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/SystemZ/and-04.ll

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

184 lines
4.0 KiB
LLVM
Raw Normal View History

[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; Test 64-bit ANDs in which the second operand is constant.
;
; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; Use RISBG for a single bit.
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
define i64 @f1(i64 %a) {
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; CHECK-LABEL: f1:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: risbg %r2, %r2, 63, 191, 0
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
%and = and i64 %a, 1
ret i64 %and
}
; Likewise 0xfffe.
define i64 @f2(i64 %a) {
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; CHECK-LABEL: f2:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: risbg %r2, %r2, 48, 190, 0
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
%and = and i64 %a, 65534
ret i64 %and
}
; ...but 0xffff is a 16-bit zero extension.
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
define i64 @f3(i64 %a, i64 %b) {
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; CHECK-LABEL: f3:
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; CHECK: llghr %r2, %r3
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
%and = and i64 %b, 65535
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; Check the next value up, which can again use RISBG.
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
define i64 @f4(i64 %a) {
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; CHECK-LABEL: f4:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: risbg %r2, %r2, 47, 175, 0
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
%and = and i64 %a, 65536
ret i64 %and
}
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; Check 0xfffffffe, which can also use RISBG.
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
define i64 @f5(i64 %a) {
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; CHECK-LABEL: f5:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: risbg %r2, %r2, 32, 190, 0
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
%and = and i64 %a, 4294967294
ret i64 %and
}
; Check the next value up, which is a 32-bit zero extension.
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
define i64 @f6(i64 %a, i64 %b) {
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; CHECK-LABEL: f6:
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; CHECK: llgfr %r2, %r3
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
%and = and i64 %b, 4294967295
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; Check the lowest useful NIHF value (0x00000001_ffffffff).
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
define i64 @f7(i64 %a) {
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; CHECK-LABEL: f7:
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; CHECK: nihf %r2, 1
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
%and = and i64 %a, 8589934591
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; ...but RISBG can be used if a three-address form is useful.
define i64 @f8(i64 %a, i64 %b) {
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; CHECK-LABEL: f8:
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; CHECK: risbg %r2, %r3, 31, 191, 0
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
%and = and i64 %b, 8589934591
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; Check the lowest NIHH value outside the RISBG range (0x0002ffff_ffffffff).
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
define i64 @f9(i64 %a) {
Mass update to CodeGen tests to use CHECK-LABEL for labels corresponding to function definitions for more informative error messages. No functionality change and all updated tests passed locally. This update was done with the following bash script: find test/CodeGen -name "*.ll" | \ while read NAME; do echo "$NAME" if ! grep -q "^; *RUN: *llc.*debug" $NAME; then TEMP=`mktemp -t temp` cp $NAME $TEMP sed -n "s/^define [^@]*@\([A-Za-z0-9_]*\)(.*$/\1/p" < $NAME | \ while read FUNC; do sed -i '' "s/;\(.*\)\([A-Za-z0-9_-]*\):\( *\)$FUNC: *\$/;\1\2-LABEL:\3$FUNC:/g" $TEMP done sed -i '' "s/;\(.*\)-LABEL-LABEL:/;\1-LABEL:/" $TEMP sed -i '' "s/;\(.*\)-NEXT-LABEL:/;\1-NEXT:/" $TEMP sed -i '' "s/;\(.*\)-NOT-LABEL:/;\1-NOT:/" $TEMP sed -i '' "s/;\(.*\)-DAG-LABEL:/;\1-DAG:/" $TEMP mv $TEMP $NAME fi done llvm-svn: 186280
2013-07-14 14:24:09 +08:00
; CHECK-LABEL: f9:
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; CHECK: nihh %r2, 2
; CHECK: br %r14
%and = and i64 %a, 844424930131967
ret i64 %and
}
; Check the highest NIHH value outside the RISBG range (0xfffaffff_ffffffff).
define i64 @f10(i64 %a) {
; CHECK-LABEL: f10:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: nihh %r2, 65530
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
%and = and i64 %a, -1407374883553281
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
; Check the highest useful NIHF value (0xfffefffe_ffffffff).
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
define i64 @f11(i64 %a) {
; CHECK-LABEL: f11:
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: nihf %r2, 4294901758
; CHECK: br %r14
%and = and i64 %a, -281479271677953
ret i64 %and
}
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; Check the lowest NIHL value outside the RISBG range (0xffff0002_ffffffff).
define i64 @f12(i64 %a) {
; CHECK-LABEL: f12:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: nihl %r2, 2
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
%and = and i64 %a, -281462091808769
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; Check the highest NIHL value outside the RISBG range (0xfffffffa_ffffffff).
define i64 @f13(i64 %a) {
; CHECK-LABEL: f13:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: nihl %r2, 65530
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
%and = and i64 %a, -21474836481
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; Check the lowest NILF value outside the RISBG range (0xffffffff_00000002).
define i64 @f14(i64 %a) {
; CHECK-LABEL: f14:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: nilf %r2, 2
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
%and = and i64 %a, -4294967294
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; Check the lowest NILH value outside the RISBG range (0xffffffff_0002ffff).
define i64 @f15(i64 %a) {
; CHECK-LABEL: f15:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: nilh %r2, 2
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
%and = and i64 %a, -4294770689
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
; Check the next value up, which must use NILF.
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
define i64 @f16(i64 %a) {
; CHECK-LABEL: f16:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: nilf %r2, 196608
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
%and = and i64 %a, -4294770688
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; Check the highest NILH value outside the RISBG range (0xffffffff_fffaffff).
define i64 @f17(i64 %a) {
; CHECK-LABEL: f17:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: nilh %r2, 65530
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
%and = and i64 %a, -327681
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; Check the maximum useful NILF value (0xffffffff_fffefffe).
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
define i64 @f18(i64 %a) {
; CHECK-LABEL: f18:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: nilf %r2, 4294901758
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
%and = and i64 %a, -65538
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; Check the lowest NILL value outside the RISBG range (0xffffffff_ffff0002).
define i64 @f19(i64 %a) {
; CHECK-LABEL: f19:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: nill %r2, 2
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
%and = and i64 %a, -65534
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}
[SystemZ] Postpone NI->RISBG conversion to convertToThreeAddress() r186399 aggressively used the RISBG instruction for immediate ANDs, both because it can handle some values that AND IMMEDIATE can't, and because it allows the destination register to be different from the source. I realized later while implementing the distinct-ops support that it would be better to leave the choice up to convertToThreeAddress() instead. The AND IMMEDIATE form is shorter and is less likely to be cracked. This is a problem for 32-bit ANDs because we assume that all 32-bit operations will leave the high word untouched, whereas RISBG used in this way will either clear the high word or copy it from the source register. The patch uses the z196 instruction RISBLG for this instead. This means that z10 will be restricted to NILL, NILH and NILF for 32-bit ANDs, but I think that should be OK for now. Although we're using z10 as the base architecture, the optimization work is going to be focused more on z196 and zEC12. llvm-svn: 187492
2013-07-31 19:36:35 +08:00
; Check the highest NILL value outside the RISBG range.
define i64 @f20(i64 %a) {
; CHECK-LABEL: f20:
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
; CHECK: nill %r2, 65530
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
; CHECK: br %r14
[SystemZ] Use zeroing form of RISBG for some AND sequences RISBG can handle some ANDs for which no AND IMMEDIATE exists. It also acts as a three-operand AND for some cases where an AND IMMEDIATE could be used instead. It might be worth adding a pass to replace RISBG with AND IMMEDIATE in cases where the register operands end up being the same and where AND IMMEDIATE is smaller. llvm-svn: 186072
2013-07-11 16:59:12 +08:00
%and = and i64 %a, -6
[SystemZ] Add CodeGen test cases This adds all CodeGen tests for the SystemZ target. This version of the patch incorporates feedback from a review by Sean Silva. Thanks to all reviewers! Patch by Richard Sandiford. llvm-svn: 181204
2013-05-07 00:17:29 +08:00
ret i64 %and
}