[X86][AVX] Add v4i64 shift-by-32 tests

AVX1 could perform this as a v8f32 shuffle instead of splitting - based off PR46621

diff --git a/llvm/test/CodeGen/X86/vector-shift-ashr-256.ll b/llvm/test/CodeGen/X86/vector-shift-ashr-256.ll
index 81cadf7..0f1f23e 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-shift-ashr-256.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-ashr-256.ll
@@ -1659,3 +1659,77 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) nounwind {
   %shift = ashr <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
 }
+
+;
+; Special Cases
+;
+
+define <4 x i64> @shift32_v4i64(<4 x i64> %a) nounwind {
+; AVX1-LABEL: shift32_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpsrad $31, %xmm1, %xmm2
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
+; AVX1-NEXT:    vpsrad $31, %xmm0, %xmm2
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3],xmm0[4,5],xmm2[6,7]
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: shift32_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpsrad $31, %ymm0, %ymm1
+; AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
+; AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
+; AVX2-NEXT:    retq
+;
+; XOPAVX1-LABEL: shift32_v4i64:
+; XOPAVX1:       # %bb.0:
+; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; XOPAVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [18446744073709551584,18446744073709551584]
+; XOPAVX1-NEXT:    vpshaq %xmm2, %xmm1, %xmm1
+; XOPAVX1-NEXT:    vpshaq %xmm2, %xmm0, %xmm0
+; XOPAVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; XOPAVX1-NEXT:    retq
+;
+; XOPAVX2-LABEL: shift32_v4i64:
+; XOPAVX2:       # %bb.0:
+; XOPAVX2-NEXT:    vpsrad $31, %ymm0, %ymm1
+; XOPAVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
+; XOPAVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
+; XOPAVX2-NEXT:    retq
+;
+; AVX512-LABEL: shift32_v4i64:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512-NEXT:    vpsraq $32, %zmm0, %zmm0
+; AVX512-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; AVX512-NEXT:    retq
+;
+; AVX512VL-LABEL: shift32_v4i64:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpsraq $32, %ymm0, %ymm0
+; AVX512VL-NEXT:    retq
+;
+; X86-AVX1-LABEL: shift32_v4i64:
+; X86-AVX1:       # %bb.0:
+; X86-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; X86-AVX1-NEXT:    vpsrad $31, %xmm1, %xmm2
+; X86-AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
+; X86-AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,3],xmm1[4,5],xmm2[6,7]
+; X86-AVX1-NEXT:    vpsrad $31, %xmm0, %xmm2
+; X86-AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
+; X86-AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm2[2,3],xmm0[4,5],xmm2[6,7]
+; X86-AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; X86-AVX1-NEXT:    retl
+;
+; X86-AVX2-LABEL: shift32_v4i64:
+; X86-AVX2:       # %bb.0:
+; X86-AVX2-NEXT:    vpsrad $31, %ymm0, %ymm1
+; X86-AVX2-NEXT:    vpshufd {{.*#+}} ymm0 = ymm0[1,1,3,3,5,5,7,7]
+; X86-AVX2-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
+; X86-AVX2-NEXT:    retl
+  %shift = ashr <4 x i64> %a, <i64 32, i64 32, i64 32, i64 32>
+  ret <4 x i64> %shift
+}

diff --git a/llvm/test/CodeGen/X86/vector-shift-lshr-256.ll b/llvm/test/CodeGen/X86/vector-shift-lshr-256.ll
index 36d2470..9fd0960 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-shift-lshr-256.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-lshr-256.ll
@@ -1390,6 +1390,63 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) nounwind {
   ret <32 x i8> %shift
 }
 
+;
+; Special Cases
+;
+
+define <4 x i64> @shift32_v4i64(<4 x i64> %a) nounwind {
+; AVX1-LABEL: shift32_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpsrlq $32, %xmm0, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: shift32_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpsrlq $32, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; XOPAVX1-LABEL: shift32_v4i64:
+; XOPAVX1:       # %bb.0:
+; XOPAVX1-NEXT:    vpsrlq $32, %xmm0, %xmm1
+; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; XOPAVX1-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; XOPAVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; XOPAVX1-NEXT:    retq
+;
+; XOPAVX2-LABEL: shift32_v4i64:
+; XOPAVX2:       # %bb.0:
+; XOPAVX2-NEXT:    vpsrlq $32, %ymm0, %ymm0
+; XOPAVX2-NEXT:    retq
+;
+; AVX512-LABEL: shift32_v4i64:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpsrlq $32, %ymm0, %ymm0
+; AVX512-NEXT:    retq
+;
+; AVX512VL-LABEL: shift32_v4i64:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpsrlq $32, %ymm0, %ymm0
+; AVX512VL-NEXT:    retq
+;
+; X86-AVX1-LABEL: shift32_v4i64:
+; X86-AVX1:       # %bb.0:
+; X86-AVX1-NEXT:    vpsrlq $32, %xmm0, %xmm1
+; X86-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; X86-AVX1-NEXT:    vpsrlq $32, %xmm0, %xmm0
+; X86-AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; X86-AVX1-NEXT:    retl
+;
+; X86-AVX2-LABEL: shift32_v4i64:
+; X86-AVX2:       # %bb.0:
+; X86-AVX2-NEXT:    vpsrlq $32, %ymm0, %ymm0
+; X86-AVX2-NEXT:    retl
+  %shift = lshr <4 x i64> %a, <i64 32, i64 32, i64 32, i64 32>
+  ret <4 x i64> %shift
+}
+
 define <4 x i32> @sh_trunc_sh_vec(<4 x i64> %x) {
 ; AVX1-LABEL: sh_trunc_sh_vec:
 ; AVX1:       # %bb.0:

diff --git a/llvm/test/CodeGen/X86/vector-shift-shl-256.ll b/llvm/test/CodeGen/X86/vector-shift-shl-256.ll
index da3cebc..0af2398 100644 (file)
--- a/llvm/test/CodeGen/X86/vector-shift-shl-256.ll
+++ b/llvm/test/CodeGen/X86/vector-shift-shl-256.ll
@@ -1298,3 +1298,60 @@ define <32 x i8> @splatconstant_shift_v32i8(<32 x i8> %a) nounwind {
   %shift = shl <32 x i8> %a, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
   ret <32 x i8> %shift
 }
+
+;
+; Special Cases
+;
+
+define <4 x i64> @shift32_v4i64(<4 x i64> %a) nounwind {
+; AVX1-LABEL: shift32_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpsllq $32, %xmm0, %xmm1
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpsllq $32, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: shift32_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpsllq $32, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; XOPAVX1-LABEL: shift32_v4i64:
+; XOPAVX1:       # %bb.0:
+; XOPAVX1-NEXT:    vpsllq $32, %xmm0, %xmm1
+; XOPAVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; XOPAVX1-NEXT:    vpsllq $32, %xmm0, %xmm0
+; XOPAVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; XOPAVX1-NEXT:    retq
+;
+; XOPAVX2-LABEL: shift32_v4i64:
+; XOPAVX2:       # %bb.0:
+; XOPAVX2-NEXT:    vpsllq $32, %ymm0, %ymm0
+; XOPAVX2-NEXT:    retq
+;
+; AVX512-LABEL: shift32_v4i64:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpsllq $32, %ymm0, %ymm0
+; AVX512-NEXT:    retq
+;
+; AVX512VL-LABEL: shift32_v4i64:
+; AVX512VL:       # %bb.0:
+; AVX512VL-NEXT:    vpsllq $32, %ymm0, %ymm0
+; AVX512VL-NEXT:    retq
+;
+; X86-AVX1-LABEL: shift32_v4i64:
+; X86-AVX1:       # %bb.0:
+; X86-AVX1-NEXT:    vpsllq $32, %xmm0, %xmm1
+; X86-AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; X86-AVX1-NEXT:    vpsllq $32, %xmm0, %xmm0
+; X86-AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; X86-AVX1-NEXT:    retl
+;
+; X86-AVX2-LABEL: shift32_v4i64:
+; X86-AVX2:       # %bb.0:
+; X86-AVX2-NEXT:    vpsllq $32, %ymm0, %ymm0
+; X86-AVX2-NEXT:    retl
+  %shift = shl <4 x i64> %a, <i64 32, i64 32, i64 32, i64 32>
+  ret <4 x i64> %shift
+}