[x86] Teach the "generic" x86 CPU to avoid patterns that are slow on widely used processors.

[chandlerc]

Summary

This occured to me when I saw that we were generating 'inc' and 'dec'
when for Haswell and newer we shouldn't. However, there were a few "X is
slow" things that we should probably just set.

I've avoided any of the "X is fast" features because most of those would
be pretty serious regressions on processors where X isn't actually fast.
The slow things are likely to be negligible costs on processors where
these aren't slow and a significant win when they are slow.

In retrospect this seems somewhat obvious. Not sure why we didn't do this a long time ago.

[craig.topper]

This seems reasonable to me. @zvi and @RKSimon, what do you think?

[echristo]

Seems reasonable to add a comment as to what microarch features we're attempting to target as modern here.

Definitely starting to hit the point where we should verify for amd processors though - I'm not sure how any of the new zen based fare here.

-eric

[chandlerc]

In D36947#847139, @echristo wrote:

Seems reasonable to add a comment as to what microarch features we're attempting to target as modern here.

Definitely starting to hit the point where we should verify for amd processors though - I'm not sure how any of the new zen based fare here.

Yeah, I actually looked and this is as good as LLVM's existing information allows.

For example, btver2 up through znver1 have SlowSHLD, and so not including FastSHLDRotate is probably good here to produce "least bad" x86 across architectures.

Other than that though, I couldn't find any FastFoo or SlowFoo features in LLVM's AMD processor feature sets that would make sense here and aren't already covered. Really just SlowBTMem on Barcelona and fam10, and that's been long covered.

[chandlerc]

This seems really uncontroversial as it essentially just avoids some patterns. Going ahead and landing for now. Happy to revisit or enhance this as desired by others of course.

Thanks for the reviews so far (and glad you mentioned AMD processors Eric!)

llvm/trunk/lib/Target/X86/X86.td

 // basic SSE2 and 64-bit ones. It disables slow things from any mainstream and
 // modern 64-bit x86 chip, and enables features that are generally beneficial.
 //
 // We currently use the Sandy Bridge model as the default scheduling model as
 // we use it across Nehalem, Westmere, Sandy Bridge, and Ivy Bridge which
 // covers a huge swath of x86 processors. If there are specific scheduling
 // knobs which need to be tuned differently for AMD chips, we might consider
 // forming a common base for them.
-def : ProcessorModel<"x86-64", SandyBridgeModel,
-                     [FeatureX87, FeatureMMX, FeatureSSE2, FeatureFXSR,
-                      Feature64Bit, FeatureSlowBTMem ]>;
+def : ProcessorModel<"x86-64", SandyBridgeModel, [
+  FeatureX87,
+  FeatureMMX,
+  FeatureSSE2,
+  FeatureFXSR,
+  Feature64Bit,
+  FeatureSlow3OpsLEA,
+  FeatureSlowBTMem,
+  FeatureSlowIncDec,
+  FeatureSlowUAMem32
+]>;
 
 //===----------------------------------------------------------------------===//
 // Register File Description
 //===----------------------------------------------------------------------===//
 
 include "X86RegisterInfo.td"
 include "X86RegisterBanks.td"
 

llvm/trunk/test/CodeGen/X86/avx-schedule.ll

; ZNVER1-NEXT:    retq # sched: [5:0.50]
   %4 = fadd <4 x double> %1, %3
   ret <4 x double> %4
 }
 
 define <8 x float> @test_cvtdq2ps(<8 x i32> %a0, <8 x i32> *%a1) {
 ; GENERIC-LABEL: test_cvtdq2ps:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    vcvtdq2ps %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT:    vcvtdq2ps (%rdi), %ymm1 # sched: [10:1.00]
+; GENERIC-NEXT:    vmovaps (%rdi), %xmm1 # sched: [6:0.50]
+; GENERIC-NEXT:    vinsertf128 $1, 16(%rdi), %ymm1, %ymm1 # sched: [7:0.50]
+; GENERIC-NEXT:    vcvtdq2ps %ymm1, %ymm1 # sched: [3:1.00]
 ; GENERIC-NEXT:    vaddps %ymm1, %ymm0, %ymm0 # sched: [3:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_cvtdq2ps:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vcvtdq2ps %ymm0, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vmovaps (%rdi), %xmm1 # sched: [6:0.50]
 ; SANDY-NEXT:    vinsertf128 $1, 16(%rdi), %ymm1, %ymm1 # sched: [7:0.50]
; ZNVER1-NEXT:    retq # sched: [5:0.50]
   %3 = shufflevector <8 x float> %2, <8 x float> undef, <8 x i32> <i32 0, i32 0, i32 2, i32 2, i32 4, i32 4, i32 6, i32 6>
   %4 = fadd <8 x float> %1, %3
   ret <8 x float> %4
 }
 
 define <4 x double> @test_movupd(<4 x double> *%a0, <4 x double> *%a1) {
 ; GENERIC-LABEL: test_movupd:
 ; GENERIC:       # BB#0:
-; GENERIC-NEXT:    vmovupd (%rdi), %ymm0 # sched: [7:0.50]
+; GENERIC-NEXT:    vmovups (%rdi), %xmm0 # sched: [6:0.50]
+; GENERIC-NEXT:    vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 # sched: [7:0.50]
 ; GENERIC-NEXT:    vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT:    vmovupd %ymm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    vextractf128 $1, %ymm0, 16(%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovupd %xmm0, (%rsi) # sched: [5:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movupd:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vmovups (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-NEXT:    vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 # sched: [7:0.50]
 ; SANDY-NEXT:    vaddpd %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vextractf128 $1, %ymm0, 16(%rsi) # sched: [5:1.00]
; ZNVER1-NEXT:    retq # sched: [5:0.50]
   %2 = fadd <4 x double> %1, %1
   store <4 x double> %2, <4 x double> *%a1, align 1
   ret <4 x double> %2
 }
 
 define <8 x float> @test_movups(<8 x float> *%a0, <8 x float> *%a1) {
 ; GENERIC-LABEL: test_movups:
 ; GENERIC:       # BB#0:
-; GENERIC-NEXT:    vmovups (%rdi), %ymm0 # sched: [7:0.50]
+; GENERIC-NEXT:    vmovups (%rdi), %xmm0 # sched: [6:0.50]
+; GENERIC-NEXT:    vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 # sched: [7:0.50]
 ; GENERIC-NEXT:    vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
-; GENERIC-NEXT:    vmovups %ymm0, (%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    vextractf128 $1, %ymm0, 16(%rsi) # sched: [5:1.00]
+; GENERIC-NEXT:    vmovups %xmm0, (%rsi) # sched: [5:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SANDY-LABEL: test_movups:
 ; SANDY:       # BB#0:
 ; SANDY-NEXT:    vmovups (%rdi), %xmm0 # sched: [6:0.50]
 ; SANDY-NEXT:    vinsertf128 $1, 16(%rdi), %ymm0, %ymm0 # sched: [7:0.50]
 ; SANDY-NEXT:    vaddps %ymm0, %ymm0, %ymm0 # sched: [3:1.00]
 ; SANDY-NEXT:    vextractf128 $1, %ymm0, 16(%rsi) # sched: [5:1.00]

llvm/trunk/test/CodeGen/X86/lea32-schedule.ll

; ZNVER1-NEXT:    retq # sched: [5:0.50]
   ret i32 %3
 }
 
 define i32 @test_lea_add_offset(i32, i32) {
 ; GENERIC-LABEL: test_lea_add_offset:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; GENERIC-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; GENERIC-NEXT:    leal 16(%rdi,%rsi), %eax # sched: [1:0.50]
+; GENERIC-NEXT:    leal (%rdi,%rsi), %eax # sched: [1:0.50]
+; GENERIC-NEXT:    addl $16, %eax # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_lea_add_offset:
 ; ATOM:       # BB#0:
 ; ATOM-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; ATOM-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ATOM-NEXT:    leal 16(%rdi,%rsi), %eax # sched: [1:1.00]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
; ZNVER1-NEXT:    retq # sched: [5:0.50]
   ret i32 %4
 }
 
 define i32 @test_lea_add_offset_big(i32, i32) {
 ; GENERIC-LABEL: test_lea_add_offset_big:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; GENERIC-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; GENERIC-NEXT:    leal -4096(%rdi,%rsi), %eax # sched: [1:0.50]
+; GENERIC-NEXT:    leal (%rdi,%rsi), %eax # sched: [1:0.50]
+; GENERIC-NEXT:    addl $-4096, %eax # imm = 0xF000
+; GENERIC-NEXT:    # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_lea_add_offset_big:
 ; ATOM:       # BB#0:
 ; ATOM-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; ATOM-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ATOM-NEXT:    leal -4096(%rdi,%rsi), %eax # sched: [1:1.00]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
; ZNVER1-NEXT:    retq # sched: [5:0.50]
   %2 = mul nsw i32 %0, 3
   ret i32 %2
 }
 
 define i32 @test_lea_mul_offset(i32) {
 ; GENERIC-LABEL: test_lea_mul_offset:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; GENERIC-NEXT:    leal -32(%rdi,%rdi,2), %eax # sched: [1:0.50]
+; GENERIC-NEXT:    leal (%rdi,%rdi,2), %eax # sched: [1:0.50]
+; GENERIC-NEXT:    addl $-32, %eax # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_lea_mul_offset:
 ; ATOM:       # BB#0:
 ; ATOM-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ATOM-NEXT:    leal -32(%rdi,%rdi,2), %eax # sched: [1:1.00]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
; ZNVER1-NEXT:    retq # sched: [5:0.50]
   %3 = add nsw i32 %2, -32
   ret i32 %3
 }
 
 define i32 @test_lea_mul_offset_big(i32) {
 ; GENERIC-LABEL: test_lea_mul_offset_big:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; GENERIC-NEXT:    leal 10000(%rdi,%rdi,8), %eax # sched: [1:0.50]
+; GENERIC-NEXT:    leal (%rdi,%rdi,8), %eax # sched: [1:0.50]
+; GENERIC-NEXT:    addl $10000, %eax # imm = 0x2710
+; GENERIC-NEXT:    # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_lea_mul_offset_big:
 ; ATOM:       # BB#0:
 ; ATOM-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ATOM-NEXT:    leal 10000(%rdi,%rdi,8), %eax # sched: [1:1.00]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
; ZNVER1-NEXT:    retq # sched: [5:0.50]
   ret i32 %4
 }
 
 define i32 @test_lea_add_scale_offset(i32, i32) {
 ; GENERIC-LABEL: test_lea_add_scale_offset:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; GENERIC-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; GENERIC-NEXT:    leal 96(%rdi,%rsi,4), %eax # sched: [1:0.50]
+; GENERIC-NEXT:    leal (%rdi,%rsi,4), %eax # sched: [1:0.50]
+; GENERIC-NEXT:    addl $96, %eax # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_lea_add_scale_offset:
 ; ATOM:       # BB#0:
 ; ATOM-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; ATOM-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ATOM-NEXT:    leal 96(%rdi,%rsi,4), %eax # sched: [1:1.00]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
; ZNVER1-NEXT:    retq # sched: [5:0.50]
   ret i32 %5
 }
 
 define i32 @test_lea_add_scale_offset_big(i32, i32) {
 ; GENERIC-LABEL: test_lea_add_scale_offset_big:
 ; GENERIC:       # BB#0:
 ; GENERIC-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; GENERIC-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
-; GENERIC-NEXT:    leal -1200(%rdi,%rsi,8), %eax # sched: [1:0.50]
+; GENERIC-NEXT:    leal (%rdi,%rsi,8), %eax # sched: [1:0.50]
+; GENERIC-NEXT:    addl $-1200, %eax # imm = 0xFB50
+; GENERIC-NEXT:    # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_lea_add_scale_offset_big:
 ; ATOM:       # BB#0:
 ; ATOM-NEXT:    # kill: %ESI<def> %ESI<kill> %RSI<def>
 ; ATOM-NEXT:    # kill: %EDI<def> %EDI<kill> %RDI<def>
 ; ATOM-NEXT:    leal -1200(%rdi,%rsi,8), %eax # sched: [1:1.00]
 ; ATOM-NEXT:    nop # sched: [1:0.50]

llvm/trunk/test/CodeGen/X86/lea64-schedule.ll

 ; ZNVER1-NEXT:    retq # sched: [5:0.50]
   %3 = add nsw i64 %1, %0
   ret i64 %3
 }
 
 define i64 @test_lea_add_offset(i64, i64) {
 ; GENERIC-LABEL: test_lea_add_offset:
 ; GENERIC:       # BB#0:
-; GENERIC-NEXT:    leaq 16(%rdi,%rsi), %rax # sched: [1:0.50]
+; GENERIC-NEXT:    leaq (%rdi,%rsi), %rax # sched: [1:0.50]
+; GENERIC-NEXT:    addq $16, %rax # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_lea_add_offset:
 ; ATOM:       # BB#0:
 ; ATOM-NEXT:    leaq 16(%rdi,%rsi), %rax # sched: [1:1.00]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
; ZNVER1-NEXT:    retq # sched: [5:0.50]
   %3 = add i64 %0, 16
   %4 = add i64 %3, %1
   ret i64 %4
 }
 
 define i64 @test_lea_add_offset_big(i64, i64) {
 ; GENERIC-LABEL: test_lea_add_offset_big:
 ; GENERIC:       # BB#0:
-; GENERIC-NEXT:    leaq -4096(%rdi,%rsi), %rax # sched: [1:0.50]
+; GENERIC-NEXT:    leaq (%rdi,%rsi), %rax # sched: [1:0.50]
+; GENERIC-NEXT:    addq $-4096, %rax # imm = 0xF000
+; GENERIC-NEXT:    # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_lea_add_offset_big:
 ; ATOM:       # BB#0:
 ; ATOM-NEXT:    leaq -4096(%rdi,%rsi), %rax # sched: [1:1.00]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]

 ; ZNVER1-NEXT:    retq # sched: [5:0.50]
   %2 = mul nsw i64 %0, 3
   ret i64 %2
 }
 
 define i64 @test_lea_mul_offset(i64) {
 ; GENERIC-LABEL: test_lea_mul_offset:
 ; GENERIC:       # BB#0:
-; GENERIC-NEXT:    leaq -32(%rdi,%rdi,2), %rax # sched: [1:0.50]
+; GENERIC-NEXT:    leaq (%rdi,%rdi,2), %rax # sched: [1:0.50]
+; GENERIC-NEXT:    addq $-32, %rax # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_lea_mul_offset:
 ; ATOM:       # BB#0:
 ; ATOM-NEXT:    leaq -32(%rdi,%rdi,2), %rax # sched: [1:1.00]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
; ZNVER1-NEXT:    retq # sched: [5:0.50]
   %2 = mul nsw i64 %0, 3
   %3 = add nsw i64 %2, -32
   ret i64 %3
 }
 
 define i64 @test_lea_mul_offset_big(i64) {
 ; GENERIC-LABEL: test_lea_mul_offset_big:
 ; GENERIC:       # BB#0:
-; GENERIC-NEXT:    leaq 10000(%rdi,%rdi,8), %rax # sched: [1:0.50]
+; GENERIC-NEXT:    leaq (%rdi,%rdi,8), %rax # sched: [1:0.50]
+; GENERIC-NEXT:    addq $10000, %rax # imm = 0x2710
+; GENERIC-NEXT:    # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_lea_mul_offset_big:
 ; ATOM:       # BB#0:
 ; ATOM-NEXT:    leaq 10000(%rdi,%rdi,8), %rax # sched: [1:1.00]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
; ZNVER1-NEXT:    retq # sched: [5:0.50]
   %3 = shl i64 %1, 1
   %4 = add nsw i64 %3, %0
   ret i64 %4
 }
 
 define i64 @test_lea_add_scale_offset(i64, i64) {
 ; GENERIC-LABEL: test_lea_add_scale_offset:
 ; GENERIC:       # BB#0:
-; GENERIC-NEXT:    leaq 96(%rdi,%rsi,4), %rax # sched: [1:0.50]
+; GENERIC-NEXT:    leaq (%rdi,%rsi,4), %rax # sched: [1:0.50]
+; GENERIC-NEXT:    addq $96, %rax # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_lea_add_scale_offset:
 ; ATOM:       # BB#0:
 ; ATOM-NEXT:    leaq 96(%rdi,%rsi,4), %rax # sched: [1:1.00]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
; ZNVER1-NEXT:    retq # sched: [5:0.50]
   %4 = add i64 %0, 96
   %5 = add i64 %4, %3
   ret i64 %5
 }
 
 define i64 @test_lea_add_scale_offset_big(i64, i64) {
 ; GENERIC-LABEL: test_lea_add_scale_offset_big:
 ; GENERIC:       # BB#0:
-; GENERIC-NEXT:    leaq -1200(%rdi,%rsi,8), %rax # sched: [1:0.50]
+; GENERIC-NEXT:    leaq (%rdi,%rsi,8), %rax # sched: [1:0.50]
+; GENERIC-NEXT:    addq $-1200, %rax # imm = 0xFB50
+; GENERIC-NEXT:    # sched: [1:0.33]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; ATOM-LABEL: test_lea_add_scale_offset_big:
 ; ATOM:       # BB#0:
 ; ATOM-NEXT:    leaq -1200(%rdi,%rsi,8), %rax # sched: [1:1.00]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]
 ; ATOM-NEXT:    nop # sched: [1:0.50]

llvm/trunk/test/CodeGen/X86/vector-shuffle-512-v16.ll

 ; ALL-NEXT:    retq
   %res = shufflevector <16 x i32> %v, <16 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 12>
   ret <4 x i32> %res
 }
 
 define <8 x float> @shuffle_v16f32_extract_256(float* %RET, float* %a) {
 ; ALL-LABEL: shuffle_v16f32_extract_256:
 ; ALL:       # BB#0:
-; ALL-NEXT:    vmovups 32(%rsi), %ymm0
+; ALL-NEXT:    vmovups 32(%rsi), %xmm0
+; ALL-NEXT:    vinsertf128 $1, 48(%rsi), %ymm0, %ymm0
 ; ALL-NEXT:    retq
   %ptr_a = bitcast float* %a to <16 x float>*
   %v_a = load <16 x float>, <16 x float>* %ptr_a, align 4
   %v2 = shufflevector <16 x float> %v_a, <16 x float> undef, <8 x i32>  <i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
   ret <8 x float> %v2
 }
 
 ;FIXME: can do better with vcompressp

llvm/trunk/test/DebugInfo/COFF/register-variables.ll

 ; ASM:         #DEBUG_VALUE: c <- %EAX
 ; ASM:         testl   %esi, %esi
 ; ASM:         je      .LBB0_2
 ; ASM: [[after_je:\.Ltmp.*]]:
 ; ASM: # BB#1:                                 # %if.then
 ; ASM-DAG:     #DEBUG_VALUE: inlineinc:a <- %EAX
 ; ASM-DAG:     #DEBUG_VALUE: a <- %EAX
 ; ASM-DAG:     #DEBUG_VALUE: f:p <- %ESI
-; ASM:         incl    %eax
+; ASM:         addl    $1, %eax
 ; ASM: [[after_inc_eax:\.Ltmp.*]]:
 ; ASM:         #DEBUG_VALUE: inlineinc:b <- %EAX
 ; ASM:         #DEBUG_VALUE: b <- %EAX
 ; ASM:         incl    x(%rip)
 ; ASM: [[after_if:\.Ltmp.*]]:
 ; ASM: .LBB0_2:                                # %if.else
 ; ASM:         #DEBUG_VALUE: f:p <- %ESI
 ; ASM:         movl    %eax, %ecx

 ; OBJ:       Range: 0x7
 ; OBJ:     }
 ; OBJ:   }
 ; OBJ:   DefRangeRegisterSym {
 ; OBJ:     Register: 23
 ; OBJ:     LocalVariableAddrRange {
 ; OBJ:       OffsetStart: .text+0x7
 ; OBJ:       ISectStart: 0x0
-; OBJ:       Range: 0x18
+; OBJ:       Range: 0x19
 ; OBJ:     }
 ; OBJ:   }
 ; OBJ:   LocalSym {
 ; OBJ:     Type: int (0x74)
 ; OBJ:     Flags [ (0x0)
 ; OBJ:     ]
 ; OBJ:     VarName: a
 ; OBJ:   }
 ; OBJ:   DefRangeRegisterSym {
 ; OBJ:     Register: 17
 ; OBJ:     LocalVariableAddrRange {
 ; OBJ:       OffsetStart: .text+0xC
 ; OBJ:       ISectStart: 0x0
-; OBJ:       Range: 0x6
+; OBJ:       Range: 0x7
 ; OBJ:     }
 ; OBJ:   }
 ; OBJ:   LocalSym {
 ; OBJ:     Type: int (0x74)
 ; OBJ:     Flags [ (0x0)
 ; OBJ:     ]
 ; OBJ:     VarName: c
 ; OBJ:   }

 ; OBJ:     Type: int (0x74)
 ; OBJ:     Flags [ (0x0)
 ; OBJ:     ]
 ; OBJ:     VarName: b
 ; OBJ:   }
 ; OBJ:   DefRangeRegisterSym {
 ; OBJ:     Register: 17
 ; OBJ:     MayHaveNoName: 0
-; OBJ:       OffsetStart: .text+0x12
+; OBJ:       OffsetStart: .text+0x13
 ; OBJ:       ISectStart: 0x0
 ; OBJ:       Range: 0x6
 ; OBJ:     }
 ; OBJ:   }
 ; OBJ:   InlineSiteSym {
 ; OBJ:     PtrParent: 0x0
 ; OBJ:     PtrEnd: 0x0
 ; OBJ:     Inlinee: inlineinc (0x1002)
 ; OBJ:   }
 ; OBJ:   LocalSym {
 ; OBJ:     Type: int (0x74)
 ; OBJ:     Flags [ (0x1)
 ; OBJ:       IsParameter (0x1)
 ; OBJ:     ]
 ; OBJ:     VarName: a
 ; OBJ:   }
 ; OBJ:   DefRangeRegisterSym {
 ; OBJ:     Register: 17
 ; OBJ:     LocalVariableAddrRange {
 ; OBJ:       OffsetStart: .text+0xC
 ; OBJ:       ISectStart: 0x0
-; OBJ:       Range: 0x6
+; OBJ:       Range: 0x7
 ; OBJ:     }
 ; OBJ:   }
 ; OBJ:   LocalSym {
 ; OBJ:     Type: int (0x74)
 ; OBJ:     Flags [ (0x0)
 ; OBJ:     ]
 ; OBJ:     VarName: b
 ; OBJ:   }
 ; OBJ:   DefRangeRegisterSym {
 ; OBJ:     Register: 17
 ; OBJ:     LocalVariableAddrRange {
-; OBJ:       OffsetStart: .text+0x12
+; OBJ:       OffsetStart: .text+0x13
 ; OBJ:       ISectStart: 0x0
 ; OBJ:       Range: 0x6
 ; OBJ:     }
 ; OBJ:   }
 ; OBJ:   InlineSiteEnd {
 ; OBJ:   }
 ; OBJ:   ProcEnd
 ; OBJ: ]