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/WebAssembly/byval.ll

133 lines
5.5 KiB
LLVM
Raw Normal View History

[WebAssembly] Optimize away return instructions using fallthroughs. This saves a small amount of code size, and is a first small step toward passing values on the stack across block boundaries. Differential Review: http://reviews.llvm.org/D20450 llvm-svn: 270294
2016-05-21 08:21:56 +08:00
; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -verify-machineinstrs | FileCheck %s
; RUN: llc < %s -asm-verbose=false -disable-wasm-fallthrough-return-opt -verify-machineinstrs -fast-isel | FileCheck %s
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
target datalayout = "e-m:e-p:32:32-i64:64-n32:64-S128"
target triple = "wasm32-unknown-unknown"
%SmallStruct = type { i32 }
%OddStruct = type { i32, i8, i32 }
%AlignedStruct = type { double, double }
%BigStruct = type { double, double, double, double, double, double, double, double, double, double, double, i8, i8, i8 }
[WebAssembly] Fix byval for empty types. llvm-svn: 260740
2016-02-13 05:30:18 +08:00
%EmptyStruct = type { }
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
%BigArray = type { [33 x i8] }
declare void @ext_func(%SmallStruct*)
[WebAssembly] Fix byval for empty types. llvm-svn: 260740
2016-02-13 05:30:18 +08:00
declare void @ext_func_empty(%EmptyStruct* byval)
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
declare void @ext_byval_func(%SmallStruct* byval)
declare void @ext_byval_func_align8(%SmallStruct* byval align 8)
declare void @ext_byval_func_alignedstruct(%AlignedStruct* byval)
declare void @ext_byval_func_bigarray(%BigArray* byval)
[WebAssembly] Fix byval for empty types. llvm-svn: 260740
2016-02-13 05:30:18 +08:00
declare void @ext_byval_func_empty(%EmptyStruct* byval)
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
; CHECK-LABEL: byval_arg
define void @byval_arg(%SmallStruct* %ptr) {
; CHECK: .param i32
[WebAssembly] Put __stack_pointer in the offset field of loads and stores. Instead of this: i32.const $push10=, __stack_pointer i32.load $push11=, 0($pop10) Emit this: i32.const $push10=, 0 i32.load $push11=, __stack_pointer($pop10) It's not currently clear which is better, though there's a chance the second form may be better at overall compression. We can revisit this when we have more data; for now it makes sense to make PEI consistent with isel. Differential Revision: http://reviews.llvm.org/D20411 llvm-svn: 270635
2016-05-25 07:47:41 +08:00
; CHECK: i32.const $push[[L4:.+]]=, 0
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
; Subtract 16 from SP (SP is 16-byte aligned)
[WebAssembly] Put __stack_pointer in the offset field of loads and stores. Instead of this: i32.const $push10=, __stack_pointer i32.load $push11=, 0($pop10) Emit this: i32.const $push10=, 0 i32.load $push11=, __stack_pointer($pop10) It's not currently clear which is better, though there's a chance the second form may be better at overall compression. We can revisit this when we have more data; for now it makes sense to make PEI consistent with isel. Differential Revision: http://reviews.llvm.org/D20411 llvm-svn: 270635
2016-05-25 07:47:41 +08:00
; CHECK: i32.const $push[[L1:.+]]=, 0
; CHECK-NEXT: i32.load $push[[L2:.+]]=, __stack_pointer($pop[[L1]])
[WebAssembly] Stackify function prologs and epilogs The instructions are the same, but fewer locals are used. Differential Revision: http://reviews.llvm.org/D17428 llvm-svn: 261452
2016-02-21 05:46:50 +08:00
; CHECK-NEXT: i32.const $push[[L3:.+]]=, 16
[WebAssembly] Disable the store-results optimization. The WebAssemly spec removing the return value from store instructions, so remove the associated optimization from LLVM. This patch leaves the store instruction operands in place for now, so stores now always write to "$drop"; these will be removed in a seperate patch. llvm-svn: 279100
2016-08-19 01:51:27 +08:00
; CHECK-NEXT: i32.sub $push[[L11:.+]]=, $pop[[L2]], $pop[[L3]]
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
; Ensure SP is stored back before the call
[WebAssembly] Disable the store-results optimization. The WebAssemly spec removing the return value from store instructions, so remove the associated optimization from LLVM. This patch leaves the store instruction operands in place for now, so stores now always write to "$drop"; these will be removed in a seperate patch. llvm-svn: 279100
2016-08-19 01:51:27 +08:00
; CHECK-NEXT: tee_local $push[[L10:.+]]=, $[[SP:.+]]=, $pop[[L11]]{{$}}
[WebAssembly] Remove the output operand from stores. Per spec changes, store instructions in WebAssembly no longer have a return value. Update the instruction descriptions. Differential Revision: https://reviews.llvm.org/D25122 llvm-svn: 283501
2016-10-07 06:08:28 +08:00
; CHECK-NEXT: i32.store __stack_pointer($pop[[L4]]), $pop[[L10]]{{$}}
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
; Copy the SmallStruct argument to the stack (SP+12, original SP-4)
[WebAssembly] Move register stackification and coloring to a late phase. Move the register stackification and coloring passes to run very late, after PEI, tail duplication, and most other passes. This means that all code emitted and expanded by those passes is now exposed to these passes. This also eliminates the need for prologue/epilogue code to be manually stackified, which significantly simplifies the code. This does require running LiveIntervals a second time. It's useful to think of these late passes not as late optimization passes, but as a domain-specific compression algorithm based on knowledge of liveness information. It's used to compress the code after all conventional optimizations are complete, which is why it uses LiveIntervals at a phase when actual optimization passes don't typically need it. Differential Revision: http://reviews.llvm.org/D20075 llvm-svn: 269012
2016-05-10 12:24:02 +08:00
; CHECK-NEXT: i32.load $push[[L0:.+]]=, 0($0)
[WebAssembly] Remove the output operand from stores. Per spec changes, store instructions in WebAssembly no longer have a return value. Update the instruction descriptions. Differential Revision: https://reviews.llvm.org/D25122 llvm-svn: 283501
2016-10-07 06:08:28 +08:00
; CHECK-NEXT: i32.store 12($[[SP]]), $pop[[L0]]
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
; Pass a pointer to the stack slot to the function
[WebAssembly] Move register stackification and coloring to a late phase. Move the register stackification and coloring passes to run very late, after PEI, tail duplication, and most other passes. This means that all code emitted and expanded by those passes is now exposed to these passes. This also eliminates the need for prologue/epilogue code to be manually stackified, which significantly simplifies the code. This does require running LiveIntervals a second time. It's useful to think of these late passes not as late optimization passes, but as a domain-specific compression algorithm based on knowledge of liveness information. It's used to compress the code after all conventional optimizations are complete, which is why it uses LiveIntervals at a phase when actual optimization passes don't typically need it. Differential Revision: http://reviews.llvm.org/D20075 llvm-svn: 269012
2016-05-10 12:24:02 +08:00
; CHECK-NEXT: i32.const $push[[L5:.+]]=, 12{{$}}
; CHECK-NEXT: i32.add $push[[ARG:.+]]=, $[[SP]], $pop[[L5]]{{$}}
; CHECK-NEXT: call ext_byval_func@FUNCTION, $pop[[ARG]]{{$}}
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
call void @ext_byval_func(%SmallStruct* byval %ptr)
; Restore the stack
[WebAssembly] Put __stack_pointer in the offset field of loads and stores. Instead of this: i32.const $push10=, __stack_pointer i32.load $push11=, 0($pop10) Emit this: i32.const $push10=, 0 i32.load $push11=, __stack_pointer($pop10) It's not currently clear which is better, though there's a chance the second form may be better at overall compression. We can revisit this when we have more data; for now it makes sense to make PEI consistent with isel. Differential Revision: http://reviews.llvm.org/D20411 llvm-svn: 270635
2016-05-25 07:47:41 +08:00
; CHECK-NEXT: i32.const $push[[L7:.+]]=, 0
[WebAssembly] Stackify code emitted by eliminateFrameIndex and SP writeback Summary: MRI::eliminateFrameIndex can emit several instructions to do address calculations; these can usually be stackified. Because instructions with FI operands can have subsequent operands which may be expression trees, find the top of the leftmost tree and insert the code before it, to keep the LIFO property. Also use stackified registers when writing back the SP value to memory in the epilog; it's unnecessary because SP will not be used after the epilog, and it results in better code. Differential Revision: http://reviews.llvm.org/D18234 llvm-svn: 263725
2016-03-18 01:00:29 +08:00
; CHECK-NEXT: i32.const $push[[L6:.+]]=, 16
[WebAssembly] Move register stackification and coloring to a late phase. Move the register stackification and coloring passes to run very late, after PEI, tail duplication, and most other passes. This means that all code emitted and expanded by those passes is now exposed to these passes. This also eliminates the need for prologue/epilogue code to be manually stackified, which significantly simplifies the code. This does require running LiveIntervals a second time. It's useful to think of these late passes not as late optimization passes, but as a domain-specific compression algorithm based on knowledge of liveness information. It's used to compress the code after all conventional optimizations are complete, which is why it uses LiveIntervals at a phase when actual optimization passes don't typically need it. Differential Revision: http://reviews.llvm.org/D20075 llvm-svn: 269012
2016-05-10 12:24:02 +08:00
; CHECK-NEXT: i32.add $push[[L8:.+]]=, $[[SP]], $pop[[L6]]
[WebAssembly] Remove the output operand from stores. Per spec changes, store instructions in WebAssembly no longer have a return value. Update the instruction descriptions. Differential Revision: https://reviews.llvm.org/D25122 llvm-svn: 283501
2016-10-07 06:08:28 +08:00
; CHECK-NEXT: i32.store __stack_pointer($pop[[L7]]), $pop[[L8]]
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
; CHECK-NEXT: return
ret void
}
; CHECK-LABEL: byval_arg_align8
define void @byval_arg_align8(%SmallStruct* %ptr) {
; CHECK: .param i32
; Don't check the entire SP sequence, just enough to get the alignment.
[WebAssembly] Stackify function prologs and epilogs The instructions are the same, but fewer locals are used. Differential Revision: http://reviews.llvm.org/D17428 llvm-svn: 261452
2016-02-21 05:46:50 +08:00
; CHECK: i32.const $push[[L1:.+]]=, 16
[WebAssembly] Disable the store-results optimization. The WebAssemly spec removing the return value from store instructions, so remove the associated optimization from LLVM. This patch leaves the store instruction operands in place for now, so stores now always write to "$drop"; these will be removed in a seperate patch. llvm-svn: 279100
2016-08-19 01:51:27 +08:00
; CHECK-NEXT: i32.sub $push[[L11:.+]]=, {{.+}}, $pop[[L1]]
; CHECK-NEXT: tee_local $push[[L10:.+]]=, $[[SP:.+]]=, $pop[[L11]]{{$}}
[WebAssembly] Remove the output operand from stores. Per spec changes, store instructions in WebAssembly no longer have a return value. Update the instruction descriptions. Differential Revision: https://reviews.llvm.org/D25122 llvm-svn: 283501
2016-10-07 06:08:28 +08:00
; CHECK-NEXT: i32.store __stack_pointer($pop{{.+}}), $pop[[L10]]{{$}}
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
; Copy the SmallStruct argument to the stack (SP+8, original SP-8)
[WebAssembly] Move register stackification and coloring to a late phase. Move the register stackification and coloring passes to run very late, after PEI, tail duplication, and most other passes. This means that all code emitted and expanded by those passes is now exposed to these passes. This also eliminates the need for prologue/epilogue code to be manually stackified, which significantly simplifies the code. This does require running LiveIntervals a second time. It's useful to think of these late passes not as late optimization passes, but as a domain-specific compression algorithm based on knowledge of liveness information. It's used to compress the code after all conventional optimizations are complete, which is why it uses LiveIntervals at a phase when actual optimization passes don't typically need it. Differential Revision: http://reviews.llvm.org/D20075 llvm-svn: 269012
2016-05-10 12:24:02 +08:00
; CHECK-NEXT: i32.load $push[[L0:.+]]=, 0($0){{$}}
[WebAssembly] Remove the output operand from stores. Per spec changes, store instructions in WebAssembly no longer have a return value. Update the instruction descriptions. Differential Revision: https://reviews.llvm.org/D25122 llvm-svn: 283501
2016-10-07 06:08:28 +08:00
; CHECK-NEXT: i32.store 8($[[SP]]), $pop[[L0]]{{$}}
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
; Pass a pointer to the stack slot to the function
[WebAssembly] Move register stackification and coloring to a late phase. Move the register stackification and coloring passes to run very late, after PEI, tail duplication, and most other passes. This means that all code emitted and expanded by those passes is now exposed to these passes. This also eliminates the need for prologue/epilogue code to be manually stackified, which significantly simplifies the code. This does require running LiveIntervals a second time. It's useful to think of these late passes not as late optimization passes, but as a domain-specific compression algorithm based on knowledge of liveness information. It's used to compress the code after all conventional optimizations are complete, which is why it uses LiveIntervals at a phase when actual optimization passes don't typically need it. Differential Revision: http://reviews.llvm.org/D20075 llvm-svn: 269012
2016-05-10 12:24:02 +08:00
; CHECK-NEXT: i32.const $push[[L5:.+]]=, 8{{$}}
; CHECK-NEXT: i32.add $push[[ARG:.+]]=, $[[SP]], $pop[[L5]]{{$}}
; CHECK-NEXT: call ext_byval_func_align8@FUNCTION, $pop[[ARG]]{{$}}
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
call void @ext_byval_func_align8(%SmallStruct* byval align 8 %ptr)
ret void
}
; CHECK-LABEL: byval_arg_double
define void @byval_arg_double(%AlignedStruct* %ptr) {
; CHECK: .param i32
; Subtract 16 from SP (SP is 16-byte aligned)
[WebAssembly] Stackify function prologs and epilogs The instructions are the same, but fewer locals are used. Differential Revision: http://reviews.llvm.org/D17428 llvm-svn: 261452
2016-02-21 05:46:50 +08:00
; CHECK: i32.const $push[[L1:.+]]=, 16
[WebAssembly] Disable the store-results optimization. The WebAssemly spec removing the return value from store instructions, so remove the associated optimization from LLVM. This patch leaves the store instruction operands in place for now, so stores now always write to "$drop"; these will be removed in a seperate patch. llvm-svn: 279100
2016-08-19 01:51:27 +08:00
; CHECK-NEXT: i32.sub $push[[L14:.+]]=, {{.+}}, $pop[[L1]]
; CHECK-NEXT: tee_local $push[[L13:.+]]=, $[[SP:.+]]=, $pop[[L14]]
[WebAssembly] Remove the output operand from stores. Per spec changes, store instructions in WebAssembly no longer have a return value. Update the instruction descriptions. Differential Revision: https://reviews.llvm.org/D25122 llvm-svn: 283501
2016-10-07 06:08:28 +08:00
; CHECK-NEXT: i32.store {{.+}}, $pop[[L13]]
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
; Copy the AlignedStruct argument to the stack (SP+0, original SP-16)
; Just check the last load/store pair of the memcpy
; CHECK: i64.load $push[[L4:.+]]=, 0($0)
[WebAssembly] Remove the output operand from stores. Per spec changes, store instructions in WebAssembly no longer have a return value. Update the instruction descriptions. Differential Revision: https://reviews.llvm.org/D25122 llvm-svn: 283501
2016-10-07 06:08:28 +08:00
; CHECK-NEXT: i64.store 0($[[SP]]), $pop[[L4]]
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
; Pass a pointer to the stack slot to the function
[WebAssembly] Move register stackification and coloring to a late phase. Move the register stackification and coloring passes to run very late, after PEI, tail duplication, and most other passes. This means that all code emitted and expanded by those passes is now exposed to these passes. This also eliminates the need for prologue/epilogue code to be manually stackified, which significantly simplifies the code. This does require running LiveIntervals a second time. It's useful to think of these late passes not as late optimization passes, but as a domain-specific compression algorithm based on knowledge of liveness information. It's used to compress the code after all conventional optimizations are complete, which is why it uses LiveIntervals at a phase when actual optimization passes don't typically need it. Differential Revision: http://reviews.llvm.org/D20075 llvm-svn: 269012
2016-05-10 12:24:02 +08:00
; CHECK-NEXT: call ext_byval_func_alignedstruct@FUNCTION, $[[SP]]
[WebAssembly] Implement byval arguments Summary: Just does the simple allocation of a stack object and passes a pointer to the callee. Differential Revision: http://reviews.llvm.org/D16610 llvm-svn: 258989
2016-01-28 05:17:39 +08:00
tail call void @ext_byval_func_alignedstruct(%AlignedStruct* byval %ptr)
ret void
}
; CHECK-LABEL: byval_param
define void @byval_param(%SmallStruct* byval align 32 %ptr) {
; CHECK: .param i32
; %ptr is just a pointer to a struct, so pass it directly through
; CHECK: call ext_func@FUNCTION, $0
call void @ext_func(%SmallStruct* %ptr)
ret void
}
[WebAssembly] Fix byval for empty types. llvm-svn: 260740
2016-02-13 05:30:18 +08:00
; CHECK-LABEL: byval_empty_caller
define void @byval_empty_caller(%EmptyStruct* %ptr) {
; CHECK: .param i32
; CHECK: call ext_byval_func_empty@FUNCTION, $0
call void @ext_byval_func_empty(%EmptyStruct* byval %ptr)
ret void
}
; CHECK-LABEL: byval_empty_callee
define void @byval_empty_callee(%EmptyStruct* byval %ptr) {
; CHECK: .param i32
; CHECK: call ext_func_empty@FUNCTION, $0
call void @ext_func_empty(%EmptyStruct* %ptr)
ret void
}
[WebAssembly] Call memcpy for large byval copies. This fixes very slow compilation on test/CodeGen/Generic/2010-11-04-BigByval.ll . Note that MaxStoresPerMemcpy and friends are not yet carefully tuned so the cutoff point is currently somewhat arbitrary. However, it's important that there be a cutoff point so that we don't emit unbounded quantities of loads and stores. llvm-svn: 261050
2016-02-17 09:43:37 +08:00
; Call memcpy for "big" byvals.
; CHECK-LABEL: big_byval:
[WebAssembly] Put __stack_pointer in the offset field of loads and stores. Instead of this: i32.const $push10=, __stack_pointer i32.load $push11=, 0($pop10) Emit this: i32.const $push10=, 0 i32.load $push11=, __stack_pointer($pop10) It's not currently clear which is better, though there's a chance the second form may be better at overall compression. We can revisit this when we have more data; for now it makes sense to make PEI consistent with isel. Differential Revision: http://reviews.llvm.org/D20411 llvm-svn: 270635
2016-05-25 07:47:41 +08:00
; CHECK: i32.const $push[[L4:.+]]=, 0
; CHECK: i32.const $push[[L1:.+]]=, 0
; CHECK-NEXT: i32.load $push[[L2:.+]]=, __stack_pointer($pop[[L1]])
[WebAssembly] Stackify function prologs and epilogs The instructions are the same, but fewer locals are used. Differential Revision: http://reviews.llvm.org/D17428 llvm-svn: 261452
2016-02-21 05:46:50 +08:00
; CHECK-NEXT: i32.const $push[[L3:.+]]=, 131072
[WebAssembly] Disable the store-results optimization. The WebAssemly spec removing the return value from store instructions, so remove the associated optimization from LLVM. This patch leaves the store instruction operands in place for now, so stores now always write to "$drop"; these will be removed in a seperate patch. llvm-svn: 279100
2016-08-19 01:51:27 +08:00
; CHECK-NEXT: i32.sub $push[[L11:.+]]=, $pop[[L2]], $pop[[L3]]
; CHECK-NEXT: tee_local $push[[L10:.+]]=, $[[SP:.+]]=, $pop[[L11]]{{$}}
[WebAssembly] Remove the output operand from stores. Per spec changes, store instructions in WebAssembly no longer have a return value. Update the instruction descriptions. Differential Revision: https://reviews.llvm.org/D25122 llvm-svn: 283501
2016-10-07 06:08:28 +08:00
; CHECK-NEXT: i32.store __stack_pointer($pop[[L4]]), $pop[[L10]]{{$}}
[WebAssembly] Move register stackification and coloring to a late phase. Move the register stackification and coloring passes to run very late, after PEI, tail duplication, and most other passes. This means that all code emitted and expanded by those passes is now exposed to these passes. This also eliminates the need for prologue/epilogue code to be manually stackified, which significantly simplifies the code. This does require running LiveIntervals a second time. It's useful to think of these late passes not as late optimization passes, but as a domain-specific compression algorithm based on knowledge of liveness information. It's used to compress the code after all conventional optimizations are complete, which is why it uses LiveIntervals at a phase when actual optimization passes don't typically need it. Differential Revision: http://reviews.llvm.org/D20075 llvm-svn: 269012
2016-05-10 12:24:02 +08:00
; CHECK-NEXT: i32.const $push[[L0:.+]]=, 131072
[WebAssembly] Disable the store-results optimization. The WebAssemly spec removing the return value from store instructions, so remove the associated optimization from LLVM. This patch leaves the store instruction operands in place for now, so stores now always write to "$drop"; these will be removed in a seperate patch. llvm-svn: 279100
2016-08-19 01:51:27 +08:00
; CHECK-NEXT: i32.call $push[[L11:.+]]=, memcpy@FUNCTION, $[[SP]], ${{.+}}, $pop{{.+}}
[WebAssembly] Move register stackification and coloring to a late phase. Move the register stackification and coloring passes to run very late, after PEI, tail duplication, and most other passes. This means that all code emitted and expanded by those passes is now exposed to these passes. This also eliminates the need for prologue/epilogue code to be manually stackified, which significantly simplifies the code. This does require running LiveIntervals a second time. It's useful to think of these late passes not as late optimization passes, but as a domain-specific compression algorithm based on knowledge of liveness information. It's used to compress the code after all conventional optimizations are complete, which is why it uses LiveIntervals at a phase when actual optimization passes don't typically need it. Differential Revision: http://reviews.llvm.org/D20075 llvm-svn: 269012
2016-05-10 12:24:02 +08:00
; CHECK-NEXT: tee_local $push[[L9:.+]]=, $[[SP:.+]]=, $pop[[L11]]{{$}}
[WebAssembly] Call memcpy for large byval copies. This fixes very slow compilation on test/CodeGen/Generic/2010-11-04-BigByval.ll . Note that MaxStoresPerMemcpy and friends are not yet carefully tuned so the cutoff point is currently somewhat arbitrary. However, it's important that there be a cutoff point so that we don't emit unbounded quantities of loads and stores. llvm-svn: 261050
2016-02-17 09:43:37 +08:00
; CHECK-NEXT: call big_byval_callee@FUNCTION,
%big = type [131072 x i8]
declare void @big_byval_callee(%big* byval align 1)
define void @big_byval(%big* byval align 1 %x) {
call void @big_byval_callee(%big* byval align 1 %x)
ret void
}