This is an archive of the discontinued LLVM Phabricator instance.

Differential D96674

[AArch64][GlobalISel] Make overflow legalization use clampScalar
ClosedPublic

Authored by porglezomp on Feb 14 2021, 3:41 PM.

Download Raw Diff

Details

Reviewers

aemerson
paquette
arsenm

Commits

rG8b10aa67ad5c: [AArch64][GlobalISel] Make overflow legalization use clampScalar

Diff Detail

Repository: rG LLVM Github Monorepo

Unit TestsFailed

	Time	Test
	300 ms	x64 debian > LLVM.CodeGen/AArch64/GlobalISel::legalize-uaddo.mir
	310 ms	x64 debian > LLVM.CodeGen/AArch64/GlobalISel::legalize-usubo.mir
	220 ms	x64 windows > LLVM.CodeGen/AArch64/GlobalISel::legalize-uaddo.mir
	190 ms	x64 windows > LLVM.CodeGen/AArch64/GlobalISel::legalize-usubo.mir

Event Timeline

porglezomp created this revision.Feb 14 2021, 3:41 PM

Herald added subscribers: danielkiss, hiraditya, kristof.beyls, rovka. · View Herald TranscriptFeb 14 2021, 3:41 PM

porglezomp requested review of this revision.Feb 14 2021, 3:41 PM

Herald added a project: Restricted Project. · View Herald TranscriptFeb 14 2021, 3:41 PM

Herald added subscribers: llvm-commits, wdng. · View Herald Transcript

porglezomp added a parent revision: D96671: [GlobalISel] Implement narrowScalar for UADDO/USUBO.Feb 14 2021, 3:41 PM

porglezomp added a child revision: D96675: [Test][AArch64] Test SADDO/SSUBO narrowing legalization.Feb 14 2021, 4:32 PM

porglezomp added a child revision: D96676: [Test][AArch64] Test SADDE/SSUBE/UADDE/USUBE narrowing legalization.

Harbormaster completed remote builds in B89171: Diff 323641.Feb 14 2021, 4:38 PM

arsenm accepted this revision.Feb 18 2021, 2:43 PM

This revision is now accepted and ready to land.Feb 18 2021, 2:43 PM

This revision was landed with ongoing or failed builds.Feb 22 2021, 5:00 PM

Closed by commit rG8b10aa67ad5c: [AArch64][GlobalISel] Make overflow legalization use clampScalar (authored by porglezomp). · Explain Why

This revision was automatically updated to reflect the committed changes.

porglezomp added a commit: rG8b10aa67ad5c: [AArch64][GlobalISel] Make overflow legalization use clampScalar.

Revision Contents

Path

Size

llvm/

lib/

Target/

AArch64/

GISel/

AArch64LegalizerInfo.cpp

3 lines

test/

CodeGen/

AArch64/

GlobalISel/

legalize-uaddo.mir

61 lines

legalize-usubo.mir

61 lines

Diff 323641

llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp

	Show First 20 Lines • Show All 162 Lines • ▼ Show 20 Lines		AArch64LegalizerInfo::AArch64LegalizerInfo(const AArch64Subtarget &ST)

	getActionDefinitionsBuilder({G_SMULO, G_UMULO}).lowerFor({{s64, s1}});			getActionDefinitionsBuilder({G_SMULO, G_UMULO}).lowerFor({{s64, s1}});

	getActionDefinitionsBuilder({G_SMULH, G_UMULH}).legalFor({s32, s64});			getActionDefinitionsBuilder({G_SMULH, G_UMULH}).legalFor({s32, s64});

	getActionDefinitionsBuilder(			getActionDefinitionsBuilder(
	{G_SADDE, G_SSUBE, G_UADDE, G_USUBE, G_SADDO, G_SSUBO, G_UADDO, G_USUBO})			{G_SADDE, G_SSUBE, G_UADDE, G_USUBE, G_SADDO, G_SSUBO, G_UADDO, G_USUBO})
	.legalFor({{s32, s1}, {s64, s1}})			.legalFor({{s32, s1}, {s64, s1}})
	.minScalar(0, s32);			.clampScalar(0, s32, s64)
				.widenScalarToNextPow2(0);

	getActionDefinitionsBuilder({G_FADD, G_FSUB, G_FMUL, G_FDIV, G_FNEG})			getActionDefinitionsBuilder({G_FADD, G_FSUB, G_FMUL, G_FDIV, G_FNEG})
	.legalFor({s32, s64, v2s64, v4s32, v2s32})			.legalFor({s32, s64, v2s64, v4s32, v2s32})
	.clampNumElements(0, v2s32, v4s32)			.clampNumElements(0, v2s32, v4s32)
	.clampNumElements(0, v2s64, v2s64);			.clampNumElements(0, v2s64, v2s64);

	getActionDefinitionsBuilder(G_FREM).libcallFor({s32, s64});			getActionDefinitionsBuilder(G_FREM).libcallFor({s32, s64});

	▲ Show 20 Lines • Show All 763 Lines • Show Last 20 Lines

llvm/test/CodeGen/AArch64/GlobalISel/legalize-uaddo.mir

	# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py			# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
	# RUN: llc -mtriple aarch64 -verify-machineinstrs -run-pass=legalizer -debugify-and-strip-all-safe %s -o - \| FileCheck %s			# RUN: llc -mtriple aarch64 -verify-machineinstrs -run-pass=legalizer -debugify-and-strip-all-safe %s -o - \| FileCheck %s

	---			---
				name: narrow_scalar_uaddo_s128
				body: \|
				bb.0.entry:
				; CHECK-LABEL: name: narrow_scalar_uaddo_s128
				; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x0
				; CHECK: [[COPY1:%[0-9]+]]:_(s64) = COPY $x1
				; CHECK: [[COPY2:%[0-9]+]]:_(s64) = COPY $x2
				; CHECK: [[COPY3:%[0-9]+]]:_(s64) = COPY $x3
				; CHECK: [[UADDO:%[0-9]+]]:_(s64), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[COPY]], [[COPY2]]
				; CHECK: %17:_(s64), %carry_out:_(s1) = G_UADDE [[COPY1]], [[COPY3]], [[UADDO1]]
				; CHECK: %carry_out_ext:_(s64) = G_ANYEXT %carry_out(s1)
				; CHECK: $x0 = COPY [[UADDO]](s64)
				; CHECK: $x1 = COPY %17(s64)
				; CHECK: $x2 = COPY %carry_out_ext(s64)
				%0:_(s64) = COPY $x0
				%1:_(s64) = COPY $x1
				%2:_(s64) = COPY $x2
				%3:_(s64) = COPY $x3
				%lhs:_(s128) = G_MERGE_VALUES %0, %1
				%rhs:_(s128) = G_MERGE_VALUES %2, %3
				%add:_(s128), %carry_out:_(s1) = G_UADDO %lhs, %rhs
				%add0:_(s64), %add1:_(s64) = G_UNMERGE_VALUES %add
				%carry_out_ext:_(s64) = G_ANYEXT %carry_out
				$x0 = COPY %add0
				$x1 = COPY %add1
				$x2 = COPY %carry_out_ext

				...
				---
				name: narrow_scalar_uaddo_big_nonpow2
				body: \|
				bb.0.entry:
				; CHECK-LABEL: name: narrow_scalar_uaddo_big_nonpow2
				; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x0
				; CHECK: [[COPY1:%[0-9]+]]:_(s64) = COPY $x1
				; CHECK: [[COPY2:%[0-9]+]]:_(s64) = COPY $x2
				; CHECK: [[COPY3:%[0-9]+]]:_(s64) = COPY $x3
				; CHECK: [[UADDO:%[0-9]+]]:_(s64), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[COPY]], [[COPY1]]
				; CHECK: [[UADDE:%[0-9]+]]:_(s64), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[COPY1]], [[COPY2]], [[UADDO1]]
				; CHECK: %22:_(s64), %carry_out:_(s1) = G_UADDE [[COPY2]], [[COPY3]], [[UADDE1]]
				; CHECK: %carry_out_ext:_(s64) = G_ANYEXT %carry_out(s1)
				; CHECK: $x0 = COPY [[UADDO]](s64)
				; CHECK: $x1 = COPY [[UADDE]](s64)
				; CHECK: $x2 = COPY %22(s64)
				; CHECK: $x3 = COPY %carry_out_ext(s64)
				%0:_(s64) = COPY $x0
				%1:_(s64) = COPY $x1
				%2:_(s64) = COPY $x2
				%3:_(s64) = COPY $x3
				%lhs:_(s192) = G_MERGE_VALUES %0, %1, %2
				%rhs:_(s192) = G_MERGE_VALUES %1, %2, %3
				%add:_(s192), %carry_out:_(s1) = G_UADDO %lhs, %rhs
				%add0:_(s64), %add1:_(s64), %add2:_(s64) = G_UNMERGE_VALUES %add
				%carry_out_ext:_(s64) = G_ANYEXT %carry_out
				$x0 = COPY %add0
				$x1 = COPY %add1
				$x2 = COPY %add2
				$x3 = COPY %carry_out_ext

				...
				---
	name: test_scalar_uaddo_small			name: test_scalar_uaddo_small
	body: \|			body: \|
	bb.0.entry:			bb.0.entry:
	; CHECK-LABEL: name: test_scalar_uaddo_small			; CHECK-LABEL: name: test_scalar_uaddo_small
	; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x0			; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x0
	; CHECK: [[COPY1:%[0-9]+]]:_(s64) = COPY $x1			; CHECK: [[COPY1:%[0-9]+]]:_(s64) = COPY $x1
	; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 255			; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
	; CHECK: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY]](s64)			; CHECK: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY]](s64)
	▲ Show 20 Lines • Show All 43 Lines • Show Last 20 Lines

llvm/test/CodeGen/AArch64/GlobalISel/legalize-usubo.mir

	# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py			# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
	# RUN: llc -mtriple aarch64 -verify-machineinstrs -run-pass=legalizer -debugify-and-strip-all-safe %s -o - \| FileCheck %s			# RUN: llc -mtriple aarch64 -verify-machineinstrs -run-pass=legalizer -debugify-and-strip-all-safe %s -o - \| FileCheck %s

	---			---
				name: narrow_scalar_usubo_s128
				body: \|
				bb.0.entry:
				; CHECK-LABEL: name: narrow_scalar_usubo_s128
				; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x0
				; CHECK: [[COPY1:%[0-9]+]]:_(s64) = COPY $x1
				; CHECK: [[COPY2:%[0-9]+]]:_(s64) = COPY $x2
				; CHECK: [[COPY3:%[0-9]+]]:_(s64) = COPY $x3
				; CHECK: [[USUBO:%[0-9]+]]:_(s64), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[COPY]], [[COPY2]]
				; CHECK: %17:_(s64), %carry_out:_(s1) = G_USUBE [[COPY1]], [[COPY3]], [[USUBO1]]
				; CHECK: %carry_out_ext:_(s64) = G_ANYEXT %carry_out(s1)
				; CHECK: $x0 = COPY [[USUBO]](s64)
				; CHECK: $x1 = COPY %17(s64)
				; CHECK: $x2 = COPY %carry_out_ext(s64)
				%0:_(s64) = COPY $x0
				%1:_(s64) = COPY $x1
				%2:_(s64) = COPY $x2
				%3:_(s64) = COPY $x3
				%lhs:_(s128) = G_MERGE_VALUES %0, %1
				%rhs:_(s128) = G_MERGE_VALUES %2, %3
				%sub:_(s128), %carry_out:_(s1) = G_USUBO %lhs, %rhs
				%sub0:_(s64), %sub1:_(s64) = G_UNMERGE_VALUES %sub
				%carry_out_ext:_(s64) = G_ANYEXT %carry_out
				$x0 = COPY %sub0
				$x1 = COPY %sub1
				$x2 = COPY %carry_out_ext

				...
				---
				name: narrow_scalar_usubo_big_nonpow2
				body: \|
				bb.0.entry:
				; CHECK-LABEL: name: narrow_scalar_usubo_big_nonpow2
				; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x0
				; CHECK: [[COPY1:%[0-9]+]]:_(s64) = COPY $x1
				; CHECK: [[COPY2:%[0-9]+]]:_(s64) = COPY $x2
				; CHECK: [[COPY3:%[0-9]+]]:_(s64) = COPY $x3
				; CHECK: [[USUBO:%[0-9]+]]:_(s64), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[COPY]], [[COPY1]]
				; CHECK: [[USUBE:%[0-9]+]]:_(s64), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[COPY1]], [[COPY2]], [[USUBO1]]
				; CHECK: %22:_(s64), %carry_out:_(s1) = G_USUBE [[COPY2]], [[COPY3]], [[USUBE1]]
				; CHECK: %carry_out_ext:_(s64) = G_ANYEXT %carry_out(s1)
				; CHECK: $x0 = COPY [[USUBO]](s64)
				; CHECK: $x1 = COPY [[USUBE]](s64)
				; CHECK: $x2 = COPY %22(s64)
				; CHECK: $x3 = COPY %carry_out_ext(s64)
				%0:_(s64) = COPY $x0
				%1:_(s64) = COPY $x1
				%2:_(s64) = COPY $x2
				%3:_(s64) = COPY $x3
				%lhs:_(s192) = G_MERGE_VALUES %0, %1, %2
				%rhs:_(s192) = G_MERGE_VALUES %1, %2, %3
				%sub:_(s192), %carry_out:_(s1) = G_USUBO %lhs, %rhs
				%sub0:_(s64), %sub1:_(s64), %sub2:_(s64) = G_UNMERGE_VALUES %sub
				%carry_out_ext:_(s64) = G_ANYEXT %carry_out
				$x0 = COPY %sub0
				$x1 = COPY %sub1
				$x2 = COPY %sub2
				$x3 = COPY %carry_out_ext

				...
				---
	name: test_scalar_usubo_small			name: test_scalar_usubo_small
	body: \|			body: \|
	bb.0.entry:			bb.0.entry:
	; CHECK-LABEL: name: test_scalar_usubo_small			; CHECK-LABEL: name: test_scalar_usubo_small
	; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x0			; CHECK: [[COPY:%[0-9]+]]:_(s64) = COPY $x0
	; CHECK: [[COPY1:%[0-9]+]]:_(s64) = COPY $x1			; CHECK: [[COPY1:%[0-9]+]]:_(s64) = COPY $x1
	; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 255			; CHECK: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 255
	; CHECK: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY]](s64)			; CHECK: [[TRUNC:%[0-9]+]]:_(s32) = G_TRUNC [[COPY]](s64)
	▲ Show 20 Lines • Show All 43 Lines • Show Last 20 Lines