[llvm] e417e61 - [RISCV][LegalizeTypes] Add additional test coverage for type promotion of VP_FSHL/FSHR. NFC

[Craig Topper via llvm-commits]

Author: Craig Topper
Date: 2024-05-14T16:25:07-07:00
New Revision: e417e61532ac373e7b0708262dedefcdaf6ced9c

URL: https://github.com/llvm/llvm-project/commit/e417e61532ac373e7b0708262dedefcdaf6ced9c
DIFF: https://github.com/llvm/llvm-project/commit/e417e61532ac373e7b0708262dedefcdaf6ced9c.diff

LOG: [RISCV][LegalizeTypes] Add additional test coverage for type promotion of VP_FSHL/FSHR. NFC

There's a special path when the promoted type has an element size
more than twice the size of the original type.

Added: 
    

Modified: 
    llvm/test/CodeGen/RISCV/rvv/fshr-fshl-vp.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/CodeGen/RISCV/rvv/fshr-fshl-vp.ll b/llvm/test/CodeGen/RISCV/rvv/fshr-fshl-vp.ll
index f9f085dcc1614..277cd7dcdabce 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fshr-fshl-vp.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fshr-fshl-vp.ll
@@ -1370,3 +1370,56 @@ define <vscale x 1 x i9> @fshl_v1i9(<vscale x 1 x i9> %a, <vscale x 1 x i9> %b,
   %res = call <vscale x 1 x i9> @llvm.vp.fshl.nxv1i9(<vscale x 1 x i9> %a, <vscale x 1 x i9> %b, <vscale x 1 x i9> %c, <vscale x 1 x i1> %m, i32 %evl)
   ret <vscale x 1 x i9> %res
 }
+
+declare <vscale x 1 x i4> @llvm.vp.trunc.nxv1i4.nxv1i8(<vscale x 1 x i8>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x i8> @llvm.vp.zext.nxv1i8.nxv1i4(<vscale x 1 x i4>, <vscale x 1 x i1>, i32)
+declare <vscale x 1 x i4> @llvm.vp.fshr.nxv1i4(<vscale x 1 x i4>, <vscale x 1 x i4>, <vscale x 1 x i4>, <vscale x 1 x i1>, i32)
+define <vscale x 1 x i8> @fshr_v1i4(<vscale x 1 x i8> %a, <vscale x 1 x i8> %b, <vscale x 1 x i8> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: fshr_v1i4:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, mf8, ta, ma
+; CHECK-NEXT:    vand.vi v10, v10, 15
+; CHECK-NEXT:    li a1, 4
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf8, ta, ma
+; CHECK-NEXT:    vremu.vx v10, v10, a1, v0.t
+; CHECK-NEXT:    vsll.vi v8, v8, 4, v0.t
+; CHECK-NEXT:    vsetvli a1, zero, e8, mf8, ta, ma
+; CHECK-NEXT:    vand.vi v9, v9, 15
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf8, ta, ma
+; CHECK-NEXT:    vor.vv v8, v8, v9, v0.t
+; CHECK-NEXT:    vsrl.vv v8, v8, v10, v0.t
+; CHECK-NEXT:    vand.vi v8, v8, 15, v0.t
+; CHECK-NEXT:    ret
+  %trunca = call <vscale x 1 x i4> @llvm.vp.trunc.nxv1i4.nxv1i8(<vscale x 1 x i8> %a, <vscale x 1 x i1> %m, i32 zeroext %evl)
+  %truncb = call <vscale x 1 x i4> @llvm.vp.trunc.nxv1i4.nxv1i8(<vscale x 1 x i8> %b, <vscale x 1 x i1> %m, i32 zeroext %evl)
+  %truncc = call <vscale x 1 x i4> @llvm.vp.trunc.nxv1i4.nxv1i8(<vscale x 1 x i8> %c, <vscale x 1 x i1> %m, i32 zeroext %evl)
+  %fshr = call <vscale x 1 x i4> @llvm.vp.fshr.nxv1i4(<vscale x 1 x i4> %trunca, <vscale x 1 x i4> %truncb, <vscale x 1 x i4> %truncc, <vscale x 1 x i1> %m, i32 %evl)
+  %res = call <vscale x 1 x i8> @llvm.vp.zext.nxv1i8.nxv1i4(<vscale x 1 x i4> %fshr, <vscale x 1 x i1> %m, i32 zeroext %evl)
+  ret <vscale x 1 x i8> %res
+}
+
+declare <vscale x 1 x i4> @llvm.vp.fshl.nxv1i4(<vscale x 1 x i4>, <vscale x 1 x i4>, <vscale x 1 x i4>, <vscale x 1 x i1>, i32)
+define <vscale x 1 x i8> @fshl_v1i4(<vscale x 1 x i8> %a, <vscale x 1 x i8> %b, <vscale x 1 x i8> %c, <vscale x 1 x i1> %m, i32 zeroext %evl) {
+; CHECK-LABEL: fshl_v1i4:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vsetvli a1, zero, e8, mf8, ta, ma
+; CHECK-NEXT:    vand.vi v10, v10, 15
+; CHECK-NEXT:    li a1, 4
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf8, ta, ma
+; CHECK-NEXT:    vremu.vx v10, v10, a1, v0.t
+; CHECK-NEXT:    vsll.vi v8, v8, 4, v0.t
+; CHECK-NEXT:    vsetvli a1, zero, e8, mf8, ta, ma
+; CHECK-NEXT:    vand.vi v9, v9, 15
+; CHECK-NEXT:    vsetvli zero, a0, e8, mf8, ta, ma
+; CHECK-NEXT:    vor.vv v8, v8, v9, v0.t
+; CHECK-NEXT:    vsll.vv v8, v8, v10, v0.t
+; CHECK-NEXT:    vsrl.vi v8, v8, 4, v0.t
+; CHECK-NEXT:    vand.vi v8, v8, 15, v0.t
+; CHECK-NEXT:    ret
+  %trunca = call <vscale x 1 x i4> @llvm.vp.trunc.nxv1i4.nxv1i8(<vscale x 1 x i8> %a, <vscale x 1 x i1> %m, i32 zeroext %evl)
+  %truncb = call <vscale x 1 x i4> @llvm.vp.trunc.nxv1i4.nxv1i8(<vscale x 1 x i8> %b, <vscale x 1 x i1> %m, i32 zeroext %evl)
+  %truncc = call <vscale x 1 x i4> @llvm.vp.trunc.nxv1i4.nxv1i8(<vscale x 1 x i8> %c, <vscale x 1 x i1> %m, i32 zeroext %evl)
+  %fshl = call <vscale x 1 x i4> @llvm.vp.fshl.nxv1i4(<vscale x 1 x i4> %trunca, <vscale x 1 x i4> %truncb, <vscale x 1 x i4> %truncc, <vscale x 1 x i1> %m, i32 %evl)
+  %res = call <vscale x 1 x i8> @llvm.vp.zext.nxv1i8.nxv1i4(<vscale x 1 x i4> %fshl, <vscale x 1 x i1> %m, i32 zeroext %evl)
+  ret <vscale x 1 x i8> %res
+}