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/X86/3dnow-schedule.ll

395 lines
16 KiB
LLVM
Raw Normal View History

[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -print-schedule -mcpu=x86-64 -mattr=+3dnowa | FileCheck %s --check-prefix=CHECK --check-prefix=GENERIC
define void @test_femms() optsize {
; CHECK-LABEL: test_femms:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: femms
; CHECK-NEXT: retq # sched: [1:1.00]
call void @llvm.x86.mmx.femms()
ret void
}
declare void @llvm.x86.mmx.femms() nounwind readnone
define i64 @test_pavgusb(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pavgusb:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pavgusb %mm1, %mm0 # sched: [5:1.00]
; CHECK-NEXT: pavgusb (%rdi), %mm0 # sched: [9:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pavgusb(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pavgusb(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pavgusb(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pf2id(x86_mmx* %a0) optsize {
; CHECK-LABEL: test_pf2id:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pf2id (%rdi), %mm0 # sched: [7:1.00]
; CHECK-NEXT: pf2id %mm0, %mm0 # sched: [3:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = load x86_mmx, x86_mmx *%a0, align 8
%2 = call x86_mmx @llvm.x86.3dnow.pf2id(x86_mmx %1)
%3 = call x86_mmx @llvm.x86.3dnow.pf2id(x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pf2id(x86_mmx) nounwind readnone
define i64 @test_pf2iw(x86_mmx* %a0) optsize {
; CHECK-LABEL: test_pf2iw:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pf2iw (%rdi), %mm0 # sched: [7:1.00]
; CHECK-NEXT: pf2iw %mm0, %mm0 # sched: [3:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = load x86_mmx, x86_mmx *%a0, align 8
%2 = call x86_mmx @llvm.x86.3dnowa.pf2iw(x86_mmx %1)
%3 = call x86_mmx @llvm.x86.3dnowa.pf2iw(x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnowa.pf2iw(x86_mmx) nounwind readnone
define i64 @test_pfacc(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfacc:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfacc %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfacc (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfacc(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfacc(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfacc(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfadd(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfadd:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfadd %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfadd (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfadd(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfadd(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfadd(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfcmpeq(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfcmpeq:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfcmpeq %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfcmpeq (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfcmpeq(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfcmpeq(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfcmpeq(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfcmpge(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfcmpge:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfcmpge %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfcmpge (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfcmpge(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfcmpge(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfcmpge(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfcmpgt(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfcmpgt:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfcmpgt %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfcmpgt (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfcmpgt(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfcmpgt(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfcmpgt(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfmax(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfmax:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfmax %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfmax (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfmax(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfmax(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfmax(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfmin(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfmin:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfmin %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfmin (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfmin(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfmin(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfmin(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfmul(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfmul:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfmul %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfmul (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfmul(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfmul(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfmul(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfnacc(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfnacc:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfnacc %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfnacc (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnowa.pfnacc(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnowa.pfnacc(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnowa.pfnacc(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfpnacc(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfpnacc:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfpnacc %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfpnacc (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnowa.pfpnacc(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnowa.pfpnacc(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnowa.pfpnacc(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfrcp(x86_mmx* %a0) optsize {
; CHECK-LABEL: test_pfrcp:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfrcp (%rdi), %mm0 # sched: [7:1.00]
; CHECK-NEXT: pfrcp %mm0, %mm0 # sched: [3:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = load x86_mmx, x86_mmx *%a0, align 8
%2 = call x86_mmx @llvm.x86.3dnow.pfrcp(x86_mmx %1)
%3 = call x86_mmx @llvm.x86.3dnow.pfrcp(x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfrcp(x86_mmx) nounwind readnone
define i64 @test_pfrcpit1(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfrcpit1:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfrcpit1 %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfrcpit1 (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfrcpit1(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfrcpit1(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfrcpit1(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfrcpit2(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfrcpit2:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfrcpit2 %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfrcpit2 (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfrcpit2(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfrcpit2(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfrcpit2(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfrsqit1(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfrsqit1:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfrsqit1 %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfrsqit1 (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfrsqit1(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfrsqit1(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfrsqit1(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfrsqrt(x86_mmx* %a0) optsize {
; CHECK-LABEL: test_pfrsqrt:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfrsqrt (%rdi), %mm0 # sched: [7:1.00]
; CHECK-NEXT: pfrsqrt %mm0, %mm0 # sched: [3:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = load x86_mmx, x86_mmx *%a0, align 8
%2 = call x86_mmx @llvm.x86.3dnow.pfrsqrt(x86_mmx %1)
%3 = call x86_mmx @llvm.x86.3dnow.pfrsqrt(x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfrsqrt(x86_mmx) nounwind readnone
define i64 @test_pfsub(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfsub:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfsub %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfsub (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfsub(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfsub(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfsub(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pfsubr(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pfsubr:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pfsubr %mm1, %mm0 # sched: [3:1.00]
; CHECK-NEXT: pfsubr (%rdi), %mm0 # sched: [7:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pfsubr(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pfsubr(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pfsubr(x86_mmx, x86_mmx) nounwind readnone
define i64 @test_pi2fd(x86_mmx* %a0) optsize {
; CHECK-LABEL: test_pi2fd:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pi2fd (%rdi), %mm0 # sched: [8:1.00]
; CHECK-NEXT: pi2fd %mm0, %mm0 # sched: [4:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = load x86_mmx, x86_mmx *%a0, align 8
%2 = call x86_mmx @llvm.x86.3dnow.pi2fd(x86_mmx %1)
%3 = call x86_mmx @llvm.x86.3dnow.pi2fd(x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pi2fd(x86_mmx) nounwind readnone
define i64 @test_pi2fw(x86_mmx* %a0) optsize {
; CHECK-LABEL: test_pi2fw:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pi2fw (%rdi), %mm0 # sched: [8:1.00]
; CHECK-NEXT: pi2fw %mm0, %mm0 # sched: [4:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = load x86_mmx, x86_mmx *%a0, align 8
%2 = call x86_mmx @llvm.x86.3dnowa.pi2fw(x86_mmx %1)
%3 = call x86_mmx @llvm.x86.3dnowa.pi2fw(x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnowa.pi2fw(x86_mmx) nounwind readnone
define i64 @test_pmulhrw(x86_mmx %a0, x86_mmx %a1, x86_mmx* %a2) optsize {
; CHECK-LABEL: test_pmulhrw:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pmulhrw %mm1, %mm0 # sched: [5:1.00]
; CHECK-NEXT: pmulhrw (%rdi), %mm0 # sched: [9:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = call x86_mmx @llvm.x86.3dnow.pmulhrw(x86_mmx %a0, x86_mmx %a1)
%2 = load x86_mmx, x86_mmx *%a2, align 8
%3 = call x86_mmx @llvm.x86.3dnow.pmulhrw(x86_mmx %1, x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnow.pmulhrw(x86_mmx, x86_mmx) nounwind readnone
[X86][3DNow] Add instruction itinerary and scheduling classes for femms/prefetch/prefetchw llvm-svn: 319167
2017-11-28 20:37:35 +08:00
define void @test_prefetch(i8* %a0) optsize {
; CHECK-LABEL: test_prefetch:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add instruction itinerary and scheduling classes for femms/prefetch/prefetchw llvm-svn: 319167
2017-11-28 20:37:35 +08:00
; CHECK-NEXT: #APP
; CHECK-NEXT: prefetch (%rdi) # sched: [5:0.50]
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: retq # sched: [1:1.00]
tail call void asm sideeffect "prefetch $0", "*m"(i8 *%a0) nounwind
ret void
}
define void @test_prefetchw(i8* %a0) optsize {
; CHECK-LABEL: test_prefetchw:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add instruction itinerary and scheduling classes for femms/prefetch/prefetchw llvm-svn: 319167
2017-11-28 20:37:35 +08:00
; CHECK-NEXT: #APP
; CHECK-NEXT: prefetchw (%rdi) # sched: [5:0.50]
; CHECK-NEXT: #NO_APP
; CHECK-NEXT: retq # sched: [1:1.00]
tail call void asm sideeffect "prefetchw $0", "*m"(i8 *%a0) nounwind
ret void
}
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
define i64 @test_pswapd(x86_mmx* %a0) optsize {
; CHECK-LABEL: test_pswapd:
[CodeGen] Unify MBB reference format in both MIR and debug output As part of the unification of the debug format and the MIR format, print MBB references as '%bb.5'. The MIR printer prints the IR name of a MBB only for block definitions. * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)->getNumber\(\)/" << printMBBReference(*\1)/g' * find . \( -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#" << ([a-zA-Z0-9_]+)\.getNumber\(\)/" << printMBBReference(\1)/g' * find . \( -name "*.txt" -o -name "*.s" -o -name "*.mir" -o -name "*.cpp" -o -name "*.h" -o -name "*.ll" \) -type f -print0 | xargs -0 sed -i '' -E 's/BB#([0-9]+)/%bb.\1/g' * grep -nr 'BB#' and fix Differential Revision: https://reviews.llvm.org/D40422 llvm-svn: 319665
2017-12-05 01:18:51 +08:00
; CHECK: # %bb.0:
[X86][3DNow] Add 3DNow! instruction itinerary and scheduling classes llvm-svn: 319005
2017-11-27 04:50:29 +08:00
; CHECK-NEXT: pswapd (%rdi), %mm0 # mm0 = mem[1,0] sched: [5:1.00]
; CHECK-NEXT: pswapd %mm0, %mm0 # mm0 = mm0[1,0] sched: [1:1.00]
[X86] Stop printing moves between VR64 and GR64 with 'movd' mnemonic. Use 'movq' instead. This behavior existed to work with an old version of the gnu assembler on MacOS that only accepted this form. Newer versions of GNU assembler and the current LLVM derived version of the assembler on MacOS support movq as well. llvm-svn: 321898
2018-01-06 04:55:12 +08:00
; CHECK-NEXT: movq %mm0, %rax # sched: [1:0.33]
[X86][3DNow] Add scheduling latency/throughput tests for 3DNow! instructions llvm-svn: 315942
2017-10-17 05:55:09 +08:00
; CHECK-NEXT: retq # sched: [1:1.00]
%1 = load x86_mmx, x86_mmx *%a0, align 8
%2 = call x86_mmx @llvm.x86.3dnowa.pswapd(x86_mmx %1)
%3 = call x86_mmx @llvm.x86.3dnowa.pswapd(x86_mmx %2)
%4 = bitcast x86_mmx %3 to i64
ret i64 %4
}
declare x86_mmx @llvm.x86.3dnowa.pswapd(x86_mmx) nounwind readnone