jump Fixes: <rdar://problem/9262453> Scheduler unnecessary break of cmp/jump fusion llvm-svn: 129508
2011-04-14 07:15:06 +02:00
; CHECK-NEXT: ucomiss %xmm{{[0-2]}}, %xmm{{[0-2]}}
Start function numbering at 0. llvm-svn: 101638
2010-04-17 18:29:15 +02:00
; CHECK-NEXT: ja .LBB2_4
; CHECK-NEXT: .LBB2_2:
Add support for tail duplication to BranchFolding, and extend tail merging support to handle more cases. - Recognize several cases where tail merging is beneficial even when the tail size is smaller than the generic threshold. - Make use of MachineInstrDesc::isBarrier to help detect non-fallthrough blocks. - Check for and avoid disrupting fall-through edges in more cases. llvm-svn: 86871
2009-11-11 20:48:59 +01:00
; CHECK-NEXT: movb $1, %al
; CHECK-NEXT: ret
Start function numbering at 0. llvm-svn: 101638
2010-04-17 18:29:15 +02:00
; CHECK-NEXT: .LBB2_3:
In the pre-RA scheduler, maintain cmp+br proximity. This is done by pushing physical register definitions close to their use, which happens to handle flag definitions if they're not glued to the branch. This seems to be generally a good thing though, so I didn't need to add a target hook yet. The primary motivation is to generate code closer to what people expect and rule out missed opportunity from enabling macro-op fusion. As a side benefit, we get several 2-5% gains on x86 benchmarks. There is one regression: SingleSource/Benchmarks/Shootout/lists slows down be -10%. But this is an independent scheduler bug that will be tracked separately. See rdar://problem/9283108. Incidentally, pre-RA scheduling is only half the solution. Fixing the later passes is tracked by: <rdar://problem/8932804> [pre-RA-sched] on x86, attempt to schedule CMP/TEST adjacent with condition jump Fixes: <rdar://problem/9262453> Scheduler unnecessary break of cmp/jump fusion llvm-svn: 129508
2011-04-14 07:15:06 +02:00
; CHECK-NEXT: ucomiss %xmm{{[0-2]}}, %xmm{{[0-2]}}
Start function numbering at 0. llvm-svn: 101638
2010-04-17 18:29:15 +02:00
; CHECK-NEXT: jbe .LBB2_2
; CHECK-NEXT: .LBB2_4:
Add support for tail duplication to BranchFolding, and extend tail merging support to handle more cases. - Recognize several cases where tail merging is beneficial even when the tail size is smaller than the generic threshold. - Make use of MachineInstrDesc::isBarrier to help detect non-fallthrough blocks. - Check for and avoid disrupting fall-through edges in more cases. llvm-svn: 86871
2009-11-11 20:48:59 +01:00
; CHECK-NEXT: xorb %al, %al
; CHECK-NEXT: ret
define i1 @dont_merge_oddly(float* %result) nounwind {
entry:
%tmp4 = getelementptr float* %result, i32 2
%tmp5 = load float* %tmp4, align 4
%tmp7 = getelementptr float* %result, i32 4
%tmp8 = load float* %tmp7, align 4
%tmp10 = getelementptr float* %result, i32 6
%tmp11 = load float* %tmp10, align 4
%tmp12 = fcmp olt float %tmp8, %tmp11
br i1 %tmp12, label %bb, label %bb21
bb:
%tmp23469 = fcmp olt float %tmp5, %tmp8
br i1 %tmp23469, label %bb26, label %bb30
bb21:
%tmp23 = fcmp olt float %tmp5, %tmp11
br i1 %tmp23, label %bb26, label %bb30
bb26:
ret i1 0
bb30:
ret i1 1
}
; Do any-size tail-merging when two candidate blocks will both require
; an unconditional jump to complete a two-way conditional branch.
; CHECK: c_expand_expr_stmt:
Fix register-dependent X86 tests. llvm-svn: 128867
2011-04-05 02:32:44 +02:00
;
; This test only works when register allocation happens to use %rax for both
; load addresses.
;
; CHE: jmp .LBB3_11
; CHE-NEXT: .LBB3_9:
; CHE-NEXT: movq 8(%rax), %rax
; CHE-NEXT: xorb %dl, %dl
; CHE-NEXT: movb 16(%rax), %al
; CHE-NEXT: cmpb $16, %al
; CHE-NEXT: je .LBB3_11
; CHE-NEXT: cmpb $23, %al
; CHE-NEXT: jne .LBB3_14
; CHE-NEXT: .LBB3_11:
Add support for tail duplication to BranchFolding, and extend tail merging support to handle more cases. - Recognize several cases where tail merging is beneficial even when the tail size is smaller than the generic threshold. - Make use of MachineInstrDesc::isBarrier to help detect non-fallthrough blocks. - Check for and avoid disrupting fall-through edges in more cases. llvm-svn: 86871
2009-11-11 20:48:59 +01:00
%0 = type { %struct.rtx_def* }
%struct.lang_decl = type opaque
%struct.rtx_def = type { i16, i8, i8, [1 x %union.rtunion] }
%struct.tree_decl = type { [24 x i8], i8*, i32, %union.tree_node*, i32, i8, i8, i8, i8, %union.tree_node*, %union.tree_node*, %union.tree_node*, %union.tree_node*, %union.tree_node*, %union.tree_node*, %union.tree_node*, %union.tree_node*, %union.tree_node*, %struct.rtx_def*, %union..2anon, %0, %union.tree_node*, %struct.lang_decl* }
%union..2anon = type { i32 }
%union.rtunion = type { i8* }
%union.tree_node = type { %struct.tree_decl }
define fastcc void @c_expand_expr_stmt(%union.tree_node* %expr) nounwind {
entry:
%tmp4 = load i8* null, align 8 ; <i8> [#uses=3]
switch i8 %tmp4, label %bb3 [
i8 18, label %bb
]
bb: ; preds = %entry
switch i32 undef, label %bb1 [
i32 0, label %bb2.i
i32 37, label %bb.i
]
bb.i: ; preds = %bb
switch i32 undef, label %bb1 [
i32 0, label %lvalue_p.exit
]
bb2.i: ; preds = %bb
br label %bb3
lvalue_p.exit: ; preds = %bb.i
%tmp21 = load %union.tree_node** null, align 8 ; <%union.tree_node*> [#uses=3]
%tmp22 = getelementptr inbounds %union.tree_node* %tmp21, i64 0, i32 0, i32 0, i64 0 ; <i8*> [#uses=1]
%tmp23 = load i8* %tmp22, align 8 ; <i8> [#uses=1]
%tmp24 = zext i8 %tmp23 to i32 ; <i32> [#uses=1]
switch i32 %tmp24, label %lvalue_p.exit4 [
i32 0, label %bb2.i3
i32 2, label %bb.i1
]
bb.i1: ; preds = %lvalue_p.exit
%tmp25 = getelementptr inbounds %union.tree_node* %tmp21, i64 0, i32 0, i32 2 ; <i32*> [#uses=1]
%tmp26 = bitcast i32* %tmp25 to %union.tree_node** ; <%union.tree_node**> [#uses=1]
%tmp27 = load %union.tree_node** %tmp26, align 8 ; <%union.tree_node*> [#uses=2]
%tmp28 = getelementptr inbounds %union.tree_node* %tmp27, i64 0, i32 0, i32 0, i64 16 ; <i8*> [#uses=1]
%tmp29 = load i8* %tmp28, align 8 ; <i8> [#uses=1]
%tmp30 = zext i8 %tmp29 to i32 ; <i32> [#uses=1]
switch i32 %tmp30, label %lvalue_p.exit4 [
i32 0, label %bb2.i.i2
i32 2, label %bb.i.i
]
bb.i.i: ; preds = %bb.i1
%tmp34 = tail call fastcc i32 @lvalue_p(%union.tree_node* null) nounwind ; <i32> [#uses=1]
%phitmp = icmp ne i32 %tmp34, 0 ; <i1> [#uses=1]
br label %lvalue_p.exit4
bb2.i.i2: ; preds = %bb.i1
%tmp35 = getelementptr inbounds %union.tree_node* %tmp27, i64 0, i32 0, i32 0, i64 8 ; <i8*> [#uses=1]
%tmp36 = bitcast i8* %tmp35 to %union.tree_node** ; <%union.tree_node**> [#uses=1]
%tmp37 = load %union.tree_node** %tmp36, align 8 ; <%union.tree_node*> [#uses=1]
%tmp38 = getelementptr inbounds %union.tree_node* %tmp37, i64 0, i32 0, i32 0, i64 16 ; <i8*> [#uses=1]
%tmp39 = load i8* %tmp38, align 8 ; <i8> [#uses=1]
switch i8 %tmp39, label %bb2 [
i8 16, label %lvalue_p.exit4
i8 23, label %lvalue_p.exit4
]
bb2.i3: ; preds = %lvalue_p.exit
%tmp40 = getelementptr inbounds %union.tree_node* %tmp21, i64 0, i32 0, i32 0, i64 8 ; <i8*> [#uses=1]
%tmp41 = bitcast i8* %tmp40 to %union.tree_node** ; <%union.tree_node**> [#uses=1]
%tmp42 = load %union.tree_node** %tmp41, align 8 ; <%union.tree_node*> [#uses=1]
%tmp43 = getelementptr inbounds %union.tree_node* %tmp42, i64 0, i32 0, i32 0, i64 16 ; <i8*> [#uses=1]
%tmp44 = load i8* %tmp43, align 8 ; <i8> [#uses=1]
switch i8 %tmp44, label %bb2 [
i8 16, label %lvalue_p.exit4
i8 23, label %lvalue_p.exit4
]
lvalue_p.exit4: ; preds = %bb2.i3, %bb2.i3, %bb2.i.i2, %bb2.i.i2, %bb.i.i, %bb.i1, %lvalue_p.exit
%tmp45 = phi i1 [ %phitmp, %bb.i.i ], [ false, %bb2.i.i2 ], [ false, %bb2.i.i2 ], [ false, %bb.i1 ], [ false, %bb2.i3 ], [ false, %bb2.i3 ], [ false, %lvalue_p.exit ] ; <i1> [#uses=1]
%tmp46 = icmp eq i8 %tmp4, 0 ; <i1> [#uses=1]
%or.cond = or i1 %tmp45, %tmp46 ; <i1> [#uses=1]
br i1 %or.cond, label %bb2, label %bb3
bb1: ; preds = %bb2.i.i, %bb.i, %bb
%.old = icmp eq i8 %tmp4, 23 ; <i1> [#uses=1]
br i1 %.old, label %bb2, label %bb3
bb2: ; preds = %bb1, %lvalue_p.exit4, %bb2.i3, %bb2.i.i2
br label %bb3
bb3: ; preds = %bb2, %bb1, %lvalue_p.exit4, %bb2.i, %entry
%expr_addr.0 = phi %union.tree_node* [ null, %bb2 ], [ %expr, %bb2.i ], [ %expr, %entry ], [ %expr, %bb1 ], [ %expr, %lvalue_p.exit4 ] ; <%union.tree_node*> [#uses=0]
unreachable
}
declare fastcc i32 @lvalue_p(%union.tree_node* nocapture) nounwind readonly
declare fastcc %union.tree_node* @default_conversion(%union.tree_node*) nounwind
Tail merge at any size when there are two potentials blocks and one can be made to fall through into the other. llvm-svn: 86909
2009-11-12 01:39:10 +01:00
; If one tail merging candidate falls through into the other,
; tail merging is likely profitable regardless of how few
; instructions are involved. This function should have only
; one ret instruction.
; CHECK: foo:
Instruction fixes, added instructions, and AsmString changes in the X86 instruction tables. Also (while I was at it) cleaned up the X86 tables, removing tabs and 80-line violations. This patch was reviewed by Chris Lattner, but please let me know if there are any problems. * X86*.td Removed tabs and fixed 80-line violations * X86Instr64bit.td (IRET, POPCNT, BT_, LSL, SWPGS, PUSH_S, POP_S, L_S, SMSW) Added (CALL, CMOV) Added qualifiers (JMP) Added PC-relative jump instruction (POPFQ/PUSHFQ) Added qualifiers; renamed PUSHFQ to indicate that it is 64-bit only (ambiguous since it has no REX prefix) (MOV) Added rr form going the other way, which is encoded differently (MOV) Changed immediates to offsets, which is more correct; also fixed MOV64o64a to have to a 64-bit offset (MOV) Fixed qualifiers (MOV) Added debug-register and condition-register moves (MOVZX) Added more forms (ADC, SUB, SBB, AND, OR, XOR) Added reverse forms, which (as with MOV) are encoded differently (ROL) Made REX.W required (BT) Uncommented mr form for disassembly only (CVT__2__) Added several missing non-intrinsic forms (LXADD, XCHG) Reordered operands to make more sense for MRMSrcMem (XCHG) Added register-to-register forms (XADD, CMPXCHG, XCHG) Added non-locked forms * X86InstrSSE.td (CVTSS2SI, COMISS, CVTTPS2DQ, CVTPS2PD, CVTPD2PS, MOVQ) Added * X86InstrFPStack.td (COM_FST0, COMP_FST0, COM_FI, COM_FIP, FFREE, FNCLEX, FNOP, FXAM, FLDL2T, FLDL2E, FLDPI, FLDLG2, FLDLN2, F2XM1, FYL2X, FPTAN, FPATAN, FXTRACT, FPREM1, FDECSTP, FINCSTP, FPREM, FYL2XP1, FSINCOS, FRNDINT, FSCALE, FCOMPP, FXSAVE, FXRSTOR) Added (FCOM, FCOMP) Added qualifiers (FSTENV, FSAVE, FSTSW) Fixed opcode names (FNSTSW) Added implicit register operand * X86InstrInfo.td (opaque512mem) Added for FXSAVE/FXRSTOR (offset8, offset16, offset32, offset64) Added for MOV (NOOPW, IRET, POPCNT, IN, BTC, BTR, BTS, LSL, INVLPG, STR, LTR, PUSHFS, PUSHGS, POPFS, POPGS, LDS, LSS, LES, LFS, LGS, VERR, VERW, SGDT, SIDT, SLDT, LGDT, LIDT, LLDT, LODSD, OUTSB, OUTSW, OUTSD, HLT, RSM, FNINIT, CLC, STC, CLI, STI, CLD, STD, CMC, CLTS, XLAT, WRMSR, RDMSR, RDPMC, SMSW, LMSW, CPUID, INVD, WBINVD, INVEPT, INVVPID, VMCALL, VMCLEAR, VMLAUNCH, VMRESUME, VMPTRLD, VMPTRST, VMREAD, VMWRITE, VMXOFF, VMXON) Added (NOOPL, POPF, POPFD, PUSHF, PUSHFD) Added qualifier (JO, JNO, JB, JAE, JE, JNE, JBE, JA, JS, JNS, JP, JNP, JL, JGE, JLE, JG, JCXZ) Added 32-bit forms (MOV) Changed some immediate forms to offset forms (MOV) Added reversed reg-reg forms, which are encoded differently (MOV) Added debug-register and condition-register moves (CMOV) Added qualifiers (AND, OR, XOR, ADC, SUB, SBB) Added reverse forms, like MOV (BT) Uncommented memory-register forms for disassembler (MOVSX, MOVZX) Added forms (XCHG, LXADD) Made operand order make sense for MRMSrcMem (XCHG) Added register-register forms (XADD, CMPXCHG) Added unlocked forms * X86InstrMMX.td (MMX_MOVD, MMV_MOVQ) Added forms * X86InstrInfo.cpp: Changed PUSHFQ to PUSHFQ64 to reflect table change * X86RegisterInfo.td: Added debug and condition register sets * x86-64-pic-3.ll: Fixed testcase to reflect call qualifier * peep-test-3.ll: Fixed testcase to reflect test qualifier * cmov.ll: Fixed testcase to reflect cmov qualifier * loop-blocks.ll: Fixed testcase to reflect call qualifier * x86-64-pic-11.ll: Fixed testcase to reflect call qualifier * 2009-11-04-SubregCoalescingBug.ll: Fixed testcase to reflect call qualifier * x86-64-pic-2.ll: Fixed testcase to reflect call qualifier * live-out-reg-info.ll: Fixed testcase to reflect test qualifier * tail-opts.ll: Fixed testcase to reflect call qualifiers * x86-64-pic-10.ll: Fixed testcase to reflect call qualifier * bss-pagealigned.ll: Fixed testcase to reflect call qualifier * x86-64-pic-1.ll: Fixed testcase to reflect call qualifier * widen_load-1.ll: Fixed testcase to reflect call qualifier llvm-svn: 91638
2009-12-18 01:01:26 +01:00
; CHECK: callq func
Start function numbering at 0. llvm-svn: 101638
2010-04-17 18:29:15 +02:00
; CHECK-NEXT: .LBB4_2:
Use pushq / popq instead of subq $8, %rsp / addq $8, %rsp to adjust stack in prologue and epilogue if the adjustment is 8. Similarly, use pushl / popl if the adjustment is 4 in 32-bit mode. In the epilogue, takes care to pop to a caller-saved register that's not live at the exit (either return or tailcall instruction). rdar://8771137 llvm-svn: 122783
2011-01-03 23:53:22 +01:00
; CHECK-NEXT: popq
Tail merge at any size when there are two potentials blocks and one can be made to fall through into the other. llvm-svn: 86909
2009-11-12 01:39:10 +01:00
; CHECK-NEXT: ret
define void @foo(i1* %V) nounwind {
entry:
%t0 = icmp eq i1* %V, null
br i1 %t0, label %return, label %bb
bb:
call void @func()
ret void
return:
ret void
}
declare void @func()
When optimizing for size, don't tail-merge unless it's likely to be a code-size win, and not when it's only likely to be code-size neutral, such as when only a single instruction would be eliminated and a new branch would be required. This fixes rdar://7392894. llvm-svn: 88692
2009-11-13 22:02:15 +01:00
; one - One instruction may be tail-duplicated even with optsize.
; CHECK: one:
; CHECK: movl $0, XYZ(%rip)
; CHECK: movl $0, XYZ(%rip)
@XYZ = external global i32
define void @one() nounwind optsize {
entry:
%0 = icmp eq i32 undef, 0
br i1 %0, label %bbx, label %bby
bby:
switch i32 undef, label %bb7 [
i32 16, label %return
]
bb7:
Upgrade syntax of tests using volatile instructions to use 'load volatile' instead of 'volatile load', which is archaic. llvm-svn: 145171
2011-11-27 07:54:59 +01:00
store volatile i32 0, i32* @XYZ
When optimizing for size, don't tail-merge unless it's likely to be a code-size win, and not when it's only likely to be code-size neutral, such as when only a single instruction would be eliminated and a new branch would be required. This fixes rdar://7392894. llvm-svn: 88692
2009-11-13 22:02:15 +01:00
unreachable
bbx:
switch i32 undef, label %bb12 [
i32 128, label %return
]
bb12:
Upgrade syntax of tests using volatile instructions to use 'load volatile' instead of 'volatile load', which is archaic. llvm-svn: 145171
2011-11-27 07:54:59 +01:00
store volatile i32 0, i32* @XYZ
When optimizing for size, don't tail-merge unless it's likely to be a code-size win, and not when it's only likely to be code-size neutral, such as when only a single instruction would be eliminated and a new branch would be required. This fixes rdar://7392894. llvm-svn: 88692
2009-11-13 22:02:15 +01:00
unreachable
return:
ret void
}
; two - Same as one, but with two instructions in the common
; tail instead of one. This is too much to be merged, given
; the optsize attribute.
; CHECK: two:
; CHECK-NOT: XYZ
; CHECK: movl $0, XYZ(%rip)
; CHECK: movl $1, XYZ(%rip)
; CHECK-NOT: XYZ
; CHECK: ret
define void @two() nounwind optsize {
entry:
%0 = icmp eq i32 undef, 0
br i1 %0, label %bbx, label %bby
bby:
switch i32 undef, label %bb7 [
i32 16, label %return
]
bb7:
Upgrade syntax of tests using volatile instructions to use 'load volatile' instead of 'volatile load', which is archaic. llvm-svn: 145171
2011-11-27 07:54:59 +01:00
store volatile i32 0, i32* @XYZ
store volatile i32 1, i32* @XYZ
When optimizing for size, don't tail-merge unless it's likely to be a code-size win, and not when it's only likely to be code-size neutral, such as when only a single instruction would be eliminated and a new branch would be required. This fixes rdar://7392894. llvm-svn: 88692
2009-11-13 22:02:15 +01:00
unreachable
bbx:
switch i32 undef, label %bb12 [
i32 128, label %return
]
bb12:
Upgrade syntax of tests using volatile instructions to use 'load volatile' instead of 'volatile load', which is archaic. llvm-svn: 145171
2011-11-27 07:54:59 +01:00
store volatile i32 0, i32* @XYZ
store volatile i32 1, i32* @XYZ
When optimizing for size, don't tail-merge unless it's likely to be a code-size win, and not when it's only likely to be code-size neutral, such as when only a single instruction would be eliminated and a new branch would be required. This fixes rdar://7392894. llvm-svn: 88692
2009-11-13 22:02:15 +01:00
unreachable
return:
ret void
}
; two_nosize - Same as two, but without the optsize attribute.
; Now two instructions are enough to be tail-duplicated.
; CHECK: two_nosize:
; CHECK: movl $0, XYZ(%rip)
; CHECK: movl $1, XYZ(%rip)
; CHECK: movl $0, XYZ(%rip)
; CHECK: movl $1, XYZ(%rip)
define void @two_nosize() nounwind {
entry:
%0 = icmp eq i32 undef, 0
br i1 %0, label %bbx, label %bby
bby:
switch i32 undef, label %bb7 [
i32 16, label %return
]
bb7:
Upgrade syntax of tests using volatile instructions to use 'load volatile' instead of 'volatile load', which is archaic. llvm-svn: 145171
2011-11-27 07:54:59 +01:00
store volatile i32 0, i32* @XYZ
store volatile i32 1, i32* @XYZ
When optimizing for size, don't tail-merge unless it's likely to be a code-size win, and not when it's only likely to be code-size neutral, such as when only a single instruction would be eliminated and a new branch would be required. This fixes rdar://7392894. llvm-svn: 88692
2009-11-13 22:02:15 +01:00
unreachable
bbx:
switch i32 undef, label %bb12 [
i32 128, label %return
]
bb12:
Upgrade syntax of tests using volatile instructions to use 'load volatile' instead of 'volatile load', which is archaic. llvm-svn: 145171
2011-11-27 07:54:59 +01:00
store volatile i32 0, i32* @XYZ
store volatile i32 1, i32* @XYZ
When optimizing for size, don't tail-merge unless it's likely to be a code-size win, and not when it's only likely to be code-size neutral, such as when only a single instruction would be eliminated and a new branch would be required. This fixes rdar://7392894. llvm-svn: 88692
2009-11-13 22:02:15 +01:00
unreachable
return:
ret void
}
Fix a bug which prevented tail merging of return instructions in beneficial cases. See the changes in test/CodeGen/X86/tail-opts.ll and test/CodeGen/ARM/ifcvt2.ll for details. The fix is to change HashEndOfMBB to hash at most one instruction, instead of trying to apply heuristics about when it will be profitable to consider more than one instruction. The regular tail-merging heuristics are already prepared to handle the same cases, and they're more precise. Also, make test/CodeGen/ARM/ifcvt5.ll and test/CodeGen/Thumb2/thumb2-branch.ll slightly more complex so that they continue to test what they're intended to test. And, this eliminates the problem in test/CodeGen/Thumb2/2009-10-15-ITBlockBranch.ll, the testcase from PR5204. Update it accordingly. llvm-svn: 102907
2010-05-03 16:35:47 +02:00
; Tail-merging should merge the two ret instructions since one side
; can fall-through into the ret and the other side has to branch anyway.
; CHECK: TESTE:
Fix a batch of x86 tests to be coalescer independent. Most of these tests require a single mov instruction that can come either before or after a 2-addr instruction. -join-physregs changes the behavior, but the results are equivalent. llvm-svn: 130891
2011-05-05 01:54:51 +02:00
; CHECK: ret
; CHECK-NOT: ret
; CHECK: size TESTE
Fix a bug which prevented tail merging of return instructions in beneficial cases. See the changes in test/CodeGen/X86/tail-opts.ll and test/CodeGen/ARM/ifcvt2.ll for details. The fix is to change HashEndOfMBB to hash at most one instruction, instead of trying to apply heuristics about when it will be profitable to consider more than one instruction. The regular tail-merging heuristics are already prepared to handle the same cases, and they're more precise. Also, make test/CodeGen/ARM/ifcvt5.ll and test/CodeGen/Thumb2/thumb2-branch.ll slightly more complex so that they continue to test what they're intended to test. And, this eliminates the problem in test/CodeGen/Thumb2/2009-10-15-ITBlockBranch.ll, the testcase from PR5204. Update it accordingly. llvm-svn: 102907
2010-05-03 16:35:47 +02:00
define i64 @TESTE(i64 %parami, i64 %paraml) nounwind readnone {
entry:
%cmp = icmp slt i64 %parami, 1 ; <i1> [#uses=1]
%varx.0 = select i1 %cmp, i64 1, i64 %parami ; <i64> [#uses=1]
%cmp410 = icmp slt i64 %paraml, 1 ; <i1> [#uses=1]
br i1 %cmp410, label %for.end, label %bb.nph
bb.nph: ; preds = %entry
%tmp15 = mul i64 %paraml, %parami ; <i64> [#uses=1]
ret i64 %tmp15
for.end: ; preds = %entry
ret i64 %varx.0
}
Reference in New Issue Copy Permalink