Skip to content

[GlobalISel] Add support to moreElementsVector for G_SEXT, G_ZEXT and G_ANYEXT #85038

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Mar 18, 2024
Merged

[GlobalISel] Add support to moreElementsVector for G_SEXT, G_ZEXT and G_ANYEXT #85038

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e6c1c9f
[𝘀𝗽𝗿] initial version
dc03-work Mar 13, 2024
beec23e
Refactor code to reuse G_TRUNC et al. cases
dc03-work Mar 13, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5411,6 +5411,9 @@ LegalizerHelper::moreElementsVector(MachineInstr &MI, unsigned TypeIdx,
MI.eraseFromParent();
return Legalized;
}
case TargetOpcode::G_SEXT:
case TargetOpcode::G_ZEXT:
case TargetOpcode::G_ANYEXT:
case TargetOpcode::G_TRUNC:
case TargetOpcode::G_FPTRUNC:
case TargetOpcode::G_FPEXT:
Expand Down
2 changes: 1 addition & 1 deletion llvm/lib/Target/AArch64/GISel/AArch64LegalizerInfo.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -600,7 +600,7 @@ AArch64LegalizerInfo::AArch64LegalizerInfo(const AArch64Subtarget &ST)
.legalIf(ExtLegalFunc)
.legalFor({{v2s64, v2s32}, {v4s32, v4s16}, {v8s16, v8s8}})
.clampScalar(0, s64, s64) // Just for s128, others are handled above.
.moreElementsToNextPow2(1)
.moreElementsToNextPow2(0)
.clampMaxNumElements(1, s8, 8)
.clampMaxNumElements(1, s16, 4)
.clampMaxNumElements(1, s32, 2)
Expand Down
81 changes: 37 additions & 44 deletions llvm/test/CodeGen/AArch64/fcmp.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1108,61 +1108,54 @@ define <7 x i32> @v7f16_i32(<7 x half> %a, <7 x half> %b, <7 x i32> %d, <7 x i32
;
; CHECK-GI-FP16-LABEL: v7f16_i32:
; CHECK-GI-FP16: // %bb.0: // %entry
; CHECK-GI-FP16-NEXT: fcmgt v1.8h, v1.8h, v0.8h
; CHECK-GI-FP16-NEXT: mov w12, #31 // =0x1f
; CHECK-GI-FP16-NEXT: ldr s4, [sp]
; CHECK-GI-FP16-NEXT: fmov s2, w12
; CHECK-GI-FP16-NEXT: fcmgt v0.8h, v1.8h, v0.8h
; CHECK-GI-FP16-NEXT: mov w10, #31 // =0x1f
; CHECK-GI-FP16-NEXT: ldr s3, [sp]
; CHECK-GI-FP16-NEXT: fmov s1, w10
; CHECK-GI-FP16-NEXT: fmov s6, w0
; CHECK-GI-FP16-NEXT: ldr s5, [sp, #8]
; CHECK-GI-FP16-NEXT: ldr s4, [sp, #8]
; CHECK-GI-FP16-NEXT: ldr s7, [sp, #24]
; CHECK-GI-FP16-NEXT: ldr s16, [sp, #32]
; CHECK-GI-FP16-NEXT: umov w9, v1.h[4]
; CHECK-GI-FP16-NEXT: umov w8, v1.h[0]
; CHECK-GI-FP16-NEXT: umov w11, v1.h[5]
; CHECK-GI-FP16-NEXT: umov w10, v1.h[1]
; CHECK-GI-FP16-NEXT: mov v2.s[1], w12
; CHECK-GI-FP16-NEXT: umov w13, v1.h[2]
; CHECK-GI-FP16-NEXT: umov w8, v0.h[4]
; CHECK-GI-FP16-NEXT: umov w9, v0.h[5]
; CHECK-GI-FP16-NEXT: mov v1.s[1], w10
; CHECK-GI-FP16-NEXT: mov v6.s[1], w1
; CHECK-GI-FP16-NEXT: mov v7.s[1], v16.s[0]
; CHECK-GI-FP16-NEXT: ldr s16, [sp, #40]
; CHECK-GI-FP16-NEXT: fmov s3, w9
; CHECK-GI-FP16-NEXT: fmov s0, w8
; CHECK-GI-FP16-NEXT: umov w8, v1.h[6]
; CHECK-GI-FP16-NEXT: mov v2.s[2], w12
; CHECK-GI-FP16-NEXT: umov w9, v1.h[3]
; CHECK-GI-FP16-NEXT: fmov s2, w8
; CHECK-GI-FP16-NEXT: umov w8, v0.h[6]
; CHECK-GI-FP16-NEXT: mov v1.s[2], w10
; CHECK-GI-FP16-NEXT: ushll v0.4s, v0.4h, #0
; CHECK-GI-FP16-NEXT: mov v6.s[2], w2
; CHECK-GI-FP16-NEXT: mov v7.s[2], v16.s[0]
; CHECK-GI-FP16-NEXT: mov v3.s[1], w11
; CHECK-GI-FP16-NEXT: mov v0.s[1], w10
; CHECK-GI-FP16-NEXT: mov w10, #-1 // =0xffffffff
; CHECK-GI-FP16-NEXT: fmov s1, w10
; CHECK-GI-FP16-NEXT: neg v17.4s, v2.4s
; CHECK-GI-FP16-NEXT: mov v2.s[1], w9
; CHECK-GI-FP16-NEXT: mov w9, #-1 // =0xffffffff
; CHECK-GI-FP16-NEXT: fmov s5, w9
; CHECK-GI-FP16-NEXT: neg v17.4s, v1.4s
; CHECK-GI-FP16-NEXT: shl v0.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: mov v6.s[3], w3
; CHECK-GI-FP16-NEXT: mov v2.s[2], w8
; CHECK-GI-FP16-NEXT: fmov w8, s3
; CHECK-GI-FP16-NEXT: fmov s3, w7
; CHECK-GI-FP16-NEXT: mov v5.s[1], w9
; CHECK-GI-FP16-NEXT: sshr v0.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: mov v3.s[1], w8
; CHECK-GI-FP16-NEXT: fmov w8, s4
; CHECK-GI-FP16-NEXT: ldr s4, [sp, #16]
; CHECK-GI-FP16-NEXT: ushl v1.4s, v2.4s, v1.4s
; CHECK-GI-FP16-NEXT: fmov s2, w4
; CHECK-GI-FP16-NEXT: mov v5.s[2], w9
; CHECK-GI-FP16-NEXT: mov v2.s[1], w5
; CHECK-GI-FP16-NEXT: mov v3.s[2], w8
; CHECK-GI-FP16-NEXT: sshl v1.4s, v1.4s, v17.4s
; CHECK-GI-FP16-NEXT: fmov w8, s4
; CHECK-GI-FP16-NEXT: fmov s4, w7
; CHECK-GI-FP16-NEXT: mov v0.s[2], w13
; CHECK-GI-FP16-NEXT: mov v1.s[1], w10
; CHECK-GI-FP16-NEXT: mov v4.s[1], w8
; CHECK-GI-FP16-NEXT: fmov w8, s5
; CHECK-GI-FP16-NEXT: ldr s5, [sp, #16]
; CHECK-GI-FP16-NEXT: ushl v2.4s, v3.4s, v2.4s
; CHECK-GI-FP16-NEXT: fmov s3, w4
; CHECK-GI-FP16-NEXT: mov v0.s[3], w9
; CHECK-GI-FP16-NEXT: mov v1.s[2], w10
; CHECK-GI-FP16-NEXT: mov v3.s[1], w5
; CHECK-GI-FP16-NEXT: mov v4.s[2], w8
; CHECK-GI-FP16-NEXT: sshl v2.4s, v2.4s, v17.4s
; CHECK-GI-FP16-NEXT: fmov w8, s5
; CHECK-GI-FP16-NEXT: shl v0.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: eor v1.16b, v2.16b, v1.16b
; CHECK-GI-FP16-NEXT: mov v3.s[2], w6
; CHECK-GI-FP16-NEXT: mov v4.s[3], w8
; CHECK-GI-FP16-NEXT: sshr v0.4s, v0.4s, #31
; CHECK-GI-FP16-NEXT: and v1.16b, v7.16b, v1.16b
; CHECK-GI-FP16-NEXT: and v2.16b, v3.16b, v2.16b
; CHECK-GI-FP16-NEXT: bsl v0.16b, v6.16b, v4.16b
; CHECK-GI-FP16-NEXT: orr v1.16b, v2.16b, v1.16b
; CHECK-GI-FP16-NEXT: eor v4.16b, v1.16b, v5.16b
; CHECK-GI-FP16-NEXT: mov v2.s[2], w6
; CHECK-GI-FP16-NEXT: mov v3.s[3], w8
; CHECK-GI-FP16-NEXT: and v1.16b, v2.16b, v1.16b
; CHECK-GI-FP16-NEXT: and v2.16b, v7.16b, v4.16b
; CHECK-GI-FP16-NEXT: bsl v0.16b, v6.16b, v3.16b
; CHECK-GI-FP16-NEXT: orr v1.16b, v1.16b, v2.16b
; CHECK-GI-FP16-NEXT: mov s2, v0.s[1]
; CHECK-GI-FP16-NEXT: mov s3, v0.s[2]
; CHECK-GI-FP16-NEXT: mov s4, v0.s[3]
Expand Down
14 changes: 6 additions & 8 deletions llvm/test/CodeGen/AArch64/sext.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -280,13 +280,12 @@ define <3 x i64> @sext_v3i8_v3i64(<3 x i8> %a) {
;
; CHECK-GI-LABEL: sext_v3i8_v3i64:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: // kill: def $w0 killed $w0 def $x0
; CHECK-GI-NEXT: fmov d0, x0
; CHECK-GI-NEXT: // kill: def $w1 killed $w1 def $x1
; CHECK-GI-NEXT: fmov s0, w0
; CHECK-GI-NEXT: // kill: def $w2 killed $w2 def $x2
; CHECK-GI-NEXT: sxtb x8, w2
; CHECK-GI-NEXT: fmov d2, x8
; CHECK-GI-NEXT: mov v0.d[1], x1
; CHECK-GI-NEXT: mov v0.s[1], w1
; CHECK-GI-NEXT: ushll v0.2d, v0.2s, #0
; CHECK-GI-NEXT: shl v0.2d, v0.2d, #56
; CHECK-GI-NEXT: sshr v0.2d, v0.2d, #56
; CHECK-GI-NEXT: mov d1, v0.d[1]
Expand Down Expand Up @@ -444,13 +443,12 @@ define <3 x i64> @sext_v3i10_v3i64(<3 x i10> %a) {
;
; CHECK-GI-LABEL: sext_v3i10_v3i64:
; CHECK-GI: // %bb.0: // %entry
; CHECK-GI-NEXT: // kill: def $w0 killed $w0 def $x0
; CHECK-GI-NEXT: fmov d0, x0
; CHECK-GI-NEXT: // kill: def $w1 killed $w1 def $x1
; CHECK-GI-NEXT: fmov s0, w0
; CHECK-GI-NEXT: // kill: def $w2 killed $w2 def $x2
; CHECK-GI-NEXT: sbfx x8, x2, #0, #10
; CHECK-GI-NEXT: fmov d2, x8
; CHECK-GI-NEXT: mov v0.d[1], x1
; CHECK-GI-NEXT: mov v0.s[1], w1
; CHECK-GI-NEXT: ushll v0.2d, v0.2s, #0
; CHECK-GI-NEXT: shl v0.2d, v0.2d, #54
; CHECK-GI-NEXT: sshr v0.2d, v0.2d, #54
; CHECK-GI-NEXT: mov d1, v0.d[1]
Expand Down
Loading