diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/masked_store_trunc_usat.ll')
| -rw-r--r-- | llvm/test/CodeGen/X86/masked_store_trunc_usat.ll | 183 |
1 files changed, 87 insertions, 96 deletions
diff --git a/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll b/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll index f28929589b6..085d488e086 100644 --- a/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll +++ b/llvm/test/CodeGen/X86/masked_store_trunc_usat.ll @@ -248,30 +248,28 @@ define void @truncstore_v8i64_v8i32(<8 x i64> %x, <8 x i32>* %p, <8 x i32> %mask ; AVX1-NEXT: vpxor %xmm5, %xmm3, %xmm3 ; AVX1-NEXT: vpcmpeqd %xmm4, %xmm2, %xmm2 ; AVX1-NEXT: vpxor %xmm5, %xmm2, %xmm2 -; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm2 -; AVX1-NEXT: vmovapd {{.*#+}} ymm3 = [4294967295,4294967295,4294967295,4294967295] -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm4 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372036854775808,9223372036854775808] -; AVX1-NEXT: vpxor %xmm5, %xmm4, %xmm4 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [9223372041149743103,9223372041149743103] -; AVX1-NEXT: vpcmpgtq %xmm4, %xmm6, %xmm4 -; AVX1-NEXT: vpxor %xmm5, %xmm0, %xmm7 -; AVX1-NEXT: vpcmpgtq %xmm7, %xmm6, %xmm7 -; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm7, %ymm4 -; AVX1-NEXT: vblendvpd %ymm4, %ymm0, %ymm3, %ymm0 +; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm2, %ymm8 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [9223372036854775808,9223372036854775808] +; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm4 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372041149743103,9223372041149743103] +; AVX1-NEXT: vpcmpgtq %xmm4, %xmm5, %xmm9 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm6 +; AVX1-NEXT: vpxor %xmm3, %xmm6, %xmm7 +; AVX1-NEXT: vpcmpgtq %xmm7, %xmm5, %xmm7 +; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm2 +; AVX1-NEXT: vpcmpgtq %xmm2, %xmm5, %xmm2 ; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm4 -; AVX1-NEXT: vpxor %xmm5, %xmm4, %xmm4 -; AVX1-NEXT: vpcmpgtq %xmm4, %xmm6, %xmm4 -; AVX1-NEXT: vpxor %xmm5, %xmm1, %xmm5 -; AVX1-NEXT: vpcmpgtq %xmm5, %xmm6, %xmm5 -; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm5, %ymm4 -; AVX1-NEXT: vblendvpd %ymm4, %ymm1, %ymm3, %ymm1 -; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3 +; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm3 +; AVX1-NEXT: vpcmpgtq %xmm3, %xmm5, %xmm3 +; AVX1-NEXT: vmovapd {{.*#+}} xmm5 = [4294967295,4294967295] +; AVX1-NEXT: vblendvpd %xmm3, %xmm4, %xmm5, %xmm3 +; AVX1-NEXT: vblendvpd %xmm2, %xmm1, %xmm5, %xmm1 ; AVX1-NEXT: vshufps {{.*#+}} xmm1 = xmm1[0,2],xmm3[0,2] -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm3[0,2] +; AVX1-NEXT: vblendvpd %xmm7, %xmm6, %xmm5, %xmm2 +; AVX1-NEXT: vblendvpd %xmm9, %xmm0, %xmm5, %xmm0 +; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2] ; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0 -; AVX1-NEXT: vmaskmovps %ymm0, %ymm2, (%rdi) +; AVX1-NEXT: vmaskmovps %ymm0, %ymm8, (%rdi) ; AVX1-NEXT: vzeroupper ; AVX1-NEXT: retq ; @@ -580,30 +578,28 @@ define void @truncstore_v8i64_v8i16(<8 x i64> %x, <8 x i16>* %p, <8 x i32> %mask ; AVX1-NEXT: vpxor %xmm8, %xmm8, %xmm8 ; AVX1-NEXT: vpcmpeqd %xmm8, %xmm2, %xmm5 ; AVX1-NEXT: vpcmpeqd %xmm9, %xmm9, %xmm9 -; AVX1-NEXT: vpxor %xmm9, %xmm5, %xmm5 -; AVX1-NEXT: vmovapd {{.*#+}} ymm10 = [65535,65535,65535,65535] -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm7 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [9223372036854775808,9223372036854775808] -; AVX1-NEXT: vpxor %xmm3, %xmm7, %xmm7 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [9223372036854841343,9223372036854841343] -; AVX1-NEXT: vpcmpgtq %xmm7, %xmm4, %xmm7 -; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm6 -; AVX1-NEXT: vpcmpgtq %xmm6, %xmm4, %xmm6 -; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6 -; AVX1-NEXT: vblendvpd %ymm6, %ymm0, %ymm10, %ymm0 -; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm6 -; AVX1-NEXT: vpxor %xmm3, %xmm6, %xmm6 -; AVX1-NEXT: vpcmpgtq %xmm6, %xmm4, %xmm6 -; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm3 -; AVX1-NEXT: vpcmpgtq %xmm3, %xmm4, %xmm3 -; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm3, %ymm3 -; AVX1-NEXT: vblendvpd %ymm3, %ymm1, %ymm10, %ymm1 -; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3 +; AVX1-NEXT: vpxor %xmm9, %xmm5, %xmm10 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [9223372036854775808,9223372036854775808] +; AVX1-NEXT: vpxor %xmm6, %xmm0, %xmm7 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [9223372036854841343,9223372036854841343] +; AVX1-NEXT: vpcmpgtq %xmm7, %xmm3, %xmm11 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm4 +; AVX1-NEXT: vpxor %xmm6, %xmm4, %xmm5 +; AVX1-NEXT: vpcmpgtq %xmm5, %xmm3, %xmm12 +; AVX1-NEXT: vpxor %xmm6, %xmm1, %xmm7 +; AVX1-NEXT: vpcmpgtq %xmm7, %xmm3, %xmm7 +; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm5 +; AVX1-NEXT: vpxor %xmm6, %xmm5, %xmm6 +; AVX1-NEXT: vpcmpgtq %xmm6, %xmm3, %xmm3 +; AVX1-NEXT: vmovapd {{.*#+}} xmm6 = [65535,65535] +; AVX1-NEXT: vblendvpd %xmm3, %xmm5, %xmm6, %xmm3 +; AVX1-NEXT: vblendvpd %xmm7, %xmm1, %xmm6, %xmm1 ; AVX1-NEXT: vpackusdw %xmm3, %xmm1, %xmm1 -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vblendvpd %xmm12, %xmm4, %xmm6, %xmm3 +; AVX1-NEXT: vblendvpd %xmm11, %xmm0, %xmm6, %xmm0 ; AVX1-NEXT: vpackusdw %xmm3, %xmm0, %xmm0 ; AVX1-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 -; AVX1-NEXT: vpextrb $0, %xmm5, %eax +; AVX1-NEXT: vpextrb $0, %xmm10, %eax ; AVX1-NEXT: testb $1, %al ; AVX1-NEXT: je .LBB1_2 ; AVX1-NEXT: # %bb.1: # %cond.store @@ -1092,30 +1088,28 @@ define void @truncstore_v8i64_v8i8(<8 x i64> %x, <8 x i8>* %p, <8 x i32> %mask) ; AVX1-NEXT: vpxor %xmm8, %xmm8, %xmm8 ; AVX1-NEXT: vpcmpeqd %xmm8, %xmm2, %xmm5 ; AVX1-NEXT: vpcmpeqd %xmm9, %xmm9, %xmm9 -; AVX1-NEXT: vpxor %xmm9, %xmm5, %xmm5 -; AVX1-NEXT: vmovapd {{.*#+}} ymm10 = [255,255,255,255] -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm7 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [9223372036854775808,9223372036854775808] -; AVX1-NEXT: vpxor %xmm3, %xmm7, %xmm7 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [9223372036854776063,9223372036854776063] -; AVX1-NEXT: vpcmpgtq %xmm7, %xmm4, %xmm7 -; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm6 -; AVX1-NEXT: vpcmpgtq %xmm6, %xmm4, %xmm6 -; AVX1-NEXT: vinsertf128 $1, %xmm7, %ymm6, %ymm6 -; AVX1-NEXT: vblendvpd %ymm6, %ymm0, %ymm10, %ymm0 -; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm6 -; AVX1-NEXT: vpxor %xmm3, %xmm6, %xmm6 -; AVX1-NEXT: vpcmpgtq %xmm6, %xmm4, %xmm6 -; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm3 -; AVX1-NEXT: vpcmpgtq %xmm3, %xmm4, %xmm3 -; AVX1-NEXT: vinsertf128 $1, %xmm6, %ymm3, %ymm3 -; AVX1-NEXT: vblendvpd %ymm3, %ymm1, %ymm10, %ymm1 -; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3 +; AVX1-NEXT: vpxor %xmm9, %xmm5, %xmm10 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [9223372036854775808,9223372036854775808] +; AVX1-NEXT: vpxor %xmm6, %xmm0, %xmm7 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [9223372036854776063,9223372036854776063] +; AVX1-NEXT: vpcmpgtq %xmm7, %xmm3, %xmm11 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm4 +; AVX1-NEXT: vpxor %xmm6, %xmm4, %xmm5 +; AVX1-NEXT: vpcmpgtq %xmm5, %xmm3, %xmm12 +; AVX1-NEXT: vpxor %xmm6, %xmm1, %xmm7 +; AVX1-NEXT: vpcmpgtq %xmm7, %xmm3, %xmm7 +; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm5 +; AVX1-NEXT: vpxor %xmm6, %xmm5, %xmm6 +; AVX1-NEXT: vpcmpgtq %xmm6, %xmm3, %xmm3 +; AVX1-NEXT: vmovapd {{.*#+}} xmm6 = [255,255] +; AVX1-NEXT: vblendvpd %xmm3, %xmm5, %xmm6, %xmm3 +; AVX1-NEXT: vblendvpd %xmm7, %xmm1, %xmm6, %xmm1 ; AVX1-NEXT: vpackusdw %xmm3, %xmm1, %xmm1 -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vblendvpd %xmm12, %xmm4, %xmm6, %xmm3 +; AVX1-NEXT: vblendvpd %xmm11, %xmm0, %xmm6, %xmm0 ; AVX1-NEXT: vpackusdw %xmm3, %xmm0, %xmm0 ; AVX1-NEXT: vpackusdw %xmm1, %xmm0, %xmm0 -; AVX1-NEXT: vpextrb $0, %xmm5, %eax +; AVX1-NEXT: vpextrb $0, %xmm10, %eax ; AVX1-NEXT: testb $1, %al ; AVX1-NEXT: je .LBB2_2 ; AVX1-NEXT: # %bb.1: # %cond.store @@ -1492,17 +1486,16 @@ define void @truncstore_v4i64_v4i32(<4 x i64> %x, <4 x i32>* %p, <4 x i32> %mask ; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1 ; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2 ; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1 -; AVX1-NEXT: vmovapd {{.*#+}} ymm2 = [4294967295,4294967295,4294967295,4294967295] -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [9223372036854775808,9223372036854775808] -; AVX1-NEXT: vpxor %xmm4, %xmm3, %xmm3 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372041149743103,9223372041149743103] -; AVX1-NEXT: vpcmpgtq %xmm3, %xmm5, %xmm3 -; AVX1-NEXT: vpxor %xmm4, %xmm0, %xmm4 -; AVX1-NEXT: vpcmpgtq %xmm4, %xmm5, %xmm4 -; AVX1-NEXT: vinsertf128 $1, %xmm3, %ymm4, %ymm3 -; AVX1-NEXT: vblendvpd %ymm3, %ymm0, %ymm2, %ymm0 -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm2 = [9223372036854775808,9223372036854775808] +; AVX1-NEXT: vpxor %xmm2, %xmm0, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [9223372041149743103,9223372041149743103] +; AVX1-NEXT: vpcmpgtq %xmm3, %xmm4, %xmm3 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5 +; AVX1-NEXT: vpxor %xmm2, %xmm5, %xmm2 +; AVX1-NEXT: vpcmpgtq %xmm2, %xmm4, %xmm2 +; AVX1-NEXT: vmovapd {{.*#+}} xmm4 = [4294967295,4294967295] +; AVX1-NEXT: vblendvpd %xmm2, %xmm5, %xmm4, %xmm2 +; AVX1-NEXT: vblendvpd %xmm3, %xmm0, %xmm4, %xmm0 ; AVX1-NEXT: vshufps {{.*#+}} xmm0 = xmm0[0,2],xmm2[0,2] ; AVX1-NEXT: vmaskmovps %xmm0, %xmm1, (%rdi) ; AVX1-NEXT: vzeroupper @@ -1698,17 +1691,16 @@ define void @truncstore_v4i64_v4i16(<4 x i64> %x, <4 x i16>* %p, <4 x i32> %mask ; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2 ; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3 ; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2 -; AVX1-NEXT: vmovapd {{.*#+}} ymm3 = [65535,65535,65535,65535] -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm4 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372036854775808,9223372036854775808] -; AVX1-NEXT: vpxor %xmm5, %xmm4, %xmm4 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [9223372036854841343,9223372036854841343] -; AVX1-NEXT: vpcmpgtq %xmm4, %xmm6, %xmm4 -; AVX1-NEXT: vpxor %xmm5, %xmm0, %xmm5 -; AVX1-NEXT: vpcmpgtq %xmm5, %xmm6, %xmm5 -; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm5, %ymm4 -; AVX1-NEXT: vblendvpd %ymm4, %ymm0, %ymm3, %ymm0 -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [9223372036854775808,9223372036854775808] +; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm4 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372036854841343,9223372036854841343] +; AVX1-NEXT: vpcmpgtq %xmm4, %xmm5, %xmm4 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm6 +; AVX1-NEXT: vpxor %xmm3, %xmm6, %xmm3 +; AVX1-NEXT: vpcmpgtq %xmm3, %xmm5, %xmm3 +; AVX1-NEXT: vmovapd {{.*#+}} xmm5 = [65535,65535] +; AVX1-NEXT: vblendvpd %xmm3, %xmm6, %xmm5, %xmm3 +; AVX1-NEXT: vblendvpd %xmm4, %xmm0, %xmm5, %xmm0 ; AVX1-NEXT: vpackusdw %xmm3, %xmm0, %xmm0 ; AVX1-NEXT: vpextrb $0, %xmm2, %eax ; AVX1-NEXT: testb $1, %al @@ -1986,17 +1978,16 @@ define void @truncstore_v4i64_v4i8(<4 x i64> %x, <4 x i8>* %p, <4 x i32> %mask) ; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2 ; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3 ; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2 -; AVX1-NEXT: vmovapd {{.*#+}} ymm3 = [255,255,255,255] -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm4 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372036854775808,9223372036854775808] -; AVX1-NEXT: vpxor %xmm5, %xmm4, %xmm4 -; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [9223372036854776063,9223372036854776063] -; AVX1-NEXT: vpcmpgtq %xmm4, %xmm6, %xmm4 -; AVX1-NEXT: vpxor %xmm5, %xmm0, %xmm5 -; AVX1-NEXT: vpcmpgtq %xmm5, %xmm6, %xmm5 -; AVX1-NEXT: vinsertf128 $1, %xmm4, %ymm5, %ymm4 -; AVX1-NEXT: vblendvpd %ymm4, %ymm0, %ymm3, %ymm0 -; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [9223372036854775808,9223372036854775808] +; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm4 +; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372036854776063,9223372036854776063] +; AVX1-NEXT: vpcmpgtq %xmm4, %xmm5, %xmm4 +; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm6 +; AVX1-NEXT: vpxor %xmm3, %xmm6, %xmm3 +; AVX1-NEXT: vpcmpgtq %xmm3, %xmm5, %xmm3 +; AVX1-NEXT: vmovapd {{.*#+}} xmm5 = [255,255] +; AVX1-NEXT: vblendvpd %xmm3, %xmm6, %xmm5, %xmm3 +; AVX1-NEXT: vblendvpd %xmm4, %xmm0, %xmm5, %xmm0 ; AVX1-NEXT: vpackusdw %xmm3, %xmm0, %xmm0 ; AVX1-NEXT: vpextrb $0, %xmm2, %eax ; AVX1-NEXT: testb $1, %al |
