maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/llvm/test/CodeGen/X86/win64_alloca_dynalloca.ll

139 lines
3.2 KiB
LLVM
Raw Normal View History

Enable MI Sched for x86. This changes the SelectionDAG scheduling preference to source order. Soon, the SelectionDAG scheduler can be bypassed saving a nice chunk of compile time. Performance differences that result from this change are often a consequence of register coalescing. The register coalescer is far from perfect. Bugs can be filed for deficiencies. On x86 SandyBridge/Haswell, the source order schedule is often preserved, particularly for small blocks. Register pressure is generally improved over the SD scheduler's ILP mode. However, we are still able to handle large blocks that require latency hiding, unlike the SD scheduler's BURR mode. MI scheduler also attempts to discover the critical path in single-block loops and adjust heuristics accordingly. The MI scheduler relies on the new machine model. This is currently unimplemented for AVX, so we may not be generating the best code yet. Unit tests are updated so they don't depend on SD scheduling heuristics. llvm-svn: 192750
2013-10-16 07:33:07 +08:00
; RUN: llc < %s -mcpu=generic -enable-misched=false -mtriple=x86_64-mingw32 | FileCheck %s -check-prefix=M64
; RUN: llc < %s -mcpu=generic -enable-misched=false -mtriple=x86_64-win32 | FileCheck %s -check-prefix=W64
x86: Fix large model calls to __chkstk for dynamic allocas In the large code model, we now put __chkstk in %r11 before calling it. Refactor the code so that we only do this once. Simplify things by using __chkstk_ms instead of __chkstk on cygming. We already use that symbol in the prolog emission, and it simplifies our logic. Second half of PR18582. llvm-svn: 227519
2015-01-30 07:58:04 +08:00
; RUN: llc < %s -mcpu=generic -enable-misched=false -mtriple=x86_64-win32 -code-model=large | FileCheck %s -check-prefix=L64
Enable MI Sched for x86. This changes the SelectionDAG scheduling preference to source order. Soon, the SelectionDAG scheduler can be bypassed saving a nice chunk of compile time. Performance differences that result from this change are often a consequence of register coalescing. The register coalescer is far from perfect. Bugs can be filed for deficiencies. On x86 SandyBridge/Haswell, the source order schedule is often preserved, particularly for small blocks. Register pressure is generally improved over the SD scheduler's ILP mode. However, we are still able to handle large blocks that require latency hiding, unlike the SD scheduler's BURR mode. MI scheduler also attempts to discover the critical path in single-block loops and adjust heuristics accordingly. The MI scheduler relies on the new machine model. This is currently unimplemented for AVX, so we may not be generating the best code yet. Unit tests are updated so they don't depend on SD scheduling heuristics. llvm-svn: 192750
2013-10-16 07:33:07 +08:00
; RUN: llc < %s -mcpu=generic -enable-misched=false -mtriple=x86_64-win32-macho | FileCheck %s -check-prefix=EFI
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
; PR8777
; PR8778
Fixed a bug in dynamic allocation memory on stack. The alignment of allocated space was wrong, see Bugzila 17345. Done by Zvi Rackover <zvi.rackover@intel.com>. llvm-svn: 192573
2013-10-14 15:26:51 +08:00
define i64 @unaligned(i64 %n, i64 %x) nounwind {
; M64-LABEL: unaligned:
; W64-LABEL: unaligned:
; EFI-LABEL: unaligned:
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
entry:
%buf0 = alloca i8, i64 4096, align 1
Change stack probing code for MingW. Since gcc 4.6 the compiler uses ___chkstk_ms which has the same semantics as the MS CRT function __chkstk. This simplifies the prologue generation a bit. Reviewed by Rafael Espíndola. llvm-svn: 197205
2013-12-13 13:37:05 +08:00
; ___chkstk_ms does not adjust %rsp.
X86: Use a smaller 'mov' instruction for stack probe calls Prologue emission, in some cases, requires calls to a stack probe helper function. The amount of stack to probe is passed as a register argument in the Win64 ABI but the instruction sequence used is pessimistic: it assumes that the number of bytes to probe is greater than 4 GB. Instead, select a more appropriate opcode depending on the number of bytes we are going to probe. llvm-svn: 230270
2015-02-24 05:50:30 +08:00
; M64: $4096, %eax
Change stack probing code for MingW. Since gcc 4.6 the compiler uses ___chkstk_ms which has the same semantics as the MS CRT function __chkstk. This simplifies the prologue generation a bit. Reviewed by Rafael Espíndola. llvm-svn: 197205
2013-12-13 13:37:05 +08:00
; M64: callq ___chkstk_ms
; M64: subq %rax, %rsp
X86: Emit an ABI compliant prologue and epilogue for Win64 Win64 has specific contraints on what valid prologues and epilogues look like. This constraint is born from the flexibility and descriptiveness of Win64's unwind opcodes. Prologues previously emitted by LLVM could not be represented by the unwind opcodes, preventing operations powered by stack unwinding to successfully work. Differential Revision: http://reviews.llvm.org/D7520 llvm-svn: 228641
2015-02-10 08:57:42 +08:00
; M64: leaq 128(%rsp), %rbp
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
; __chkstk does not adjust %rsp.
X86: Use a smaller 'mov' instruction for stack probe calls Prologue emission, in some cases, requires calls to a stack probe helper function. The amount of stack to probe is passed as a register argument in the Win64 ABI but the instruction sequence used is pessimistic: it assumes that the number of bytes to probe is greater than 4 GB. Instead, select a more appropriate opcode depending on the number of bytes we are going to probe. llvm-svn: 230270
2015-02-24 05:50:30 +08:00
; W64: $4096, %eax
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
; W64: callq __chkstk
Reuse %rax after calling __chkstk on win64 Reapply this as I reverted the wrong commit. llvm-svn: 185807
2013-07-08 19:20:11 +08:00
; W64: subq %rax, %rsp
X86: Emit an ABI compliant prologue and epilogue for Win64 Win64 has specific contraints on what valid prologues and epilogues look like. This constraint is born from the flexibility and descriptiveness of Win64's unwind opcodes. Prologues previously emitted by LLVM could not be represented by the unwind opcodes, preventing operations powered by stack unwinding to successfully work. Differential Revision: http://reviews.llvm.org/D7520 llvm-svn: 228641
2015-02-10 08:57:42 +08:00
; W64: leaq 128(%rsp), %rbp
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
x86: Fix large model calls to __chkstk for dynamic allocas In the large code model, we now put __chkstk in %r11 before calling it. Refactor the code so that we only do this once. Simplify things by using __chkstk_ms instead of __chkstk on cygming. We already use that symbol in the prolog emission, and it simplifies our logic. Second half of PR18582. llvm-svn: 227519
2015-01-30 07:58:04 +08:00
; Use %r11 for the large model.
X86: Use a smaller 'mov' instruction for stack probe calls Prologue emission, in some cases, requires calls to a stack probe helper function. The amount of stack to probe is passed as a register argument in the Win64 ABI but the instruction sequence used is pessimistic: it assumes that the number of bytes to probe is greater than 4 GB. Instead, select a more appropriate opcode depending on the number of bytes we are going to probe. llvm-svn: 230270
2015-02-24 05:50:30 +08:00
; L64: $4096, %eax
x86: Fix large model calls to __chkstk for dynamic allocas In the large code model, we now put __chkstk in %r11 before calling it. Refactor the code so that we only do this once. Simplify things by using __chkstk_ms instead of __chkstk on cygming. We already use that symbol in the prolog emission, and it simplifies our logic. Second half of PR18582. llvm-svn: 227519
2015-01-30 07:58:04 +08:00
; L64: movabsq $__chkstk, %r11
; L64: callq *%r11
; L64: subq %rax, %rsp
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
; Freestanding
; EFI: $[[B0OFS:4096|4104]], %rsp
; EFI-NOT: call
%buf1 = alloca i8, i64 %n, align 1
Remove -join-physregs from the test suite. This option has been disabled for a while, and it is going away so I can clean up the coalescer code. The tests that required physreg joining to be enabled were almost all of the form "tiny function with interference between arguments and return value". Such functions are usually inlined in the real world. The problem exposed by phys_subreg_coalesce-3.ll is real, but fairly rare. llvm-svn: 157027
2012-05-18 07:44:19 +08:00
; M64: leaq 15(%{{.*}}), %rax
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
; M64: andq $-16, %rax
x86: Fix large model calls to __chkstk for dynamic allocas In the large code model, we now put __chkstk in %r11 before calling it. Refactor the code so that we only do this once. Simplify things by using __chkstk_ms instead of __chkstk on cygming. We already use that symbol in the prolog emission, and it simplifies our logic. Second half of PR18582. llvm-svn: 227519
2015-01-30 07:58:04 +08:00
; M64: callq ___chkstk_ms
; M64: subq %rax, %rsp
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
; M64: movq %rsp, %rax
Remove -join-physregs from the test suite. This option has been disabled for a while, and it is going away so I can clean up the coalescer code. The tests that required physreg joining to be enabled were almost all of the form "tiny function with interference between arguments and return value". Such functions are usually inlined in the real world. The problem exposed by phys_subreg_coalesce-3.ll is real, but fairly rare. llvm-svn: 157027
2012-05-18 07:44:19 +08:00
; W64: leaq 15(%{{.*}}), %rax
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
; W64: andq $-16, %rax
; W64: callq __chkstk
; W64: subq %rax, %rsp
; W64: movq %rsp, %rax
x86: Fix large model calls to __chkstk for dynamic allocas In the large code model, we now put __chkstk in %r11 before calling it. Refactor the code so that we only do this once. Simplify things by using __chkstk_ms instead of __chkstk on cygming. We already use that symbol in the prolog emission, and it simplifies our logic. Second half of PR18582. llvm-svn: 227519
2015-01-30 07:58:04 +08:00
; L64: leaq 15(%{{.*}}), %rax
; L64: andq $-16, %rax
; L64: movabsq $__chkstk, %r11
; L64: callq *%r11
; L64: subq %rax, %rsp
; L64: movq %rsp, %rax
Remove -join-physregs from the test suite. This option has been disabled for a while, and it is going away so I can clean up the coalescer code. The tests that required physreg joining to be enabled were almost all of the form "tiny function with interference between arguments and return value". Such functions are usually inlined in the real world. The problem exposed by phys_subreg_coalesce-3.ll is real, but fairly rare. llvm-svn: 157027
2012-05-18 07:44:19 +08:00
; EFI: leaq 15(%{{.*}}), [[R1:%r.*]]
Fix register-dependent X86 tests. llvm-svn: 128867
2011-04-05 08:32:44 +08:00
; EFI: andq $-16, [[R1]]
Revert "Temporarily enable MI-Sched on X86." This reverts commit 98a9b72e8c56dc13a2617de84503a3d78352789c. llvm-svn: 184823
2013-06-25 10:48:58 +08:00
; EFI: movq %rsp, [[R64:%r.*]]
Fix register-dependent X86 tests. llvm-svn: 128867
2011-04-05 08:32:44 +08:00
; EFI: subq [[R1]], [[R64]]
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
; EFI: movq [[R64]], %rsp
%r = call i64 @bar(i64 %n, i64 %x, i64 %n, i8* %buf0, i8* %buf1) nounwind
; M64: subq $48, %rsp
Revert "Temporarily enable MI-Sched on X86." This reverts commit 98a9b72e8c56dc13a2617de84503a3d78352789c. llvm-svn: 184823
2013-06-25 10:48:58 +08:00
; M64: movq %rax, 32(%rsp)
X86: Emit an ABI compliant prologue and epilogue for Win64 Win64 has specific contraints on what valid prologues and epilogues look like. This constraint is born from the flexibility and descriptiveness of Win64's unwind opcodes. Prologues previously emitted by LLVM could not be represented by the unwind opcodes, preventing operations powered by stack unwinding to successfully work. Differential Revision: http://reviews.llvm.org/D7520 llvm-svn: 228641
2015-02-10 08:57:42 +08:00
; M64: leaq -128(%rbp), %r9
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
; M64: callq bar
; W64: subq $48, %rsp
Revert "Temporarily enable MI-Sched on X86." This reverts commit 98a9b72e8c56dc13a2617de84503a3d78352789c. llvm-svn: 184823
2013-06-25 10:48:58 +08:00
; W64: movq %rax, 32(%rsp)
X86: Emit an ABI compliant prologue and epilogue for Win64 Win64 has specific contraints on what valid prologues and epilogues look like. This constraint is born from the flexibility and descriptiveness of Win64's unwind opcodes. Prologues previously emitted by LLVM could not be represented by the unwind opcodes, preventing operations powered by stack unwinding to successfully work. Differential Revision: http://reviews.llvm.org/D7520 llvm-svn: 228641
2015-02-10 08:57:42 +08:00
; W64: leaq -128(%rbp), %r9
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
; W64: callq bar
; EFI: subq $48, %rsp
Revert "Temporarily enable MI-Sched on X86." This reverts commit 98a9b72e8c56dc13a2617de84503a3d78352789c. llvm-svn: 184823
2013-06-25 10:48:58 +08:00
; EFI: movq [[R64]], 32(%rsp)
Enable MI Sched for x86. This changes the SelectionDAG scheduling preference to source order. Soon, the SelectionDAG scheduler can be bypassed saving a nice chunk of compile time. Performance differences that result from this change are often a consequence of register coalescing. The register coalescer is far from perfect. Bugs can be filed for deficiencies. On x86 SandyBridge/Haswell, the source order schedule is often preserved, particularly for small blocks. Register pressure is generally improved over the SD scheduler's ILP mode. However, we are still able to handle large blocks that require latency hiding, unlike the SD scheduler's BURR mode. MI scheduler also attempts to discover the critical path in single-block loops and adjust heuristics accordingly. The MI scheduler relies on the new machine model. This is currently unimplemented for AVX, so we may not be generating the best code yet. Unit tests are updated so they don't depend on SD scheduling heuristics. llvm-svn: 192750
2013-10-16 07:33:07 +08:00
; EFI: leaq -[[B0OFS]](%rbp), %r9
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
; EFI: callq _bar
ret i64 %r
X86: Emit an ABI compliant prologue and epilogue for Win64 Win64 has specific contraints on what valid prologues and epilogues look like. This constraint is born from the flexibility and descriptiveness of Win64's unwind opcodes. Prologues previously emitted by LLVM could not be represented by the unwind opcodes, preventing operations powered by stack unwinding to successfully work. Differential Revision: http://reviews.llvm.org/D7520 llvm-svn: 228641
2015-02-10 08:57:42 +08:00
; M64: leaq 3968(%rbp), %rsp
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
X86: Emit an ABI compliant prologue and epilogue for Win64 Win64 has specific contraints on what valid prologues and epilogues look like. This constraint is born from the flexibility and descriptiveness of Win64's unwind opcodes. Prologues previously emitted by LLVM could not be represented by the unwind opcodes, preventing operations powered by stack unwinding to successfully work. Differential Revision: http://reviews.llvm.org/D7520 llvm-svn: 228641
2015-02-10 08:57:42 +08:00
; W64: leaq 3968(%rbp), %rsp
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
}
Fixed a bug in dynamic allocation memory on stack. The alignment of allocated space was wrong, see Bugzila 17345. Done by Zvi Rackover <zvi.rackover@intel.com>. llvm-svn: 192573
2013-10-14 15:26:51 +08:00
define i64 @aligned(i64 %n, i64 %x) nounwind {
; M64-LABEL: aligned:
; W64-LABEL: aligned:
; EFI-LABEL: aligned:
entry:
%buf1 = alloca i8, i64 %n, align 128
; M64: leaq 15(%{{.*}}), %rax
; M64: andq $-16, %rax
x86: Fix large model calls to __chkstk for dynamic allocas In the large code model, we now put __chkstk in %r11 before calling it. Refactor the code so that we only do this once. Simplify things by using __chkstk_ms instead of __chkstk on cygming. We already use that symbol in the prolog emission, and it simplifies our logic. Second half of PR18582. llvm-svn: 227519
2015-01-30 07:58:04 +08:00
; M64: callq ___chkstk_ms
; M64: subq %rax, %rsp
Fixed a bug in dynamic allocation memory on stack. The alignment of allocated space was wrong, see Bugzila 17345. Done by Zvi Rackover <zvi.rackover@intel.com>. llvm-svn: 192573
2013-10-14 15:26:51 +08:00
; M64: movq %rsp, [[R2:%r.*]]
; M64: andq $-128, [[R2]]
; M64: movq [[R2]], %rsp
; W64: leaq 15(%{{.*}}), %rax
; W64: andq $-16, %rax
; W64: callq __chkstk
; W64: subq %rax, %rsp
; W64: movq %rsp, [[R2:%r.*]]
; W64: andq $-128, [[R2]]
; W64: movq [[R2]], %rsp
; EFI: leaq 15(%{{.*}}), [[R1:%r.*]]
; EFI: andq $-16, [[R1]]
; EFI: movq %rsp, [[R64:%r.*]]
; EFI: subq [[R1]], [[R64]]
; EFI: andq $-128, [[R64]]
; EFI: movq [[R64]], %rsp
%r = call i64 @bar(i64 %n, i64 %x, i64 %n, i8* undef, i8* %buf1) nounwind
; M64: subq $48, %rsp
; M64: movq [[R2]], 32(%rsp)
; M64: callq bar
; W64: subq $48, %rsp
; W64: movq [[R2]], 32(%rsp)
; W64: callq bar
; EFI: subq $48, %rsp
; EFI: movq [[R64]], 32(%rsp)
; EFI: callq _bar
ret i64 %r
}
Target/X86: [PR8777][PR8778] Tweak alloca/chkstk for Windows targets. FIXME: Some cleanups would be needed. llvm-svn: 128206
2011-03-24 15:07:00 +08:00
declare i64 @bar(i64, i64, i64, i8* nocapture, i8* nocapture) nounwind