Author: Juneyoung Lee Date: 2020-12-30T23:50:37+09:00 New Revision: c6035a7bdf207dc3c1e12c041d5658770893fc68
URL: https://github.com/llvm/llvm-project/commit/c6035a7bdf207dc3c1e12c041d5658770893fc68 DIFF: https://github.com/llvm/llvm-project/commit/c6035a7bdf207dc3c1e12c041d5658770893fc68.diff LOG: Remove functions from *-inseltpoison.ll tests if unnecessary X-inseltpoison.ll is a copy of X.ll with insertelement/shufflevector's placeholder replaced with poison. This commit removes a few redundant functions which do not contain any shufflevector/insertelement. Added: Modified: llvm/test/Transforms/InstCombine/assume-inseltpoison.ll llvm/test/Transforms/InstCombine/bswap-inseltpoison.ll llvm/test/Transforms/InstCombine/fmul-inseltpoison.ll Removed: ################################################################################ diff --git a/llvm/test/Transforms/InstCombine/assume-inseltpoison.ll b/llvm/test/Transforms/InstCombine/assume-inseltpoison.ll index 8c04c4af28ced..f49755570ddc9 100644 --- a/llvm/test/Transforms/InstCombine/assume-inseltpoison.ll +++ b/llvm/test/Transforms/InstCombine/assume-inseltpoison.ll @@ -4,368 +4,10 @@ target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" target triple = "x86_64-unknown-linux-gnu" +; A copy of assume.ll, with undef at insertelement/shufflevector replaced with +; poison. declare void @llvm.assume(i1) #1 -; Check that the alignment has been upgraded and that the assume has not -; been removed: - -define i32 @foo1(i32* %a) #0 { -; CHECK-LABEL: @foo1( -; CHECK-NEXT: [[T0:%.*]] = load i32, i32* [[A:%.*]], align 32 -; CHECK-NEXT: [[PTRINT:%.*]] = ptrtoint i32* [[A]] to i64 -; CHECK-NEXT: [[MASKEDPTR:%.*]] = and i64 [[PTRINT]], 31 -; CHECK-NEXT: [[MASKCOND:%.*]] = icmp eq i64 [[MASKEDPTR]], 0 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[MASKCOND]]) -; CHECK-NEXT: ret i32 [[T0]] -; - %t0 = load i32, i32* %a, align 4 - %ptrint = ptrtoint i32* %a to i64 - %maskedptr = and i64 %ptrint, 31 - %maskcond = icmp eq i64 %maskedptr, 0 - tail call void @llvm.assume(i1 %maskcond) - ret i32 %t0 -} - -; Same check as in @foo1, but make sure it works if the assume is first too. - -define i32 @foo2(i32* %a) #0 { -; CHECK-LABEL: @foo2( -; CHECK-NEXT: [[PTRINT:%.*]] = ptrtoint i32* [[A:%.*]] to i64 -; CHECK-NEXT: [[MASKEDPTR:%.*]] = and i64 [[PTRINT]], 31 -; CHECK-NEXT: [[MASKCOND:%.*]] = icmp eq i64 [[MASKEDPTR]], 0 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[MASKCOND]]) -; CHECK-NEXT: [[T0:%.*]] = load i32, i32* [[A]], align 32 -; CHECK-NEXT: ret i32 [[T0]] -; - %ptrint = ptrtoint i32* %a to i64 - %maskedptr = and i64 %ptrint, 31 - %maskcond = icmp eq i64 %maskedptr, 0 - tail call void @llvm.assume(i1 %maskcond) - %t0 = load i32, i32* %a, align 4 - ret i32 %t0 -} - -define i32 @simple(i32 %a) #1 { -; CHECK-LABEL: @simple( -; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[A:%.*]], 4 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: ret i32 4 -; - %cmp = icmp eq i32 %a, 4 - tail call void @llvm.assume(i1 %cmp) - ret i32 %a -} - -define i32 @can1(i1 %a, i1 %b, i1 %c) { -; CHECK-LABEL: @can1( -; CHECK-NEXT: call void @llvm.assume(i1 [[A:%.*]]) -; CHECK-NEXT: call void @llvm.assume(i1 [[B:%.*]]) -; CHECK-NEXT: call void @llvm.assume(i1 [[C:%.*]]) -; CHECK-NEXT: ret i32 5 -; - %and1 = and i1 %a, %b - %and = and i1 %and1, %c - tail call void @llvm.assume(i1 %and) - ret i32 5 -} - -define i32 @can2(i1 %a, i1 %b, i1 %c) { -; CHECK-LABEL: @can2( -; CHECK-NEXT: [[TMP1:%.*]] = xor i1 [[A:%.*]], true -; CHECK-NEXT: call void @llvm.assume(i1 [[TMP1]]) -; CHECK-NEXT: [[TMP2:%.*]] = xor i1 [[B:%.*]], true -; CHECK-NEXT: call void @llvm.assume(i1 [[TMP2]]) -; CHECK-NEXT: ret i32 5 -; - %v = or i1 %a, %b - %w = xor i1 %v, 1 - tail call void @llvm.assume(i1 %w) - ret i32 5 -} - -define i32 @bar1(i32 %a) #0 { -; CHECK-LABEL: @bar1( -; CHECK-NEXT: [[AND:%.*]] = and i32 [[A:%.*]], 7 -; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[AND]], 1 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: ret i32 1 -; - %and1 = and i32 %a, 3 - %and = and i32 %a, 7 - %cmp = icmp eq i32 %and, 1 - tail call void @llvm.assume(i1 %cmp) - ret i32 %and1 -} - -define i32 @bar2(i32 %a) #0 { -; CHECK-LABEL: @bar2( -; CHECK-NEXT: [[AND:%.*]] = and i32 [[A:%.*]], 7 -; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[AND]], 1 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: ret i32 1 -; - %and = and i32 %a, 7 - %cmp = icmp eq i32 %and, 1 - tail call void @llvm.assume(i1 %cmp) - %and1 = and i32 %a, 3 - ret i32 %and1 -} - -define i32 @bar3(i32 %a, i1 %x, i1 %y) #0 { -; CHECK-LABEL: @bar3( -; CHECK-NEXT: entry: -; CHECK-NEXT: tail call void @llvm.assume(i1 [[X:%.*]]) -; CHECK-NEXT: [[AND:%.*]] = and i32 [[A:%.*]], 7 -; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[AND]], 1 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: tail call void @llvm.assume(i1 [[Y:%.*]]) -; CHECK-NEXT: ret i32 1 -; -entry: - %and1 = and i32 %a, 3 - -; Don't be fooled by other assumes around. - - tail call void @llvm.assume(i1 %x) - - %and = and i32 %a, 7 - %cmp = icmp eq i32 %and, 1 - tail call void @llvm.assume(i1 %cmp) - - tail call void @llvm.assume(i1 %y) - - ret i32 %and1 -} - -define i32 @bar4(i32 %a, i32 %b) { -; CHECK-LABEL: @bar4( -; CHECK-NEXT: entry: -; CHECK-NEXT: [[AND:%.*]] = and i32 [[A:%.*]], 7 -; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[AND]], 1 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: [[CMP2:%.*]] = icmp eq i32 [[A]], [[B:%.*]] -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP2]]) -; CHECK-NEXT: ret i32 1 -; -entry: - %and1 = and i32 %b, 3 - %and = and i32 %a, 7 - %cmp = icmp eq i32 %and, 1 - tail call void @llvm.assume(i1 %cmp) - %cmp2 = icmp eq i32 %a, %b - tail call void @llvm.assume(i1 %cmp2) - ret i32 %and1 -} - -define i32 @icmp1(i32 %a) #0 { -; CHECK-LABEL: @icmp1( -; CHECK-NEXT: [[CMP:%.*]] = icmp sgt i32 [[A:%.*]], 5 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: ret i32 1 -; - %cmp = icmp sgt i32 %a, 5 - tail call void @llvm.assume(i1 %cmp) - %conv = zext i1 %cmp to i32 - ret i32 %conv -} - -define i32 @icmp2(i32 %a) #0 { -; CHECK-LABEL: @icmp2( -; CHECK-NEXT: [[CMP:%.*]] = icmp sgt i32 [[A:%.*]], 5 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: ret i32 0 -; - %cmp = icmp sgt i32 %a, 5 - tail call void @llvm.assume(i1 %cmp) - %t0 = zext i1 %cmp to i32 - %lnot.ext = xor i32 %t0, 1 - ret i32 %lnot.ext -} - -; If the 'not' of a condition is known true, then the condition must be false. - -define i1 @assume_not(i1 %cond) { -; CHECK-LABEL: @assume_not( -; CHECK-NEXT: [[NOTCOND:%.*]] = xor i1 [[COND:%.*]], true -; CHECK-NEXT: call void @llvm.assume(i1 [[NOTCOND]]) -; CHECK-NEXT: ret i1 false -; - %notcond = xor i1 %cond, true - call void @llvm.assume(i1 %notcond) - ret i1 %cond -} - -declare void @escape(i32* %a) - -; Canonicalize a nonnull assumption on a load into metadata form. - -define i32 @bundle1(i32* %P) { -; CHECK-LABEL: @bundle1( -; CHECK-NEXT: tail call void @llvm.assume(i1 true) [ "nonnull"(i32* [[P:%.*]]) ] -; CHECK-NEXT: [[LOAD:%.*]] = load i32, i32* [[P]], align 4 -; CHECK-NEXT: ret i32 [[LOAD]] -; - tail call void @llvm.assume(i1 true) ["nonnull"(i32* %P)] - %load = load i32, i32* %P - ret i32 %load -} - -define i32 @bundle2(i32* %P) { -; CHECK-LABEL: @bundle2( -; CHECK-NEXT: [[LOAD:%.*]] = load i32, i32* [[P:%.*]], align 4 -; CHECK-NEXT: ret i32 [[LOAD]] -; - tail call void @llvm.assume(i1 true) ["ignore"(i32* undef)] - %load = load i32, i32* %P - ret i32 %load -} - -define i1 @nonnull1(i32** %a) { -; CHECK-LABEL: @nonnull1( -; CHECK-NEXT: [[LOAD:%.*]] = load i32*, i32** [[A:%.*]], align 8, !nonnull !6 -; CHECK-NEXT: tail call void @escape(i32* nonnull [[LOAD]]) -; CHECK-NEXT: ret i1 false -; - %load = load i32*, i32** %a - %cmp = icmp ne i32* %load, null - tail call void @llvm.assume(i1 %cmp) - tail call void @escape(i32* %load) - %rval = icmp eq i32* %load, null - ret i1 %rval -} - -; Make sure the above canonicalization applies only -; to pointer types. Doing otherwise would be illegal. - -define i1 @nonnull2(i32* %a) { -; CHECK-LABEL: @nonnull2( -; CHECK-NEXT: [[LOAD:%.*]] = load i32, i32* [[A:%.*]], align 4 -; CHECK-NEXT: [[CMP:%.*]] = icmp ne i32 [[LOAD]], 0 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: ret i1 false -; - %load = load i32, i32* %a - %cmp = icmp ne i32 %load, 0 - tail call void @llvm.assume(i1 %cmp) - %rval = icmp eq i32 %load, 0 - ret i1 %rval -} - -; Make sure the above canonicalization does not trigger -; if the assume is control dependent on something else - -define i1 @nonnull3(i32** %a, i1 %control) { -; CHECK-LABEL: @nonnull3( -; CHECK-NEXT: entry: -; CHECK-NEXT: [[LOAD:%.*]] = load i32*, i32** [[A:%.*]], align 8 -; CHECK-NEXT: [[CMP:%.*]] = icmp ne i32* [[LOAD]], null -; CHECK-NEXT: br i1 [[CONTROL:%.*]], label [[TAKEN:%.*]], label [[NOT_TAKEN:%.*]] -; CHECK: taken: -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: ret i1 false -; CHECK: not_taken: -; CHECK-NEXT: [[RVAL_2:%.*]] = icmp sgt i32* [[LOAD]], null -; CHECK-NEXT: ret i1 [[RVAL_2]] -; -entry: - %load = load i32*, i32** %a - %cmp = icmp ne i32* %load, null - br i1 %control, label %taken, label %not_taken -taken: - tail call void @llvm.assume(i1 %cmp) - %rval = icmp eq i32* %load, null - ret i1 %rval -not_taken: - %rval.2 = icmp sgt i32* %load, null - ret i1 %rval.2 -} - -; Make sure the above canonicalization does not trigger -; if the path from the load to the assume is potentially -; interrupted by an exception being thrown - -define i1 @nonnull4(i32** %a) { -; CHECK-LABEL: @nonnull4( -; CHECK-NEXT: [[LOAD:%.*]] = load i32*, i32** [[A:%.*]], align 8 -; CHECK-NEXT: tail call void @escape(i32* [[LOAD]]) -; CHECK-NEXT: [[CMP:%.*]] = icmp ne i32* [[LOAD]], null -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: ret i1 false -; - %load = load i32*, i32** %a - ;; This call may throw! - tail call void @escape(i32* %load) - %cmp = icmp ne i32* %load, null - tail call void @llvm.assume(i1 %cmp) - %rval = icmp eq i32* %load, null - ret i1 %rval -} -define i1 @nonnull5(i32** %a) { -; CHECK-LABEL: @nonnull5( -; CHECK-NEXT: [[LOAD:%.*]] = load i32*, i32** [[A:%.*]], align 8 -; CHECK-NEXT: tail call void @escape(i32* [[LOAD]]) -; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32* [[LOAD]], null -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: ret i1 false -; - %load = load i32*, i32** %a - ;; This call may throw! - tail call void @escape(i32* %load) - %integral = ptrtoint i32* %load to i64 - %cmp = icmp slt i64 %integral, 0 - tail call void @llvm.assume(i1 %cmp) ; %load has at least highest bit set - %rval = icmp eq i32* %load, null - ret i1 %rval -} - -; PR35846 - https://bugs.llvm.org/show_bug.cgi?id=35846 - -define i32 @assumption_conflicts_with_known_bits(i32 %a, i32 %b) { -; CHECK-LABEL: @assumption_conflicts_with_known_bits( -; CHECK-NEXT: [[AND1:%.*]] = and i32 [[B:%.*]], 3 -; CHECK-NEXT: tail call void @llvm.assume(i1 false) -; CHECK-NEXT: [[CMP2:%.*]] = icmp eq i32 [[AND1]], 0 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP2]]) -; CHECK-NEXT: ret i32 0 -; - %and1 = and i32 %b, 3 - %B1 = lshr i32 %and1, %and1 - %B3 = shl nuw nsw i32 %and1, %B1 - %cmp = icmp eq i32 %B3, 1 - tail call void @llvm.assume(i1 %cmp) - %cmp2 = icmp eq i32 %B1, %B3 - tail call void @llvm.assume(i1 %cmp2) - ret i32 %and1 -} - -; PR37726 - https://bugs.llvm.org/show_bug.cgi?id=37726 -; There's a loophole in eliminating a redundant assumption when -; we have conflicting assumptions. Verify that debuginfo doesn't -; get in the way of the fold. - -define void @debug_interference(i8 %x) { -; CHECK-LABEL: @debug_interference( -; CHECK-NEXT: [[CMP2:%.*]] = icmp ne i8 [[X:%.*]], 0 -; CHECK-NEXT: tail call void @llvm.assume(i1 false) -; CHECK-NEXT: tail call void @llvm.dbg.value(metadata i32 5, [[META7:metadata !.*]], metadata !DIExpression()), [[DBG9:!dbg !.*]] -; CHECK-NEXT: tail call void @llvm.assume(i1 false) -; CHECK-NEXT: tail call void @llvm.dbg.value(metadata i32 5, [[META7]], metadata !DIExpression()), [[DBG9]] -; CHECK-NEXT: tail call void @llvm.dbg.value(metadata i32 5, [[META7]], metadata !DIExpression()), [[DBG9]] -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP2]]) -; CHECK-NEXT: ret void -; - %cmp1 = icmp eq i8 %x, 0 - %cmp2 = icmp ne i8 %x, 0 - tail call void @llvm.assume(i1 %cmp1) - tail call void @llvm.dbg.value(metadata i32 5, metadata !1, metadata !DIExpression()), !dbg !9 - tail call void @llvm.assume(i1 %cmp1) - tail call void @llvm.dbg.value(metadata i32 5, metadata !1, metadata !DIExpression()), !dbg !9 - tail call void @llvm.assume(i1 %cmp2) - tail call void @llvm.dbg.value(metadata i32 5, metadata !1, metadata !DIExpression()), !dbg !9 - tail call void @llvm.assume(i1 %cmp2) - ret void -} - ; This would crash. ; Does it ever make sense to peek through a bitcast of the icmp operand? @@ -383,274 +25,3 @@ define i32 @PR40940(<4 x i8> %x) { call void @llvm.assume(i1 %t3) ret i32 %t2 } - -define i1 @nonnull3A(i32** %a, i1 %control) { -; CHECK-LABEL: @nonnull3A( -; CHECK-NEXT: entry: -; CHECK-NEXT: [[LOAD:%.*]] = load i32*, i32** [[A:%.*]], align 8 -; CHECK-NEXT: br i1 [[CONTROL:%.*]], label [[TAKEN:%.*]], label [[NOT_TAKEN:%.*]] -; CHECK: taken: -; CHECK-NEXT: [[CMP:%.*]] = icmp ne i32* [[LOAD]], null -; CHECK-NEXT: call void @llvm.assume(i1 [[CMP]]) -; CHECK-NEXT: ret i1 true -; CHECK: not_taken: -; CHECK-NEXT: [[RVAL_2:%.*]] = icmp sgt i32* [[LOAD]], null -; CHECK-NEXT: ret i1 [[RVAL_2]] -; -entry: - %load = load i32*, i32** %a - %cmp = icmp ne i32* %load, null - br i1 %control, label %taken, label %not_taken -taken: - call void @llvm.assume(i1 %cmp) - ret i1 %cmp -not_taken: - call void @llvm.assume(i1 %cmp) - %rval.2 = icmp sgt i32* %load, null - ret i1 %rval.2 -} - -define i1 @nonnull3B(i32** %a, i1 %control) { -; CHECK-LABEL: @nonnull3B( -; CHECK-NEXT: entry: -; CHECK-NEXT: br i1 [[CONTROL:%.*]], label [[TAKEN:%.*]], label [[NOT_TAKEN:%.*]] -; CHECK: taken: -; CHECK-NEXT: [[LOAD:%.*]] = load i32*, i32** [[A:%.*]], align 8 -; CHECK-NEXT: [[CMP:%.*]] = icmp ne i32* [[LOAD]], null -; CHECK-NEXT: call void @llvm.assume(i1 [[CMP]]) [ "nonnull"(i32* [[LOAD]]), "nonnull"(i1 [[CMP]]) ] -; CHECK-NEXT: ret i1 true -; CHECK: not_taken: -; CHECK-NEXT: ret i1 [[CONTROL]] -; -entry: - %load = load i32*, i32** %a - %cmp = icmp ne i32* %load, null - br i1 %control, label %taken, label %not_taken -taken: - call void @llvm.assume(i1 %cmp) ["nonnull"(i32* %load), "nonnull"(i1 %cmp)] - ret i1 %cmp -not_taken: - call void @llvm.assume(i1 %cmp) ["nonnull"(i32* %load), "nonnull"(i1 %cmp)] - ret i1 %control -} - -declare i1 @tmp1(i1) - -define i1 @nonnull3C(i32** %a, i1 %control) { -; CHECK-LABEL: @nonnull3C( -; CHECK-NEXT: entry: -; CHECK-NEXT: br i1 [[CONTROL:%.*]], label [[TAKEN:%.*]], label [[NOT_TAKEN:%.*]] -; CHECK: taken: -; CHECK-NEXT: [[LOAD:%.*]] = load i32*, i32** [[A:%.*]], align 8 -; CHECK-NEXT: [[CMP:%.*]] = icmp ne i32* [[LOAD]], null -; CHECK-NEXT: [[CMP2:%.*]] = call i1 @tmp1(i1 [[CMP]]) -; CHECK-NEXT: br label [[EXIT:%.*]] -; CHECK: exit: -; CHECK-NEXT: ret i1 [[CMP2]] -; CHECK: not_taken: -; CHECK-NEXT: ret i1 [[CONTROL]] -; -entry: - %load = load i32*, i32** %a - %cmp = icmp ne i32* %load, null - br i1 %control, label %taken, label %not_taken -taken: - %cmp2 = call i1 @tmp1(i1 %cmp) - br label %exit -exit: - ; FIXME: this shouldn't be dropped because it is still dominated by the new position of %load - call void @llvm.assume(i1 %cmp) ["nonnull"(i32* %load), "nonnull"(i1 %cmp)] - ret i1 %cmp2 -not_taken: - call void @llvm.assume(i1 %cmp) - ret i1 %control -} - -define i1 @nonnull3D(i32** %a, i1 %control) { -; CHECK-LABEL: @nonnull3D( -; CHECK-NEXT: entry: -; CHECK-NEXT: br i1 [[CONTROL:%.*]], label [[TAKEN:%.*]], label [[NOT_TAKEN:%.*]] -; CHECK: taken: -; CHECK-NEXT: [[LOAD:%.*]] = load i32*, i32** [[A:%.*]], align 8 -; CHECK-NEXT: [[CMP:%.*]] = icmp ne i32* [[LOAD]], null -; CHECK-NEXT: [[CMP2:%.*]] = call i1 @tmp1(i1 [[CMP]]) -; CHECK-NEXT: br label [[EXIT:%.*]] -; CHECK: exit: -; CHECK-NEXT: ret i1 [[CMP2]] -; CHECK: not_taken: -; CHECK-NEXT: call void @llvm.assume(i1 true) [ "ignore"(i32* undef), "ignore"(i1 undef), "nonnull"(i1 [[CONTROL]]) ] -; CHECK-NEXT: ret i1 [[CONTROL]] -; -entry: - %load = load i32*, i32** %a - %cmp = icmp ne i32* %load, null - br i1 %control, label %taken, label %not_taken -taken: - %cmp2 = call i1 @tmp1(i1 %cmp) - br label %exit -exit: - ret i1 %cmp2 -not_taken: - call void @llvm.assume(i1 %cmp) ["nonnull"(i32* %load), "nonnull"(i1 %cmp), "nonnull"(i1 %control)] - ret i1 %control -} - - -define void @always_true_assumption() { -; CHECK-LABEL: @always_true_assumption( -; CHECK-NEXT: ret void -; - call void @llvm.assume(i1 true) - ret void -} - -; The alloca guarantees that the low bits of %a are zero because of alignment. -; The assume says the opposite. Make sure we don't crash. - -define i64 @PR31809() { -; CHECK-LABEL: @PR31809( -; CHECK-NEXT: [[A:%.*]] = alloca i32, align 4 -; CHECK-NEXT: [[T1:%.*]] = ptrtoint i32* [[A]] to i64 -; CHECK-NEXT: call void @llvm.assume(i1 false) -; CHECK-NEXT: ret i64 [[T1]] -; - %a = alloca i32 - %t1 = ptrtoint i32* %a to i64 - %cond = icmp eq i64 %t1, 3 - call void @llvm.assume(i1 %cond) - ret i64 %t1 -} - -; Similar to above: there's no way to know which assumption is truthful, -; so just don't crash. - -define i8 @conflicting_assumptions(i8 %x){ -; CHECK-LABEL: @conflicting_assumptions( -; CHECK-NEXT: call void @llvm.assume(i1 false) -; CHECK-NEXT: [[COND2:%.*]] = icmp eq i8 [[X:%.*]], 4 -; CHECK-NEXT: call void @llvm.assume(i1 [[COND2]]) -; CHECK-NEXT: ret i8 5 -; - %add = add i8 %x, 1 - %cond1 = icmp eq i8 %x, 3 - call void @llvm.assume(i1 %cond1) - %cond2 = icmp eq i8 %x, 4 - call void @llvm.assume(i1 %cond2) - ret i8 %add -} - -; Another case of conflicting assumptions. This would crash because we'd -; try to set more known bits than existed in the known bits struct. - -define void @PR36270(i32 %b) { -; CHECK-LABEL: @PR36270( -; CHECK-NEXT: unreachable -; - %B7 = xor i32 -1, 2147483647 - %and1 = and i32 %b, 3 - %B12 = lshr i32 %B7, %and1 - %C1 = icmp ult i32 %and1, %B12 - tail call void @llvm.assume(i1 %C1) - %cmp2 = icmp eq i32 0, %B12 - tail call void @llvm.assume(i1 %cmp2) - unreachable -} - -; PR47416 - -define i32 @unreachable_assume(i32 %x, i32 %y) { -; CHECK-LABEL: @unreachable_assume( -; CHECK-NEXT: entry: -; CHECK-NEXT: [[CMP0:%.*]] = icmp sgt i32 [[X:%.*]], 1 -; CHECK-NEXT: [[CMP1:%.*]] = icmp eq i32 [[Y:%.*]], 1 -; CHECK-NEXT: [[OR:%.*]] = or i1 [[CMP0]], [[CMP1]] -; CHECK-NEXT: tail call void @llvm.assume(i1 [[OR]]) -; CHECK-NEXT: [[CMP2:%.*]] = icmp eq i32 [[X]], 1 -; CHECK-NEXT: br i1 [[CMP2]], label [[IF:%.*]], label [[EXIT:%.*]] -; CHECK: if: -; CHECK-NEXT: [[A:%.*]] = and i32 [[Y]], -2 -; CHECK-NEXT: [[CMP3:%.*]] = icmp ne i32 [[A]], 104 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP3]]) -; CHECK-NEXT: br label [[EXIT]] -; CHECK: exit: -; CHECK-NEXT: unreachable -; -entry: - %cmp0 = icmp sgt i32 %x, 1 - %cmp1 = icmp eq i32 %y, 1 - %or = or i1 %cmp0, %cmp1 - tail call void @llvm.assume(i1 %or) - %cmp2 = icmp eq i32 %x, 1 - br i1 %cmp2, label %if, label %exit - -if: - %a = and i32 %y, -2 - %cmp3 = icmp ne i32 %a, 104 - tail call void @llvm.assume(i1 %cmp3) - br label %exit - -exit: - %cmp4 = icmp eq i32 %x, 2 - tail call void @llvm.assume(i1 %cmp4) - unreachable -} - -define i32 @unreachable_assumes_and_store(i32 %x, i32 %y, i32* %p) { -; CHECK-LABEL: @unreachable_assumes_and_store( -; CHECK-NEXT: entry: -; CHECK-NEXT: [[CMP0:%.*]] = icmp sgt i32 [[X:%.*]], 1 -; CHECK-NEXT: [[CMP1:%.*]] = icmp eq i32 [[Y:%.*]], 1 -; CHECK-NEXT: [[OR:%.*]] = or i1 [[CMP0]], [[CMP1]] -; CHECK-NEXT: tail call void @llvm.assume(i1 [[OR]]) -; CHECK-NEXT: [[CMP2:%.*]] = icmp eq i32 [[X]], 1 -; CHECK-NEXT: br i1 [[CMP2]], label [[IF:%.*]], label [[EXIT:%.*]] -; CHECK: if: -; CHECK-NEXT: [[A:%.*]] = and i32 [[Y]], -2 -; CHECK-NEXT: [[CMP3:%.*]] = icmp ne i32 [[A]], 104 -; CHECK-NEXT: tail call void @llvm.assume(i1 [[CMP3]]) -; CHECK-NEXT: br label [[EXIT]] -; CHECK: exit: -; CHECK-NEXT: unreachable -; -entry: - %cmp0 = icmp sgt i32 %x, 1 - %cmp1 = icmp eq i32 %y, 1 - %or = or i1 %cmp0, %cmp1 - tail call void @llvm.assume(i1 %or) - %cmp2 = icmp eq i32 %x, 1 - br i1 %cmp2, label %if, label %exit - -if: - %a = and i32 %y, -2 - %cmp3 = icmp ne i32 %a, 104 - tail call void @llvm.assume(i1 %cmp3) - br label %exit - -exit: - %cmp4 = icmp eq i32 %x, 2 - tail call void @llvm.assume(i1 %cmp4) - %cmp5 = icmp ugt i32 %y, 42 - tail call void @llvm.assume(i1 %cmp5) - store i32 %x, i32* %p - unreachable -} - -declare void @llvm.dbg.value(metadata, metadata, metadata) - -!llvm.dbg.cu = !{!0} -!llvm.module.flags = !{!5, !6, !7, !8} - -!0 = distinct !DICompileUnit(language: DW_LANG_C, file: !3, producer: "Me", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: null, retainedTypes: null, imports: null) -!1 = !DILocalVariable(name: "", arg: 1, scope: !2, file: null, line: 1, type: null) -!2 = distinct !DISubprogram(name: "debug", linkageName: "debug", scope: null, file: null, line: 0, type: null, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, unit: !0) -!3 = !DIFile(filename: "consecutive-fences.ll", directory: "") -!5 = !{i32 2, !"Dwarf Version", i32 4} -!6 = !{i32 2, !"Debug Info Version", i32 3} -!7 = !{i32 1, !"wchar_size", i32 4} -!8 = !{i32 7, !"PIC Level", i32 2} -!9 = !DILocation(line: 0, column: 0, scope: !2) - - -attributes #0 = { nounwind uwtable } -attributes #1 = { nounwind } - diff --git a/llvm/test/Transforms/InstCombine/bswap-inseltpoison.ll b/llvm/test/Transforms/InstCombine/bswap-inseltpoison.ll index 3730496ba6e33..7d5bcbed1447d 100644 --- a/llvm/test/Transforms/InstCombine/bswap-inseltpoison.ll +++ b/llvm/test/Transforms/InstCombine/bswap-inseltpoison.ll @@ -3,629 +3,8 @@ target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32" -define i32 @test1(i32 %i) { -; CHECK-LABEL: @test1( -; CHECK-NEXT: [[T12:%.*]] = call i32 @llvm.bswap.i32(i32 [[I:%.*]]) -; CHECK-NEXT: ret i32 [[T12]] -; - %t1 = lshr i32 %i, 24 - %t3 = lshr i32 %i, 8 - %t4 = and i32 %t3, 65280 - %t5 = or i32 %t1, %t4 - %t7 = shl i32 %i, 8 - %t8 = and i32 %t7, 16711680 - %t9 = or i32 %t5, %t8 - %t11 = shl i32 %i, 24 - %t12 = or i32 %t9, %t11 - ret i32 %t12 -} - -define <2 x i32> @test1_vector(<2 x i32> %i) { -; CHECK-LABEL: @test1_vector( -; CHECK-NEXT: [[T12:%.*]] = call <2 x i32> @llvm.bswap.v2i32(<2 x i32> [[I:%.*]]) -; CHECK-NEXT: ret <2 x i32> [[T12]] -; - %t1 = lshr <2 x i32> %i, <i32 24, i32 24> - %t3 = lshr <2 x i32> %i, <i32 8, i32 8> - %t4 = and <2 x i32> %t3, <i32 65280, i32 65280> - %t5 = or <2 x i32> %t1, %t4 - %t7 = shl <2 x i32> %i, <i32 8, i32 8> - %t8 = and <2 x i32> %t7, <i32 16711680, i32 16711680> - %t9 = or <2 x i32> %t5, %t8 - %t11 = shl <2 x i32> %i, <i32 24, i32 24> - %t12 = or <2 x i32> %t9, %t11 - ret <2 x i32> %t12 -} - -define i32 @test2(i32 %arg) { -; CHECK-LABEL: @test2( -; CHECK-NEXT: [[T14:%.*]] = call i32 @llvm.bswap.i32(i32 [[ARG:%.*]]) -; CHECK-NEXT: ret i32 [[T14]] -; - %t2 = shl i32 %arg, 24 - %t4 = shl i32 %arg, 8 - %t5 = and i32 %t4, 16711680 - %t6 = or i32 %t2, %t5 - %t8 = lshr i32 %arg, 8 - %t9 = and i32 %t8, 65280 - %t10 = or i32 %t6, %t9 - %t12 = lshr i32 %arg, 24 - %t14 = or i32 %t10, %t12 - ret i32 %t14 -} - -define <2 x i32> @test2_vector(<2 x i32> %arg) { -; CHECK-LABEL: @test2_vector( -; CHECK-NEXT: [[T14:%.*]] = call <2 x i32> @llvm.bswap.v2i32(<2 x i32> [[ARG:%.*]]) -; CHECK-NEXT: ret <2 x i32> [[T14]] -; - %t2 = shl <2 x i32> %arg, <i32 24, i32 24> - %t4 = shl <2 x i32> %arg, <i32 8, i32 8> - %t5 = and <2 x i32> %t4, <i32 16711680, i32 16711680> - %t6 = or <2 x i32> %t2, %t5 - %t8 = lshr <2 x i32> %arg, <i32 8, i32 8> - %t9 = and <2 x i32> %t8, <i32 65280, i32 65280> - %t10 = or <2 x i32> %t6, %t9 - %t12 = lshr <2 x i32> %arg, <i32 24, i32 24> - %t14 = or <2 x i32> %t10, %t12 - ret <2 x i32> %t14 -} - -define <2 x i32> @test2_vector_undef(<2 x i32> %arg) { -; CHECK-LABEL: @test2_vector_undef( -; CHECK-NEXT: [[T2:%.*]] = shl <2 x i32> [[ARG:%.*]], <i32 24, i32 undef> -; CHECK-NEXT: [[T4:%.*]] = shl <2 x i32> [[ARG]], <i32 8, i32 8> -; CHECK-NEXT: [[T5:%.*]] = and <2 x i32> [[T4]], <i32 16711680, i32 undef> -; CHECK-NEXT: [[T6:%.*]] = or <2 x i32> [[T2]], [[T5]] -; CHECK-NEXT: [[T8:%.*]] = lshr <2 x i32> [[ARG]], <i32 8, i32 8> -; CHECK-NEXT: [[T9:%.*]] = and <2 x i32> [[T8]], <i32 65280, i32 undef> -; CHECK-NEXT: [[T10:%.*]] = or <2 x i32> [[T6]], [[T9]] -; CHECK-NEXT: [[T12:%.*]] = lshr <2 x i32> [[ARG]], <i32 24, i32 undef> -; CHECK-NEXT: [[T14:%.*]] = or <2 x i32> [[T10]], [[T12]] -; CHECK-NEXT: ret <2 x i32> [[T14]] -; - %t2 = shl <2 x i32> %arg, <i32 24, i32 undef> - %t4 = shl <2 x i32> %arg, <i32 8, i32 8> - %t5 = and <2 x i32> %t4, <i32 16711680, i32 undef> - %t6 = or <2 x i32> %t2, %t5 - %t8 = lshr <2 x i32> %arg, <i32 8, i32 8> - %t9 = and <2 x i32> %t8, <i32 65280, i32 undef> - %t10 = or <2 x i32> %t6, %t9 - %t12 = lshr <2 x i32> %arg, <i32 24, i32 undef> - %t14 = or <2 x i32> %t10, %t12 - ret <2 x i32> %t14 -} - -define i16 @test3(i16 %s) { -; CHECK-LABEL: @test3( -; CHECK-NEXT: [[T5:%.*]] = call i16 @llvm.bswap.i16(i16 [[S:%.*]]) -; CHECK-NEXT: ret i16 [[T5]] -; - %t2 = lshr i16 %s, 8 - %t4 = shl i16 %s, 8 - %t5 = or i16 %t2, %t4 - ret i16 %t5 -} - -define <2 x i16> @test3_vector(<2 x i16> %s) { -; CHECK-LABEL: @test3_vector( -; CHECK-NEXT: [[T5:%.*]] = call <2 x i16> @llvm.bswap.v2i16(<2 x i16> [[S:%.*]]) -; CHECK-NEXT: ret <2 x i16> [[T5]] -; - %t2 = lshr <2 x i16> %s, <i16 8, i16 8> - %t4 = shl <2 x i16> %s, <i16 8, i16 8> - %t5 = or <2 x i16> %t2, %t4 - ret <2 x i16> %t5 -} - -define <2 x i16> @test3_vector_undef(<2 x i16> %s) { -; CHECK-LABEL: @test3_vector_undef( -; CHECK-NEXT: [[T5:%.*]] = call <2 x i16> @llvm.bswap.v2i16(<2 x i16> [[S:%.*]]) -; CHECK-NEXT: ret <2 x i16> [[T5]] -; - %t2 = lshr <2 x i16> %s, <i16 undef, i16 8> - %t4 = shl <2 x i16> %s, <i16 8, i16 undef> - %t5 = or <2 x i16> %t2, %t4 - ret <2 x i16> %t5 -} - -define i16 @test4(i16 %s) { -; CHECK-LABEL: @test4( -; CHECK-NEXT: [[T5:%.*]] = call i16 @llvm.bswap.i16(i16 [[S:%.*]]) -; CHECK-NEXT: ret i16 [[T5]] -; - %t2 = lshr i16 %s, 8 - %t4 = shl i16 %s, 8 - %t5 = or i16 %t4, %t2 - ret i16 %t5 -} - -define <2 x i16> @test4_vector(<2 x i16> %s) { -; CHECK-LABEL: @test4_vector( -; CHECK-NEXT: [[T5:%.*]] = call <2 x i16> @llvm.bswap.v2i16(<2 x i16> [[S:%.*]]) -; CHECK-NEXT: ret <2 x i16> [[T5]] -; - %t2 = lshr <2 x i16> %s, <i16 8, i16 8> - %t4 = shl <2 x i16> %s, <i16 8, i16 8> - %t5 = or <2 x i16> %t4, %t2 - ret <2 x i16> %t5 -} - -define i16 @test5(i16 %a) { -; CHECK-LABEL: @test5( -; CHECK-NEXT: [[T_UPGRD_3:%.*]] = call i16 @llvm.bswap.i16(i16 [[A:%.*]]) -; CHECK-NEXT: ret i16 [[T_UPGRD_3]] -; - %t = zext i16 %a to i32 - %t1 = and i32 %t, 65280 - %t2 = ashr i32 %t1, 8 - %t2.upgrd.1 = trunc i32 %t2 to i16 - %t4 = and i32 %t, 255 - %t5 = shl i32 %t4, 8 - %t5.upgrd.2 = trunc i32 %t5 to i16 - %t.upgrd.3 = or i16 %t2.upgrd.1, %t5.upgrd.2 - %t6 = bitcast i16 %t.upgrd.3 to i16 - %t6.upgrd.4 = zext i16 %t6 to i32 - %retval = trunc i32 %t6.upgrd.4 to i16 - ret i16 %retval -} - -define <2 x i16> @test5_vector(<2 x i16> %a) { -; CHECK-LABEL: @test5_vector( -; CHECK-NEXT: [[T_UPGRD_3:%.*]] = call <2 x i16> @llvm.bswap.v2i16(<2 x i16> [[A:%.*]]) -; CHECK-NEXT: ret <2 x i16> [[T_UPGRD_3]] -; - %t = zext <2 x i16> %a to <2 x i32> - %t1 = and <2 x i32> %t, <i32 65280, i32 65280> - %t2 = ashr <2 x i32> %t1, <i32 8, i32 8> - %t2.upgrd.1 = trunc <2 x i32> %t2 to <2 x i16> - %t4 = and <2 x i32> %t, <i32 255, i32 255> - %t5 = shl <2 x i32> %t4, <i32 8, i32 8> - %t5.upgrd.2 = trunc <2 x i32> %t5 to <2 x i16> - %t.upgrd.3 = or <2 x i16> %t2.upgrd.1, %t5.upgrd.2 - %t6 = bitcast <2 x i16> %t.upgrd.3 to <2 x i16> - %t6.upgrd.4 = zext <2 x i16> %t6 to <2 x i32> - %retval = trunc <2 x i32> %t6.upgrd.4 to <2 x i16> - ret <2 x i16> %retval -} - -; PR2842 -define i32 @test6(i32 %x) nounwind readnone { -; CHECK-LABEL: @test6( -; CHECK-NEXT: [[T7:%.*]] = call i32 @llvm.bswap.i32(i32 [[X:%.*]]) -; CHECK-NEXT: ret i32 [[T7]] -; - %t = shl i32 %x, 16 - %x.mask = and i32 %x, 65280 - %t1 = lshr i32 %x, 16 - %t2 = and i32 %t1, 255 - %t3 = or i32 %x.mask, %t - %t4 = or i32 %t3, %t2 - %t5 = shl i32 %t4, 8 - %t6 = lshr i32 %x, 24 - %t7 = or i32 %t5, %t6 - ret i32 %t7 -} - -define <2 x i32> @test6_vector(<2 x i32> %x) nounwind readnone { -; CHECK-LABEL: @test6_vector( -; CHECK-NEXT: [[T7:%.*]] = call <2 x i32> @llvm.bswap.v2i32(<2 x i32> [[X:%.*]]) -; CHECK-NEXT: ret <2 x i32> [[T7]] -; - %t = shl <2 x i32> %x, <i32 16, i32 16> - %x.mask = and <2 x i32> %x, <i32 65280, i32 65280> - %t1 = lshr <2 x i32> %x, <i32 16, i32 16> - %t2 = and <2 x i32> %t1, <i32 255, i32 255> - %t3 = or <2 x i32> %x.mask, %t - %t4 = or <2 x i32> %t3, %t2 - %t5 = shl <2 x i32> %t4, <i32 8, i32 8> - %t6 = lshr <2 x i32> %x, <i32 24, i32 24> - %t7 = or <2 x i32> %t5, %t6 - ret <2 x i32> %t7 -} - -declare void @extra_use(i32) - -; swaphalf = (x << 16 | x >> 16) -; ((swaphalf & 0x00ff00ff) << 8) | ((swaphalf >> 8) & 0x00ff00ff) - -define i32 @bswap32_and_first(i32 %x) { -; CHECK-LABEL: @bswap32_and_first( -; CHECK-NEXT: [[BSWAP:%.*]] = call i32 @llvm.bswap.i32(i32 [[X:%.*]]) -; CHECK-NEXT: ret i32 [[BSWAP]] -; - %shl = shl i32 %x, 16 - %shr = lshr i32 %x, 16 - %swaphalf = or i32 %shl, %shr - %t = and i32 %swaphalf, 16711935 - %tshl = shl nuw i32 %t, 8 - %b = lshr i32 %swaphalf, 8 - %band = and i32 %b, 16711935 - %bswap = or i32 %tshl, %band - ret i32 %bswap -} - -; Extra use should not prevent matching to bswap. -; swaphalf = (x << 16 | x >> 16) -; ((swaphalf & 0x00ff00ff) << 8) | ((swaphalf >> 8) & 0x00ff00ff) - -define i32 @bswap32_and_first_extra_use(i32 %x) { -; CHECK-LABEL: @bswap32_and_first_extra_use( -; CHECK-NEXT: [[SWAPHALF:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 16) -; CHECK-NEXT: [[T:%.*]] = and i32 [[SWAPHALF]], 16711935 -; CHECK-NEXT: [[BSWAP:%.*]] = call i32 @llvm.bswap.i32(i32 [[X]]) -; CHECK-NEXT: call void @extra_use(i32 [[T]]) -; CHECK-NEXT: ret i32 [[BSWAP]] -; - %shl = shl i32 %x, 16 - %shr = lshr i32 %x, 16 - %swaphalf = or i32 %shl, %shr - %t = and i32 %swaphalf, 16711935 - %tshl = shl nuw i32 %t, 8 - %b = lshr i32 %swaphalf, 8 - %band = and i32 %b, 16711935 - %bswap = or i32 %tshl, %band - call void @extra_use(i32 %t) - ret i32 %bswap -} - -; swaphalf = (x << 16 | x >> 16) -; ((swaphalf << 8) & 0xff00ff00) | ((swaphalf >> 8) & 0x00ff00ff) - -; PR23863 -define i32 @bswap32_shl_first(i32 %x) { -; CHECK-LABEL: @bswap32_shl_first( -; CHECK-NEXT: [[BSWAP:%.*]] = call i32 @llvm.bswap.i32(i32 [[X:%.*]]) -; CHECK-NEXT: ret i32 [[BSWAP]] -; - %shl = shl i32 %x, 16 - %shr = lshr i32 %x, 16 - %swaphalf = or i32 %shl, %shr - %t = shl i32 %swaphalf, 8 - %tand = and i32 %t, -16711936 - %b = lshr i32 %swaphalf, 8 - %band = and i32 %b, 16711935 - %bswap = or i32 %tand, %band - ret i32 %bswap -} - -; Extra use should not prevent matching to bswap. -; swaphalf = (x << 16 | x >> 16) -; ((swaphalf << 8) & 0xff00ff00) | ((swaphalf >> 8) & 0x00ff00ff) - -define i32 @bswap32_shl_first_extra_use(i32 %x) { -; CHECK-LABEL: @bswap32_shl_first_extra_use( -; CHECK-NEXT: [[SWAPHALF:%.*]] = call i32 @llvm.fshl.i32(i32 [[X:%.*]], i32 [[X]], i32 16) -; CHECK-NEXT: [[T:%.*]] = shl i32 [[SWAPHALF]], 8 -; CHECK-NEXT: [[BSWAP:%.*]] = call i32 @llvm.bswap.i32(i32 [[X]]) -; CHECK-NEXT: call void @extra_use(i32 [[T]]) -; CHECK-NEXT: ret i32 [[BSWAP]] -; - %shl = shl i32 %x, 16 - %shr = lshr i32 %x, 16 - %swaphalf = or i32 %shl, %shr - %t = shl i32 %swaphalf, 8 - %tand = and i32 %t, -16711936 - %b = lshr i32 %swaphalf, 8 - %band = and i32 %b, 16711935 - %bswap = or i32 %tand, %band - call void @extra_use(i32 %t) - ret i32 %bswap -} - -define i16 @test8(i16 %a) { -; CHECK-LABEL: @test8( -; CHECK-NEXT: [[OR:%.*]] = call i16 @llvm.bswap.i16(i16 [[A:%.*]]) -; CHECK-NEXT: ret i16 [[OR]] -; - %conv = zext i16 %a to i32 - %shr = lshr i16 %a, 8 - %shl = shl i32 %conv, 8 - %conv1 = zext i16 %shr to i32 - %or = or i32 %conv1, %shl - %conv2 = trunc i32 %or to i16 - ret i16 %conv2 -} - -define i16 @test9(i16 %a) { -; CHECK-LABEL: @test9( -; CHECK-NEXT: [[OR:%.*]] = call i16 @llvm.bswap.i16(i16 [[A:%.*]]) -; CHECK-NEXT: ret i16 [[OR]] -; - %conv = zext i16 %a to i32 - %shr = lshr i32 %conv, 8 - %shl = shl i32 %conv, 8 - %or = or i32 %shr, %shl - %conv2 = trunc i32 %or to i16 - ret i16 %conv2 -} - -define i16 @test10(i32 %a) { -; CHECK-LABEL: @test10( -; CHECK-NEXT: [[TRUNC:%.*]] = trunc i32 [[A:%.*]] to i16 -; CHECK-NEXT: [[REV:%.*]] = call i16 @llvm.bswap.i16(i16 [[TRUNC]]) -; CHECK-NEXT: ret i16 [[REV]] -; - %shr1 = lshr i32 %a, 8 - %and1 = and i32 %shr1, 255 - %and2 = shl i32 %a, 8 - %shl1 = and i32 %and2, 65280 - %or = or i32 %and1, %shl1 - %conv = trunc i32 %or to i16 - ret i16 %conv -} - -define <2 x i16> @test10_vector(<2 x i32> %a) { -; CHECK-LABEL: @test10_vector( -; CHECK-NEXT: [[TRUNC:%.*]] = trunc <2 x i32> [[A:%.*]] to <2 x i16> -; CHECK-NEXT: [[REV:%.*]] = call <2 x i16> @llvm.bswap.v2i16(<2 x i16> [[TRUNC]]) -; CHECK-NEXT: ret <2 x i16> [[REV]] -; - %shr1 = lshr <2 x i32> %a, <i32 8, i32 8> - %and1 = and <2 x i32> %shr1, <i32 255, i32 255> - %and2 = shl <2 x i32> %a, <i32 8, i32 8> - %shl1 = and <2 x i32> %and2, <i32 65280, i32 65280> - %or = or <2 x i32> %and1, %shl1 - %conv = trunc <2 x i32> %or to <2 x i16> - ret <2 x i16> %conv -} - -define i64 @PR39793_bswap_u64_as_u32(i64 %0) { -; CHECK-LABEL: @PR39793_bswap_u64_as_u32( -; CHECK-NEXT: [[TRUNC:%.*]] = trunc i64 [[TMP0:%.*]] to i32 -; CHECK-NEXT: [[REV:%.*]] = call i32 @llvm.bswap.i32(i32 [[TRUNC]]) -; CHECK-NEXT: [[TMP2:%.*]] = zext i32 [[REV]] to i64 -; CHECK-NEXT: ret i64 [[TMP2]] -; - %2 = lshr i64 %0, 24 - %3 = and i64 %2, 255 - %4 = lshr i64 %0, 8 - %5 = and i64 %4, 65280 - %6 = or i64 %3, %5 - %7 = shl i64 %0, 8 - %8 = and i64 %7, 16711680 - %9 = or i64 %6, %8 - %10 = shl i64 %0, 24 - %11 = and i64 %10, 4278190080 - %12 = or i64 %9, %11 - ret i64 %12 -} - -define i16 @PR39793_bswap_u64_as_u32_trunc(i64 %0) { -; CHECK-LABEL: @PR39793_bswap_u64_as_u32_trunc( -; CHECK-NEXT: [[TRUNC:%.*]] = trunc i64 [[TMP0:%.*]] to i32 -; CHECK-NEXT: [[REV:%.*]] = call i32 @llvm.bswap.i32(i32 [[TRUNC]]) -; CHECK-NEXT: [[TMP2:%.*]] = trunc i32 [[REV]] to i16 -; CHECK-NEXT: ret i16 [[TMP2]] -; - %2 = lshr i64 %0, 24 - %3 = and i64 %2, 255 - %4 = lshr i64 %0, 8 - %5 = and i64 %4, 65280 - %6 = or i64 %3, %5 - %7 = shl i64 %0, 8 - %8 = and i64 %7, 16711680 - %9 = or i64 %6, %8 - %10 = shl i64 %0, 24 - %11 = and i64 %10, 4278190080 - %12 = or i64 %9, %11 - %13 = trunc i64 %12 to i16 - ret i16 %13 -} - -define i64 @PR39793_bswap_u64_as_u16(i64 %0) { -; CHECK-LABEL: @PR39793_bswap_u64_as_u16( -; CHECK-NEXT: [[TRUNC:%.*]] = trunc i64 [[TMP0:%.*]] to i16 -; CHECK-NEXT: [[REV:%.*]] = call i16 @llvm.bswap.i16(i16 [[TRUNC]]) -; CHECK-NEXT: [[TMP2:%.*]] = zext i16 [[REV]] to i64 -; CHECK-NEXT: ret i64 [[TMP2]] -; - %2 = lshr i64 %0, 8 - %3 = and i64 %2, 255 - %4 = shl i64 %0, 8 - %5 = and i64 %4, 65280 - %6 = or i64 %3, %5 - ret i64 %6 -} - -define <2 x i64> @PR39793_bswap_u64_as_u16_vector(<2 x i64> %0) { -; CHECK-LABEL: @PR39793_bswap_u64_as_u16_vector( -; CHECK-NEXT: [[TRUNC:%.*]] = trunc <2 x i64> [[TMP0:%.*]] to <2 x i16> -; CHECK-NEXT: [[REV:%.*]] = call <2 x i16> @llvm.bswap.v2i16(<2 x i16> [[TRUNC]]) -; CHECK-NEXT: [[TMP2:%.*]] = zext <2 x i16> [[REV]] to <2 x i64> -; CHECK-NEXT: ret <2 x i64> [[TMP2]] -; - %2 = lshr <2 x i64> %0, <i64 8, i64 8> - %3 = and <2 x i64> %2, <i64 255, i64 255> - %4 = shl <2 x i64> %0, <i64 8, i64 8> - %5 = and <2 x i64> %4, <i64 65280, i64 65280> - %6 = or <2 x i64> %3, %5 - ret <2 x i64> %6 -} - -define i8 @PR39793_bswap_u64_as_u16_trunc(i64 %0) { -; CHECK-LABEL: @PR39793_bswap_u64_as_u16_trunc( -; CHECK-NEXT: [[REV1:%.*]] = lshr i64 [[TMP0:%.*]], 8 -; CHECK-NEXT: [[TMP2:%.*]] = trunc i64 [[REV1]] to i8 -; CHECK-NEXT: ret i8 [[TMP2]] -; - %2 = lshr i64 %0, 8 - %3 = and i64 %2, 255 - %4 = shl i64 %0, 8 - %5 = and i64 %4, 65280 - %6 = or i64 %3, %5 - %7 = trunc i64 %6 to i8 - ret i8 %7 -} - -define i50 @PR39793_bswap_u50_as_u16(i50 %0) { -; CHECK-LABEL: @PR39793_bswap_u50_as_u16( -; CHECK-NEXT: [[TRUNC:%.*]] = trunc i50 [[TMP0:%.*]] to i16 -; CHECK-NEXT: [[REV:%.*]] = call i16 @llvm.bswap.i16(i16 [[TRUNC]]) -; CHECK-NEXT: [[TMP2:%.*]] = zext i16 [[REV]] to i50 -; CHECK-NEXT: ret i50 [[TMP2]] -; - %2 = lshr i50 %0, 8 - %3 = and i50 %2, 255 - %4 = shl i50 %0, 8 - %5 = and i50 %4, 65280 - %6 = or i50 %3, %5 - ret i50 %6 -} - -define i32 @PR39793_bswap_u32_as_u16(i32 %0) { -; CHECK-LABEL: @PR39793_bswap_u32_as_u16( -; CHECK-NEXT: [[TRUNC:%.*]] = trunc i32 [[TMP0:%.*]] to i16 -; CHECK-NEXT: [[REV:%.*]] = call i16 @llvm.bswap.i16(i16 [[TRUNC]]) -; CHECK-NEXT: [[TMP2:%.*]] = zext i16 [[REV]] to i32 -; CHECK-NEXT: ret i32 [[TMP2]] -; - %2 = lshr i32 %0, 8 - %3 = and i32 %2, 255 - %4 = shl i32 %0, 8 - %5 = and i32 %4, 65280 - %6 = or i32 %3, %5 - ret i32 %6 -} - -define i8 @PR39793_bswap_u32_as_u16_trunc(i32 %0) { -; CHECK-LABEL: @PR39793_bswap_u32_as_u16_trunc( -; CHECK-NEXT: [[REV1:%.*]] = lshr i32 [[TMP0:%.*]], 8 -; CHECK-NEXT: [[TMP2:%.*]] = trunc i32 [[REV1]] to i8 -; CHECK-NEXT: ret i8 [[TMP2]] -; - %2 = lshr i32 %0, 8 - %3 = and i32 %2, 255 - %4 = shl i32 %0, 8 - %5 = and i32 %4, 65280 - %6 = or i32 %3, %5 - %7 = trunc i32 %6 to i8 - ret i8 %7 -} - -define i32 @partial_bswap(i32 %x) { -; CHECK-LABEL: @partial_bswap( -; CHECK-NEXT: [[TMP1:%.*]] = call i32 @llvm.bswap.i32(i32 [[X:%.*]]) -; CHECK-NEXT: ret i32 [[TMP1]] -; - %x3 = shl i32 %x, 24 - %a2 = shl i32 %x, 8 - %x2 = and i32 %a2, 16711680 - %x32 = or i32 %x3, %x2 - %t1 = and i32 %x, -65536 - %t2 = call i32 @llvm.bswap.i32(i32 %t1) - %r = or i32 %x32, %t2 - ret i32 %r -} -declare i32 @llvm.bswap.i32(i32) - -define <2 x i32> @partial_bswap_vector(<2 x i32> %x) { -; CHECK-LABEL: @partial_bswap_vector( -; CHECK-NEXT: [[TMP1:%.*]] = call <2 x i32> @llvm.bswap.v2i32(<2 x i32> [[X:%.*]]) -; CHECK-NEXT: ret <2 x i32> [[TMP1]] -; - %x3 = shl <2 x i32> %x, <i32 24, i32 24> - %a2 = shl <2 x i32> %x, <i32 8, i32 8> - %x2 = and <2 x i32> %a2, <i32 16711680, i32 16711680> - %x32 = or <2 x i32> %x3, %x2 - %t1 = and <2 x i32> %x, <i32 -65536, i32 -65536> - %t2 = call <2 x i32> @llvm.bswap.v2i32(<2 x i32> %t1) - %r = or <2 x i32> %x32, %t2 - ret <2 x i32> %r -} -declare <2 x i32> @llvm.bswap.v2i32(<2 x i32>) - -define i16 @partial_bitreverse(i16 %x) { -; CHECK-LABEL: @partial_bitreverse( -; CHECK-NEXT: [[OR:%.*]] = call i16 @llvm.bswap.i16(i16 [[X:%.*]]) -; CHECK-NEXT: ret i16 [[OR]] -; - %rev= call i16 @llvm.bitreverse.i16(i16 %x) - %lo = and i16 %rev, 255 - %hi = and i16 %rev, -256 - %revlo = call i16 @llvm.bitreverse.i16(i16 %lo) - %revhi = call i16 @llvm.bitreverse.i16(i16 %hi) - %newlo = lshr i16 %revlo, 8 - %newhi = shl i16 %revhi, 8 - %or = or i16 %newlo, %newhi - ret i16 %or -} -declare i16 @llvm.bitreverse.i16(i16) - -define i64 @bswap_and_mask_0(i64 %0) { -; CHECK-LABEL: @bswap_and_mask_0( -; CHECK-NEXT: [[TMP2:%.*]] = and i64 [[TMP0:%.*]], -72057594037927681 -; CHECK-NEXT: [[TMP3:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP2]]) -; CHECK-NEXT: ret i64 [[TMP3]] -; - %2 = lshr i64 %0, 56 - %3 = shl i64 %0, 56 - %4 = or i64 %2, %3 - ret i64 %4 -} - -define i64 @bswap_and_mask_1(i64 %0) { -; CHECK-LABEL: @bswap_and_mask_1( -; CHECK-NEXT: [[TMP2:%.*]] = lshr i64 [[TMP0:%.*]], 56 -; CHECK-NEXT: [[TMP3:%.*]] = lshr i64 [[TMP0]], 40 -; CHECK-NEXT: [[TMP4:%.*]] = and i64 [[TMP3]], 65280 -; CHECK-NEXT: [[TMP5:%.*]] = or i64 [[TMP4]], [[TMP2]] -; CHECK-NEXT: ret i64 [[TMP5]] -; - %2 = lshr i64 %0, 56 - %3 = lshr i64 %0, 40 - %4 = and i64 %3, 65280 - %5 = or i64 %4, %2 - ret i64 %5 -} - -define i64 @bswap_and_mask_2(i64 %0) { -; CHECK-LABEL: @bswap_and_mask_2( -; CHECK-NEXT: [[TMP2:%.*]] = and i64 [[TMP0:%.*]], -72057594037862401 -; CHECK-NEXT: [[TMP3:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP2]]) -; CHECK-NEXT: ret i64 [[TMP3]] -; - %2 = lshr i64 %0, 56 - %3 = shl i64 %0, 56 - %4 = or i64 %2, %3 - %5 = shl i64 %0, 40 - %6 = and i64 %5, 71776119061217280 - %7 = or i64 %4, %6 - ret i64 %7 -} - -define i64 @bswap_trunc(i64 %x01234567) { -; CHECK-LABEL: @bswap_trunc( -; CHECK-NEXT: [[X7ZZZZZZZ:%.*]] = shl i64 [[X01234567:%.*]], 56 -; CHECK-NEXT: [[XZ0123456:%.*]] = lshr i64 [[X01234567]], 8 -; CHECK-NEXT: [[XZZZZZ012:%.*]] = lshr i64 [[X01234567]], 40 -; CHECK-NEXT: [[X3456:%.*]] = trunc i64 [[XZ0123456]] to i32 -; CHECK-NEXT: [[XZ012:%.*]] = trunc i64 [[XZZZZZ012]] to i32 -; CHECK-NEXT: [[X6543:%.*]] = call i32 @llvm.bswap.i32(i32 [[X3456]]) -; CHECK-NEXT: [[X210Z:%.*]] = call i32 @llvm.bswap.i32(i32 [[XZ012]]) -; CHECK-NEXT: [[XZ210:%.*]] = lshr exact i32 [[X210Z]], 8 -; CHECK-NEXT: [[XZZZZ6543:%.*]] = zext i32 [[X6543]] to i64 -; CHECK-NEXT: [[XZZZZZ210:%.*]] = zext i32 [[XZ210]] to i64 -; CHECK-NEXT: [[XZ6543ZZZ:%.*]] = shl nuw nsw i64 [[XZZZZ6543]], 24 -; CHECK-NEXT: [[XZ6543210:%.*]] = or i64 [[XZ6543ZZZ]], [[XZZZZZ210]] -; CHECK-NEXT: [[X76543210:%.*]] = or i64 [[XZ6543210]], [[X7ZZZZZZZ]] -; CHECK-NEXT: ret i64 [[X76543210]] -; - %x7zzzzzzz = shl i64 %x01234567, 56 - %xz0123456 = lshr i64 %x01234567, 8 - %xzzzzz012 = lshr i64 %x01234567, 40 - %x3456 = trunc i64 %xz0123456 to i32 - %xz012 = trunc i64 %xzzzzz012 to i32 - %x6543 = call i32 @llvm.bswap.i32(i32 %x3456) - %x210z = call i32 @llvm.bswap.i32(i32 %xz012) - %xz210 = lshr i32 %x210z, 8 - %xzzzz6543 = zext i32 %x6543 to i64 - %xzzzzz210 = zext i32 %xz210 to i64 - %xz6543zzz = shl i64 %xzzzz6543, 24 - %xz6543210 = or i64 %xzzzzz210, %xz6543zzz - %x76543210 = or i64 %xz6543210, %x7zzzzzzz - ret i64 %x76543210 -} +; A copy of bswap.ll, with undef at insertelement/shufflevector replaced with +; poison define i32 @shuf_4bytes(<4 x i8> %x) { ; CHECK-LABEL: @shuf_4bytes( @@ -706,162 +85,3 @@ define i32 @shuf_2bytes_widening(<2 x i8> %x) { declare i32 @llvm.fshl.i32(i32, i32, i32) declare i32 @llvm.fshr.i32(i32, i32, i32) - -define i32 @funnel_unary(i32 %abcd) { -; CHECK-LABEL: @funnel_unary( -; CHECK-NEXT: [[DCBA:%.*]] = call i32 @llvm.bswap.i32(i32 [[ABCD:%.*]]) -; CHECK-NEXT: ret i32 [[DCBA]] -; - %dabc = call i32 @llvm.fshl.i32(i32 %abcd, i32 %abcd, i32 24) - %bcda = call i32 @llvm.fshr.i32(i32 %abcd, i32 %abcd, i32 24) - %dzbz = and i32 %dabc, -16711936 - %zcza = and i32 %bcda, 16711935 - %dcba = or i32 %dzbz, %zcza - ret i32 %dcba -} - -define i32 @funnel_binary(i32 %abcd) { -; CHECK-LABEL: @funnel_binary( -; CHECK-NEXT: [[DCBA:%.*]] = call i32 @llvm.bswap.i32(i32 [[ABCD:%.*]]) -; CHECK-NEXT: ret i32 [[DCBA]] -; - %cdzz = shl i32 %abcd, 16 - %dcdz = call i32 @llvm.fshl.i32(i32 %abcd, i32 %cdzz, i32 24) - %zzab = lshr i32 %abcd, 16 - %zaba = call i32 @llvm.fshr.i32(i32 %zzab, i32 %abcd, i32 24) - %dczz = and i32 %dcdz, -65536 - %zzba = and i32 %zaba, 65535 - %dcba = or i32 %dczz, %zzba - ret i32 %dcba -} - -define i32 @funnel_and(i32 %abcd) { -; CHECK-LABEL: @funnel_and( -; CHECK-NEXT: [[DCBA:%.*]] = call i32 @llvm.bswap.i32(i32 [[ABCD:%.*]]) -; CHECK-NEXT: ret i32 [[DCBA]] -; - %zzcz = and i32 %abcd, 65280 - %zcza = call i32 @llvm.fshl.i32(i32 %zzcz, i32 %abcd, i32 8) - %zbzz = and i32 %abcd, 16711680 - %dzbz = call i32 @llvm.fshl.i32(i32 %abcd, i32 %zbzz, i32 24) - %dcba = or i32 %zcza, %dzbz - ret i32 %dcba -} - -; PR47191 - deep IR trees prevent ADD/XOR instructions being simplified to OR. - -define i64 @PR47191_problem1(i64 %0) { -; CHECK-LABEL: @PR47191_problem1( -; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP0:%.*]]) -; CHECK-NEXT: ret i64 [[TMP2]] -; - %2 = lshr i64 %0, 56 - %3 = lshr i64 %0, 40 - %4 = and i64 %3, 65280 - %5 = lshr i64 %0, 24 - %6 = and i64 %5, 16711680 - %7 = lshr i64 %0, 8 - %8 = and i64 %7, 4278190080 - %9 = shl i64 %0, 56 - %10 = shl i64 %0, 40 - %11 = and i64 %10, 71776119061217280 - %12 = shl i64 %0, 24 - %13 = and i64 %12, 280375465082880 - %14 = or i64 %9, %2 - %15 = or i64 %14, %4 - %16 = or i64 %15, %6 - %17 = or i64 %16, %8 - %18 = or i64 %17, %11 - %19 = or i64 %18, %13 - %20 = shl i64 %0, 8 - %21 = and i64 %20, 1095216660480 - %22 = add i64 %19, %21 - ret i64 %22 -} - -define i64 @PR47191_problem2(i64 %0) { -; CHECK-LABEL: @PR47191_problem2( -; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP0:%.*]]) -; CHECK-NEXT: ret i64 [[TMP2]] -; - %2 = lshr i64 %0, 56 - %3 = lshr i64 %0, 40 - %4 = and i64 %3, 65280 - %5 = lshr i64 %0, 24 - %6 = and i64 %5, 16711680 - %7 = lshr i64 %0, 8 - %8 = and i64 %7, 4278190080 - %9 = shl i64 %0, 56 - %10 = shl i64 %0, 40 - %11 = and i64 %10, 71776119061217280 - %12 = or i64 %9, %2 - %13 = or i64 %12, %4 - %14 = or i64 %13, %6 - %15 = or i64 %14, %8 - %16 = or i64 %15, %11 - %17 = shl i64 %0, 24 - %18 = and i64 %17, 280375465082880 - %19 = shl i64 %0, 8 - %20 = and i64 %19, 1095216660480 - %21 = or i64 %20, %18 - %22 = xor i64 %21, %16 - ret i64 %22 -} - -define i64 @PR47191_problem3(i64 %0) { -; CHECK-LABEL: @PR47191_problem3( -; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP0:%.*]]) -; CHECK-NEXT: ret i64 [[TMP2]] -; - %2 = lshr i64 %0, 56 - %3 = lshr i64 %0, 40 - %4 = and i64 %3, 65280 - %5 = lshr i64 %0, 24 - %6 = and i64 %5, 16711680 - %7 = lshr i64 %0, 8 - %8 = and i64 %7, 4278190080 - %9 = shl i64 %0, 56 - %10 = shl i64 %0, 40 - %11 = and i64 %10, 71776119061217280 - %12 = or i64 %9, %2 - %13 = or i64 %12, %4 - %14 = or i64 %13, %6 - %15 = or i64 %14, %8 - %16 = or i64 %15, %11 - %17 = shl i64 %0, 24 - %18 = and i64 %17, 280375465082880 - %19 = shl i64 %0, 8 - %20 = and i64 %19, 1095216660480 - %21 = or i64 %20, %18 - %22 = xor i64 %21, %16 - ret i64 %22 -} - -define i64 @PR47191_problem4(i64 %0) { -; CHECK-LABEL: @PR47191_problem4( -; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.bswap.i64(i64 [[TMP0:%.*]]) -; CHECK-NEXT: ret i64 [[TMP2]] -; - %2 = lshr i64 %0, 56 - %3 = shl i64 %0, 56 - %4 = or i64 %2, %3 - %5 = lshr i64 %0, 40 - %6 = and i64 %5, 65280 - %7 = or i64 %4, %6 - %8 = shl i64 %0, 40 - %9 = and i64 %8, 71776119061217280 - %10 = or i64 %7, %9 - %11 = lshr i64 %0, 24 - %12 = and i64 %11, 16711680 - %13 = or i64 %10, %12 - %14 = shl i64 %0, 24 - %15 = and i64 %14, 280375465082880 - %16 = or i64 %13, %15 - %17 = lshr i64 %0, 8 - %18 = and i64 %17, 4278190080 - %19 = or i64 %16, %18 - %20 = shl i64 %0, 8 - %21 = and i64 %20, 1095216660480 - %22 = add i64 %19, %21 - ret i64 %22 -} diff --git a/llvm/test/Transforms/InstCombine/fmul-inseltpoison.ll b/llvm/test/Transforms/InstCombine/fmul-inseltpoison.ll index 083b7a6ad6ad8..41d486856b6a5 100644 --- a/llvm/test/Transforms/InstCombine/fmul-inseltpoison.ll +++ b/llvm/test/Transforms/InstCombine/fmul-inseltpoison.ll @@ -1,375 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -S -instcombine < %s | FileCheck %s -; (-0.0 - X) * C => X * -C -define float @neg_constant(float %x) { -; CHECK-LABEL: @neg_constant( -; CHECK-NEXT: [[MUL:%.*]] = fmul ninf float [[X:%.*]], -2.000000e+01 -; CHECK-NEXT: ret float [[MUL]] -; - %sub = fsub float -0.0, %x - %mul = fmul ninf float %sub, 2.0e+1 - ret float %mul -} - -define float @unary_neg_constant(float %x) { -; CHECK-LABEL: @unary_neg_constant( -; CHECK-NEXT: [[MUL:%.*]] = fmul ninf float [[X:%.*]], -2.000000e+01 -; CHECK-NEXT: ret float [[MUL]] -; - %sub = fneg float %x - %mul = fmul ninf float %sub, 2.0e+1 - ret float %mul -} - -define <2 x float> @neg_constant_vec(<2 x float> %x) { -; CHECK-LABEL: @neg_constant_vec( -; CHECK-NEXT: [[MUL:%.*]] = fmul ninf <2 x float> [[X:%.*]], <float -2.000000e+00, float -3.000000e+00> -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sub = fsub <2 x float> <float -0.0, float -0.0>, %x - %mul = fmul ninf <2 x float> %sub, <float 2.0, float 3.0> - ret <2 x float> %mul -} - -define <2 x float> @unary_neg_constant_vec(<2 x float> %x) { -; CHECK-LABEL: @unary_neg_constant_vec( -; CHECK-NEXT: [[MUL:%.*]] = fmul ninf <2 x float> [[X:%.*]], <float -2.000000e+00, float -3.000000e+00> -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sub = fneg <2 x float> %x - %mul = fmul ninf <2 x float> %sub, <float 2.0, float 3.0> - ret <2 x float> %mul -} - -define <2 x float> @neg_constant_vec_undef(<2 x float> %x) { -; CHECK-LABEL: @neg_constant_vec_undef( -; CHECK-NEXT: [[MUL:%.*]] = fmul ninf <2 x float> [[X:%.*]], <float -2.000000e+00, float -3.000000e+00> -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sub = fsub <2 x float> <float undef, float -0.0>, %x - %mul = fmul ninf <2 x float> %sub, <float 2.0, float 3.0> - ret <2 x float> %mul -} - -; (0.0 - X) * C => X * -C -define float @neg_nsz_constant(float %x) { -; CHECK-LABEL: @neg_nsz_constant( -; CHECK-NEXT: [[MUL:%.*]] = fmul nnan float [[X:%.*]], -2.000000e+01 -; CHECK-NEXT: ret float [[MUL]] -; - %sub = fsub nsz float 0.0, %x - %mul = fmul nnan float %sub, 2.0e+1 - ret float %mul -} - -define float @unary_neg_nsz_constant(float %x) { -; CHECK-LABEL: @unary_neg_nsz_constant( -; CHECK-NEXT: [[MUL:%.*]] = fmul nnan float [[X:%.*]], -2.000000e+01 -; CHECK-NEXT: ret float [[MUL]] -; - %sub = fneg nsz float %x - %mul = fmul nnan float %sub, 2.0e+1 - ret float %mul -} - -; (-0.0 - X) * (-0.0 - Y) => X * Y -define float @neg_neg(float %x, float %y) { -; CHECK-LABEL: @neg_neg( -; CHECK-NEXT: [[MUL:%.*]] = fmul arcp float [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL]] -; - %sub1 = fsub float -0.0, %x - %sub2 = fsub float -0.0, %y - %mul = fmul arcp float %sub1, %sub2 - ret float %mul -} - -define float @unary_neg_unary_neg(float %x, float %y) { -; CHECK-LABEL: @unary_neg_unary_neg( -; CHECK-NEXT: [[MUL:%.*]] = fmul arcp float [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL]] -; - %sub1 = fneg float %x - %sub2 = fneg float %y - %mul = fmul arcp float %sub1, %sub2 - ret float %mul -} - -define float @unary_neg_neg(float %x, float %y) { -; CHECK-LABEL: @unary_neg_neg( -; CHECK-NEXT: [[MUL:%.*]] = fmul arcp float [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL]] -; - %sub1 = fneg float %x - %sub2 = fsub float -0.0, %y - %mul = fmul arcp float %sub1, %sub2 - ret float %mul -} - -define float @neg_unary_neg(float %x, float %y) { -; CHECK-LABEL: @neg_unary_neg( -; CHECK-NEXT: [[MUL:%.*]] = fmul arcp float [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL]] -; - %sub1 = fsub float -0.0, %x - %sub2 = fneg float %y - %mul = fmul arcp float %sub1, %sub2 - ret float %mul -} - -define <2 x float> @neg_neg_vec(<2 x float> %x, <2 x float> %y) { -; CHECK-LABEL: @neg_neg_vec( -; CHECK-NEXT: [[MUL:%.*]] = fmul arcp <2 x float> [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sub1 = fsub <2 x float> <float -0.0, float -0.0>, %x - %sub2 = fsub <2 x float> <float -0.0, float -0.0>, %y - %mul = fmul arcp <2 x float> %sub1, %sub2 - ret <2 x float> %mul -} - -define <2 x float> @unary_neg_unary_neg_vec(<2 x float> %x, <2 x float> %y) { -; CHECK-LABEL: @unary_neg_unary_neg_vec( -; CHECK-NEXT: [[MUL:%.*]] = fmul arcp <2 x float> [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sub1 = fneg <2 x float> %x - %sub2 = fneg <2 x float> %y - %mul = fmul arcp <2 x float> %sub1, %sub2 - ret <2 x float> %mul -} - -define <2 x float> @unary_neg_neg_vec(<2 x float> %x, <2 x float> %y) { -; CHECK-LABEL: @unary_neg_neg_vec( -; CHECK-NEXT: [[MUL:%.*]] = fmul arcp <2 x float> [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sub1 = fneg <2 x float> %x - %sub2 = fsub <2 x float> <float -0.0, float -0.0>, %y - %mul = fmul arcp <2 x float> %sub1, %sub2 - ret <2 x float> %mul -} - -define <2 x float> @neg_unary_neg_vec(<2 x float> %x, <2 x float> %y) { -; CHECK-LABEL: @neg_unary_neg_vec( -; CHECK-NEXT: [[MUL:%.*]] = fmul arcp <2 x float> [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sub1 = fsub <2 x float> <float -0.0, float -0.0>, %x - %sub2 = fneg <2 x float> %y - %mul = fmul arcp <2 x float> %sub1, %sub2 - ret <2 x float> %mul -} - -define <2 x float> @neg_neg_vec_undef(<2 x float> %x, <2 x float> %y) { -; CHECK-LABEL: @neg_neg_vec_undef( -; CHECK-NEXT: [[MUL:%.*]] = fmul arcp <2 x float> [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sub1 = fsub <2 x float> <float -0.0, float undef>, %x - %sub2 = fsub <2 x float> <float undef, float -0.0>, %y - %mul = fmul arcp <2 x float> %sub1, %sub2 - ret <2 x float> %mul -} - -define <2 x float> @unary_neg_neg_vec_undef(<2 x float> %x, <2 x float> %y) { -; CHECK-LABEL: @unary_neg_neg_vec_undef( -; CHECK-NEXT: [[MUL:%.*]] = fmul arcp <2 x float> [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %neg = fneg <2 x float> %x - %sub = fsub <2 x float> <float undef, float -0.0>, %y - %mul = fmul arcp <2 x float> %neg, %sub - ret <2 x float> %mul -} - -define <2 x float> @neg_unary_neg_vec_undef(<2 x float> %x, <2 x float> %y) { -; CHECK-LABEL: @neg_unary_neg_vec_undef( -; CHECK-NEXT: [[MUL:%.*]] = fmul arcp <2 x float> [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sub = fsub <2 x float> <float -0.0, float undef>, %x - %neg = fneg <2 x float> %y - %mul = fmul arcp <2 x float> %sub, %neg - ret <2 x float> %mul -} - -; (0.0 - X) * (0.0 - Y) => X * Y -define float @neg_neg_nsz(float %x, float %y) { -; CHECK-LABEL: @neg_neg_nsz( -; CHECK-NEXT: [[MUL:%.*]] = fmul afn float [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL]] -; - %sub1 = fsub nsz float 0.0, %x - %sub2 = fsub nsz float 0.0, %y - %mul = fmul afn float %sub1, %sub2 - ret float %mul -} - -declare void @use_f32(float) - -define float @neg_neg_multi_use(float %x, float %y) { -; CHECK-LABEL: @neg_neg_multi_use( -; CHECK-NEXT: [[NX:%.*]] = fneg float [[X:%.*]] -; CHECK-NEXT: [[NY:%.*]] = fneg float [[Y:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fmul afn float [[X]], [[Y]] -; CHECK-NEXT: call void @use_f32(float [[NX]]) -; CHECK-NEXT: call void @use_f32(float [[NY]]) -; CHECK-NEXT: ret float [[MUL]] -; - %nx = fsub float -0.0, %x - %ny = fsub float -0.0, %y - %mul = fmul afn float %nx, %ny - call void @use_f32(float %nx) - call void @use_f32(float %ny) - ret float %mul -} - -define float @unary_neg_unary_neg_multi_use(float %x, float %y) { -; CHECK-LABEL: @unary_neg_unary_neg_multi_use( -; CHECK-NEXT: [[NX:%.*]] = fneg float [[X:%.*]] -; CHECK-NEXT: [[NY:%.*]] = fneg float [[Y:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fmul afn float [[X]], [[Y]] -; CHECK-NEXT: call void @use_f32(float [[NX]]) -; CHECK-NEXT: call void @use_f32(float [[NY]]) -; CHECK-NEXT: ret float [[MUL]] -; - %nx = fneg float %x - %ny = fneg float %y - %mul = fmul afn float %nx, %ny - call void @use_f32(float %nx) - call void @use_f32(float %ny) - ret float %mul -} - -define float @unary_neg_neg_multi_use(float %x, float %y) { -; CHECK-LABEL: @unary_neg_neg_multi_use( -; CHECK-NEXT: [[NX:%.*]] = fneg float [[X:%.*]] -; CHECK-NEXT: [[NY:%.*]] = fneg float [[Y:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fmul afn float [[X]], [[Y]] -; CHECK-NEXT: call void @use_f32(float [[NX]]) -; CHECK-NEXT: call void @use_f32(float [[NY]]) -; CHECK-NEXT: ret float [[MUL]] -; - %nx = fneg float %x - %ny = fsub float -0.0, %y - %mul = fmul afn float %nx, %ny - call void @use_f32(float %nx) - call void @use_f32(float %ny) - ret float %mul -} - -define float @neg_unary_neg_multi_use(float %x, float %y) { -; CHECK-LABEL: @neg_unary_neg_multi_use( -; CHECK-NEXT: [[NX:%.*]] = fneg float [[X:%.*]] -; CHECK-NEXT: [[NY:%.*]] = fneg float [[Y:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fmul afn float [[X]], [[Y]] -; CHECK-NEXT: call void @use_f32(float [[NX]]) -; CHECK-NEXT: call void @use_f32(float [[NY]]) -; CHECK-NEXT: ret float [[MUL]] -; - %nx = fsub float -0.0, %x - %ny = fneg float %y - %mul = fmul afn float %nx, %ny - call void @use_f32(float %nx) - call void @use_f32(float %ny) - ret float %mul -} - -; (-0.0 - X) * Y -define float @neg_mul(float %x, float %y) { -; CHECK-LABEL: @neg_mul( -; CHECK-NEXT: [[SUB:%.*]] = fneg float [[X:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fmul float [[SUB]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL]] -; - %sub = fsub float -0.0, %x - %mul = fmul float %sub, %y - ret float %mul -} - -define float @unary_neg_mul(float %x, float %y) { -; CHECK-LABEL: @unary_neg_mul( -; CHECK-NEXT: [[NEG:%.*]] = fneg float [[X:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fmul float [[NEG]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL]] -; - %neg = fneg float %x - %mul = fmul float %neg, %y - ret float %mul -} - -define <2 x float> @neg_mul_vec(<2 x float> %x, <2 x float> %y) { -; CHECK-LABEL: @neg_mul_vec( -; CHECK-NEXT: [[SUB:%.*]] = fneg <2 x float> [[X:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fmul <2 x float> [[SUB]], [[Y:%.*]] -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sub = fsub <2 x float> <float -0.0, float -0.0>, %x - %mul = fmul <2 x float> %sub, %y - ret <2 x float> %mul -} - -define <2 x float> @unary_neg_mul_vec(<2 x float> %x, <2 x float> %y) { -; CHECK-LABEL: @unary_neg_mul_vec( -; CHECK-NEXT: [[SUB:%.*]] = fneg <2 x float> [[X:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fmul <2 x float> [[SUB]], [[Y:%.*]] -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sub = fneg <2 x float> %x - %mul = fmul <2 x float> %sub, %y - ret <2 x float> %mul -} - -define <2 x float> @neg_mul_vec_undef(<2 x float> %x, <2 x float> %y) { -; CHECK-LABEL: @neg_mul_vec_undef( -; CHECK-NEXT: [[SUB:%.*]] = fneg <2 x float> [[X:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fmul <2 x float> [[SUB]], [[Y:%.*]] -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sub = fsub <2 x float> <float undef, float -0.0>, %x - %mul = fmul <2 x float> %sub, %y - ret <2 x float> %mul -} - -; (0.0 - X) * Y -define float @neg_sink_nsz(float %x, float %y) { -; CHECK-LABEL: @neg_sink_nsz( -; CHECK-NEXT: [[SUB1:%.*]] = fneg nsz float [[X:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fmul float [[SUB1]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL]] -; - %sub1 = fsub nsz float 0.0, %x - %mul = fmul float %sub1, %y - ret float %mul -} - -define float @neg_sink_multi_use(float %x, float %y) { -; CHECK-LABEL: @neg_sink_multi_use( -; CHECK-NEXT: [[SUB1:%.*]] = fneg float [[X:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fmul float [[SUB1]], [[Y:%.*]] -; CHECK-NEXT: [[MUL2:%.*]] = fmul float [[MUL]], [[SUB1]] -; CHECK-NEXT: ret float [[MUL2]] -; - %sub1 = fsub float -0.0, %x - %mul = fmul float %sub1, %y - %mul2 = fmul float %mul, %sub1 - ret float %mul2 -} - -define float @unary_neg_mul_multi_use(float %x, float %y) { -; CHECK-LABEL: @unary_neg_mul_multi_use( -; CHECK-NEXT: [[SUB1:%.*]] = fneg float [[X:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fmul float [[SUB1]], [[Y:%.*]] -; CHECK-NEXT: [[MUL2:%.*]] = fmul float [[MUL]], [[SUB1]] -; CHECK-NEXT: ret float [[MUL2]] -; - %sub1 = fneg float %x - %mul = fmul float %sub1, %y - %mul2 = fmul float %mul, %sub1 - ret float %mul2 -} +; A copy of fmul.ll, with undef at insertelement/shufflevector replaced with +; poison ; Don't crash when attempting to cast a constant FMul to an instruction. define void @test8(i32* %inout) { @@ -406,764 +39,6 @@ for.end: ; preds = %for.cond ret void } -; X * -1.0 => -0.0 - X -define float @test9(float %x) { -; CHECK-LABEL: @test9( -; CHECK-NEXT: [[MUL:%.*]] = fneg float [[X:%.*]] -; CHECK-NEXT: ret float [[MUL]] -; - %mul = fmul float %x, -1.0 - ret float %mul -} - -; PR18532 -define <4 x float> @test10(<4 x float> %x) { -; CHECK-LABEL: @test10( -; CHECK-NEXT: [[MUL:%.*]] = fneg arcp afn <4 x float> [[X:%.*]] -; CHECK-NEXT: ret <4 x float> [[MUL]] -; - %mul = fmul arcp afn <4 x float> %x, <float -1.0, float -1.0, float -1.0, float -1.0> - ret <4 x float> %mul -} - -define float @test11(float %x, float %y) { -; CHECK-LABEL: @test11( -; CHECK-NEXT: [[B:%.*]] = fadd fast float [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: [[C:%.*]] = fadd fast float [[B]], 3.000000e+00 -; CHECK-NEXT: ret float [[C]] -; - %a = fadd fast float %x, 1.0 - %b = fadd fast float %y, 2.0 - %c = fadd fast float %a, %b - ret float %c -} - -declare double @llvm.sqrt.f64(double) - -; With unsafe/fast math, sqrt(X) * sqrt(X) is just X, -; but make sure another use of the sqrt is intact. -; Note that the remaining fmul is altered but is not 'fast' -; itself because it was not marked 'fast' originally. -; Thus, we have an overall fast result, but no more indication of -; 'fast'ness in the code. -define double @sqrt_squared2(double %f) { -; CHECK-LABEL: @sqrt_squared2( -; CHECK-NEXT: [[SQRT:%.*]] = call double @llvm.sqrt.f64(double [[F:%.*]]) -; CHECK-NEXT: [[MUL2:%.*]] = fmul double [[SQRT]], [[F]] -; CHECK-NEXT: ret double [[MUL2]] -; - %sqrt = call double @llvm.sqrt.f64(double %f) - %mul1 = fmul fast double %sqrt, %sqrt - %mul2 = fmul double %mul1, %sqrt - ret double %mul2 -} - -declare float @llvm.fabs.f32(float) nounwind readnone - -define float @fabs_squared(float %x) { -; CHECK-LABEL: @fabs_squared( -; CHECK-NEXT: [[MUL:%.*]] = fmul float [[X:%.*]], [[X]] -; CHECK-NEXT: ret float [[MUL]] -; - %x.fabs = call float @llvm.fabs.f32(float %x) - %mul = fmul float %x.fabs, %x.fabs - ret float %mul -} - -define float @fabs_squared_fast(float %x) { -; CHECK-LABEL: @fabs_squared_fast( -; CHECK-NEXT: [[MUL:%.*]] = fmul fast float [[X:%.*]], [[X]] -; CHECK-NEXT: ret float [[MUL]] -; - %x.fabs = call float @llvm.fabs.f32(float %x) - %mul = fmul fast float %x.fabs, %x.fabs - ret float %mul -} - -define float @fabs_fabs(float %x, float %y) { -; CHECK-LABEL: @fabs_fabs( -; CHECK-NEXT: [[TMP1:%.*]] = fmul float [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = call float @llvm.fabs.f32(float [[TMP1]]) -; CHECK-NEXT: ret float [[MUL]] -; - %x.fabs = call float @llvm.fabs.f32(float %x) - %y.fabs = call float @llvm.fabs.f32(float %y) - %mul = fmul float %x.fabs, %y.fabs - ret float %mul -} - -define float @fabs_fabs_extra_use1(float %x, float %y) { -; CHECK-LABEL: @fabs_fabs_extra_use1( -; CHECK-NEXT: [[X_FABS:%.*]] = call float @llvm.fabs.f32(float [[X:%.*]]) -; CHECK-NEXT: call void @use_f32(float [[X_FABS]]) -; CHECK-NEXT: [[TMP1:%.*]] = fmul ninf float [[X]], [[Y:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = call ninf float @llvm.fabs.f32(float [[TMP1]]) -; CHECK-NEXT: ret float [[MUL]] -; - %x.fabs = call float @llvm.fabs.f32(float %x) - call void @use_f32(float %x.fabs) - %y.fabs = call float @llvm.fabs.f32(float %y) - %mul = fmul ninf float %x.fabs, %y.fabs - ret float %mul -} - -define float @fabs_fabs_extra_use2(float %x, float %y) { -; CHECK-LABEL: @fabs_fabs_extra_use2( -; CHECK-NEXT: [[Y_FABS:%.*]] = call fast float @llvm.fabs.f32(float [[Y:%.*]]) -; CHECK-NEXT: call void @use_f32(float [[Y_FABS]]) -; CHECK-NEXT: [[TMP1:%.*]] = fmul reassoc ninf float [[X:%.*]], [[Y]] -; CHECK-NEXT: [[MUL:%.*]] = call reassoc ninf float @llvm.fabs.f32(float [[TMP1]]) -; CHECK-NEXT: ret float [[MUL]] -; - %x.fabs = call fast float @llvm.fabs.f32(float %x) - %y.fabs = call fast float @llvm.fabs.f32(float %y) - call void @use_f32(float %y.fabs) - %mul = fmul reassoc ninf float %x.fabs, %y.fabs - ret float %mul -} - -; negative test - don't create an extra instruction - -define float @fabs_fabs_extra_use3(float %x, float %y) { -; CHECK-LABEL: @fabs_fabs_extra_use3( -; CHECK-NEXT: [[X_FABS:%.*]] = call float @llvm.fabs.f32(float [[X:%.*]]) -; CHECK-NEXT: call void @use_f32(float [[X_FABS]]) -; CHECK-NEXT: [[Y_FABS:%.*]] = call float @llvm.fabs.f32(float [[Y:%.*]]) -; CHECK-NEXT: call void @use_f32(float [[Y_FABS]]) -; CHECK-NEXT: [[MUL:%.*]] = fmul float [[X_FABS]], [[Y_FABS]] -; CHECK-NEXT: ret float [[MUL]] -; - %x.fabs = call float @llvm.fabs.f32(float %x) - call void @use_f32(float %x.fabs) - %y.fabs = call float @llvm.fabs.f32(float %y) - call void @use_f32(float %y.fabs) - %mul = fmul float %x.fabs, %y.fabs - ret float %mul -} - -; (X*Y) * X => (X*X) * Y -; The transform only requires 'reassoc', but test other FMF in -; the commuted variants to make sure FMF propagates as expected. - -define float @reassoc_common_operand1(float %x, float %y) { -; CHECK-LABEL: @reassoc_common_operand1( -; CHECK-NEXT: [[TMP1:%.*]] = fmul reassoc float [[X:%.*]], [[X]] -; CHECK-NEXT: [[MUL2:%.*]] = fmul reassoc float [[TMP1]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL2]] -; - %mul1 = fmul float %x, %y - %mul2 = fmul reassoc float %mul1, %x - ret float %mul2 -} - -; (Y*X) * X => (X*X) * Y - -define float @reassoc_common_operand2(float %x, float %y) { -; CHECK-LABEL: @reassoc_common_operand2( -; CHECK-NEXT: [[TMP1:%.*]] = fmul fast float [[X:%.*]], [[X]] -; CHECK-NEXT: [[MUL2:%.*]] = fmul fast float [[TMP1]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL2]] -; - %mul1 = fmul float %y, %x - %mul2 = fmul fast float %mul1, %x - ret float %mul2 -} - -; X * (X*Y) => (X*X) * Y - -define float @reassoc_common_operand3(float %x1, float %y) { -; CHECK-LABEL: @reassoc_common_operand3( -; CHECK-NEXT: [[X:%.*]] = fdiv float [[X1:%.*]], 3.000000e+00 -; CHECK-NEXT: [[TMP1:%.*]] = fmul reassoc nnan float [[X]], [[X]] -; CHECK-NEXT: [[MUL2:%.*]] = fmul reassoc nnan float [[TMP1]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL2]] -; - %x = fdiv float %x1, 3.0 ; thwart complexity-based canonicalization - %mul1 = fmul float %x, %y - %mul2 = fmul reassoc nnan float %x, %mul1 - ret float %mul2 -} - -; X * (Y*X) => (X*X) * Y - -define float @reassoc_common_operand4(float %x1, float %y) { -; CHECK-LABEL: @reassoc_common_operand4( -; CHECK-NEXT: [[X:%.*]] = fdiv float [[X1:%.*]], 3.000000e+00 -; CHECK-NEXT: [[TMP1:%.*]] = fmul reassoc ninf float [[X]], [[X]] -; CHECK-NEXT: [[MUL2:%.*]] = fmul reassoc ninf float [[TMP1]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL2]] -; - %x = fdiv float %x1, 3.0 ; thwart complexity-based canonicalization - %mul1 = fmul float %y, %x - %mul2 = fmul reassoc ninf float %x, %mul1 - ret float %mul2 -} - -; No change if the first fmul has another use. - -define float @reassoc_common_operand_multi_use(float %x, float %y) { -; CHECK-LABEL: @reassoc_common_operand_multi_use( -; CHECK-NEXT: [[MUL1:%.*]] = fmul float [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: [[MUL2:%.*]] = fmul fast float [[MUL1]], [[X]] -; CHECK-NEXT: call void @use_f32(float [[MUL1]]) -; CHECK-NEXT: ret float [[MUL2]] -; - %mul1 = fmul float %x, %y - %mul2 = fmul fast float %mul1, %x - call void @use_f32(float %mul1) - ret float %mul2 -} - -declare float @llvm.log2.f32(float) - -; log2(Y * 0.5) * X = log2(Y) * X - X - -define float @log2half(float %x, float %y) { -; CHECK-LABEL: @log2half( -; CHECK-NEXT: [[TMP1:%.*]] = call fast float @llvm.log2.f32(float [[Y:%.*]]) -; CHECK-NEXT: [[TMP2:%.*]] = fmul fast float [[TMP1]], [[X:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fsub fast float [[TMP2]], [[X]] -; CHECK-NEXT: ret float [[MUL]] -; - %halfy = fmul float %y, 0.5 - %log2 = call float @llvm.log2.f32(float %halfy) - %mul = fmul fast float %log2, %x - ret float %mul -} - -define float @log2half_commute(float %x1, float %y) { -; CHECK-LABEL: @log2half_commute( -; CHECK-NEXT: [[TMP1:%.*]] = call fast float @llvm.log2.f32(float [[Y:%.*]]) -; CHECK-NEXT: [[TMP2:%.*]] = fmul fast float [[TMP1]], [[X1:%.*]] -; CHECK-NEXT: [[TMP3:%.*]] = fsub fast float [[TMP2]], [[X1]] -; CHECK-NEXT: [[MUL:%.*]] = fmul fast float [[TMP3]], 0x3FC24924A0000000 -; CHECK-NEXT: ret float [[MUL]] -; - %x = fdiv float %x1, 7.0 ; thwart complexity-based canonicalization - %halfy = fmul float %y, 0.5 - %log2 = call float @llvm.log2.f32(float %halfy) - %mul = fmul fast float %x, %log2 - ret float %mul -} - -; C1/X * C2 => (C1*C2) / X - -define float @fdiv_constant_numerator_fmul(float %x) { -; CHECK-LABEL: @fdiv_constant_numerator_fmul( -; CHECK-NEXT: [[T3:%.*]] = fdiv reassoc float 1.200000e+07, [[X:%.*]] -; CHECK-NEXT: ret float [[T3]] -; - %t1 = fdiv float 2.0e+3, %x - %t3 = fmul reassoc float %t1, 6.0e+3 - ret float %t3 -} - -; C1/X * C2 => (C1*C2) / X is disabled if C1/X has multiple uses - -@fmul2_external = external global float - -define float @fdiv_constant_numerator_fmul_extra_use(float %x) { -; CHECK-LABEL: @fdiv_constant_numerator_fmul_extra_use( -; CHECK-NEXT: [[DIV:%.*]] = fdiv fast float 1.000000e+00, [[X:%.*]] -; CHECK-NEXT: store float [[DIV]], float* @fmul2_external, align 4 -; CHECK-NEXT: [[MUL:%.*]] = fmul fast float [[DIV]], 2.000000e+00 -; CHECK-NEXT: ret float [[MUL]] -; - %div = fdiv fast float 1.0, %x - store float %div, float* @fmul2_external - %mul = fmul fast float %div, 2.0 - ret float %mul -} - -; X/C1 * C2 => X * (C2/C1) (if C2/C1 is normal FP) - -define float @fdiv_constant_denominator_fmul(float %x) { -; CHECK-LABEL: @fdiv_constant_denominator_fmul( -; CHECK-NEXT: [[T3:%.*]] = fmul reassoc float [[X:%.*]], 3.000000e+00 -; CHECK-NEXT: ret float [[T3]] -; - %t1 = fdiv float %x, 2.0e+3 - %t3 = fmul reassoc float %t1, 6.0e+3 - ret float %t3 -} - -define <4 x float> @fdiv_constant_denominator_fmul_vec(<4 x float> %x) { -; CHECK-LABEL: @fdiv_constant_denominator_fmul_vec( -; CHECK-NEXT: [[T3:%.*]] = fmul reassoc <4 x float> [[X:%.*]], <float 3.000000e+00, float 2.000000e+00, float 1.000000e+00, float 1.000000e+00> -; CHECK-NEXT: ret <4 x float> [[T3]] -; - %t1 = fdiv <4 x float> %x, <float 2.0e+3, float 3.0e+3, float 2.0e+3, float 1.0e+3> - %t3 = fmul reassoc <4 x float> %t1, <float 6.0e+3, float 6.0e+3, float 2.0e+3, float 1.0e+3> - ret <4 x float> %t3 -} - -; Make sure fmul with constant expression doesn't assert. - -define <4 x float> @fdiv_constant_denominator_fmul_vec_constexpr(<4 x float> %x) { -; CHECK-LABEL: @fdiv_constant_denominator_fmul_vec_constexpr( -; CHECK-NEXT: [[T3:%.*]] = fmul reassoc <4 x float> [[X:%.*]], <float 3.000000e+00, float 2.000000e+00, float 1.000000e+00, float 1.000000e+00> -; CHECK-NEXT: ret <4 x float> [[T3]] -; - %constExprMul = bitcast i128 trunc (i160 bitcast (<5 x float> <float 6.0e+3, float 6.0e+3, float 2.0e+3, float 1.0e+3, float undef> to i160) to i128) to <4 x float> - %t1 = fdiv <4 x float> %x, <float 2.0e+3, float 3.0e+3, float 2.0e+3, float 1.0e+3> - %t3 = fmul reassoc <4 x float> %t1, %constExprMul - ret <4 x float> %t3 -} - -; This shows that at least part of instcombine does not check constant -; values to see if it is creating denorms (0x3800000000000000 is a denorm -; for 32-bit float), so protecting against denorms in other parts is -; probably not doing the intended job. - -define float @fmul_constant_reassociation(float %x) { -; CHECK-LABEL: @fmul_constant_reassociation( -; CHECK-NEXT: [[R:%.*]] = fmul reassoc nsz float [[X:%.*]], 0x3800000000000000 -; CHECK-NEXT: ret float [[R]] -; - %mul_flt_min = fmul reassoc nsz float %x, 0x3810000000000000 - %r = fmul reassoc nsz float %mul_flt_min, 0.5 - ret float %r -} - -; Canonicalization "X/C1 * C2 => X * (C2/C1)" still applies if C2/C1 is denormal -; (otherwise, we should not have allowed the reassociation in the previous test). -; 0x3810000000000000 == FLT_MIN - -define float @fdiv_constant_denominator_fmul_denorm(float %x) { -; CHECK-LABEL: @fdiv_constant_denominator_fmul_denorm( -; CHECK-NEXT: [[T3:%.*]] = fmul fast float [[X:%.*]], 0x3760620000000000 -; CHECK-NEXT: ret float [[T3]] -; - %t1 = fdiv float %x, 2.0e+3 - %t3 = fmul fast float %t1, 0x3810000000000000 - ret float %t3 -} - -; X / C1 * C2 => X / (C2/C1) if C1/C2 is abnormal, but C2/C1 is a normal value. -; TODO: We don't convert the fast fdiv to fmul because that would be multiplication -; by a denormal, but we could do better when we know that denormals are not a problem. - -define float @fdiv_constant_denominator_fmul_denorm_try_harder(float %x) { -; CHECK-LABEL: @fdiv_constant_denominator_fmul_denorm_try_harder( -; CHECK-NEXT: [[T3:%.*]] = fdiv reassoc float [[X:%.*]], 0x47E8000000000000 -; CHECK-NEXT: ret float [[T3]] -; - %t1 = fdiv float %x, 3.0 - %t3 = fmul reassoc float %t1, 0x3810000000000000 - ret float %t3 -} - -; Negative test: we should not have 2 divisions instead of the 1 we started with. - -define float @fdiv_constant_denominator_fmul_denorm_try_harder_extra_use(float %x) { -; CHECK-LABEL: @fdiv_constant_denominator_fmul_denorm_try_harder_extra_use( -; CHECK-NEXT: [[T1:%.*]] = fdiv float [[X:%.*]], 3.000000e+00 -; CHECK-NEXT: [[T3:%.*]] = fmul fast float [[T1]], 0x3810000000000000 -; CHECK-NEXT: [[R:%.*]] = fadd float [[T1]], [[T3]] -; CHECK-NEXT: ret float [[R]] -; - %t1 = fdiv float %x, 3.0e+0 - %t3 = fmul fast float %t1, 0x3810000000000000 - %r = fadd float %t1, %t3 - ret float %r -} - -; (X + C1) * C2 --> (X * C2) + C1*C2 - -define float @fmul_fadd_distribute(float %x) { -; CHECK-LABEL: @fmul_fadd_distribute( -; CHECK-NEXT: [[TMP1:%.*]] = fmul reassoc float [[X:%.*]], 3.000000e+00 -; CHECK-NEXT: [[T3:%.*]] = fadd reassoc float [[TMP1]], 6.000000e+00 -; CHECK-NEXT: ret float [[T3]] -; - %t2 = fadd float %x, 2.0 - %t3 = fmul reassoc float %t2, 3.0 - ret float %t3 -} - -; (X - C1) * C2 --> (X * C2) - C1*C2 - -define float @fmul_fsub_distribute1(float %x) { -; CHECK-LABEL: @fmul_fsub_distribute1( -; CHECK-NEXT: [[TMP1:%.*]] = fmul reassoc float [[X:%.*]], 3.000000e+00 -; CHECK-NEXT: [[T3:%.*]] = fadd reassoc float [[TMP1]], -6.000000e+00 -; CHECK-NEXT: ret float [[T3]] -; - %t2 = fsub float %x, 2.0 - %t3 = fmul reassoc float %t2, 3.0 - ret float %t3 -} - -; (C1 - X) * C2 --> C1*C2 - (X * C2) - -define float @fmul_fsub_distribute2(float %x) { -; CHECK-LABEL: @fmul_fsub_distribute2( -; CHECK-NEXT: [[TMP1:%.*]] = fmul reassoc float [[X:%.*]], 3.000000e+00 -; CHECK-NEXT: [[T3:%.*]] = fsub reassoc float 6.000000e+00, [[TMP1]] -; CHECK-NEXT: ret float [[T3]] -; - %t2 = fsub float 2.0, %x - %t3 = fmul reassoc float %t2, 3.0 - ret float %t3 -} - -; FIXME: This should only need 'reassoc'. -; ((X*C1) + C2) * C3 => (X * (C1*C3)) + (C2*C3) - -define float @fmul_fadd_fmul_distribute(float %x) { -; CHECK-LABEL: @fmul_fadd_fmul_distribute( -; CHECK-NEXT: [[TMP1:%.*]] = fmul fast float [[X:%.*]], 3.000000e+01 -; CHECK-NEXT: [[T3:%.*]] = fadd fast float [[TMP1]], 1.000000e+01 -; CHECK-NEXT: ret float [[T3]] -; - %t1 = fmul float %x, 6.0 - %t2 = fadd float %t1, 2.0 - %t3 = fmul fast float %t2, 5.0 - ret float %t3 -} - -define float @fmul_fadd_distribute_extra_use(float %x) { -; CHECK-LABEL: @fmul_fadd_distribute_extra_use( -; CHECK-NEXT: [[T1:%.*]] = fmul float [[X:%.*]], 6.000000e+00 -; CHECK-NEXT: [[T2:%.*]] = fadd float [[T1]], 2.000000e+00 -; CHECK-NEXT: [[T3:%.*]] = fmul fast float [[T2]], 5.000000e+00 -; CHECK-NEXT: call void @use_f32(float [[T2]]) -; CHECK-NEXT: ret float [[T3]] -; - %t1 = fmul float %x, 6.0 - %t2 = fadd float %t1, 2.0 - %t3 = fmul fast float %t2, 5.0 - call void @use_f32(float %t2) - ret float %t3 -} - -; (X/C1 + C2) * C3 => X/(C1/C3) + C2*C3 -; 0x10000000000000 = DBL_MIN -; TODO: We don't convert the fast fdiv to fmul because that would be multiplication -; by a denormal, but we could do better when we know that denormals are not a problem. - -define double @fmul_fadd_fdiv_distribute2(double %x) { -; CHECK-LABEL: @fmul_fadd_fdiv_distribute2( -; CHECK-NEXT: [[TMP1:%.*]] = fdiv reassoc double [[X:%.*]], 0x7FE8000000000000 -; CHECK-NEXT: [[T3:%.*]] = fadd reassoc double [[TMP1]], 0x34000000000000 -; CHECK-NEXT: ret double [[T3]] -; - %t1 = fdiv double %x, 3.0 - %t2 = fadd double %t1, 5.0 - %t3 = fmul reassoc double %t2, 0x10000000000000 - ret double %t3 -} - -; 5.0e-1 * DBL_MIN yields denormal, so "(f1*3.0 + 5.0e-1) * DBL_MIN" cannot -; be simplified into f1 * (3.0*DBL_MIN) + (5.0e-1*DBL_MIN) - -define double @fmul_fadd_fdiv_distribute3(double %x) { -; CHECK-LABEL: @fmul_fadd_fdiv_distribute3( -; CHECK-NEXT: [[TMP1:%.*]] = fdiv reassoc double [[X:%.*]], 0x7FE8000000000000 -; CHECK-NEXT: [[T3:%.*]] = fadd reassoc double [[TMP1]], 0x34000000000000 -; CHECK-NEXT: ret double [[T3]] -; - %t1 = fdiv double %x, 3.0 - %t2 = fadd double %t1, 5.0 - %t3 = fmul reassoc double %t2, 0x10000000000000 - ret double %t3 -} - -; FIXME: This should only need 'reassoc'. -; (C2 - (X*C1)) * C3 => (C2*C3) - (X * (C1*C3)) - -define float @fmul_fsub_fmul_distribute(float %x) { -; CHECK-LABEL: @fmul_fsub_fmul_distribute( -; CHECK-NEXT: [[TMP1:%.*]] = fmul fast float [[X:%.*]], 3.000000e+01 -; CHECK-NEXT: [[T3:%.*]] = fsub fast float 1.000000e+01, [[TMP1]] -; CHECK-NEXT: ret float [[T3]] -; - %t1 = fmul float %x, 6.0 - %t2 = fsub float 2.0, %t1 - %t3 = fmul fast float %t2, 5.0 - ret float %t3 -} - -define float @fmul_fsub_fmul_distribute_extra_use(float %x) { -; CHECK-LABEL: @fmul_fsub_fmul_distribute_extra_use( -; CHECK-NEXT: [[T1:%.*]] = fmul float [[X:%.*]], 6.000000e+00 -; CHECK-NEXT: [[T2:%.*]] = fsub float 2.000000e+00, [[T1]] -; CHECK-NEXT: [[T3:%.*]] = fmul fast float [[T2]], 5.000000e+00 -; CHECK-NEXT: call void @use_f32(float [[T2]]) -; CHECK-NEXT: ret float [[T3]] -; - %t1 = fmul float %x, 6.0 - %t2 = fsub float 2.0, %t1 - %t3 = fmul fast float %t2, 5.0 - call void @use_f32(float %t2) - ret float %t3 -} - -; FIXME: This should only need 'reassoc'. -; ((X*C1) - C2) * C3 => (X * (C1*C3)) - C2*C3 - -define float @fmul_fsub_fmul_distribute2(float %x) { -; CHECK-LABEL: @fmul_fsub_fmul_distribute2( -; CHECK-NEXT: [[TMP1:%.*]] = fmul fast float [[X:%.*]], 3.000000e+01 -; CHECK-NEXT: [[T3:%.*]] = fadd fast float [[TMP1]], -1.000000e+01 -; CHECK-NEXT: ret float [[T3]] -; - %t1 = fmul float %x, 6.0 - %t2 = fsub float %t1, 2.0 - %t3 = fmul fast float %t2, 5.0 - ret float %t3 -} - -define float @fmul_fsub_fmul_distribute2_extra_use(float %x) { -; CHECK-LABEL: @fmul_fsub_fmul_distribute2_extra_use( -; CHECK-NEXT: [[T1:%.*]] = fmul float [[X:%.*]], 6.000000e+00 -; CHECK-NEXT: [[T2:%.*]] = fsub float 2.000000e+00, [[T1]] -; CHECK-NEXT: [[T3:%.*]] = fmul fast float [[T2]], 5.000000e+00 -; CHECK-NEXT: call void @use_f32(float [[T2]]) -; CHECK-NEXT: ret float [[T3]] -; - %t1 = fmul float %x, 6.0 - %t2 = fsub float 2.0, %t1 - %t3 = fmul fast float %t2, 5.0 - call void @use_f32(float %t2) - ret float %t3 -} - -; "(X*Y) * X => (X*X) * Y" is disabled if "X*Y" has multiple uses - -define float @common_factor(float %x, float %y) { -; CHECK-LABEL: @common_factor( -; CHECK-NEXT: [[MUL:%.*]] = fmul float [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: [[MUL1:%.*]] = fmul fast float [[MUL]], [[X]] -; CHECK-NEXT: [[ADD:%.*]] = fadd float [[MUL1]], [[MUL]] -; CHECK-NEXT: ret float [[ADD]] -; - %mul = fmul float %x, %y - %mul1 = fmul fast float %mul, %x - %add = fadd float %mul1, %mul - ret float %add -} - -define double @fmul_fdiv_factor_squared(double %x, double %y) { -; CHECK-LABEL: @fmul_fdiv_factor_squared( -; CHECK-NEXT: [[DIV:%.*]] = fdiv fast double [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: [[SQUARED:%.*]] = fmul fast double [[DIV]], [[DIV]] -; CHECK-NEXT: ret double [[SQUARED]] -; - %div = fdiv fast double %x, %y - %squared = fmul fast double %div, %div - ret double %squared -} - -define double @fmul_fdivs_factor_common_denominator(double %x, double %y, double %z) { -; CHECK-LABEL: @fmul_fdivs_factor_common_denominator( -; CHECK-NEXT: [[TMP1:%.*]] = fmul fast double [[Y:%.*]], [[X:%.*]] -; CHECK-NEXT: [[TMP2:%.*]] = fmul fast double [[Z:%.*]], [[Z]] -; CHECK-NEXT: [[MUL:%.*]] = fdiv fast double [[TMP1]], [[TMP2]] -; CHECK-NEXT: ret double [[MUL]] -; - %div1 = fdiv double %x, %z - %div2 = fdiv double %y, %z - %mul = fmul fast double %div1, %div2 - ret double %mul -} - -define double @fmul_fdivs_factor(double %x, double %y, double %z, double %w) { -; CHECK-LABEL: @fmul_fdivs_factor( -; CHECK-NEXT: [[TMP1:%.*]] = fmul reassoc double [[Z:%.*]], [[X:%.*]] -; CHECK-NEXT: [[TMP2:%.*]] = fdiv reassoc double [[TMP1]], [[W:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fdiv reassoc double [[TMP2]], [[Y:%.*]] -; CHECK-NEXT: ret double [[MUL]] -; - %div1 = fdiv double %x, %y - %div2 = fdiv double %z, %w - %mul = fmul reassoc double %div1, %div2 - ret double %mul -} - -define double @fmul_fdiv_factor(double %x, double %y, double %z) { -; CHECK-LABEL: @fmul_fdiv_factor( -; CHECK-NEXT: [[TMP1:%.*]] = fmul reassoc double [[X:%.*]], [[Z:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fdiv reassoc double [[TMP1]], [[Y:%.*]] -; CHECK-NEXT: ret double [[MUL]] -; - %div = fdiv double %x, %y - %mul = fmul reassoc double %div, %z - ret double %mul -} - -define double @fmul_fdiv_factor_constant1(double %x, double %y) { -; CHECK-LABEL: @fmul_fdiv_factor_constant1( -; CHECK-NEXT: [[TMP1:%.*]] = fmul reassoc double [[X:%.*]], 4.200000e+01 -; CHECK-NEXT: [[MUL:%.*]] = fdiv reassoc double [[TMP1]], [[Y:%.*]] -; CHECK-NEXT: ret double [[MUL]] -; - %div = fdiv double %x, %y - %mul = fmul reassoc double %div, 42.0 - ret double %mul -} - -define <2 x float> @fmul_fdiv_factor_constant2(<2 x float> %x, <2 x float> %y) { -; CHECK-LABEL: @fmul_fdiv_factor_constant2( -; CHECK-NEXT: [[TMP1:%.*]] = fmul reassoc <2 x float> [[X:%.*]], [[Y:%.*]] -; CHECK-NEXT: [[MUL:%.*]] = fdiv reassoc <2 x float> [[TMP1]], <float 4.200000e+01, float 1.200000e+01> -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %div = fdiv <2 x float> %x, <float 42.0, float 12.0> - %mul = fmul reassoc <2 x float> %div, %y - ret <2 x float> %mul -} - -define float @fmul_fdiv_factor_extra_use(float %x, float %y) { -; CHECK-LABEL: @fmul_fdiv_factor_extra_use( -; CHECK-NEXT: [[DIV:%.*]] = fdiv float [[X:%.*]], 4.200000e+01 -; CHECK-NEXT: call void @use_f32(float [[DIV]]) -; CHECK-NEXT: [[MUL:%.*]] = fmul reassoc float [[DIV]], [[Y:%.*]] -; CHECK-NEXT: ret float [[MUL]] -; - %div = fdiv float %x, 42.0 - call void @use_f32(float %div) - %mul = fmul reassoc float %div, %y - ret float %mul -} - -; Avoid infinite looping by moving negation out of a constant expression. - -@g = external global {[2 x i8*]}, align 1 - -define double @fmul_negated_constant_expression(double %x) { -; CHECK-LABEL: @fmul_negated_constant_expression( -; CHECK-NEXT: [[R:%.*]] = fmul double [[X:%.*]], fsub (double -0.000000e+00, double bitcast (i64 ptrtoint (i8** getelementptr inbounds ({ [2 x i8*] }, { [2 x i8*] }* @g, i64 0, inrange i32 0, i64 2) to i64) to double)) -; CHECK-NEXT: ret double [[R]] -; - %r = fmul double %x, fsub (double -0.000000e+00, double bitcast (i64 ptrtoint (i8** getelementptr inbounds ({ [2 x i8*] }, { [2 x i8*] }* @g, i64 0, inrange i32 0, i64 2) to i64) to double)) - ret double %r -} - -define float @negate_if_true(float %x, i1 %cond) { -; CHECK-LABEL: @negate_if_true( -; CHECK-NEXT: [[TMP1:%.*]] = fneg float [[X:%.*]] -; CHECK-NEXT: [[TMP2:%.*]] = select i1 [[COND:%.*]], float [[TMP1]], float [[X]] -; CHECK-NEXT: ret float [[TMP2]] -; - %sel = select i1 %cond, float -1.0, float 1.0 - %r = fmul float %sel, %x - ret float %r -} - -define float @negate_if_false(float %x, i1 %cond) { -; CHECK-LABEL: @negate_if_false( -; CHECK-NEXT: [[TMP1:%.*]] = fneg arcp float [[X:%.*]] -; CHECK-NEXT: [[TMP2:%.*]] = select arcp i1 [[COND:%.*]], float [[X]], float [[TMP1]] -; CHECK-NEXT: ret float [[TMP2]] -; - %sel = select i1 %cond, float 1.0, float -1.0 - %r = fmul arcp float %sel, %x - ret float %r -} - -define <2 x double> @negate_if_true_commute(<2 x double> %px, i1 %cond) { -; CHECK-LABEL: @negate_if_true_commute( -; CHECK-NEXT: [[X:%.*]] = fdiv <2 x double> <double 4.200000e+01, double 4.200000e+01>, [[PX:%.*]] -; CHECK-NEXT: [[TMP1:%.*]] = fneg ninf <2 x double> [[X]] -; CHECK-NEXT: [[TMP2:%.*]] = select ninf i1 [[COND:%.*]], <2 x double> [[TMP1]], <2 x double> [[X]] -; CHECK-NEXT: ret <2 x double> [[TMP2]] -; - %x = fdiv <2 x double> <double 42.0, double 42.0>, %px ; thwart complexity-based canonicalization - %sel = select i1 %cond, <2 x double> <double -1.0, double -1.0>, <2 x double> <double 1.0, double 1.0> - %r = fmul ninf <2 x double> %x, %sel - ret <2 x double> %r -} - -define <2 x double> @negate_if_false_commute(<2 x double> %px, <2 x i1> %cond) { -; CHECK-LABEL: @negate_if_false_commute( -; CHECK-NEXT: [[X:%.*]] = fdiv <2 x double> <double 4.200000e+01, double 5.100000e+00>, [[PX:%.*]] -; CHECK-NEXT: [[TMP1:%.*]] = fneg <2 x double> [[X]] -; CHECK-NEXT: [[TMP2:%.*]] = select <2 x i1> [[COND:%.*]], <2 x double> [[X]], <2 x double> [[TMP1]] -; CHECK-NEXT: ret <2 x double> [[TMP2]] -; - %x = fdiv <2 x double> <double 42.0, double 5.1>, %px ; thwart complexity-based canonicalization - %sel = select <2 x i1> %cond, <2 x double> <double 1.0, double 1.0>, <2 x double> <double -1.0, double -1.0> - %r = fmul <2 x double> %x, %sel - ret <2 x double> %r -} - -; Negative test - -define float @negate_if_true_extra_use(float %x, i1 %cond) { -; CHECK-LABEL: @negate_if_true_extra_use( -; CHECK-NEXT: [[SEL:%.*]] = select i1 [[COND:%.*]], float -1.000000e+00, float 1.000000e+00 -; CHECK-NEXT: call void @use_f32(float [[SEL]]) -; CHECK-NEXT: [[R:%.*]] = fmul float [[SEL]], [[X:%.*]] -; CHECK-NEXT: ret float [[R]] -; - %sel = select i1 %cond, float -1.0, float 1.0 - call void @use_f32(float %sel) - %r = fmul float %sel, %x - ret float %r -} - -; Negative test - -define <2 x double> @negate_if_true_wrong_constant(<2 x double> %px, i1 %cond) { -; CHECK-LABEL: @negate_if_true_wrong_constant( -; CHECK-NEXT: [[X:%.*]] = fdiv <2 x double> <double 4.200000e+01, double 4.200000e+01>, [[PX:%.*]] -; CHECK-NEXT: [[SEL:%.*]] = select i1 [[COND:%.*]], <2 x double> <double -1.000000e+00, double 0.000000e+00>, <2 x double> <double 1.000000e+00, double 1.000000e+00> -; CHECK-NEXT: [[R:%.*]] = fmul <2 x double> [[X]], [[SEL]] -; CHECK-NEXT: ret <2 x double> [[R]] -; - %x = fdiv <2 x double> <double 42.0, double 42.0>, %px ; thwart complexity-based canonicalization - %sel = select i1 %cond, <2 x double> <double -1.0, double 0.0>, <2 x double> <double 1.0, double 1.0> - %r = fmul <2 x double> %x, %sel - ret <2 x double> %r -} - -; X *fast (C ? 1.0 : 0.0) -> C ? X : 0.0 -define float @fmul_select(float %x, i1 %c) { -; CHECK-LABEL: @fmul_select( -; CHECK-NEXT: [[MUL:%.*]] = select fast i1 [[C:%.*]], float [[X:%.*]], float 0.000000e+00 -; CHECK-NEXT: ret float [[MUL]] -; - %sel = select i1 %c, float 1.0, float 0.0 - %mul = fmul fast float %sel, %x - ret float %mul -} - -; X *fast (C ? 1.0 : 0.0) -> C ? X : 0.0 -define <2 x float> @fmul_select_vec(<2 x float> %x, i1 %c) { -; CHECK-LABEL: @fmul_select_vec( -; CHECK-NEXT: [[MUL:%.*]] = select fast i1 [[C:%.*]], <2 x float> [[X:%.*]], <2 x float> zeroinitializer -; CHECK-NEXT: ret <2 x float> [[MUL]] -; - %sel = select i1 %c, <2 x float> <float 1.0, float 1.0>, <2 x float> zeroinitializer - %mul = fmul fast <2 x float> %sel, %x - ret <2 x float> %mul -} - -; Without fast math flags we can't optimize X * (C ? 1.0 : 0.0) -> C ? X : 0.0 -define float @fmul_select_strict(float %x, i1 %c) { -; CHECK-LABEL: @fmul_select_strict( -; CHECK-NEXT: [[SEL:%.*]] = select i1 [[C:%.*]], float 1.000000e+00, float 0.000000e+00 -; CHECK-NEXT: [[MUL:%.*]] = fmul float [[SEL]], [[X:%.*]] -; CHECK-NEXT: ret float [[MUL]] -; - %sel = select i1 %c, float 1.0, float 0.0 - %mul = fmul float %sel, %x - ret float %mul -} - -; sqrt(X) *fast (C ? sqrt(X) : 1.0) -> C ? X : sqrt(X) -define double @fmul_sqrt_select(double %x, i1 %c) { -; CHECK-LABEL: @fmul_sqrt_select( -; CHECK-NEXT: [[SQR:%.*]] = call double @llvm.sqrt.f64(double [[X:%.*]]) -; CHECK-NEXT: [[MUL:%.*]] = select fast i1 [[C:%.*]], double [[X]], double [[SQR]] -; CHECK-NEXT: ret double [[MUL]] -; - %sqr = call double @llvm.sqrt.f64(double %x) - %sel = select i1 %c, double %sqr, double 1.0 - %mul = fmul fast double %sqr, %sel - ret double %mul -} ; fastmath => z * splat(0) = splat(0), even for scalable vectors define <vscale x 2 x float> @mul_scalable_splat_zero(<vscale x 2 x float> %z) { _______________________________________________ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
