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/combine-urem.ll

376 lines
13 KiB
LLVM
Raw Normal View History

[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
[X86] Add common CHECK prefix for tests without SSE/AVX codegen llvm-svn: 321810
2018-01-05 02:23:46 +08:00
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
[X86] Add srem/udiv/urem by one combine tests llvm-svn: 321826
2018-01-05 06:08:36 +08:00
; fold (urem x, 1) -> 0
define i32 @combine_urem_by_one(i32 %x) {
; CHECK-LABEL: combine_urem_by_one:
; CHECK: # %bb.0:
; CHECK-NEXT: xorl %eax, %eax
; CHECK-NEXT: retq
%1 = urem i32 %x, 1
ret i32 %1
}
define <4 x i32> @combine_vec_urem_by_one(<4 x i32> %x) {
; SSE-LABEL: combine_vec_urem_by_one:
; SSE: # %bb.0:
; SSE-NEXT: xorps %xmm0, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: combine_vec_urem_by_one:
; AVX: # %bb.0:
; AVX-NEXT: vxorps %xmm0, %xmm0, %xmm0
; AVX-NEXT: retq
%1 = urem <4 x i32> %x, <i32 1, i32 1, i32 1, i32 1>
ret <4 x i32> %1
}
[X86] Add srem/udiv/urem by constant tests Match the tests in combine-sdiv.ll llvm-svn: 336694
2018-07-11 00:08:28 +08:00
; TODO fold (urem x, -1) -> select((icmp eq x, -1), 0, x)
define i32 @combine_urem_by_negone(i32 %x) {
; CHECK-LABEL: combine_urem_by_negone:
; CHECK: # %bb.0:
; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: movq %rax, %rcx
; CHECK-NEXT: shlq $31, %rcx
; CHECK-NEXT: addq %rax, %rcx
; CHECK-NEXT: sarq $63, %rcx
; CHECK-NEXT: subl %ecx, %edi
; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: retq
%1 = urem i32 %x, -1
ret i32 %1
}
define <4 x i32> @combine_vec_urem_by_negone(<4 x i32> %x) {
; SSE-LABEL: combine_vec_urem_by_negone:
; SSE: # %bb.0:
; SSE-NEXT: pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
; SSE-NEXT: movdqa {{.*#+}} xmm2 = [2147483649,2147483649,2147483649,2147483649]
; SSE-NEXT: pmuludq %xmm2, %xmm1
; SSE-NEXT: pmuludq %xmm0, %xmm2
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; SSE-NEXT: pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
; SSE-NEXT: psrad $31, %xmm2
; SSE-NEXT: psubd %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX1-LABEL: combine_vec_urem_by_negone:
; AVX1: # %bb.0:
; AVX1-NEXT: vpshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
; AVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [2147483649,2147483649,2147483649,2147483649]
; AVX1-NEXT: vpmuludq %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpmuludq %xmm2, %xmm0, %xmm2
; AVX1-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,1,3,3]
; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3],xmm2[4,5],xmm1[6,7]
; AVX1-NEXT: vpsrad $31, %xmm1, %xmm1
; AVX1-NEXT: vpsubd %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_negone:
; AVX2: # %bb.0:
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [2147483649,2147483649,2147483649,2147483649]
; AVX2-NEXT: vpshufd {{.*#+}} xmm2 = xmm1[1,1,3,3]
; AVX2-NEXT: vpshufd {{.*#+}} xmm3 = xmm0[1,1,3,3]
; AVX2-NEXT: vpmuludq %xmm2, %xmm3, %xmm2
; AVX2-NEXT: vpmuludq %xmm1, %xmm0, %xmm1
; AVX2-NEXT: vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; AVX2-NEXT: vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
; AVX2-NEXT: vpsrad $31, %xmm1, %xmm1
; AVX2-NEXT: vpsubd %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
%1 = urem <4 x i32> %x, <i32 -1, i32 -1, i32 -1, i32 -1>
ret <4 x i32> %1
}
; fold (urem x, INT_MIN) -> (and x, ~INT_MIN)
define i32 @combine_urem_by_minsigned(i32 %x) {
; CHECK-LABEL: combine_urem_by_minsigned:
; CHECK: # %bb.0:
; CHECK-NEXT: andl $2147483647, %edi # imm = 0x7FFFFFFF
; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: retq
%1 = urem i32 %x, -2147483648
ret i32 %1
}
define <4 x i32> @combine_vec_urem_by_minsigned(<4 x i32> %x) {
; SSE-LABEL: combine_vec_urem_by_minsigned:
; SSE: # %bb.0:
; SSE-NEXT: andps {{.*}}(%rip), %xmm0
; SSE-NEXT: retq
;
; AVX1-LABEL: combine_vec_urem_by_minsigned:
; AVX1: # %bb.0:
; AVX1-NEXT: vandps {{.*}}(%rip), %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_minsigned:
; AVX2: # %bb.0:
; AVX2-NEXT: vbroadcastss {{.*#+}} xmm1 = [2147483647,2147483647,2147483647,2147483647]
; AVX2-NEXT: vandps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
%1 = urem <4 x i32> %x, <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648>
ret <4 x i32> %1
}
[X86] Show missed combine for X/X for SDIV/UDIV and X%X for SREM/UREM llvm-svn: 321807
2018-01-05 02:20:46 +08:00
; TODO fold (urem x, x) -> 0
define i32 @combine_urem_dupe(i32 %x) {
[X86] Add common CHECK prefix for tests without SSE/AVX codegen llvm-svn: 321810
2018-01-05 02:23:46 +08:00
; CHECK-LABEL: combine_urem_dupe:
; CHECK: # %bb.0:
; CHECK-NEXT: xorl %edx, %edx
; CHECK-NEXT: movl %edi, %eax
; CHECK-NEXT: divl %edi
; CHECK-NEXT: movl %edx, %eax
; CHECK-NEXT: retq
[X86] Show missed combine for X/X for SDIV/UDIV and X%X for SREM/UREM llvm-svn: 321807
2018-01-05 02:20:46 +08:00
%1 = urem i32 %x, %x
ret i32 %1
}
define <4 x i32> @combine_vec_urem_dupe(<4 x i32> %x) {
; SSE-LABEL: combine_vec_urem_dupe:
; SSE: # %bb.0:
; SSE-NEXT: pextrd $1, %xmm0, %eax
; SSE-NEXT: xorl %edx, %edx
; SSE-NEXT: divl %eax
; SSE-NEXT: movl %edx, %ecx
; SSE-NEXT: movd %xmm0, %eax
; SSE-NEXT: xorl %edx, %edx
; SSE-NEXT: divl %eax
; SSE-NEXT: movd %edx, %xmm1
; SSE-NEXT: pinsrd $1, %ecx, %xmm1
; SSE-NEXT: pextrd $2, %xmm0, %eax
; SSE-NEXT: xorl %edx, %edx
; SSE-NEXT: divl %eax
; SSE-NEXT: pinsrd $2, %edx, %xmm1
; SSE-NEXT: pextrd $3, %xmm0, %eax
; SSE-NEXT: xorl %edx, %edx
; SSE-NEXT: divl %eax
; SSE-NEXT: pinsrd $3, %edx, %xmm1
; SSE-NEXT: movdqa %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: combine_vec_urem_dupe:
; AVX: # %bb.0:
; AVX-NEXT: vpextrd $1, %xmm0, %eax
; AVX-NEXT: xorl %edx, %edx
; AVX-NEXT: divl %eax
; AVX-NEXT: movl %edx, %ecx
; AVX-NEXT: vmovd %xmm0, %eax
; AVX-NEXT: xorl %edx, %edx
; AVX-NEXT: divl %eax
; AVX-NEXT: vmovd %edx, %xmm1
; AVX-NEXT: vpinsrd $1, %ecx, %xmm1, %xmm1
; AVX-NEXT: vpextrd $2, %xmm0, %eax
; AVX-NEXT: xorl %edx, %edx
; AVX-NEXT: divl %eax
; AVX-NEXT: vpinsrd $2, %edx, %xmm1, %xmm1
; AVX-NEXT: vpextrd $3, %xmm0, %eax
; AVX-NEXT: xorl %edx, %edx
; AVX-NEXT: divl %eax
; AVX-NEXT: vpinsrd $3, %edx, %xmm1, %xmm0
; AVX-NEXT: retq
%1 = urem <4 x i32> %x, %x
ret <4 x i32> %1
}
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
; fold (urem x, pow2) -> (and x, (pow2-1))
define <4 x i32> @combine_vec_urem_by_pow2a(<4 x i32> %x) {
; SSE-LABEL: combine_vec_urem_by_pow2a:
[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
; SSE: # %bb.0:
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
; SSE-NEXT: andps {{.*}}(%rip), %xmm0
; SSE-NEXT: retq
;
[X86][SSE] Add AVX1 tests to sdiv/udiv srem/urem combine tests As requested on D27714 llvm-svn: 289652
2016-12-14 22:39:51 +08:00
; AVX1-LABEL: combine_vec_urem_by_pow2a:
[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
; AVX1: # %bb.0:
[X86][SSE] Add AVX1 tests to sdiv/udiv srem/urem combine tests As requested on D27714 llvm-svn: 289652
2016-12-14 22:39:51 +08:00
; AVX1-NEXT: vandps {{.*}}(%rip), %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_pow2a:
[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
; AVX2: # %bb.0:
[X86][AVX] Regenerate combine tests with constant broadcast comments llvm-svn: 308128
2017-07-16 19:36:11 +08:00
; AVX2-NEXT: vbroadcastss {{.*#+}} xmm1 = [3,3,3,3]
[X86][SSE] Add AVX1 tests to sdiv/udiv srem/urem combine tests As requested on D27714 llvm-svn: 289652
2016-12-14 22:39:51 +08:00
; AVX2-NEXT: vandps %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
%1 = urem <4 x i32> %x, <i32 4, i32 4, i32 4, i32 4>
ret <4 x i32> %1
}
define <4 x i32> @combine_vec_urem_by_pow2b(<4 x i32> %x) {
; SSE-LABEL: combine_vec_urem_by_pow2b:
[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
; SSE: # %bb.0:
[DAGCombiner] Try to use SelectionDAG::isKnownToBeAPowerOfTwo instead of just APInt::isPowerOf2 Generalize sdiv/udiv/srem/urem combines using APInt::isPowerOf2, which only works for const/splat-const values, to call SelectionDAG::isKnownToBeAPowerOfTwo instead which recognises many more cases. Added a DAGCombiner::BuildLogBase2 helper since PowerOf2 combines often involve taking the log2 of such a value. Differential Revision: https://reviews.llvm.org/D27714 llvm-svn: 289654
2016-12-14 23:08:13 +08:00
; SSE-NEXT: andps {{.*}}(%rip), %xmm0
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
; SSE-NEXT: retq
;
; AVX-LABEL: combine_vec_urem_by_pow2b:
[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
; AVX: # %bb.0:
[DAGCombiner] Try to use SelectionDAG::isKnownToBeAPowerOfTwo instead of just APInt::isPowerOf2 Generalize sdiv/udiv/srem/urem combines using APInt::isPowerOf2, which only works for const/splat-const values, to call SelectionDAG::isKnownToBeAPowerOfTwo instead which recognises many more cases. Added a DAGCombiner::BuildLogBase2 helper since PowerOf2 combines often involve taking the log2 of such a value. Differential Revision: https://reviews.llvm.org/D27714 llvm-svn: 289654
2016-12-14 23:08:13 +08:00
; AVX-NEXT: vandps {{.*}}(%rip), %xmm0, %xmm0
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
; AVX-NEXT: retq
%1 = urem <4 x i32> %x, <i32 1, i32 4, i32 8, i32 16>
ret <4 x i32> %1
}
[SelectionDAG] Recognise splat vector isKnownToBeAPowerOfTwo one/sign bit shift cases. llvm-svn: 301303
2017-04-25 20:29:07 +08:00
define <4 x i32> @combine_vec_urem_by_pow2c(<4 x i32> %x, <4 x i32> %y) {
; SSE-LABEL: combine_vec_urem_by_pow2c:
[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
; SSE: # %bb.0:
[SelectionDAG] Recognise splat vector isKnownToBeAPowerOfTwo one/sign bit shift cases. llvm-svn: 301303
2017-04-25 20:29:07 +08:00
; SSE-NEXT: pslld $23, %xmm1
; SSE-NEXT: paddd {{.*}}(%rip), %xmm1
; SSE-NEXT: cvttps2dq %xmm1, %xmm1
; SSE-NEXT: pcmpeqd %xmm2, %xmm2
; SSE-NEXT: paddd %xmm1, %xmm2
; SSE-NEXT: pand %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX1-LABEL: combine_vec_urem_by_pow2c:
[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
; AVX1: # %bb.0:
[SelectionDAG] Recognise splat vector isKnownToBeAPowerOfTwo one/sign bit shift cases. llvm-svn: 301303
2017-04-25 20:29:07 +08:00
; AVX1-NEXT: vpslld $23, %xmm1, %xmm1
; AVX1-NEXT: vpaddd {{.*}}(%rip), %xmm1, %xmm1
; AVX1-NEXT: vcvttps2dq %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_pow2c:
[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
; AVX2: # %bb.0:
[X86][AVX] Regenerate combine tests with constant broadcast comments llvm-svn: 308128
2017-07-16 19:36:11 +08:00
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm2 = [1,1,1,1]
[SelectionDAG] Recognise splat vector isKnownToBeAPowerOfTwo one/sign bit shift cases. llvm-svn: 301303
2017-04-25 20:29:07 +08:00
; AVX2-NEXT: vpsllvd %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
%1 = shl <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %y
%2 = urem <4 x i32> %x, %1
ret <4 x i32> %2
}
define <4 x i32> @combine_vec_urem_by_pow2d(<4 x i32> %x, <4 x i32> %y) {
; SSE-LABEL: combine_vec_urem_by_pow2d:
[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
; SSE: # %bb.0:
[X86][SSE] Reduce instruction/register usages for v4i32 vector shifts (PR37441) As suggested by Fabian on PR37441, use PSHUFLW to extend shift amount types for use with PSRAD/PSRLD to reduce register pressure. Some of this ideally would be done by combineTargetShuffle but its tricky to do as most of the shuffles are sharing inputs. Differential Revision: https://reviews.llvm.org/D46959 llvm-svn: 332524
2018-05-17 04:52:52 +08:00
; SSE-NEXT: pshuflw {{.*#+}} xmm2 = xmm1[2,3,3,3,4,5,6,7]
[SelectionDAG] Recognise splat vector isKnownToBeAPowerOfTwo one/sign bit shift cases. llvm-svn: 301303
2017-04-25 20:29:07 +08:00
; SSE-NEXT: movdqa {{.*#+}} xmm3 = [2147483648,2147483648,2147483648,2147483648]
; SSE-NEXT: movdqa %xmm3, %xmm4
; SSE-NEXT: psrld %xmm2, %xmm4
[X86][SSE] Reduce instruction/register usages for v4i32 vector shifts (PR37441) As suggested by Fabian on PR37441, use PSHUFLW to extend shift amount types for use with PSRAD/PSRLD to reduce register pressure. Some of this ideally would be done by combineTargetShuffle but its tricky to do as most of the shuffles are sharing inputs. Differential Revision: https://reviews.llvm.org/D46959 llvm-svn: 332524
2018-05-17 04:52:52 +08:00
; SSE-NEXT: pshufd {{.*#+}} xmm2 = xmm1[2,3,0,1]
; SSE-NEXT: pshuflw {{.*#+}} xmm5 = xmm2[2,3,3,3,4,5,6,7]
; SSE-NEXT: movdqa %xmm3, %xmm6
; SSE-NEXT: psrld %xmm5, %xmm6
; SSE-NEXT: pblendw {{.*#+}} xmm6 = xmm4[0,1,2,3],xmm6[4,5,6,7]
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm1[0,1,1,1,4,5,6,7]
; SSE-NEXT: movdqa %xmm3, %xmm4
; SSE-NEXT: psrld %xmm1, %xmm4
; SSE-NEXT: pshuflw {{.*#+}} xmm1 = xmm2[0,1,1,1,4,5,6,7]
; SSE-NEXT: psrld %xmm1, %xmm3
; SSE-NEXT: pblendw {{.*#+}} xmm3 = xmm4[0,1,2,3],xmm3[4,5,6,7]
; SSE-NEXT: pblendw {{.*#+}} xmm3 = xmm3[0,1],xmm6[2,3],xmm3[4,5],xmm6[6,7]
[SelectionDAG] Recognise splat vector isKnownToBeAPowerOfTwo one/sign bit shift cases. llvm-svn: 301303
2017-04-25 20:29:07 +08:00
; SSE-NEXT: pcmpeqd %xmm1, %xmm1
; SSE-NEXT: paddd %xmm3, %xmm1
; SSE-NEXT: pand %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX1-LABEL: combine_vec_urem_by_pow2d:
[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
; AVX1: # %bb.0:
[SelectionDAG] Recognise splat vector isKnownToBeAPowerOfTwo one/sign bit shift cases. llvm-svn: 301303
2017-04-25 20:29:07 +08:00
; AVX1-NEXT: vpsrldq {{.*#+}} xmm2 = xmm1[12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [2147483648,2147483648,2147483648,2147483648]
; AVX1-NEXT: vpsrld %xmm2, %xmm3, %xmm2
; AVX1-NEXT: vpsrlq $32, %xmm1, %xmm4
; AVX1-NEXT: vpsrld %xmm4, %xmm3, %xmm4
; AVX1-NEXT: vpblendw {{.*#+}} xmm2 = xmm4[0,1,2,3],xmm2[4,5,6,7]
; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
; AVX1-NEXT: vpunpckhdq {{.*#+}} xmm4 = xmm1[2],xmm4[2],xmm1[3],xmm4[3]
; AVX1-NEXT: vpsrld %xmm4, %xmm3, %xmm4
; AVX1-NEXT: vpmovzxdq {{.*#+}} xmm1 = xmm1[0],zero,xmm1[1],zero
; AVX1-NEXT: vpsrld %xmm1, %xmm3, %xmm1
; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1,2,3],xmm4[4,5,6,7]
; AVX1-NEXT: vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_pow2d:
[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
; AVX2: # %bb.0:
[X86][AVX] Regenerate combine tests with constant broadcast comments llvm-svn: 308128
2017-07-16 19:36:11 +08:00
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm2 = [2147483648,2147483648,2147483648,2147483648]
[SelectionDAG] Recognise splat vector isKnownToBeAPowerOfTwo one/sign bit shift cases. llvm-svn: 301303
2017-04-25 20:29:07 +08:00
; AVX2-NEXT: vpsrlvd %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
%1 = lshr <4 x i32> <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648>, %y
%2 = urem <4 x i32> %x, %1
ret <4 x i32> %2
}
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
; fold (urem x, (shl pow2, y)) -> (and x, (add (shl pow2, y), -1))
define <4 x i32> @combine_vec_urem_by_shl_pow2a(<4 x i32> %x, <4 x i32> %y) {
; SSE-LABEL: combine_vec_urem_by_shl_pow2a:
[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
; SSE: # %bb.0:
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
; SSE-NEXT: pslld $23, %xmm1
; SSE-NEXT: paddd {{.*}}(%rip), %xmm1
[DAGCombiner] Enable (urem x, (shl pow2, y)) -> (and x, (add (shl pow2, y), -1)) combine for splatted vectors llvm-svn: 285129
2016-10-26 06:01:09 +08:00
; SSE-NEXT: cvttps2dq %xmm1, %xmm1
; SSE-NEXT: pslld $2, %xmm1
; SSE-NEXT: pcmpeqd %xmm2, %xmm2
; SSE-NEXT: paddd %xmm1, %xmm2
; SSE-NEXT: pand %xmm2, %xmm0
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
; SSE-NEXT: retq
;
[X86][SSE] Add AVX1 tests to sdiv/udiv srem/urem combine tests As requested on D27714 llvm-svn: 289652
2016-12-14 22:39:51 +08:00
; AVX1-LABEL: combine_vec_urem_by_shl_pow2a:
[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
; AVX1: # %bb.0:
[X86][SSE] Add AVX1 tests to sdiv/udiv srem/urem combine tests As requested on D27714 llvm-svn: 289652
2016-12-14 22:39:51 +08:00
; AVX1-NEXT: vpslld $23, %xmm1, %xmm1
; AVX1-NEXT: vpaddd {{.*}}(%rip), %xmm1, %xmm1
; AVX1-NEXT: vcvttps2dq %xmm1, %xmm1
; AVX1-NEXT: vpslld $2, %xmm1, %xmm1
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_shl_pow2a:
[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
; AVX2: # %bb.0:
[X86][AVX] Regenerate combine tests with constant broadcast comments llvm-svn: 308128
2017-07-16 19:36:11 +08:00
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm2 = [4,4,4,4]
[X86][SSE] Add AVX1 tests to sdiv/udiv srem/urem combine tests As requested on D27714 llvm-svn: 289652
2016-12-14 22:39:51 +08:00
; AVX2-NEXT: vpsllvd %xmm1, %xmm2, %xmm1
; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX2-NEXT: retq
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
%1 = shl <4 x i32> <i32 4, i32 4, i32 4, i32 4>, %y
%2 = urem <4 x i32> %x, %1
ret <4 x i32> %2
}
define <4 x i32> @combine_vec_urem_by_shl_pow2b(<4 x i32> %x, <4 x i32> %y) {
; SSE-LABEL: combine_vec_urem_by_shl_pow2b:
[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
; SSE: # %bb.0:
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
; SSE-NEXT: pslld $23, %xmm1
; SSE-NEXT: paddd {{.*}}(%rip), %xmm1
[DAGCombiner] Try to use SelectionDAG::isKnownToBeAPowerOfTwo instead of just APInt::isPowerOf2 Generalize sdiv/udiv/srem/urem combines using APInt::isPowerOf2, which only works for const/splat-const values, to call SelectionDAG::isKnownToBeAPowerOfTwo instead which recognises many more cases. Added a DAGCombiner::BuildLogBase2 helper since PowerOf2 combines often involve taking the log2 of such a value. Differential Revision: https://reviews.llvm.org/D27714 llvm-svn: 289654
2016-12-14 23:08:13 +08:00
; SSE-NEXT: cvttps2dq %xmm1, %xmm1
; SSE-NEXT: pmulld {{.*}}(%rip), %xmm1
; SSE-NEXT: pcmpeqd %xmm2, %xmm2
; SSE-NEXT: paddd %xmm1, %xmm2
; SSE-NEXT: pand %xmm2, %xmm0
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
; SSE-NEXT: retq
;
[X86][SSE] Add AVX1 tests to sdiv/udiv srem/urem combine tests As requested on D27714 llvm-svn: 289652
2016-12-14 22:39:51 +08:00
; AVX1-LABEL: combine_vec_urem_by_shl_pow2b:
[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
; AVX1: # %bb.0:
[X86][SSE] Add AVX1 tests to sdiv/udiv srem/urem combine tests As requested on D27714 llvm-svn: 289652
2016-12-14 22:39:51 +08:00
; AVX1-NEXT: vpslld $23, %xmm1, %xmm1
; AVX1-NEXT: vpaddd {{.*}}(%rip), %xmm1, %xmm1
; AVX1-NEXT: vcvttps2dq %xmm1, %xmm1
; AVX1-NEXT: vpmulld {{.*}}(%rip), %xmm1, %xmm1
[DAGCombiner] Try to use SelectionDAG::isKnownToBeAPowerOfTwo instead of just APInt::isPowerOf2 Generalize sdiv/udiv/srem/urem combines using APInt::isPowerOf2, which only works for const/splat-const values, to call SelectionDAG::isKnownToBeAPowerOfTwo instead which recognises many more cases. Added a DAGCombiner::BuildLogBase2 helper since PowerOf2 combines often involve taking the log2 of such a value. Differential Revision: https://reviews.llvm.org/D27714 llvm-svn: 289654
2016-12-14 23:08:13 +08:00
; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX1-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX1-NEXT: vpand %xmm1, %xmm0, %xmm0
[X86][SSE] Add AVX1 tests to sdiv/udiv srem/urem combine tests As requested on D27714 llvm-svn: 289652
2016-12-14 22:39:51 +08:00
; AVX1-NEXT: retq
;
; AVX2-LABEL: combine_vec_urem_by_shl_pow2b:
[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
; AVX2: # %bb.0:
[X86][SSE] Add AVX1 tests to sdiv/udiv srem/urem combine tests As requested on D27714 llvm-svn: 289652
2016-12-14 22:39:51 +08:00
; AVX2-NEXT: vmovdqa {{.*#+}} xmm2 = [1,4,8,16]
; AVX2-NEXT: vpsllvd %xmm1, %xmm2, %xmm1
[DAGCombiner] Try to use SelectionDAG::isKnownToBeAPowerOfTwo instead of just APInt::isPowerOf2 Generalize sdiv/udiv/srem/urem combines using APInt::isPowerOf2, which only works for const/splat-const values, to call SelectionDAG::isKnownToBeAPowerOfTwo instead which recognises many more cases. Added a DAGCombiner::BuildLogBase2 helper since PowerOf2 combines often involve taking the log2 of such a value. Differential Revision: https://reviews.llvm.org/D27714 llvm-svn: 289654
2016-12-14 23:08:13 +08:00
; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
; AVX2-NEXT: vpaddd %xmm2, %xmm1, %xmm1
; AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
[X86][SSE] Add AVX1 tests to sdiv/udiv srem/urem combine tests As requested on D27714 llvm-svn: 289652
2016-12-14 22:39:51 +08:00
; AVX2-NEXT: retq
[X86][SSE] Added vector urem combine tests llvm-svn: 285119
2016-10-26 05:10:12 +08:00
%1 = shl <4 x i32> <i32 1, i32 4, i32 8, i32 16>, %y
%2 = urem <4 x i32> %x, %1
ret <4 x i32> %2
}