[llvm] r367489 - [ScalarizeMaskedMemIntrin] Bitcast the mask to the scalar domain and use scalar bit tests for the branches.
Craig Topper via llvm-commits
llvm-commits at lists.llvm.org
Wed Jul 31 15:58:16 PDT 2019
Modified: llvm/trunk/test/CodeGen/X86/masked_store_trunc_usat.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/masked_store_trunc_usat.ll?rev=367489&r1=367488&r2=367489&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/masked_store_trunc_usat.ll (original)
+++ llvm/trunk/test/CodeGen/X86/masked_store_trunc_usat.ll Wed Jul 31 15:58:15 2019
@@ -11,233 +11,220 @@
define void @truncstore_v8i64_v8i32(<8 x i64> %x, <8 x i32>* %p, <8 x i32> %mask) {
; SSE2-LABEL: truncstore_v8i64_v8i32:
; SSE2: # %bb.0:
-; SSE2-NEXT: pxor %xmm6, %xmm6
-; SSE2-NEXT: pcmpeqd %xmm4, %xmm6
-; SSE2-NEXT: pcmpeqd %xmm12, %xmm12
-; SSE2-NEXT: pxor %xmm6, %xmm12
-; SSE2-NEXT: movdqa %xmm12, %xmm9
-; SSE2-NEXT: packssdw %xmm0, %xmm9
-; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [4294967295,4294967295]
-; SSE2-NEXT: movdqa {{.*#+}} xmm13 = [9223372039002259456,9223372039002259456]
-; SSE2-NEXT: movdqa %xmm1, %xmm7
-; SSE2-NEXT: pxor %xmm13, %xmm7
-; SSE2-NEXT: movdqa {{.*#+}} xmm10 = [9223372039002259455,9223372039002259455]
-; SSE2-NEXT: movdqa %xmm10, %xmm6
-; SSE2-NEXT: pcmpgtd %xmm7, %xmm6
-; SSE2-NEXT: pshufd {{.*#+}} xmm11 = xmm6[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm10, %xmm7
-; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3]
-; SSE2-NEXT: pand %xmm11, %xmm7
+; SSE2-NEXT: pxor %xmm8, %xmm8
+; SSE2-NEXT: movdqa {{.*#+}} xmm9 = [4294967295,4294967295]
+; SSE2-NEXT: movdqa {{.*#+}} xmm12 = [9223372039002259456,9223372039002259456]
+; SSE2-NEXT: movdqa %xmm1, %xmm6
+; SSE2-NEXT: pxor %xmm12, %xmm6
+; SSE2-NEXT: movdqa {{.*#+}} xmm11 = [9223372039002259455,9223372039002259455]
+; SSE2-NEXT: movdqa %xmm11, %xmm7
+; SSE2-NEXT: pcmpgtd %xmm6, %xmm7
+; SSE2-NEXT: pshufd {{.*#+}} xmm10 = xmm7[0,0,2,2]
+; SSE2-NEXT: pcmpeqd %xmm11, %xmm6
; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
-; SSE2-NEXT: por %xmm7, %xmm6
-; SSE2-NEXT: pand %xmm6, %xmm1
-; SSE2-NEXT: pandn %xmm8, %xmm6
-; SSE2-NEXT: por %xmm1, %xmm6
+; SSE2-NEXT: pand %xmm10, %xmm6
+; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3]
+; SSE2-NEXT: por %xmm6, %xmm7
+; SSE2-NEXT: pand %xmm7, %xmm1
+; SSE2-NEXT: pandn %xmm9, %xmm7
+; SSE2-NEXT: por %xmm1, %xmm7
; SSE2-NEXT: movdqa %xmm0, %xmm1
-; SSE2-NEXT: pxor %xmm13, %xmm1
-; SSE2-NEXT: movdqa %xmm10, %xmm7
-; SSE2-NEXT: pcmpgtd %xmm1, %xmm7
-; SSE2-NEXT: pshufd {{.*#+}} xmm11 = xmm7[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm10, %xmm1
+; SSE2-NEXT: pxor %xmm12, %xmm1
+; SSE2-NEXT: movdqa %xmm11, %xmm6
+; SSE2-NEXT: pcmpgtd %xmm1, %xmm6
+; SSE2-NEXT: pshufd {{.*#+}} xmm10 = xmm6[0,0,2,2]
+; SSE2-NEXT: pcmpeqd %xmm11, %xmm1
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; SSE2-NEXT: pand %xmm11, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm14 = xmm7[1,1,3,3]
-; SSE2-NEXT: por %xmm1, %xmm14
-; SSE2-NEXT: pand %xmm14, %xmm0
-; SSE2-NEXT: pandn %xmm8, %xmm14
-; SSE2-NEXT: por %xmm0, %xmm14
-; SSE2-NEXT: shufps {{.*#+}} xmm14 = xmm14[0,2],xmm6[0,2]
+; SSE2-NEXT: pand %xmm10, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm13 = xmm6[1,1,3,3]
+; SSE2-NEXT: por %xmm1, %xmm13
+; SSE2-NEXT: pand %xmm13, %xmm0
+; SSE2-NEXT: pandn %xmm9, %xmm13
+; SSE2-NEXT: por %xmm0, %xmm13
+; SSE2-NEXT: shufps {{.*#+}} xmm13 = xmm13[0,2],xmm7[0,2]
; SSE2-NEXT: movdqa %xmm3, %xmm0
-; SSE2-NEXT: pxor %xmm13, %xmm0
-; SSE2-NEXT: movdqa %xmm10, %xmm1
+; SSE2-NEXT: pxor %xmm12, %xmm0
+; SSE2-NEXT: movdqa %xmm11, %xmm1
; SSE2-NEXT: pcmpgtd %xmm0, %xmm1
; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm1[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm10, %xmm0
+; SSE2-NEXT: pcmpeqd %xmm11, %xmm0
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; SSE2-NEXT: pand %xmm6, %xmm0
-; SSE2-NEXT: pshufd {{.*#+}} xmm11 = xmm1[1,1,3,3]
-; SSE2-NEXT: pxor %xmm2, %xmm13
-; SSE2-NEXT: movdqa %xmm10, %xmm1
-; SSE2-NEXT: pcmpgtd %xmm13, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm10 = xmm1[1,1,3,3]
+; SSE2-NEXT: pxor %xmm2, %xmm12
+; SSE2-NEXT: movdqa %xmm11, %xmm1
+; SSE2-NEXT: pcmpgtd %xmm12, %xmm1
; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm1[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm10, %xmm13
-; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm13[1,1,3,3]
+; SSE2-NEXT: pcmpeqd %xmm11, %xmm12
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm12[1,1,3,3]
; SSE2-NEXT: pand %xmm7, %xmm6
-; SSE2-NEXT: pshufd {{.*#+}} xmm10 = xmm1[1,1,3,3]
-; SSE2-NEXT: movd %xmm9, %eax
+; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm1[1,1,3,3]
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm5
+; SSE2-NEXT: pcmpeqd %xmm1, %xmm1
+; SSE2-NEXT: pxor %xmm1, %xmm5
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm4
+; SSE2-NEXT: pxor %xmm1, %xmm4
+; SSE2-NEXT: packssdw %xmm5, %xmm4
+; SSE2-NEXT: packsswb %xmm0, %xmm4
+; SSE2-NEXT: pmovmskb %xmm4, %eax
; SSE2-NEXT: testb $1, %al
; SSE2-NEXT: je .LBB0_2
; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movss %xmm14, (%rdi)
+; SSE2-NEXT: movss %xmm13, (%rdi)
; SSE2-NEXT: .LBB0_2: # %else
-; SSE2-NEXT: por %xmm11, %xmm0
-; SSE2-NEXT: por %xmm10, %xmm6
-; SSE2-NEXT: psrlq $16, %xmm12
-; SSE2-NEXT: movd %xmm12, %eax
-; SSE2-NEXT: shrl $16, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: por %xmm10, %xmm0
+; SSE2-NEXT: por %xmm7, %xmm6
+; SSE2-NEXT: testb $2, %al
; SSE2-NEXT: je .LBB0_4
; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm14[1,1,2,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm13[1,1,2,3]
; SSE2-NEXT: movd %xmm1, 4(%rdi)
; SSE2-NEXT: .LBB0_4: # %else2
; SSE2-NEXT: pand %xmm0, %xmm3
-; SSE2-NEXT: pandn %xmm8, %xmm0
+; SSE2-NEXT: pandn %xmm9, %xmm0
; SSE2-NEXT: pand %xmm6, %xmm2
-; SSE2-NEXT: pandn %xmm8, %xmm6
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm4
-; SSE2-NEXT: pcmpeqd %xmm7, %xmm7
-; SSE2-NEXT: pxor %xmm4, %xmm7
-; SSE2-NEXT: pextrw $4, %xmm7, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: pandn %xmm9, %xmm6
+; SSE2-NEXT: testb $4, %al
; SSE2-NEXT: je .LBB0_6
; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm14[2,3,0,1]
+; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm13[2,3,0,1]
; SSE2-NEXT: movd %xmm1, 8(%rdi)
; SSE2-NEXT: .LBB0_6: # %else4
; SSE2-NEXT: por %xmm0, %xmm3
; SSE2-NEXT: por %xmm6, %xmm2
-; SSE2-NEXT: pextrw $6, %xmm7, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $8, %al
; SSE2-NEXT: je .LBB0_8
; SSE2-NEXT: # %bb.7: # %cond.store5
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm14[3,1,2,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm13[3,1,2,3]
; SSE2-NEXT: movd %xmm0, 12(%rdi)
; SSE2-NEXT: .LBB0_8: # %else6
; SSE2-NEXT: shufps {{.*#+}} xmm2 = xmm2[0,2],xmm3[0,2]
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm5, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm1, %xmm0
-; SSE2-NEXT: pextrw $0, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB0_10
-; SSE2-NEXT: # %bb.9: # %cond.store7
+; SSE2-NEXT: testb $16, %al
+; SSE2-NEXT: jne .LBB0_9
+; SSE2-NEXT: # %bb.10: # %else8
+; SSE2-NEXT: testb $32, %al
+; SSE2-NEXT: jne .LBB0_11
+; SSE2-NEXT: .LBB0_12: # %else10
+; SSE2-NEXT: testb $64, %al
+; SSE2-NEXT: jne .LBB0_13
+; SSE2-NEXT: .LBB0_14: # %else12
+; SSE2-NEXT: testb $-128, %al
+; SSE2-NEXT: jne .LBB0_15
+; SSE2-NEXT: .LBB0_16: # %else14
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB0_9: # %cond.store7
; SSE2-NEXT: movss %xmm2, 16(%rdi)
-; SSE2-NEXT: .LBB0_10: # %else8
-; SSE2-NEXT: pextrw $2, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $32, %al
; SSE2-NEXT: je .LBB0_12
-; SSE2-NEXT: # %bb.11: # %cond.store9
+; SSE2-NEXT: .LBB0_11: # %cond.store9
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[1,1,2,3]
; SSE2-NEXT: movd %xmm0, 20(%rdi)
-; SSE2-NEXT: .LBB0_12: # %else10
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm5
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm5, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $64, %al
; SSE2-NEXT: je .LBB0_14
-; SSE2-NEXT: # %bb.13: # %cond.store11
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[2,3,0,1]
-; SSE2-NEXT: movd %xmm1, 24(%rdi)
-; SSE2-NEXT: .LBB0_14: # %else12
-; SSE2-NEXT: pextrw $6, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: .LBB0_13: # %cond.store11
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
+; SSE2-NEXT: movd %xmm0, 24(%rdi)
+; SSE2-NEXT: testb $-128, %al
; SSE2-NEXT: je .LBB0_16
-; SSE2-NEXT: # %bb.15: # %cond.store13
+; SSE2-NEXT: .LBB0_15: # %cond.store13
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[3,1,2,3]
; SSE2-NEXT: movd %xmm0, 28(%rdi)
-; SSE2-NEXT: .LBB0_16: # %else14
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v8i64_v8i32:
; SSE4: # %bb.0:
; SSE4-NEXT: movdqa %xmm0, %xmm8
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm4, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm10, %xmm10
-; SSE4-NEXT: pxor %xmm0, %xmm10
-; SSE4-NEXT: movapd {{.*#+}} xmm9 = [4294967295,4294967295]
+; SSE4-NEXT: pxor %xmm9, %xmm9
+; SSE4-NEXT: movapd {{.*#+}} xmm10 = [4294967295,4294967295]
; SSE4-NEXT: movdqa {{.*#+}} xmm11 = [9223372036854775808,9223372036854775808]
; SSE4-NEXT: movdqa %xmm1, %xmm6
; SSE4-NEXT: pxor %xmm11, %xmm6
; SSE4-NEXT: movdqa {{.*#+}} xmm7 = [9223372041149743103,9223372041149743103]
; SSE4-NEXT: movdqa %xmm7, %xmm0
; SSE4-NEXT: pcmpgtq %xmm6, %xmm0
-; SSE4-NEXT: movapd %xmm9, %xmm6
+; SSE4-NEXT: movapd %xmm10, %xmm6
; SSE4-NEXT: blendvpd %xmm0, %xmm1, %xmm6
; SSE4-NEXT: movdqa %xmm8, %xmm1
; SSE4-NEXT: pxor %xmm11, %xmm1
; SSE4-NEXT: movdqa %xmm7, %xmm0
; SSE4-NEXT: pcmpgtq %xmm1, %xmm0
-; SSE4-NEXT: movapd %xmm9, %xmm1
+; SSE4-NEXT: movapd %xmm10, %xmm1
; SSE4-NEXT: blendvpd %xmm0, %xmm8, %xmm1
; SSE4-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,2],xmm6[0,2]
; SSE4-NEXT: movdqa %xmm3, %xmm6
; SSE4-NEXT: pxor %xmm11, %xmm6
; SSE4-NEXT: movdqa %xmm7, %xmm0
; SSE4-NEXT: pcmpgtq %xmm6, %xmm0
-; SSE4-NEXT: movapd %xmm9, %xmm8
+; SSE4-NEXT: movapd %xmm10, %xmm8
; SSE4-NEXT: blendvpd %xmm0, %xmm3, %xmm8
; SSE4-NEXT: pxor %xmm2, %xmm11
; SSE4-NEXT: pcmpgtq %xmm11, %xmm7
; SSE4-NEXT: movdqa %xmm7, %xmm0
-; SSE4-NEXT: blendvpd %xmm0, %xmm2, %xmm9
-; SSE4-NEXT: pextrb $0, %xmm10, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB0_2
-; SSE4-NEXT: # %bb.1: # %cond.store
-; SSE4-NEXT: movss %xmm1, (%rdi)
-; SSE4-NEXT: .LBB0_2: # %else
-; SSE4-NEXT: pextrb $4, %xmm10, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB0_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
-; SSE4-NEXT: extractps $1, %xmm1, 4(%rdi)
-; SSE4-NEXT: .LBB0_4: # %else2
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm4
+; SSE4-NEXT: blendvpd %xmm0, %xmm2, %xmm10
+; SSE4-NEXT: pcmpeqd %xmm9, %xmm5
; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm4, %xmm0
-; SSE4-NEXT: pextrb $8, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB0_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
-; SSE4-NEXT: extractps $2, %xmm1, 8(%rdi)
+; SSE4-NEXT: pxor %xmm0, %xmm5
+; SSE4-NEXT: pcmpeqd %xmm9, %xmm4
+; SSE4-NEXT: pxor %xmm0, %xmm4
+; SSE4-NEXT: packssdw %xmm5, %xmm4
+; SSE4-NEXT: packsswb %xmm0, %xmm4
+; SSE4-NEXT: pmovmskb %xmm4, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB0_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB0_3
+; SSE4-NEXT: .LBB0_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB0_5
; SSE4-NEXT: .LBB0_6: # %else4
-; SSE4-NEXT: pextrb $12, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $8, %al
; SSE4-NEXT: je .LBB0_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
+; SSE4-NEXT: .LBB0_7: # %cond.store5
; SSE4-NEXT: extractps $3, %xmm1, 12(%rdi)
; SSE4-NEXT: .LBB0_8: # %else6
-; SSE4-NEXT: shufps {{.*#+}} xmm9 = xmm9[0,2],xmm8[0,2]
-; SSE4-NEXT: xorps %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm5, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm1, %xmm0
-; SSE4-NEXT: pextrb $0, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB0_10
-; SSE4-NEXT: # %bb.9: # %cond.store7
-; SSE4-NEXT: movss %xmm9, 16(%rdi)
-; SSE4-NEXT: .LBB0_10: # %else8
-; SSE4-NEXT: pextrb $4, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB0_12
-; SSE4-NEXT: # %bb.11: # %cond.store9
-; SSE4-NEXT: extractps $1, %xmm9, 20(%rdi)
+; SSE4-NEXT: shufps {{.*#+}} xmm10 = xmm10[0,2],xmm8[0,2]
+; SSE4-NEXT: testb $16, %al
+; SSE4-NEXT: jne .LBB0_9
+; SSE4-NEXT: # %bb.10: # %else8
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: jne .LBB0_11
; SSE4-NEXT: .LBB0_12: # %else10
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm5
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm5, %xmm0
-; SSE4-NEXT: pextrb $8, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB0_14
-; SSE4-NEXT: # %bb.13: # %cond.store11
-; SSE4-NEXT: extractps $2, %xmm9, 24(%rdi)
+; SSE4-NEXT: testb $64, %al
+; SSE4-NEXT: jne .LBB0_13
; SSE4-NEXT: .LBB0_14: # %else12
-; SSE4-NEXT: pextrb $12, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB0_16
-; SSE4-NEXT: # %bb.15: # %cond.store13
-; SSE4-NEXT: extractps $3, %xmm9, 28(%rdi)
+; SSE4-NEXT: testb $-128, %al
+; SSE4-NEXT: jne .LBB0_15
; SSE4-NEXT: .LBB0_16: # %else14
; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB0_1: # %cond.store
+; SSE4-NEXT: movss %xmm1, (%rdi)
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: je .LBB0_4
+; SSE4-NEXT: .LBB0_3: # %cond.store1
+; SSE4-NEXT: extractps $1, %xmm1, 4(%rdi)
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: je .LBB0_6
+; SSE4-NEXT: .LBB0_5: # %cond.store3
+; SSE4-NEXT: extractps $2, %xmm1, 8(%rdi)
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB0_7
+; SSE4-NEXT: jmp .LBB0_8
+; SSE4-NEXT: .LBB0_9: # %cond.store7
+; SSE4-NEXT: movss %xmm10, 16(%rdi)
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: je .LBB0_12
+; SSE4-NEXT: .LBB0_11: # %cond.store9
+; SSE4-NEXT: extractps $1, %xmm10, 20(%rdi)
+; SSE4-NEXT: testb $64, %al
+; SSE4-NEXT: je .LBB0_14
+; SSE4-NEXT: .LBB0_13: # %cond.store11
+; SSE4-NEXT: extractps $2, %xmm10, 24(%rdi)
+; SSE4-NEXT: testb $-128, %al
+; SSE4-NEXT: je .LBB0_16
+; SSE4-NEXT: .LBB0_15: # %cond.store13
+; SSE4-NEXT: extractps $3, %xmm10, 28(%rdi)
+; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v8i64_v8i32:
; AVX1: # %bb.0:
@@ -333,65 +320,60 @@ define void @truncstore_v8i64_v8i32(<8 x
define void @truncstore_v8i64_v8i16(<8 x i64> %x, <8 x i16>* %p, <8 x i32> %mask) {
; SSE2-LABEL: truncstore_v8i64_v8i16:
; SSE2: # %bb.0:
-; SSE2-NEXT: pxor %xmm6, %xmm6
-; SSE2-NEXT: pcmpeqd %xmm4, %xmm6
-; SSE2-NEXT: pcmpeqd %xmm10, %xmm10
-; SSE2-NEXT: pxor %xmm6, %xmm10
-; SSE2-NEXT: movdqa %xmm10, %xmm9
-; SSE2-NEXT: packssdw %xmm0, %xmm9
-; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [65535,65535]
-; SSE2-NEXT: movdqa {{.*#+}} xmm12 = [9223372039002259456,9223372039002259456]
-; SSE2-NEXT: movdqa %xmm2, %xmm7
-; SSE2-NEXT: pxor %xmm12, %xmm7
-; SSE2-NEXT: movdqa {{.*#+}} xmm11 = [9223372039002324991,9223372039002324991]
-; SSE2-NEXT: movdqa %xmm11, %xmm6
-; SSE2-NEXT: pcmpgtd %xmm7, %xmm6
-; SSE2-NEXT: pshufd {{.*#+}} xmm13 = xmm6[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm11, %xmm7
-; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3]
-; SSE2-NEXT: pand %xmm13, %xmm7
-; SSE2-NEXT: pshufd {{.*#+}} xmm14 = xmm6[1,1,3,3]
-; SSE2-NEXT: por %xmm7, %xmm14
-; SSE2-NEXT: pand %xmm14, %xmm2
-; SSE2-NEXT: pandn %xmm8, %xmm14
-; SSE2-NEXT: por %xmm2, %xmm14
+; SSE2-NEXT: pxor %xmm8, %xmm8
+; SSE2-NEXT: movdqa {{.*#+}} xmm9 = [65535,65535]
+; SSE2-NEXT: movdqa {{.*#+}} xmm11 = [9223372039002259456,9223372039002259456]
+; SSE2-NEXT: movdqa %xmm2, %xmm6
+; SSE2-NEXT: pxor %xmm11, %xmm6
+; SSE2-NEXT: movdqa {{.*#+}} xmm10 = [9223372039002324991,9223372039002324991]
+; SSE2-NEXT: movdqa %xmm10, %xmm7
+; SSE2-NEXT: pcmpgtd %xmm6, %xmm7
+; SSE2-NEXT: pshufd {{.*#+}} xmm12 = xmm7[0,0,2,2]
+; SSE2-NEXT: pcmpeqd %xmm10, %xmm6
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
+; SSE2-NEXT: pand %xmm12, %xmm6
+; SSE2-NEXT: pshufd {{.*#+}} xmm13 = xmm7[1,1,3,3]
+; SSE2-NEXT: por %xmm6, %xmm13
+; SSE2-NEXT: pand %xmm13, %xmm2
+; SSE2-NEXT: pandn %xmm9, %xmm13
+; SSE2-NEXT: por %xmm2, %xmm13
; SSE2-NEXT: movdqa %xmm3, %xmm2
-; SSE2-NEXT: pxor %xmm12, %xmm2
-; SSE2-NEXT: movdqa %xmm11, %xmm7
-; SSE2-NEXT: pcmpgtd %xmm2, %xmm7
-; SSE2-NEXT: pshufd {{.*#+}} xmm13 = xmm7[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm11, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm2[1,1,3,3]
-; SSE2-NEXT: pand %xmm13, %xmm6
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm7[1,1,3,3]
-; SSE2-NEXT: por %xmm6, %xmm2
+; SSE2-NEXT: pxor %xmm11, %xmm2
+; SSE2-NEXT: movdqa %xmm10, %xmm6
+; SSE2-NEXT: pcmpgtd %xmm2, %xmm6
+; SSE2-NEXT: pshufd {{.*#+}} xmm12 = xmm6[0,0,2,2]
+; SSE2-NEXT: pcmpeqd %xmm10, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm2[1,1,3,3]
+; SSE2-NEXT: pand %xmm12, %xmm7
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm6[1,1,3,3]
+; SSE2-NEXT: por %xmm7, %xmm2
; SSE2-NEXT: pand %xmm2, %xmm3
-; SSE2-NEXT: pandn %xmm8, %xmm2
+; SSE2-NEXT: pandn %xmm9, %xmm2
; SSE2-NEXT: por %xmm3, %xmm2
; SSE2-NEXT: movdqa %xmm0, %xmm3
-; SSE2-NEXT: pxor %xmm12, %xmm3
-; SSE2-NEXT: movdqa %xmm11, %xmm6
+; SSE2-NEXT: pxor %xmm11, %xmm3
+; SSE2-NEXT: movdqa %xmm10, %xmm6
; SSE2-NEXT: pcmpgtd %xmm3, %xmm6
; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm6[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm11, %xmm3
+; SSE2-NEXT: pcmpeqd %xmm10, %xmm3
; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm3[1,1,3,3]
; SSE2-NEXT: pand %xmm7, %xmm3
; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
; SSE2-NEXT: por %xmm3, %xmm6
; SSE2-NEXT: pand %xmm6, %xmm0
-; SSE2-NEXT: pandn %xmm8, %xmm6
+; SSE2-NEXT: pandn %xmm9, %xmm6
; SSE2-NEXT: por %xmm0, %xmm6
-; SSE2-NEXT: pxor %xmm1, %xmm12
-; SSE2-NEXT: movdqa %xmm11, %xmm0
-; SSE2-NEXT: pcmpgtd %xmm12, %xmm0
+; SSE2-NEXT: pxor %xmm1, %xmm11
+; SSE2-NEXT: movdqa %xmm10, %xmm0
+; SSE2-NEXT: pcmpgtd %xmm11, %xmm0
; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm11, %xmm12
-; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm12[1,1,3,3]
+; SSE2-NEXT: pcmpeqd %xmm10, %xmm11
+; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm11[1,1,3,3]
; SSE2-NEXT: pand %xmm3, %xmm7
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
; SSE2-NEXT: por %xmm7, %xmm0
; SSE2-NEXT: pand %xmm0, %xmm1
-; SSE2-NEXT: pandn %xmm8, %xmm0
+; SSE2-NEXT: pandn %xmm9, %xmm0
; SSE2-NEXT: por %xmm1, %xmm0
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[0,2,2,3]
; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,2,2,3,4,5,6,7]
@@ -400,89 +382,87 @@ define void @truncstore_v8i64_v8i16(<8 x
; SSE2-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[0,2,2,3]
; SSE2-NEXT: pshuflw {{.*#+}} xmm2 = xmm0[0,1,0,2,4,5,6,7]
-; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm14[0,2,2,3]
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm13[0,2,2,3]
; SSE2-NEXT: pshuflw {{.*#+}} xmm0 = xmm0[0,1,0,2,4,5,6,7]
; SSE2-NEXT: punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
; SSE2-NEXT: movsd {{.*#+}} xmm0 = xmm1[0],xmm0[1]
-; SSE2-NEXT: movd %xmm9, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB1_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movd %xmm0, %eax
-; SSE2-NEXT: movw %ax, (%rdi)
-; SSE2-NEXT: .LBB1_2: # %else
-; SSE2-NEXT: psrlq $16, %xmm10
-; SSE2-NEXT: movd %xmm10, %eax
-; SSE2-NEXT: shrl $16, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB1_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: pextrw $1, %xmm0, %eax
-; SSE2-NEXT: movw %ax, 2(%rdi)
-; SSE2-NEXT: .LBB1_4: # %else2
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm4
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm5
; SSE2-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE2-NEXT: pxor %xmm4, %xmm1
-; SSE2-NEXT: pextrw $4, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB1_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: pextrw $2, %xmm0, %eax
-; SSE2-NEXT: movw %ax, 4(%rdi)
+; SSE2-NEXT: pxor %xmm1, %xmm5
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm4
+; SSE2-NEXT: pxor %xmm1, %xmm4
+; SSE2-NEXT: packssdw %xmm5, %xmm4
+; SSE2-NEXT: packsswb %xmm0, %xmm4
+; SSE2-NEXT: pmovmskb %xmm4, %eax
+; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: jne .LBB1_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB1_3
+; SSE2-NEXT: .LBB1_4: # %else2
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB1_5
; SSE2-NEXT: .LBB1_6: # %else4
-; SSE2-NEXT: pextrw $6, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB1_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
-; SSE2-NEXT: pextrw $3, %xmm0, %eax
-; SSE2-NEXT: movw %ax, 6(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB1_7
; SSE2-NEXT: .LBB1_8: # %else6
-; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: pcmpeqd %xmm5, %xmm2
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE2-NEXT: pxor %xmm2, %xmm1
-; SSE2-NEXT: pextrw $0, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB1_10
-; SSE2-NEXT: # %bb.9: # %cond.store7
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: movw %ax, 8(%rdi)
+; SSE2-NEXT: testb $16, %al
+; SSE2-NEXT: jne .LBB1_9
; SSE2-NEXT: .LBB1_10: # %else8
-; SSE2-NEXT: pextrw $2, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB1_12
-; SSE2-NEXT: # %bb.11: # %cond.store9
-; SSE2-NEXT: pextrw $5, %xmm0, %eax
-; SSE2-NEXT: movw %ax, 10(%rdi)
+; SSE2-NEXT: testb $32, %al
+; SSE2-NEXT: jne .LBB1_11
; SSE2-NEXT: .LBB1_12: # %else10
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm5
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE2-NEXT: pxor %xmm5, %xmm1
-; SSE2-NEXT: pextrw $4, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB1_14
-; SSE2-NEXT: # %bb.13: # %cond.store11
-; SSE2-NEXT: pextrw $6, %xmm0, %eax
-; SSE2-NEXT: movw %ax, 12(%rdi)
+; SSE2-NEXT: testb $64, %al
+; SSE2-NEXT: jne .LBB1_13
; SSE2-NEXT: .LBB1_14: # %else12
-; SSE2-NEXT: pextrw $6, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $-128, %al
+; SSE2-NEXT: jne .LBB1_15
+; SSE2-NEXT: .LBB1_16: # %else14
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB1_1: # %cond.store
+; SSE2-NEXT: movd %xmm0, %ecx
+; SSE2-NEXT: movw %cx, (%rdi)
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: je .LBB1_4
+; SSE2-NEXT: .LBB1_3: # %cond.store1
+; SSE2-NEXT: pextrw $1, %xmm0, %ecx
+; SSE2-NEXT: movw %cx, 2(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: je .LBB1_6
+; SSE2-NEXT: .LBB1_5: # %cond.store3
+; SSE2-NEXT: pextrw $2, %xmm0, %ecx
+; SSE2-NEXT: movw %cx, 4(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: je .LBB1_8
+; SSE2-NEXT: .LBB1_7: # %cond.store5
+; SSE2-NEXT: pextrw $3, %xmm0, %ecx
+; SSE2-NEXT: movw %cx, 6(%rdi)
+; SSE2-NEXT: testb $16, %al
+; SSE2-NEXT: je .LBB1_10
+; SSE2-NEXT: .LBB1_9: # %cond.store7
+; SSE2-NEXT: pextrw $4, %xmm0, %ecx
+; SSE2-NEXT: movw %cx, 8(%rdi)
+; SSE2-NEXT: testb $32, %al
+; SSE2-NEXT: je .LBB1_12
+; SSE2-NEXT: .LBB1_11: # %cond.store9
+; SSE2-NEXT: pextrw $5, %xmm0, %ecx
+; SSE2-NEXT: movw %cx, 10(%rdi)
+; SSE2-NEXT: testb $64, %al
+; SSE2-NEXT: je .LBB1_14
+; SSE2-NEXT: .LBB1_13: # %cond.store11
+; SSE2-NEXT: pextrw $6, %xmm0, %ecx
+; SSE2-NEXT: movw %cx, 12(%rdi)
+; SSE2-NEXT: testb $-128, %al
; SSE2-NEXT: je .LBB1_16
-; SSE2-NEXT: # %bb.15: # %cond.store13
+; SSE2-NEXT: .LBB1_15: # %cond.store13
; SSE2-NEXT: pextrw $7, %xmm0, %eax
; SSE2-NEXT: movw %ax, 14(%rdi)
-; SSE2-NEXT: .LBB1_16: # %else14
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v8i64_v8i16:
; SSE4: # %bb.0:
-; SSE4-NEXT: movdqa %xmm0, %xmm8
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm4, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm9, %xmm9
-; SSE4-NEXT: pxor %xmm0, %xmm9
+; SSE4-NEXT: movdqa %xmm0, %xmm9
+; SSE4-NEXT: pxor %xmm8, %xmm8
; SSE4-NEXT: movapd {{.*#+}} xmm6 = [65535,65535]
; SSE4-NEXT: movdqa {{.*#+}} xmm10 = [9223372036854775808,9223372036854775808]
; SSE4-NEXT: movdqa %xmm1, %xmm7
@@ -492,12 +472,12 @@ define void @truncstore_v8i64_v8i16(<8 x
; SSE4-NEXT: pcmpgtq %xmm7, %xmm0
; SSE4-NEXT: movapd %xmm6, %xmm7
; SSE4-NEXT: blendvpd %xmm0, %xmm1, %xmm7
-; SSE4-NEXT: movdqa %xmm8, %xmm1
+; SSE4-NEXT: movdqa %xmm9, %xmm1
; SSE4-NEXT: pxor %xmm10, %xmm1
; SSE4-NEXT: movdqa %xmm11, %xmm0
; SSE4-NEXT: pcmpgtq %xmm1, %xmm0
; SSE4-NEXT: movapd %xmm6, %xmm1
-; SSE4-NEXT: blendvpd %xmm0, %xmm8, %xmm1
+; SSE4-NEXT: blendvpd %xmm0, %xmm9, %xmm1
; SSE4-NEXT: packusdw %xmm7, %xmm1
; SSE4-NEXT: movdqa %xmm3, %xmm7
; SSE4-NEXT: pxor %xmm10, %xmm7
@@ -511,246 +491,234 @@ define void @truncstore_v8i64_v8i16(<8 x
; SSE4-NEXT: blendvpd %xmm0, %xmm2, %xmm6
; SSE4-NEXT: packusdw %xmm7, %xmm6
; SSE4-NEXT: packusdw %xmm6, %xmm1
-; SSE4-NEXT: pextrb $0, %xmm9, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB1_2
-; SSE4-NEXT: # %bb.1: # %cond.store
+; SSE4-NEXT: pcmpeqd %xmm8, %xmm5
+; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
+; SSE4-NEXT: pxor %xmm0, %xmm5
+; SSE4-NEXT: pcmpeqd %xmm8, %xmm4
+; SSE4-NEXT: pxor %xmm0, %xmm4
+; SSE4-NEXT: packssdw %xmm5, %xmm4
+; SSE4-NEXT: packsswb %xmm0, %xmm4
+; SSE4-NEXT: pmovmskb %xmm4, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB1_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB1_3
+; SSE4-NEXT: .LBB1_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB1_5
+; SSE4-NEXT: .LBB1_6: # %else4
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB1_7
+; SSE4-NEXT: .LBB1_8: # %else6
+; SSE4-NEXT: testb $16, %al
+; SSE4-NEXT: jne .LBB1_9
+; SSE4-NEXT: .LBB1_10: # %else8
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: jne .LBB1_11
+; SSE4-NEXT: .LBB1_12: # %else10
+; SSE4-NEXT: testb $64, %al
+; SSE4-NEXT: jne .LBB1_13
+; SSE4-NEXT: .LBB1_14: # %else12
+; SSE4-NEXT: testb $-128, %al
+; SSE4-NEXT: jne .LBB1_15
+; SSE4-NEXT: .LBB1_16: # %else14
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB1_1: # %cond.store
; SSE4-NEXT: pextrw $0, %xmm1, (%rdi)
-; SSE4-NEXT: .LBB1_2: # %else
-; SSE4-NEXT: pextrb $4, %xmm9, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $2, %al
; SSE4-NEXT: je .LBB1_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
+; SSE4-NEXT: .LBB1_3: # %cond.store1
; SSE4-NEXT: pextrw $1, %xmm1, 2(%rdi)
-; SSE4-NEXT: .LBB1_4: # %else2
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm4
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm4, %xmm0
-; SSE4-NEXT: pextrb $8, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $4, %al
; SSE4-NEXT: je .LBB1_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
+; SSE4-NEXT: .LBB1_5: # %cond.store3
; SSE4-NEXT: pextrw $2, %xmm1, 4(%rdi)
-; SSE4-NEXT: .LBB1_6: # %else4
-; SSE4-NEXT: pextrb $12, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $8, %al
; SSE4-NEXT: je .LBB1_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
+; SSE4-NEXT: .LBB1_7: # %cond.store5
; SSE4-NEXT: pextrw $3, %xmm1, 6(%rdi)
-; SSE4-NEXT: .LBB1_8: # %else6
-; SSE4-NEXT: pxor %xmm2, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm5, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm2, %xmm0
-; SSE4-NEXT: pextrb $0, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $16, %al
; SSE4-NEXT: je .LBB1_10
-; SSE4-NEXT: # %bb.9: # %cond.store7
+; SSE4-NEXT: .LBB1_9: # %cond.store7
; SSE4-NEXT: pextrw $4, %xmm1, 8(%rdi)
-; SSE4-NEXT: .LBB1_10: # %else8
-; SSE4-NEXT: pextrb $4, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $32, %al
; SSE4-NEXT: je .LBB1_12
-; SSE4-NEXT: # %bb.11: # %cond.store9
+; SSE4-NEXT: .LBB1_11: # %cond.store9
; SSE4-NEXT: pextrw $5, %xmm1, 10(%rdi)
-; SSE4-NEXT: .LBB1_12: # %else10
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm5
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm5, %xmm0
-; SSE4-NEXT: pextrb $8, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $64, %al
; SSE4-NEXT: je .LBB1_14
-; SSE4-NEXT: # %bb.13: # %cond.store11
+; SSE4-NEXT: .LBB1_13: # %cond.store11
; SSE4-NEXT: pextrw $6, %xmm1, 12(%rdi)
-; SSE4-NEXT: .LBB1_14: # %else12
-; SSE4-NEXT: pextrb $12, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $-128, %al
; SSE4-NEXT: je .LBB1_16
-; SSE4-NEXT: # %bb.15: # %cond.store13
+; SSE4-NEXT: .LBB1_15: # %cond.store13
; SSE4-NEXT: pextrw $7, %xmm1, 14(%rdi)
-; SSE4-NEXT: .LBB1_16: # %else14
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v8i64_v8i16:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpxor %xmm8, %xmm8, %xmm8
-; AVX1-NEXT: vpcmpeqd %xmm8, %xmm2, %xmm5
-; AVX1-NEXT: vpcmpeqd %xmm9, %xmm9, %xmm9
-; 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: vmovdqa {{.*#+}} xmm3 = [9223372036854775808,9223372036854775808]
+; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm4
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372036854841343,9223372036854841343]
+; AVX1-NEXT: vpcmpgtq %xmm4, %xmm5, %xmm8
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm6
+; AVX1-NEXT: vpxor %xmm3, %xmm6, %xmm7
+; AVX1-NEXT: vpcmpgtq %xmm7, %xmm5, %xmm9
+; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm4
+; AVX1-NEXT: vpcmpgtq %xmm4, %xmm5, %xmm4
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm7
+; AVX1-NEXT: vpxor %xmm3, %xmm7, %xmm3
+; AVX1-NEXT: vpcmpgtq %xmm3, %xmm5, %xmm3
+; AVX1-NEXT: vmovapd {{.*#+}} xmm5 = [65535,65535]
+; AVX1-NEXT: vblendvpd %xmm3, %xmm7, %xmm5, %xmm3
+; AVX1-NEXT: vblendvpd %xmm4, %xmm1, %xmm5, %xmm1
; AVX1-NEXT: vpackusdw %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vblendvpd %xmm12, %xmm4, %xmm6, %xmm3
-; AVX1-NEXT: vblendvpd %xmm11, %xmm0, %xmm6, %xmm0
+; AVX1-NEXT: vblendvpd %xmm9, %xmm6, %xmm5, %xmm3
+; AVX1-NEXT: vblendvpd %xmm8, %xmm0, %xmm5, %xmm0
; AVX1-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
-; AVX1-NEXT: vpextrb $0, %xmm10, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB1_2
-; AVX1-NEXT: # %bb.1: # %cond.store
+; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm1
+; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT: vpcmpeqd %xmm3, %xmm1, %xmm1
+; AVX1-NEXT: vpcmpeqd %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-NEXT: vmovmskps %ymm1, %eax
+; AVX1-NEXT: notl %eax
+; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: jne .LBB1_1
+; AVX1-NEXT: # %bb.2: # %else
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: jne .LBB1_3
+; AVX1-NEXT: .LBB1_4: # %else2
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: jne .LBB1_5
+; AVX1-NEXT: .LBB1_6: # %else4
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: jne .LBB1_7
+; AVX1-NEXT: .LBB1_8: # %else6
+; AVX1-NEXT: testb $16, %al
+; AVX1-NEXT: jne .LBB1_9
+; AVX1-NEXT: .LBB1_10: # %else8
+; AVX1-NEXT: testb $32, %al
+; AVX1-NEXT: jne .LBB1_11
+; AVX1-NEXT: .LBB1_12: # %else10
+; AVX1-NEXT: testb $64, %al
+; AVX1-NEXT: jne .LBB1_13
+; AVX1-NEXT: .LBB1_14: # %else12
+; AVX1-NEXT: testb $-128, %al
+; AVX1-NEXT: jne .LBB1_15
+; AVX1-NEXT: .LBB1_16: # %else14
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+; AVX1-NEXT: .LBB1_1: # %cond.store
; AVX1-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX1-NEXT: .LBB1_2: # %else
-; AVX1-NEXT: vpcmpeqd %xmm8, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm9, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $4, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $2, %al
; AVX1-NEXT: je .LBB1_4
-; AVX1-NEXT: # %bb.3: # %cond.store1
+; AVX1-NEXT: .LBB1_3: # %cond.store1
; AVX1-NEXT: vpextrw $1, %xmm0, 2(%rdi)
-; AVX1-NEXT: .LBB1_4: # %else2
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $8, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $4, %al
; AVX1-NEXT: je .LBB1_6
-; AVX1-NEXT: # %bb.5: # %cond.store3
+; AVX1-NEXT: .LBB1_5: # %cond.store3
; AVX1-NEXT: vpextrw $2, %xmm0, 4(%rdi)
-; AVX1-NEXT: .LBB1_6: # %else4
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $12, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $8, %al
; AVX1-NEXT: je .LBB1_8
-; AVX1-NEXT: # %bb.7: # %cond.store5
+; AVX1-NEXT: .LBB1_7: # %cond.store5
; AVX1-NEXT: vpextrw $3, %xmm0, 6(%rdi)
-; AVX1-NEXT: .LBB1_8: # %else6
-; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $0, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $16, %al
; AVX1-NEXT: je .LBB1_10
-; AVX1-NEXT: # %bb.9: # %cond.store7
+; AVX1-NEXT: .LBB1_9: # %cond.store7
; AVX1-NEXT: vpextrw $4, %xmm0, 8(%rdi)
-; AVX1-NEXT: .LBB1_10: # %else8
-; AVX1-NEXT: vpextrb $4, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $32, %al
; AVX1-NEXT: je .LBB1_12
-; AVX1-NEXT: # %bb.11: # %cond.store9
+; AVX1-NEXT: .LBB1_11: # %cond.store9
; AVX1-NEXT: vpextrw $5, %xmm0, 10(%rdi)
-; AVX1-NEXT: .LBB1_12: # %else10
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $8, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $64, %al
; AVX1-NEXT: je .LBB1_14
-; AVX1-NEXT: # %bb.13: # %cond.store11
+; AVX1-NEXT: .LBB1_13: # %cond.store11
; AVX1-NEXT: vpextrw $6, %xmm0, 12(%rdi)
-; AVX1-NEXT: .LBB1_14: # %else12
-; AVX1-NEXT: vpextrb $12, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $-128, %al
; AVX1-NEXT: je .LBB1_16
-; AVX1-NEXT: # %bb.15: # %cond.store13
+; AVX1-NEXT: .LBB1_15: # %cond.store13
; AVX1-NEXT: vpextrw $7, %xmm0, 14(%rdi)
-; AVX1-NEXT: .LBB1_16: # %else14
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: truncstore_v8i64_v8i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm2, %xmm5
-; AVX2-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX2-NEXT: vpxor %xmm4, %xmm5, %xmm5
-; AVX2-NEXT: vbroadcastsd {{.*#+}} ymm6 = [65535,65535,65535,65535]
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm7 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
-; AVX2-NEXT: vpxor %ymm7, %ymm1, %ymm8
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm9 = [9223372036854841343,9223372036854841343,9223372036854841343,9223372036854841343]
-; AVX2-NEXT: vpcmpgtq %ymm8, %ymm9, %ymm8
-; AVX2-NEXT: vblendvpd %ymm8, %ymm1, %ymm6, %ymm1
-; AVX2-NEXT: vpxor %ymm7, %ymm0, %ymm7
-; AVX2-NEXT: vpcmpgtq %ymm7, %ymm9, %ymm7
-; AVX2-NEXT: vblendvpd %ymm7, %ymm0, %ymm6, %ymm0
+; AVX2-NEXT: vbroadcastsd {{.*#+}} ymm4 = [65535,65535,65535,65535]
+; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm5 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
+; AVX2-NEXT: vpxor %ymm5, %ymm1, %ymm6
+; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm7 = [9223372036854841343,9223372036854841343,9223372036854841343,9223372036854841343]
+; AVX2-NEXT: vpcmpgtq %ymm6, %ymm7, %ymm6
+; AVX2-NEXT: vblendvpd %ymm6, %ymm1, %ymm4, %ymm1
+; AVX2-NEXT: vpxor %ymm5, %ymm0, %ymm5
+; AVX2-NEXT: vpcmpgtq %ymm5, %ymm7, %ymm5
+; AVX2-NEXT: vblendvpd %ymm5, %ymm0, %ymm4, %ymm0
; AVX2-NEXT: vpackusdw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpextrb $0, %xmm5, %eax
-; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB1_2
-; AVX2-NEXT: # %bb.1: # %cond.store
+; AVX2-NEXT: vpcmpeqd %ymm3, %ymm2, %ymm1
+; AVX2-NEXT: vmovmskps %ymm1, %eax
+; AVX2-NEXT: notl %eax
+; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: jne .LBB1_1
+; AVX2-NEXT: # %bb.2: # %else
+; AVX2-NEXT: testb $2, %al
+; AVX2-NEXT: jne .LBB1_3
+; AVX2-NEXT: .LBB1_4: # %else2
+; AVX2-NEXT: testb $4, %al
+; AVX2-NEXT: jne .LBB1_5
+; AVX2-NEXT: .LBB1_6: # %else4
+; AVX2-NEXT: testb $8, %al
+; AVX2-NEXT: jne .LBB1_7
+; AVX2-NEXT: .LBB1_8: # %else6
+; AVX2-NEXT: testb $16, %al
+; AVX2-NEXT: jne .LBB1_9
+; AVX2-NEXT: .LBB1_10: # %else8
+; AVX2-NEXT: testb $32, %al
+; AVX2-NEXT: jne .LBB1_11
+; AVX2-NEXT: .LBB1_12: # %else10
+; AVX2-NEXT: testb $64, %al
+; AVX2-NEXT: jne .LBB1_13
+; AVX2-NEXT: .LBB1_14: # %else12
+; AVX2-NEXT: testb $-128, %al
+; AVX2-NEXT: jne .LBB1_15
+; AVX2-NEXT: .LBB1_16: # %else14
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+; AVX2-NEXT: .LBB1_1: # %cond.store
; AVX2-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX2-NEXT: .LBB1_2: # %else
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm2, %xmm1
-; AVX2-NEXT: vpxor %xmm4, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $4, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $2, %al
; AVX2-NEXT: je .LBB1_4
-; AVX2-NEXT: # %bb.3: # %cond.store1
+; AVX2-NEXT: .LBB1_3: # %cond.store1
; AVX2-NEXT: vpextrw $1, %xmm0, 2(%rdi)
-; AVX2-NEXT: .LBB1_4: # %else2
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm4
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX2-NEXT: vpextrb $8, %xmm4, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $4, %al
; AVX2-NEXT: je .LBB1_6
-; AVX2-NEXT: # %bb.5: # %cond.store3
+; AVX2-NEXT: .LBB1_5: # %cond.store3
; AVX2-NEXT: vpextrw $2, %xmm0, 4(%rdi)
-; AVX2-NEXT: .LBB1_6: # %else4
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm1
-; AVX2-NEXT: vpxor %xmm3, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $12, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $8, %al
; AVX2-NEXT: je .LBB1_8
-; AVX2-NEXT: # %bb.7: # %cond.store5
+; AVX2-NEXT: .LBB1_7: # %cond.store5
; AVX2-NEXT: vpextrw $3, %xmm0, 6(%rdi)
-; AVX2-NEXT: .LBB1_8: # %else6
-; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX2-NEXT: vpextrb $0, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $16, %al
; AVX2-NEXT: je .LBB1_10
-; AVX2-NEXT: # %bb.9: # %cond.store7
+; AVX2-NEXT: .LBB1_9: # %cond.store7
; AVX2-NEXT: vpextrw $4, %xmm0, 8(%rdi)
-; AVX2-NEXT: .LBB1_10: # %else8
-; AVX2-NEXT: vpextrb $4, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $32, %al
; AVX2-NEXT: je .LBB1_12
-; AVX2-NEXT: # %bb.11: # %cond.store9
+; AVX2-NEXT: .LBB1_11: # %cond.store9
; AVX2-NEXT: vpextrw $5, %xmm0, 10(%rdi)
-; AVX2-NEXT: .LBB1_12: # %else10
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $8, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $64, %al
; AVX2-NEXT: je .LBB1_14
-; AVX2-NEXT: # %bb.13: # %cond.store11
+; AVX2-NEXT: .LBB1_13: # %cond.store11
; AVX2-NEXT: vpextrw $6, %xmm0, 12(%rdi)
-; AVX2-NEXT: .LBB1_14: # %else12
-; AVX2-NEXT: vpextrb $12, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $-128, %al
; AVX2-NEXT: je .LBB1_16
-; AVX2-NEXT: # %bb.15: # %cond.store13
+; AVX2-NEXT: .LBB1_15: # %cond.store13
; AVX2-NEXT: vpextrw $7, %xmm0, 14(%rdi)
-; AVX2-NEXT: .LBB1_16: # %else14
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
@@ -761,66 +729,61 @@ define void @truncstore_v8i64_v8i16(<8 x
; AVX512F-NEXT: vpmovusqw %zmm0, %xmm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB1_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB1_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB1_3
+; AVX512F-NEXT: .LBB1_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB1_5
+; AVX512F-NEXT: .LBB1_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB1_7
+; AVX512F-NEXT: .LBB1_8: # %else6
+; AVX512F-NEXT: testb $16, %al
+; AVX512F-NEXT: jne .LBB1_9
+; AVX512F-NEXT: .LBB1_10: # %else8
+; AVX512F-NEXT: testb $32, %al
+; AVX512F-NEXT: jne .LBB1_11
+; AVX512F-NEXT: .LBB1_12: # %else10
+; AVX512F-NEXT: testb $64, %al
+; AVX512F-NEXT: jne .LBB1_13
+; AVX512F-NEXT: .LBB1_14: # %else12
+; AVX512F-NEXT: testb $-128, %al
+; AVX512F-NEXT: jne .LBB1_15
+; AVX512F-NEXT: .LBB1_16: # %else14
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB1_1: # %cond.store
; AVX512F-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB1_2: # %else
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB1_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB1_3: # %cond.store1
; AVX512F-NEXT: vpextrw $1, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB1_4: # %else2
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $2, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB1_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB1_5: # %cond.store3
; AVX512F-NEXT: vpextrw $2, %xmm0, 4(%rdi)
-; AVX512F-NEXT: .LBB1_6: # %else4
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $3, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB1_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB1_7: # %cond.store5
; AVX512F-NEXT: vpextrw $3, %xmm0, 6(%rdi)
-; AVX512F-NEXT: .LBB1_8: # %else6
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $4, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $16, %al
; AVX512F-NEXT: je .LBB1_10
-; AVX512F-NEXT: # %bb.9: # %cond.store7
+; AVX512F-NEXT: .LBB1_9: # %cond.store7
; AVX512F-NEXT: vpextrw $4, %xmm0, 8(%rdi)
-; AVX512F-NEXT: .LBB1_10: # %else8
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $5, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $32, %al
; AVX512F-NEXT: je .LBB1_12
-; AVX512F-NEXT: # %bb.11: # %cond.store9
+; AVX512F-NEXT: .LBB1_11: # %cond.store9
; AVX512F-NEXT: vpextrw $5, %xmm0, 10(%rdi)
-; AVX512F-NEXT: .LBB1_12: # %else10
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $6, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $64, %al
; AVX512F-NEXT: je .LBB1_14
-; AVX512F-NEXT: # %bb.13: # %cond.store11
+; AVX512F-NEXT: .LBB1_13: # %cond.store11
; AVX512F-NEXT: vpextrw $6, %xmm0, 12(%rdi)
-; AVX512F-NEXT: .LBB1_14: # %else12
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $7, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $-128, %al
; AVX512F-NEXT: je .LBB1_16
-; AVX512F-NEXT: # %bb.15: # %cond.store13
+; AVX512F-NEXT: .LBB1_15: # %cond.store13
; AVX512F-NEXT: vpextrw $7, %xmm0, 14(%rdi)
-; AVX512F-NEXT: .LBB1_16: # %else14
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -851,148 +814,141 @@ define void @truncstore_v8i64_v8i16(<8 x
define void @truncstore_v8i64_v8i8(<8 x i64> %x, <8 x i8>* %p, <8 x i32> %mask) {
; SSE2-LABEL: truncstore_v8i64_v8i8:
; SSE2: # %bb.0:
-; SSE2-NEXT: pxor %xmm7, %xmm7
-; SSE2-NEXT: pcmpeqd %xmm4, %xmm7
-; SSE2-NEXT: pcmpeqd %xmm11, %xmm11
-; SSE2-NEXT: pxor %xmm7, %xmm11
-; SSE2-NEXT: movdqa %xmm11, %xmm9
-; SSE2-NEXT: packssdw %xmm0, %xmm9
-; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [255,255]
-; SSE2-NEXT: movdqa {{.*#+}} xmm12 = [9223372039002259456,9223372039002259456]
-; SSE2-NEXT: movdqa %xmm1, %xmm6
-; SSE2-NEXT: pxor %xmm12, %xmm6
+; SSE2-NEXT: pxor %xmm8, %xmm8
+; SSE2-NEXT: movdqa {{.*#+}} xmm9 = [255,255]
+; SSE2-NEXT: movdqa {{.*#+}} xmm11 = [9223372039002259456,9223372039002259456]
+; SSE2-NEXT: movdqa %xmm1, %xmm7
+; SSE2-NEXT: pxor %xmm11, %xmm7
; SSE2-NEXT: movdqa {{.*#+}} xmm10 = [9223372039002259711,9223372039002259711]
-; SSE2-NEXT: movdqa %xmm10, %xmm7
-; SSE2-NEXT: pcmpgtd %xmm6, %xmm7
-; SSE2-NEXT: pshufd {{.*#+}} xmm13 = xmm7[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm10, %xmm6
-; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
-; SSE2-NEXT: pand %xmm13, %xmm6
+; SSE2-NEXT: movdqa %xmm10, %xmm6
+; SSE2-NEXT: pcmpgtd %xmm7, %xmm6
+; SSE2-NEXT: pshufd {{.*#+}} xmm12 = xmm6[0,0,2,2]
+; SSE2-NEXT: pcmpeqd %xmm10, %xmm7
; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3]
-; SSE2-NEXT: por %xmm6, %xmm7
-; SSE2-NEXT: pand %xmm7, %xmm1
-; SSE2-NEXT: pandn %xmm8, %xmm7
-; SSE2-NEXT: por %xmm1, %xmm7
+; SSE2-NEXT: pand %xmm12, %xmm7
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
+; SSE2-NEXT: por %xmm7, %xmm6
+; SSE2-NEXT: pand %xmm6, %xmm1
+; SSE2-NEXT: pandn %xmm9, %xmm6
+; SSE2-NEXT: por %xmm1, %xmm6
; SSE2-NEXT: movdqa %xmm0, %xmm1
-; SSE2-NEXT: pxor %xmm12, %xmm1
-; SSE2-NEXT: movdqa %xmm10, %xmm6
-; SSE2-NEXT: pcmpgtd %xmm1, %xmm6
-; SSE2-NEXT: pshufd {{.*#+}} xmm13 = xmm6[0,0,2,2]
+; SSE2-NEXT: pxor %xmm11, %xmm1
+; SSE2-NEXT: movdqa %xmm10, %xmm7
+; SSE2-NEXT: pcmpgtd %xmm1, %xmm7
+; SSE2-NEXT: pshufd {{.*#+}} xmm12 = xmm7[0,0,2,2]
; SSE2-NEXT: pcmpeqd %xmm10, %xmm1
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
-; SSE2-NEXT: pand %xmm13, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm6[1,1,3,3]
-; SSE2-NEXT: por %xmm1, %xmm6
-; SSE2-NEXT: pand %xmm6, %xmm0
-; SSE2-NEXT: pandn %xmm8, %xmm6
-; SSE2-NEXT: por %xmm0, %xmm6
-; SSE2-NEXT: packuswb %xmm7, %xmm6
+; SSE2-NEXT: pand %xmm12, %xmm1
+; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm7[1,1,3,3]
+; SSE2-NEXT: por %xmm1, %xmm7
+; SSE2-NEXT: pand %xmm7, %xmm0
+; SSE2-NEXT: pandn %xmm9, %xmm7
+; SSE2-NEXT: por %xmm0, %xmm7
+; SSE2-NEXT: packuswb %xmm6, %xmm7
; SSE2-NEXT: movdqa %xmm3, %xmm0
-; SSE2-NEXT: pxor %xmm12, %xmm0
+; SSE2-NEXT: pxor %xmm11, %xmm0
; SSE2-NEXT: movdqa %xmm10, %xmm1
; SSE2-NEXT: pcmpgtd %xmm0, %xmm1
-; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm1[0,0,2,2]
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm1[0,0,2,2]
; SSE2-NEXT: pcmpeqd %xmm10, %xmm0
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-NEXT: pand %xmm7, %xmm0
+; SSE2-NEXT: pand %xmm6, %xmm0
; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm1[1,1,3,3]
; SSE2-NEXT: por %xmm0, %xmm1
; SSE2-NEXT: pand %xmm1, %xmm3
-; SSE2-NEXT: pandn %xmm8, %xmm1
+; SSE2-NEXT: pandn %xmm9, %xmm1
; SSE2-NEXT: por %xmm3, %xmm1
-; SSE2-NEXT: pxor %xmm2, %xmm12
+; SSE2-NEXT: pxor %xmm2, %xmm11
; SSE2-NEXT: movdqa %xmm10, %xmm0
-; SSE2-NEXT: pcmpgtd %xmm12, %xmm0
+; SSE2-NEXT: pcmpgtd %xmm11, %xmm0
; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm0[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm10, %xmm12
-; SSE2-NEXT: pshufd {{.*#+}} xmm7 = xmm12[1,1,3,3]
-; SSE2-NEXT: pand %xmm3, %xmm7
+; SSE2-NEXT: pcmpeqd %xmm10, %xmm11
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm11[1,1,3,3]
+; SSE2-NEXT: pand %xmm3, %xmm6
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,1,3,3]
-; SSE2-NEXT: por %xmm7, %xmm0
+; SSE2-NEXT: por %xmm6, %xmm0
; SSE2-NEXT: pand %xmm0, %xmm2
-; SSE2-NEXT: pandn %xmm8, %xmm0
+; SSE2-NEXT: pandn %xmm9, %xmm0
; SSE2-NEXT: por %xmm2, %xmm0
; SSE2-NEXT: packuswb %xmm1, %xmm0
-; SSE2-NEXT: packuswb %xmm0, %xmm6
-; SSE2-NEXT: movd %xmm9, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: movd %xmm6, %eax
-; SSE2-NEXT: je .LBB2_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movb %al, (%rdi)
-; SSE2-NEXT: .LBB2_2: # %else
-; SSE2-NEXT: psrlq $16, %xmm11
-; SSE2-NEXT: movd %xmm11, %ecx
-; SSE2-NEXT: shrl $16, %ecx
-; SSE2-NEXT: testb $1, %cl
-; SSE2-NEXT: je .LBB2_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: shrl $16, %eax
-; SSE2-NEXT: movb %al, 1(%rdi)
-; SSE2-NEXT: .LBB2_4: # %else2
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm4
+; SSE2-NEXT: packuswb %xmm0, %xmm7
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm5
; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm4, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
+; SSE2-NEXT: pxor %xmm0, %xmm5
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm4
+; SSE2-NEXT: pxor %xmm0, %xmm4
+; SSE2-NEXT: packssdw %xmm5, %xmm4
+; SSE2-NEXT: packsswb %xmm0, %xmm4
+; SSE2-NEXT: pmovmskb %xmm4, %eax
; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB2_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: pextrw $2, %xmm6, %eax
-; SSE2-NEXT: movb %al, 2(%rdi)
+; SSE2-NEXT: movd %xmm7, %ecx
+; SSE2-NEXT: jne .LBB2_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB2_3
+; SSE2-NEXT: .LBB2_4: # %else2
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB2_5
; SSE2-NEXT: .LBB2_6: # %else4
-; SSE2-NEXT: pextrw $6, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB2_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
-; SSE2-NEXT: pextrw $3, %xmm6, %eax
-; SSE2-NEXT: movb %al, 3(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB2_7
; SSE2-NEXT: .LBB2_8: # %else6
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm5, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm1, %xmm0
-; SSE2-NEXT: pextrw $0, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB2_10
-; SSE2-NEXT: # %bb.9: # %cond.store7
-; SSE2-NEXT: pextrw $4, %xmm6, %eax
-; SSE2-NEXT: movb %al, 4(%rdi)
+; SSE2-NEXT: testb $16, %al
+; SSE2-NEXT: jne .LBB2_9
; SSE2-NEXT: .LBB2_10: # %else8
-; SSE2-NEXT: pextrw $2, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB2_12
-; SSE2-NEXT: # %bb.11: # %cond.store9
-; SSE2-NEXT: pextrw $5, %xmm6, %eax
-; SSE2-NEXT: movb %al, 5(%rdi)
+; SSE2-NEXT: testb $32, %al
+; SSE2-NEXT: jne .LBB2_11
; SSE2-NEXT: .LBB2_12: # %else10
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm5
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm5, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB2_14
-; SSE2-NEXT: # %bb.13: # %cond.store11
-; SSE2-NEXT: pextrw $6, %xmm6, %eax
-; SSE2-NEXT: movb %al, 6(%rdi)
+; SSE2-NEXT: testb $64, %al
+; SSE2-NEXT: jne .LBB2_13
; SSE2-NEXT: .LBB2_14: # %else12
-; SSE2-NEXT: pextrw $6, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $-128, %al
+; SSE2-NEXT: jne .LBB2_15
+; SSE2-NEXT: .LBB2_16: # %else14
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB2_1: # %cond.store
+; SSE2-NEXT: movb %cl, (%rdi)
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: je .LBB2_4
+; SSE2-NEXT: .LBB2_3: # %cond.store1
+; SSE2-NEXT: shrl $16, %ecx
+; SSE2-NEXT: movb %cl, 1(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: je .LBB2_6
+; SSE2-NEXT: .LBB2_5: # %cond.store3
+; SSE2-NEXT: pextrw $2, %xmm7, %ecx
+; SSE2-NEXT: movb %cl, 2(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: je .LBB2_8
+; SSE2-NEXT: .LBB2_7: # %cond.store5
+; SSE2-NEXT: pextrw $3, %xmm7, %ecx
+; SSE2-NEXT: movb %cl, 3(%rdi)
+; SSE2-NEXT: testb $16, %al
+; SSE2-NEXT: je .LBB2_10
+; SSE2-NEXT: .LBB2_9: # %cond.store7
+; SSE2-NEXT: pextrw $4, %xmm7, %ecx
+; SSE2-NEXT: movb %cl, 4(%rdi)
+; SSE2-NEXT: testb $32, %al
+; SSE2-NEXT: je .LBB2_12
+; SSE2-NEXT: .LBB2_11: # %cond.store9
+; SSE2-NEXT: pextrw $5, %xmm7, %ecx
+; SSE2-NEXT: movb %cl, 5(%rdi)
+; SSE2-NEXT: testb $64, %al
+; SSE2-NEXT: je .LBB2_14
+; SSE2-NEXT: .LBB2_13: # %cond.store11
+; SSE2-NEXT: pextrw $6, %xmm7, %ecx
+; SSE2-NEXT: movb %cl, 6(%rdi)
+; SSE2-NEXT: testb $-128, %al
; SSE2-NEXT: je .LBB2_16
-; SSE2-NEXT: # %bb.15: # %cond.store13
-; SSE2-NEXT: pextrw $7, %xmm6, %eax
+; SSE2-NEXT: .LBB2_15: # %cond.store13
+; SSE2-NEXT: pextrw $7, %xmm7, %eax
; SSE2-NEXT: movb %al, 7(%rdi)
-; SSE2-NEXT: .LBB2_16: # %else14
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v8i64_v8i8:
; SSE4: # %bb.0:
-; SSE4-NEXT: movdqa %xmm0, %xmm8
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm4, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm9, %xmm9
-; SSE4-NEXT: pxor %xmm0, %xmm9
+; SSE4-NEXT: movdqa %xmm0, %xmm9
+; SSE4-NEXT: pxor %xmm8, %xmm8
; SSE4-NEXT: movapd {{.*#+}} xmm6 = [255,255]
; SSE4-NEXT: movdqa {{.*#+}} xmm10 = [9223372036854775808,9223372036854775808]
; SSE4-NEXT: movdqa %xmm1, %xmm7
@@ -1002,12 +958,12 @@ define void @truncstore_v8i64_v8i8(<8 x
; SSE4-NEXT: pcmpgtq %xmm7, %xmm0
; SSE4-NEXT: movapd %xmm6, %xmm7
; SSE4-NEXT: blendvpd %xmm0, %xmm1, %xmm7
-; SSE4-NEXT: movdqa %xmm8, %xmm1
+; SSE4-NEXT: movdqa %xmm9, %xmm1
; SSE4-NEXT: pxor %xmm10, %xmm1
; SSE4-NEXT: movdqa %xmm11, %xmm0
; SSE4-NEXT: pcmpgtq %xmm1, %xmm0
; SSE4-NEXT: movapd %xmm6, %xmm1
-; SSE4-NEXT: blendvpd %xmm0, %xmm8, %xmm1
+; SSE4-NEXT: blendvpd %xmm0, %xmm9, %xmm1
; SSE4-NEXT: packusdw %xmm7, %xmm1
; SSE4-NEXT: movdqa %xmm3, %xmm7
; SSE4-NEXT: pxor %xmm10, %xmm7
@@ -1021,246 +977,234 @@ define void @truncstore_v8i64_v8i8(<8 x
; SSE4-NEXT: blendvpd %xmm0, %xmm2, %xmm6
; SSE4-NEXT: packusdw %xmm7, %xmm6
; SSE4-NEXT: packusdw %xmm6, %xmm1
-; SSE4-NEXT: pextrb $0, %xmm9, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB2_2
-; SSE4-NEXT: # %bb.1: # %cond.store
+; SSE4-NEXT: pcmpeqd %xmm8, %xmm5
+; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
+; SSE4-NEXT: pxor %xmm0, %xmm5
+; SSE4-NEXT: pcmpeqd %xmm8, %xmm4
+; SSE4-NEXT: pxor %xmm0, %xmm4
+; SSE4-NEXT: packssdw %xmm5, %xmm4
+; SSE4-NEXT: packsswb %xmm0, %xmm4
+; SSE4-NEXT: pmovmskb %xmm4, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB2_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB2_3
+; SSE4-NEXT: .LBB2_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB2_5
+; SSE4-NEXT: .LBB2_6: # %else4
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB2_7
+; SSE4-NEXT: .LBB2_8: # %else6
+; SSE4-NEXT: testb $16, %al
+; SSE4-NEXT: jne .LBB2_9
+; SSE4-NEXT: .LBB2_10: # %else8
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: jne .LBB2_11
+; SSE4-NEXT: .LBB2_12: # %else10
+; SSE4-NEXT: testb $64, %al
+; SSE4-NEXT: jne .LBB2_13
+; SSE4-NEXT: .LBB2_14: # %else12
+; SSE4-NEXT: testb $-128, %al
+; SSE4-NEXT: jne .LBB2_15
+; SSE4-NEXT: .LBB2_16: # %else14
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB2_1: # %cond.store
; SSE4-NEXT: pextrb $0, %xmm1, (%rdi)
-; SSE4-NEXT: .LBB2_2: # %else
-; SSE4-NEXT: pextrb $4, %xmm9, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $2, %al
; SSE4-NEXT: je .LBB2_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
+; SSE4-NEXT: .LBB2_3: # %cond.store1
; SSE4-NEXT: pextrb $2, %xmm1, 1(%rdi)
-; SSE4-NEXT: .LBB2_4: # %else2
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm4
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm4, %xmm0
-; SSE4-NEXT: pextrb $8, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $4, %al
; SSE4-NEXT: je .LBB2_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
+; SSE4-NEXT: .LBB2_5: # %cond.store3
; SSE4-NEXT: pextrb $4, %xmm1, 2(%rdi)
-; SSE4-NEXT: .LBB2_6: # %else4
-; SSE4-NEXT: pextrb $12, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $8, %al
; SSE4-NEXT: je .LBB2_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
+; SSE4-NEXT: .LBB2_7: # %cond.store5
; SSE4-NEXT: pextrb $6, %xmm1, 3(%rdi)
-; SSE4-NEXT: .LBB2_8: # %else6
-; SSE4-NEXT: pxor %xmm2, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm5, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm2, %xmm0
-; SSE4-NEXT: pextrb $0, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $16, %al
; SSE4-NEXT: je .LBB2_10
-; SSE4-NEXT: # %bb.9: # %cond.store7
+; SSE4-NEXT: .LBB2_9: # %cond.store7
; SSE4-NEXT: pextrb $8, %xmm1, 4(%rdi)
-; SSE4-NEXT: .LBB2_10: # %else8
-; SSE4-NEXT: pextrb $4, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $32, %al
; SSE4-NEXT: je .LBB2_12
-; SSE4-NEXT: # %bb.11: # %cond.store9
+; SSE4-NEXT: .LBB2_11: # %cond.store9
; SSE4-NEXT: pextrb $10, %xmm1, 5(%rdi)
-; SSE4-NEXT: .LBB2_12: # %else10
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm5
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm5, %xmm0
-; SSE4-NEXT: pextrb $8, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $64, %al
; SSE4-NEXT: je .LBB2_14
-; SSE4-NEXT: # %bb.13: # %cond.store11
+; SSE4-NEXT: .LBB2_13: # %cond.store11
; SSE4-NEXT: pextrb $12, %xmm1, 6(%rdi)
-; SSE4-NEXT: .LBB2_14: # %else12
-; SSE4-NEXT: pextrb $12, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $-128, %al
; SSE4-NEXT: je .LBB2_16
-; SSE4-NEXT: # %bb.15: # %cond.store13
+; SSE4-NEXT: .LBB2_15: # %cond.store13
; SSE4-NEXT: pextrb $14, %xmm1, 7(%rdi)
-; SSE4-NEXT: .LBB2_16: # %else14
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v8i64_v8i8:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpxor %xmm8, %xmm8, %xmm8
-; AVX1-NEXT: vpcmpeqd %xmm8, %xmm2, %xmm5
-; AVX1-NEXT: vpcmpeqd %xmm9, %xmm9, %xmm9
-; 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: vmovdqa {{.*#+}} xmm3 = [9223372036854775808,9223372036854775808]
+; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm4
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372036854776063,9223372036854776063]
+; AVX1-NEXT: vpcmpgtq %xmm4, %xmm5, %xmm8
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm6
+; AVX1-NEXT: vpxor %xmm3, %xmm6, %xmm7
+; AVX1-NEXT: vpcmpgtq %xmm7, %xmm5, %xmm9
+; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm4
+; AVX1-NEXT: vpcmpgtq %xmm4, %xmm5, %xmm4
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm7
+; AVX1-NEXT: vpxor %xmm3, %xmm7, %xmm3
+; AVX1-NEXT: vpcmpgtq %xmm3, %xmm5, %xmm3
+; AVX1-NEXT: vmovapd {{.*#+}} xmm5 = [255,255]
+; AVX1-NEXT: vblendvpd %xmm3, %xmm7, %xmm5, %xmm3
+; AVX1-NEXT: vblendvpd %xmm4, %xmm1, %xmm5, %xmm1
; AVX1-NEXT: vpackusdw %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vblendvpd %xmm12, %xmm4, %xmm6, %xmm3
-; AVX1-NEXT: vblendvpd %xmm11, %xmm0, %xmm6, %xmm0
+; AVX1-NEXT: vblendvpd %xmm9, %xmm6, %xmm5, %xmm3
+; AVX1-NEXT: vblendvpd %xmm8, %xmm0, %xmm5, %xmm0
; AVX1-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
; AVX1-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
-; AVX1-NEXT: vpextrb $0, %xmm10, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB2_2
-; AVX1-NEXT: # %bb.1: # %cond.store
+; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm1
+; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT: vpcmpeqd %xmm3, %xmm1, %xmm1
+; AVX1-NEXT: vpcmpeqd %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; AVX1-NEXT: vmovmskps %ymm1, %eax
+; AVX1-NEXT: notl %eax
+; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: jne .LBB2_1
+; AVX1-NEXT: # %bb.2: # %else
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: jne .LBB2_3
+; AVX1-NEXT: .LBB2_4: # %else2
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: jne .LBB2_5
+; AVX1-NEXT: .LBB2_6: # %else4
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: jne .LBB2_7
+; AVX1-NEXT: .LBB2_8: # %else6
+; AVX1-NEXT: testb $16, %al
+; AVX1-NEXT: jne .LBB2_9
+; AVX1-NEXT: .LBB2_10: # %else8
+; AVX1-NEXT: testb $32, %al
+; AVX1-NEXT: jne .LBB2_11
+; AVX1-NEXT: .LBB2_12: # %else10
+; AVX1-NEXT: testb $64, %al
+; AVX1-NEXT: jne .LBB2_13
+; AVX1-NEXT: .LBB2_14: # %else12
+; AVX1-NEXT: testb $-128, %al
+; AVX1-NEXT: jne .LBB2_15
+; AVX1-NEXT: .LBB2_16: # %else14
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+; AVX1-NEXT: .LBB2_1: # %cond.store
; AVX1-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX1-NEXT: .LBB2_2: # %else
-; AVX1-NEXT: vpcmpeqd %xmm8, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm9, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $4, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $2, %al
; AVX1-NEXT: je .LBB2_4
-; AVX1-NEXT: # %bb.3: # %cond.store1
+; AVX1-NEXT: .LBB2_3: # %cond.store1
; AVX1-NEXT: vpextrb $2, %xmm0, 1(%rdi)
-; AVX1-NEXT: .LBB2_4: # %else2
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $8, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $4, %al
; AVX1-NEXT: je .LBB2_6
-; AVX1-NEXT: # %bb.5: # %cond.store3
+; AVX1-NEXT: .LBB2_5: # %cond.store3
; AVX1-NEXT: vpextrb $4, %xmm0, 2(%rdi)
-; AVX1-NEXT: .LBB2_6: # %else4
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $12, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $8, %al
; AVX1-NEXT: je .LBB2_8
-; AVX1-NEXT: # %bb.7: # %cond.store5
+; AVX1-NEXT: .LBB2_7: # %cond.store5
; AVX1-NEXT: vpextrb $6, %xmm0, 3(%rdi)
-; AVX1-NEXT: .LBB2_8: # %else6
-; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $0, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $16, %al
; AVX1-NEXT: je .LBB2_10
-; AVX1-NEXT: # %bb.9: # %cond.store7
+; AVX1-NEXT: .LBB2_9: # %cond.store7
; AVX1-NEXT: vpextrb $8, %xmm0, 4(%rdi)
-; AVX1-NEXT: .LBB2_10: # %else8
-; AVX1-NEXT: vpextrb $4, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $32, %al
; AVX1-NEXT: je .LBB2_12
-; AVX1-NEXT: # %bb.11: # %cond.store9
+; AVX1-NEXT: .LBB2_11: # %cond.store9
; AVX1-NEXT: vpextrb $10, %xmm0, 5(%rdi)
-; AVX1-NEXT: .LBB2_12: # %else10
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $8, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $64, %al
; AVX1-NEXT: je .LBB2_14
-; AVX1-NEXT: # %bb.13: # %cond.store11
+; AVX1-NEXT: .LBB2_13: # %cond.store11
; AVX1-NEXT: vpextrb $12, %xmm0, 6(%rdi)
-; AVX1-NEXT: .LBB2_14: # %else12
-; AVX1-NEXT: vpextrb $12, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $-128, %al
; AVX1-NEXT: je .LBB2_16
-; AVX1-NEXT: # %bb.15: # %cond.store13
+; AVX1-NEXT: .LBB2_15: # %cond.store13
; AVX1-NEXT: vpextrb $14, %xmm0, 7(%rdi)
-; AVX1-NEXT: .LBB2_16: # %else14
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: truncstore_v8i64_v8i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm2, %xmm5
-; AVX2-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX2-NEXT: vpxor %xmm4, %xmm5, %xmm5
-; AVX2-NEXT: vbroadcastsd {{.*#+}} ymm6 = [255,255,255,255]
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm7 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
-; AVX2-NEXT: vpxor %ymm7, %ymm1, %ymm8
-; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm9 = [9223372036854776063,9223372036854776063,9223372036854776063,9223372036854776063]
-; AVX2-NEXT: vpcmpgtq %ymm8, %ymm9, %ymm8
-; AVX2-NEXT: vblendvpd %ymm8, %ymm1, %ymm6, %ymm1
-; AVX2-NEXT: vpxor %ymm7, %ymm0, %ymm7
-; AVX2-NEXT: vpcmpgtq %ymm7, %ymm9, %ymm7
-; AVX2-NEXT: vblendvpd %ymm7, %ymm0, %ymm6, %ymm0
+; AVX2-NEXT: vbroadcastsd {{.*#+}} ymm4 = [255,255,255,255]
+; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm5 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
+; AVX2-NEXT: vpxor %ymm5, %ymm1, %ymm6
+; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm7 = [9223372036854776063,9223372036854776063,9223372036854776063,9223372036854776063]
+; AVX2-NEXT: vpcmpgtq %ymm6, %ymm7, %ymm6
+; AVX2-NEXT: vblendvpd %ymm6, %ymm1, %ymm4, %ymm1
+; AVX2-NEXT: vpxor %ymm5, %ymm0, %ymm5
+; AVX2-NEXT: vpcmpgtq %ymm5, %ymm7, %ymm5
+; AVX2-NEXT: vblendvpd %ymm5, %ymm0, %ymm4, %ymm0
; AVX2-NEXT: vpackusdw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpackusdw %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpextrb $0, %xmm5, %eax
-; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB2_2
-; AVX2-NEXT: # %bb.1: # %cond.store
+; AVX2-NEXT: vpcmpeqd %ymm3, %ymm2, %ymm1
+; AVX2-NEXT: vmovmskps %ymm1, %eax
+; AVX2-NEXT: notl %eax
+; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: jne .LBB2_1
+; AVX2-NEXT: # %bb.2: # %else
+; AVX2-NEXT: testb $2, %al
+; AVX2-NEXT: jne .LBB2_3
+; AVX2-NEXT: .LBB2_4: # %else2
+; AVX2-NEXT: testb $4, %al
+; AVX2-NEXT: jne .LBB2_5
+; AVX2-NEXT: .LBB2_6: # %else4
+; AVX2-NEXT: testb $8, %al
+; AVX2-NEXT: jne .LBB2_7
+; AVX2-NEXT: .LBB2_8: # %else6
+; AVX2-NEXT: testb $16, %al
+; AVX2-NEXT: jne .LBB2_9
+; AVX2-NEXT: .LBB2_10: # %else8
+; AVX2-NEXT: testb $32, %al
+; AVX2-NEXT: jne .LBB2_11
+; AVX2-NEXT: .LBB2_12: # %else10
+; AVX2-NEXT: testb $64, %al
+; AVX2-NEXT: jne .LBB2_13
+; AVX2-NEXT: .LBB2_14: # %else12
+; AVX2-NEXT: testb $-128, %al
+; AVX2-NEXT: jne .LBB2_15
+; AVX2-NEXT: .LBB2_16: # %else14
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+; AVX2-NEXT: .LBB2_1: # %cond.store
; AVX2-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX2-NEXT: .LBB2_2: # %else
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm2, %xmm1
-; AVX2-NEXT: vpxor %xmm4, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $4, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $2, %al
; AVX2-NEXT: je .LBB2_4
-; AVX2-NEXT: # %bb.3: # %cond.store1
+; AVX2-NEXT: .LBB2_3: # %cond.store1
; AVX2-NEXT: vpextrb $2, %xmm0, 1(%rdi)
-; AVX2-NEXT: .LBB2_4: # %else2
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm4
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX2-NEXT: vpextrb $8, %xmm4, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $4, %al
; AVX2-NEXT: je .LBB2_6
-; AVX2-NEXT: # %bb.5: # %cond.store3
+; AVX2-NEXT: .LBB2_5: # %cond.store3
; AVX2-NEXT: vpextrb $4, %xmm0, 2(%rdi)
-; AVX2-NEXT: .LBB2_6: # %else4
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm1
-; AVX2-NEXT: vpxor %xmm3, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $12, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $8, %al
; AVX2-NEXT: je .LBB2_8
-; AVX2-NEXT: # %bb.7: # %cond.store5
+; AVX2-NEXT: .LBB2_7: # %cond.store5
; AVX2-NEXT: vpextrb $6, %xmm0, 3(%rdi)
-; AVX2-NEXT: .LBB2_8: # %else6
-; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX2-NEXT: vpextrb $0, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $16, %al
; AVX2-NEXT: je .LBB2_10
-; AVX2-NEXT: # %bb.9: # %cond.store7
+; AVX2-NEXT: .LBB2_9: # %cond.store7
; AVX2-NEXT: vpextrb $8, %xmm0, 4(%rdi)
-; AVX2-NEXT: .LBB2_10: # %else8
-; AVX2-NEXT: vpextrb $4, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $32, %al
; AVX2-NEXT: je .LBB2_12
-; AVX2-NEXT: # %bb.11: # %cond.store9
+; AVX2-NEXT: .LBB2_11: # %cond.store9
; AVX2-NEXT: vpextrb $10, %xmm0, 5(%rdi)
-; AVX2-NEXT: .LBB2_12: # %else10
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $8, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $64, %al
; AVX2-NEXT: je .LBB2_14
-; AVX2-NEXT: # %bb.13: # %cond.store11
+; AVX2-NEXT: .LBB2_13: # %cond.store11
; AVX2-NEXT: vpextrb $12, %xmm0, 6(%rdi)
-; AVX2-NEXT: .LBB2_14: # %else12
-; AVX2-NEXT: vpextrb $12, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $-128, %al
; AVX2-NEXT: je .LBB2_16
-; AVX2-NEXT: # %bb.15: # %cond.store13
+; AVX2-NEXT: .LBB2_15: # %cond.store13
; AVX2-NEXT: vpextrb $14, %xmm0, 7(%rdi)
-; AVX2-NEXT: .LBB2_16: # %else14
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
@@ -1272,66 +1216,61 @@ define void @truncstore_v8i64_v8i8(<8 x
; AVX512F-NEXT: vpmovqw %zmm0, %xmm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB2_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB2_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB2_3
+; AVX512F-NEXT: .LBB2_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB2_5
+; AVX512F-NEXT: .LBB2_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB2_7
+; AVX512F-NEXT: .LBB2_8: # %else6
+; AVX512F-NEXT: testb $16, %al
+; AVX512F-NEXT: jne .LBB2_9
+; AVX512F-NEXT: .LBB2_10: # %else8
+; AVX512F-NEXT: testb $32, %al
+; AVX512F-NEXT: jne .LBB2_11
+; AVX512F-NEXT: .LBB2_12: # %else10
+; AVX512F-NEXT: testb $64, %al
+; AVX512F-NEXT: jne .LBB2_13
+; AVX512F-NEXT: .LBB2_14: # %else12
+; AVX512F-NEXT: testb $-128, %al
+; AVX512F-NEXT: jne .LBB2_15
+; AVX512F-NEXT: .LBB2_16: # %else14
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB2_1: # %cond.store
; AVX512F-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB2_2: # %else
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB2_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB2_3: # %cond.store1
; AVX512F-NEXT: vpextrb $2, %xmm0, 1(%rdi)
-; AVX512F-NEXT: .LBB2_4: # %else2
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $2, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB2_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB2_5: # %cond.store3
; AVX512F-NEXT: vpextrb $4, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB2_6: # %else4
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $3, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB2_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB2_7: # %cond.store5
; AVX512F-NEXT: vpextrb $6, %xmm0, 3(%rdi)
-; AVX512F-NEXT: .LBB2_8: # %else6
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $4, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $16, %al
; AVX512F-NEXT: je .LBB2_10
-; AVX512F-NEXT: # %bb.9: # %cond.store7
+; AVX512F-NEXT: .LBB2_9: # %cond.store7
; AVX512F-NEXT: vpextrb $8, %xmm0, 4(%rdi)
-; AVX512F-NEXT: .LBB2_10: # %else8
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $5, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $32, %al
; AVX512F-NEXT: je .LBB2_12
-; AVX512F-NEXT: # %bb.11: # %cond.store9
+; AVX512F-NEXT: .LBB2_11: # %cond.store9
; AVX512F-NEXT: vpextrb $10, %xmm0, 5(%rdi)
-; AVX512F-NEXT: .LBB2_12: # %else10
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $6, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $64, %al
; AVX512F-NEXT: je .LBB2_14
-; AVX512F-NEXT: # %bb.13: # %cond.store11
+; AVX512F-NEXT: .LBB2_13: # %cond.store11
; AVX512F-NEXT: vpextrb $12, %xmm0, 6(%rdi)
-; AVX512F-NEXT: .LBB2_14: # %else12
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $7, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $-128, %al
; AVX512F-NEXT: je .LBB2_16
-; AVX512F-NEXT: # %bb.15: # %cond.store13
+; AVX512F-NEXT: .LBB2_15: # %cond.store13
; AVX512F-NEXT: vpextrb $14, %xmm0, 7(%rdi)
-; AVX512F-NEXT: .LBB2_16: # %else14
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -1363,7 +1302,6 @@ define void @truncstore_v4i64_v4i32(<4 x
; SSE2-LABEL: truncstore_v4i64_v4i32:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm3
; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [4294967295,4294967295]
; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [9223372039002259456,9223372039002259456]
; SSE2-NEXT: movdqa %xmm1, %xmm6
@@ -1393,50 +1331,45 @@ define void @truncstore_v4i64_v4i32(<4 x
; SSE2-NEXT: pandn %xmm8, %xmm1
; SSE2-NEXT: por %xmm0, %xmm1
; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,2]
-; SSE2-NEXT: movd %xmm3, %eax
-; SSE2-NEXT: notl %eax
+; SSE2-NEXT: pcmpeqd %xmm2, %xmm3
+; SSE2-NEXT: movmskps %xmm3, %eax
+; SSE2-NEXT: xorl $15, %eax
; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB3_2
-; SSE2-NEXT: # %bb.1: # %cond.store
+; SSE2-NEXT: jne .LBB3_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB3_3
+; SSE2-NEXT: .LBB3_4: # %else2
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB3_5
+; SSE2-NEXT: .LBB3_6: # %else4
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB3_7
+; SSE2-NEXT: .LBB3_8: # %else6
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB3_1: # %cond.store
; SSE2-NEXT: movss %xmm1, (%rdi)
-; SSE2-NEXT: .LBB3_2: # %else
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm3
-; SSE2-NEXT: pextrw $2, %xmm3, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $2, %al
; SSE2-NEXT: je .LBB3_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm1[1,1,2,3]
-; SSE2-NEXT: movd %xmm3, 4(%rdi)
-; SSE2-NEXT: .LBB3_4: # %else2
-; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
-; SSE2-NEXT: pxor %xmm2, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: .LBB3_3: # %cond.store1
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[1,1,2,3]
+; SSE2-NEXT: movd %xmm0, 4(%rdi)
+; SSE2-NEXT: testb $4, %al
; SSE2-NEXT: je .LBB3_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
+; SSE2-NEXT: .LBB3_5: # %cond.store3
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[2,3,0,1]
; SSE2-NEXT: movd %xmm0, 8(%rdi)
-; SSE2-NEXT: .LBB3_6: # %else4
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm2
-; SSE2-NEXT: pextrw $6, %xmm2, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $8, %al
; SSE2-NEXT: je .LBB3_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
+; SSE2-NEXT: .LBB3_7: # %cond.store5
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm1[3,1,2,3]
; SSE2-NEXT: movd %xmm0, 12(%rdi)
-; SSE2-NEXT: .LBB3_8: # %else6
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v4i64_v4i32:
; SSE4: # %bb.0:
; SSE4-NEXT: movdqa %xmm0, %xmm8
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm6, %xmm6
-; SSE4-NEXT: pxor %xmm0, %xmm6
+; SSE4-NEXT: pxor %xmm6, %xmm6
; SSE4-NEXT: movapd {{.*#+}} xmm5 = [4294967295,4294967295]
; SSE4-NEXT: movdqa {{.*#+}} xmm7 = [9223372036854775808,9223372036854775808]
; SSE4-NEXT: movdqa %xmm1, %xmm3
@@ -1451,34 +1384,36 @@ define void @truncstore_v4i64_v4i32(<4 x
; SSE4-NEXT: movdqa %xmm4, %xmm0
; SSE4-NEXT: blendvpd %xmm0, %xmm8, %xmm5
; SSE4-NEXT: shufps {{.*#+}} xmm5 = xmm5[0,2],xmm3[0,2]
-; SSE4-NEXT: pextrb $0, %xmm6, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB3_2
-; SSE4-NEXT: # %bb.1: # %cond.store
+; SSE4-NEXT: pcmpeqd %xmm2, %xmm6
+; SSE4-NEXT: movmskps %xmm6, %eax
+; SSE4-NEXT: xorl $15, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB3_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB3_3
+; SSE4-NEXT: .LBB3_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB3_5
+; SSE4-NEXT: .LBB3_6: # %else4
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB3_7
+; SSE4-NEXT: .LBB3_8: # %else6
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB3_1: # %cond.store
; SSE4-NEXT: movss %xmm5, (%rdi)
-; SSE4-NEXT: .LBB3_2: # %else
-; SSE4-NEXT: pextrb $4, %xmm6, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $2, %al
; SSE4-NEXT: je .LBB3_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
+; SSE4-NEXT: .LBB3_3: # %cond.store1
; SSE4-NEXT: extractps $1, %xmm5, 4(%rdi)
-; SSE4-NEXT: .LBB3_4: # %else2
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm2, %xmm0
-; SSE4-NEXT: pextrb $8, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $4, %al
; SSE4-NEXT: je .LBB3_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
+; SSE4-NEXT: .LBB3_5: # %cond.store3
; SSE4-NEXT: extractps $2, %xmm5, 8(%rdi)
-; SSE4-NEXT: .LBB3_6: # %else4
-; SSE4-NEXT: pextrb $12, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $8, %al
; SSE4-NEXT: je .LBB3_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
+; SSE4-NEXT: .LBB3_7: # %cond.store5
; SSE4-NEXT: extractps $3, %xmm5, 12(%rdi)
-; SSE4-NEXT: .LBB3_8: # %else6
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v4i64_v4i32:
@@ -1567,7 +1502,6 @@ define void @truncstore_v4i64_v4i16(<4 x
; SSE2-LABEL: truncstore_v4i64_v4i16:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm3
; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [65535,65535]
; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [9223372039002259456,9223372039002259456]
; SSE2-NEXT: movdqa %xmm1, %xmm6
@@ -1597,51 +1531,46 @@ define void @truncstore_v4i64_v4i16(<4 x
; SSE2-NEXT: pandn %xmm8, %xmm1
; SSE2-NEXT: por %xmm0, %xmm1
; SSE2-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,2],xmm4[0,2]
-; SSE2-NEXT: movd %xmm3, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB4_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movd %xmm1, %eax
-; SSE2-NEXT: movw %ax, (%rdi)
-; SSE2-NEXT: .LBB4_2: # %else
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm3
-; SSE2-NEXT: pextrw $2, %xmm3, %eax
+; SSE2-NEXT: pcmpeqd %xmm2, %xmm3
+; SSE2-NEXT: movmskps %xmm3, %eax
+; SSE2-NEXT: xorl $15, %eax
; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB4_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: pextrw $2, %xmm1, %eax
-; SSE2-NEXT: movw %ax, 2(%rdi)
+; SSE2-NEXT: jne .LBB4_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB4_3
; SSE2-NEXT: .LBB4_4: # %else2
-; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
-; SSE2-NEXT: pxor %xmm2, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB4_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: pextrw $4, %xmm1, %eax
-; SSE2-NEXT: movw %ax, 4(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB4_5
; SSE2-NEXT: .LBB4_6: # %else4
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm2
-; SSE2-NEXT: pextrw $6, %xmm2, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB4_7
+; SSE2-NEXT: .LBB4_8: # %else6
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB4_1: # %cond.store
+; SSE2-NEXT: movd %xmm1, %ecx
+; SSE2-NEXT: movw %cx, (%rdi)
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: je .LBB4_4
+; SSE2-NEXT: .LBB4_3: # %cond.store1
+; SSE2-NEXT: pextrw $2, %xmm1, %ecx
+; SSE2-NEXT: movw %cx, 2(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: je .LBB4_6
+; SSE2-NEXT: .LBB4_5: # %cond.store3
+; SSE2-NEXT: pextrw $4, %xmm1, %ecx
+; SSE2-NEXT: movw %cx, 4(%rdi)
+; SSE2-NEXT: testb $8, %al
; SSE2-NEXT: je .LBB4_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
+; SSE2-NEXT: .LBB4_7: # %cond.store5
; SSE2-NEXT: pextrw $6, %xmm1, %eax
; SSE2-NEXT: movw %ax, 6(%rdi)
-; SSE2-NEXT: .LBB4_8: # %else6
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v4i64_v4i16:
; SSE4: # %bb.0:
; SSE4-NEXT: movdqa %xmm0, %xmm8
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm6, %xmm6
-; SSE4-NEXT: pxor %xmm0, %xmm6
+; SSE4-NEXT: pxor %xmm6, %xmm6
; SSE4-NEXT: movapd {{.*#+}} xmm5 = [65535,65535]
; SSE4-NEXT: movdqa {{.*#+}} xmm7 = [9223372036854775808,9223372036854775808]
; SSE4-NEXT: movdqa %xmm1, %xmm3
@@ -1656,42 +1585,41 @@ define void @truncstore_v4i64_v4i16(<4 x
; SSE4-NEXT: movdqa %xmm4, %xmm0
; SSE4-NEXT: blendvpd %xmm0, %xmm8, %xmm5
; SSE4-NEXT: packusdw %xmm3, %xmm5
-; SSE4-NEXT: pextrb $0, %xmm6, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB4_2
-; SSE4-NEXT: # %bb.1: # %cond.store
+; SSE4-NEXT: pcmpeqd %xmm2, %xmm6
+; SSE4-NEXT: movmskps %xmm6, %eax
+; SSE4-NEXT: xorl $15, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB4_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB4_3
+; SSE4-NEXT: .LBB4_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB4_5
+; SSE4-NEXT: .LBB4_6: # %else4
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB4_7
+; SSE4-NEXT: .LBB4_8: # %else6
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB4_1: # %cond.store
; SSE4-NEXT: pextrw $0, %xmm5, (%rdi)
-; SSE4-NEXT: .LBB4_2: # %else
-; SSE4-NEXT: pextrb $4, %xmm6, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $2, %al
; SSE4-NEXT: je .LBB4_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
+; SSE4-NEXT: .LBB4_3: # %cond.store1
; SSE4-NEXT: pextrw $2, %xmm5, 2(%rdi)
-; SSE4-NEXT: .LBB4_4: # %else2
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm2, %xmm0
-; SSE4-NEXT: pextrb $8, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $4, %al
; SSE4-NEXT: je .LBB4_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
+; SSE4-NEXT: .LBB4_5: # %cond.store3
; SSE4-NEXT: pextrw $4, %xmm5, 4(%rdi)
-; SSE4-NEXT: .LBB4_6: # %else4
-; SSE4-NEXT: pextrb $12, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $8, %al
; SSE4-NEXT: je .LBB4_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
+; SSE4-NEXT: .LBB4_7: # %cond.store5
; SSE4-NEXT: pextrw $6, %xmm5, 6(%rdi)
-; SSE4-NEXT: .LBB4_8: # %else6
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v4i64_v4i16:
; AVX1: # %bb.0:
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [9223372036854775808,9223372036854775808]
; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm4
; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372036854841343,9223372036854841343]
@@ -1703,43 +1631,43 @@ define void @truncstore_v4i64_v4i16(<4 x
; 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: vpcmpeqd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT: vmovmskps %xmm1, %eax
+; AVX1-NEXT: xorl $15, %eax
; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB4_2
-; AVX1-NEXT: # %bb.1: # %cond.store
+; AVX1-NEXT: jne .LBB4_1
+; AVX1-NEXT: # %bb.2: # %else
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: jne .LBB4_3
+; AVX1-NEXT: .LBB4_4: # %else2
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: jne .LBB4_5
+; AVX1-NEXT: .LBB4_6: # %else4
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: jne .LBB4_7
+; AVX1-NEXT: .LBB4_8: # %else6
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+; AVX1-NEXT: .LBB4_1: # %cond.store
; AVX1-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX1-NEXT: .LBB4_2: # %else
-; AVX1-NEXT: vpextrb $4, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $2, %al
; AVX1-NEXT: je .LBB4_4
-; AVX1-NEXT: # %bb.3: # %cond.store1
+; AVX1-NEXT: .LBB4_3: # %cond.store1
; AVX1-NEXT: vpextrw $2, %xmm0, 2(%rdi)
-; AVX1-NEXT: .LBB4_4: # %else2
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $8, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $4, %al
; AVX1-NEXT: je .LBB4_6
-; AVX1-NEXT: # %bb.5: # %cond.store3
+; AVX1-NEXT: .LBB4_5: # %cond.store3
; AVX1-NEXT: vpextrw $4, %xmm0, 4(%rdi)
-; AVX1-NEXT: .LBB4_6: # %else4
-; AVX1-NEXT: vpextrb $12, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $8, %al
; AVX1-NEXT: je .LBB4_8
-; AVX1-NEXT: # %bb.7: # %cond.store5
+; AVX1-NEXT: .LBB4_7: # %cond.store5
; AVX1-NEXT: vpextrw $6, %xmm0, 6(%rdi)
-; AVX1-NEXT: .LBB4_8: # %else6
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: truncstore_v4i64_v4i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vbroadcastsd {{.*#+}} ymm3 = [65535,65535,65535,65535]
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm4 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
; AVX2-NEXT: vpxor %ymm4, %ymm0, %ymm4
@@ -1748,34 +1676,37 @@ define void @truncstore_v4i64_v4i16(<4 x
; AVX2-NEXT: vblendvpd %ymm4, %ymm0, %ymm3, %ymm0
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
-; AVX2-NEXT: vpextrb $0, %xmm2, %eax
+; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
+; AVX2-NEXT: vmovmskps %xmm1, %eax
+; AVX2-NEXT: xorl $15, %eax
; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB4_2
-; AVX2-NEXT: # %bb.1: # %cond.store
+; AVX2-NEXT: jne .LBB4_1
+; AVX2-NEXT: # %bb.2: # %else
+; AVX2-NEXT: testb $2, %al
+; AVX2-NEXT: jne .LBB4_3
+; AVX2-NEXT: .LBB4_4: # %else2
+; AVX2-NEXT: testb $4, %al
+; AVX2-NEXT: jne .LBB4_5
+; AVX2-NEXT: .LBB4_6: # %else4
+; AVX2-NEXT: testb $8, %al
+; AVX2-NEXT: jne .LBB4_7
+; AVX2-NEXT: .LBB4_8: # %else6
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+; AVX2-NEXT: .LBB4_1: # %cond.store
; AVX2-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX2-NEXT: .LBB4_2: # %else
-; AVX2-NEXT: vpextrb $4, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $2, %al
; AVX2-NEXT: je .LBB4_4
-; AVX2-NEXT: # %bb.3: # %cond.store1
+; AVX2-NEXT: .LBB4_3: # %cond.store1
; AVX2-NEXT: vpextrw $2, %xmm0, 2(%rdi)
-; AVX2-NEXT: .LBB4_4: # %else2
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $8, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $4, %al
; AVX2-NEXT: je .LBB4_6
-; AVX2-NEXT: # %bb.5: # %cond.store3
+; AVX2-NEXT: .LBB4_5: # %cond.store3
; AVX2-NEXT: vpextrw $4, %xmm0, 4(%rdi)
-; AVX2-NEXT: .LBB4_6: # %else4
-; AVX2-NEXT: vpextrb $12, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $8, %al
; AVX2-NEXT: je .LBB4_8
-; AVX2-NEXT: # %bb.7: # %cond.store5
+; AVX2-NEXT: .LBB4_7: # %cond.store5
; AVX2-NEXT: vpextrw $6, %xmm0, 6(%rdi)
-; AVX2-NEXT: .LBB4_8: # %else6
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
@@ -1784,39 +1715,38 @@ define void @truncstore_v4i64_v4i16(<4 x
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm2 = [65535,65535,65535,65535]
-; AVX512F-NEXT: vpminuq %zmm2, %zmm0, %zmm0
+; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm1 = [65535,65535,65535,65535]
+; AVX512F-NEXT: vpminuq %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: vpmovqd %zmm0, %ymm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB4_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB4_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB4_3
+; AVX512F-NEXT: .LBB4_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB4_5
+; AVX512F-NEXT: .LBB4_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB4_7
+; AVX512F-NEXT: .LBB4_8: # %else6
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB4_1: # %cond.store
; AVX512F-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB4_2: # %else
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB4_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB4_3: # %cond.store1
; AVX512F-NEXT: vpextrw $2, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB4_4: # %else2
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $2, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB4_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB4_5: # %cond.store3
; AVX512F-NEXT: vpextrw $4, %xmm0, 4(%rdi)
-; AVX512F-NEXT: .LBB4_6: # %else4
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $3, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB4_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB4_7: # %cond.store5
; AVX512F-NEXT: vpextrw $6, %xmm0, 6(%rdi)
-; AVX512F-NEXT: .LBB4_8: # %else6
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -1854,7 +1784,6 @@ define void @truncstore_v4i64_v4i8(<4 x
; SSE2-LABEL: truncstore_v4i64_v4i8:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm3
; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [255,255]
; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [9223372039002259456,9223372039002259456]
; SSE2-NEXT: movdqa %xmm1, %xmm6
@@ -1884,51 +1813,46 @@ define void @truncstore_v4i64_v4i8(<4 x
; SSE2-NEXT: pandn %xmm8, %xmm1
; SSE2-NEXT: por %xmm0, %xmm1
; SSE2-NEXT: packuswb %xmm4, %xmm1
-; SSE2-NEXT: movd %xmm3, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB5_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movd %xmm1, %eax
-; SSE2-NEXT: movb %al, (%rdi)
-; SSE2-NEXT: .LBB5_2: # %else
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm3
-; SSE2-NEXT: pextrw $2, %xmm3, %eax
+; SSE2-NEXT: pcmpeqd %xmm2, %xmm3
+; SSE2-NEXT: movmskps %xmm3, %eax
+; SSE2-NEXT: xorl $15, %eax
; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB5_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: pextrw $2, %xmm1, %eax
-; SSE2-NEXT: movb %al, 1(%rdi)
+; SSE2-NEXT: jne .LBB5_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB5_3
; SSE2-NEXT: .LBB5_4: # %else2
-; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
-; SSE2-NEXT: pxor %xmm2, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB5_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: pextrw $4, %xmm1, %eax
-; SSE2-NEXT: movb %al, 2(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB5_5
; SSE2-NEXT: .LBB5_6: # %else4
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm2
-; SSE2-NEXT: pextrw $6, %xmm2, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB5_7
+; SSE2-NEXT: .LBB5_8: # %else6
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB5_1: # %cond.store
+; SSE2-NEXT: movd %xmm1, %ecx
+; SSE2-NEXT: movb %cl, (%rdi)
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: je .LBB5_4
+; SSE2-NEXT: .LBB5_3: # %cond.store1
+; SSE2-NEXT: pextrw $2, %xmm1, %ecx
+; SSE2-NEXT: movb %cl, 1(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: je .LBB5_6
+; SSE2-NEXT: .LBB5_5: # %cond.store3
+; SSE2-NEXT: pextrw $4, %xmm1, %ecx
+; SSE2-NEXT: movb %cl, 2(%rdi)
+; SSE2-NEXT: testb $8, %al
; SSE2-NEXT: je .LBB5_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
+; SSE2-NEXT: .LBB5_7: # %cond.store5
; SSE2-NEXT: pextrw $6, %xmm1, %eax
; SSE2-NEXT: movb %al, 3(%rdi)
-; SSE2-NEXT: .LBB5_8: # %else6
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v4i64_v4i8:
; SSE4: # %bb.0:
; SSE4-NEXT: movdqa %xmm0, %xmm8
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm6, %xmm6
-; SSE4-NEXT: pxor %xmm0, %xmm6
+; SSE4-NEXT: pxor %xmm6, %xmm6
; SSE4-NEXT: movapd {{.*#+}} xmm5 = [255,255]
; SSE4-NEXT: movdqa {{.*#+}} xmm7 = [9223372036854775808,9223372036854775808]
; SSE4-NEXT: movdqa %xmm1, %xmm3
@@ -1943,42 +1867,41 @@ define void @truncstore_v4i64_v4i8(<4 x
; SSE4-NEXT: movdqa %xmm4, %xmm0
; SSE4-NEXT: blendvpd %xmm0, %xmm8, %xmm5
; SSE4-NEXT: packusdw %xmm3, %xmm5
-; SSE4-NEXT: pextrb $0, %xmm6, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB5_2
-; SSE4-NEXT: # %bb.1: # %cond.store
+; SSE4-NEXT: pcmpeqd %xmm2, %xmm6
+; SSE4-NEXT: movmskps %xmm6, %eax
+; SSE4-NEXT: xorl $15, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB5_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB5_3
+; SSE4-NEXT: .LBB5_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB5_5
+; SSE4-NEXT: .LBB5_6: # %else4
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB5_7
+; SSE4-NEXT: .LBB5_8: # %else6
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB5_1: # %cond.store
; SSE4-NEXT: pextrb $0, %xmm5, (%rdi)
-; SSE4-NEXT: .LBB5_2: # %else
-; SSE4-NEXT: pextrb $4, %xmm6, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $2, %al
; SSE4-NEXT: je .LBB5_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
+; SSE4-NEXT: .LBB5_3: # %cond.store1
; SSE4-NEXT: pextrb $4, %xmm5, 1(%rdi)
-; SSE4-NEXT: .LBB5_4: # %else2
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm2, %xmm0
-; SSE4-NEXT: pextrb $8, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $4, %al
; SSE4-NEXT: je .LBB5_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
+; SSE4-NEXT: .LBB5_5: # %cond.store3
; SSE4-NEXT: pextrb $8, %xmm5, 2(%rdi)
-; SSE4-NEXT: .LBB5_6: # %else4
-; SSE4-NEXT: pextrb $12, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $8, %al
; SSE4-NEXT: je .LBB5_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
+; SSE4-NEXT: .LBB5_7: # %cond.store5
; SSE4-NEXT: pextrb $12, %xmm5, 3(%rdi)
-; SSE4-NEXT: .LBB5_8: # %else6
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v4i64_v4i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [9223372036854775808,9223372036854775808]
; AVX1-NEXT: vpxor %xmm3, %xmm0, %xmm4
; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372036854776063,9223372036854776063]
@@ -1990,43 +1913,43 @@ define void @truncstore_v4i64_v4i8(<4 x
; 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: vpcmpeqd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT: vmovmskps %xmm1, %eax
+; AVX1-NEXT: xorl $15, %eax
; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB5_2
-; AVX1-NEXT: # %bb.1: # %cond.store
+; AVX1-NEXT: jne .LBB5_1
+; AVX1-NEXT: # %bb.2: # %else
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: jne .LBB5_3
+; AVX1-NEXT: .LBB5_4: # %else2
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: jne .LBB5_5
+; AVX1-NEXT: .LBB5_6: # %else4
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: jne .LBB5_7
+; AVX1-NEXT: .LBB5_8: # %else6
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+; AVX1-NEXT: .LBB5_1: # %cond.store
; AVX1-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX1-NEXT: .LBB5_2: # %else
-; AVX1-NEXT: vpextrb $4, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $2, %al
; AVX1-NEXT: je .LBB5_4
-; AVX1-NEXT: # %bb.3: # %cond.store1
+; AVX1-NEXT: .LBB5_3: # %cond.store1
; AVX1-NEXT: vpextrb $4, %xmm0, 1(%rdi)
-; AVX1-NEXT: .LBB5_4: # %else2
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $8, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $4, %al
; AVX1-NEXT: je .LBB5_6
-; AVX1-NEXT: # %bb.5: # %cond.store3
+; AVX1-NEXT: .LBB5_5: # %cond.store3
; AVX1-NEXT: vpextrb $8, %xmm0, 2(%rdi)
-; AVX1-NEXT: .LBB5_6: # %else4
-; AVX1-NEXT: vpextrb $12, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $8, %al
; AVX1-NEXT: je .LBB5_8
-; AVX1-NEXT: # %bb.7: # %cond.store5
+; AVX1-NEXT: .LBB5_7: # %cond.store5
; AVX1-NEXT: vpextrb $12, %xmm0, 3(%rdi)
-; AVX1-NEXT: .LBB5_8: # %else6
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: truncstore_v4i64_v4i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vbroadcastsd {{.*#+}} ymm3 = [255,255,255,255]
; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm4 = [9223372036854775808,9223372036854775808,9223372036854775808,9223372036854775808]
; AVX2-NEXT: vpxor %ymm4, %ymm0, %ymm4
@@ -2035,34 +1958,37 @@ define void @truncstore_v4i64_v4i8(<4 x
; AVX2-NEXT: vblendvpd %ymm4, %ymm0, %ymm3, %ymm0
; AVX2-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX2-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
-; AVX2-NEXT: vpextrb $0, %xmm2, %eax
+; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
+; AVX2-NEXT: vmovmskps %xmm1, %eax
+; AVX2-NEXT: xorl $15, %eax
; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB5_2
-; AVX2-NEXT: # %bb.1: # %cond.store
+; AVX2-NEXT: jne .LBB5_1
+; AVX2-NEXT: # %bb.2: # %else
+; AVX2-NEXT: testb $2, %al
+; AVX2-NEXT: jne .LBB5_3
+; AVX2-NEXT: .LBB5_4: # %else2
+; AVX2-NEXT: testb $4, %al
+; AVX2-NEXT: jne .LBB5_5
+; AVX2-NEXT: .LBB5_6: # %else4
+; AVX2-NEXT: testb $8, %al
+; AVX2-NEXT: jne .LBB5_7
+; AVX2-NEXT: .LBB5_8: # %else6
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+; AVX2-NEXT: .LBB5_1: # %cond.store
; AVX2-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX2-NEXT: .LBB5_2: # %else
-; AVX2-NEXT: vpextrb $4, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $2, %al
; AVX2-NEXT: je .LBB5_4
-; AVX2-NEXT: # %bb.3: # %cond.store1
+; AVX2-NEXT: .LBB5_3: # %cond.store1
; AVX2-NEXT: vpextrb $4, %xmm0, 1(%rdi)
-; AVX2-NEXT: .LBB5_4: # %else2
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $8, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $4, %al
; AVX2-NEXT: je .LBB5_6
-; AVX2-NEXT: # %bb.5: # %cond.store3
+; AVX2-NEXT: .LBB5_5: # %cond.store3
; AVX2-NEXT: vpextrb $8, %xmm0, 2(%rdi)
-; AVX2-NEXT: .LBB5_6: # %else4
-; AVX2-NEXT: vpextrb $12, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $8, %al
; AVX2-NEXT: je .LBB5_8
-; AVX2-NEXT: # %bb.7: # %cond.store5
+; AVX2-NEXT: .LBB5_7: # %cond.store5
; AVX2-NEXT: vpextrb $12, %xmm0, 3(%rdi)
-; AVX2-NEXT: .LBB5_8: # %else6
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
@@ -2071,39 +1997,38 @@ define void @truncstore_v4i64_v4i8(<4 x
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $ymm0 killed $ymm0 def $zmm0
; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm2 = [255,255,255,255]
-; AVX512F-NEXT: vpminuq %zmm2, %zmm0, %zmm0
+; AVX512F-NEXT: vpbroadcastq {{.*#+}} ymm1 = [255,255,255,255]
+; AVX512F-NEXT: vpminuq %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: vpmovqd %zmm0, %ymm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB5_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB5_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB5_3
+; AVX512F-NEXT: .LBB5_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB5_5
+; AVX512F-NEXT: .LBB5_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB5_7
+; AVX512F-NEXT: .LBB5_8: # %else6
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB5_1: # %cond.store
; AVX512F-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB5_2: # %else
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB5_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB5_3: # %cond.store1
; AVX512F-NEXT: vpextrb $4, %xmm0, 1(%rdi)
-; AVX512F-NEXT: .LBB5_4: # %else2
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $2, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB5_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB5_5: # %cond.store3
; AVX512F-NEXT: vpextrb $8, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB5_6: # %else4
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $3, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB5_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB5_7: # %cond.store5
; AVX512F-NEXT: vpextrb $12, %xmm0, 3(%rdi)
-; AVX512F-NEXT: .LBB5_8: # %else6
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -2140,68 +2065,69 @@ define void @truncstore_v4i64_v4i8(<4 x
define void @truncstore_v2i64_v2i32(<2 x i64> %x, <2 x i32>* %p, <2 x i64> %mask) {
; SSE2-LABEL: truncstore_v2i64_v2i32:
; SSE2: # %bb.0:
-; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,0,3,2]
-; SSE2-NEXT: pand %xmm2, %xmm1
+; SSE2-NEXT: pxor %xmm3, %xmm3
; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
; SSE2-NEXT: pxor %xmm0, %xmm2
-; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [9223372039002259455,9223372039002259455]
-; SSE2-NEXT: movdqa %xmm3, %xmm4
-; SSE2-NEXT: pcmpgtd %xmm2, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; SSE2-NEXT: pand %xmm5, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
-; SSE2-NEXT: por %xmm3, %xmm2
+; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [9223372039002259455,9223372039002259455]
+; SSE2-NEXT: movdqa %xmm4, %xmm5
+; SSE2-NEXT: pcmpgtd %xmm2, %xmm5
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
+; SSE2-NEXT: pcmpeqd %xmm4, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
+; SSE2-NEXT: pand %xmm6, %xmm4
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm5[1,1,3,3]
+; SSE2-NEXT: por %xmm4, %xmm2
; SSE2-NEXT: pand %xmm2, %xmm0
; SSE2-NEXT: pandn {{.*}}(%rip), %xmm2
; SSE2-NEXT: por %xmm0, %xmm2
-; SSE2-NEXT: movd %xmm1, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB6_2
-; SSE2-NEXT: # %bb.1: # %cond.store
+; SSE2-NEXT: pcmpeqd %xmm1, %xmm3
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,0,3,2]
+; SSE2-NEXT: pand %xmm3, %xmm0
+; SSE2-NEXT: movmskpd %xmm0, %eax
+; SSE2-NEXT: xorl $3, %eax
+; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: jne .LBB6_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB6_3
+; SSE2-NEXT: .LBB6_4: # %else2
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB6_1: # %cond.store
; SSE2-NEXT: movd %xmm2, (%rdi)
-; SSE2-NEXT: .LBB6_2: # %else
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm1
-; SSE2-NEXT: pextrw $4, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $2, %al
; SSE2-NEXT: je .LBB6_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
+; SSE2-NEXT: .LBB6_3: # %cond.store1
; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm2[2,3,0,1]
; SSE2-NEXT: movd %xmm0, 4(%rdi)
-; SSE2-NEXT: .LBB6_4: # %else2
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v2i64_v2i32:
; SSE4: # %bb.0:
; SSE4-NEXT: movdqa %xmm0, %xmm2
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqq %xmm1, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm3, %xmm3
-; SSE4-NEXT: pxor %xmm0, %xmm3
-; SSE4-NEXT: movapd {{.*#+}} xmm1 = [4294967295,4294967295]
-; SSE4-NEXT: movdqa {{.*#+}} xmm4 = [9223372036854775808,9223372036854775808]
-; SSE4-NEXT: pxor %xmm2, %xmm4
+; SSE4-NEXT: pxor %xmm4, %xmm4
+; SSE4-NEXT: movapd {{.*#+}} xmm3 = [4294967295,4294967295]
+; SSE4-NEXT: movdqa {{.*#+}} xmm5 = [9223372036854775808,9223372036854775808]
+; SSE4-NEXT: pxor %xmm0, %xmm5
; SSE4-NEXT: movdqa {{.*#+}} xmm0 = [9223372041149743103,9223372041149743103]
-; SSE4-NEXT: pcmpgtq %xmm4, %xmm0
-; SSE4-NEXT: blendvpd %xmm0, %xmm2, %xmm1
-; SSE4-NEXT: pextrb $0, %xmm3, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB6_2
-; SSE4-NEXT: # %bb.1: # %cond.store
-; SSE4-NEXT: movss %xmm1, (%rdi)
-; SSE4-NEXT: .LBB6_2: # %else
-; SSE4-NEXT: pextrb $8, %xmm3, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB6_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
-; SSE4-NEXT: extractps $2, %xmm1, 4(%rdi)
+; SSE4-NEXT: pcmpgtq %xmm5, %xmm0
+; SSE4-NEXT: blendvpd %xmm0, %xmm2, %xmm3
+; SSE4-NEXT: pcmpeqq %xmm1, %xmm4
+; SSE4-NEXT: movmskpd %xmm4, %eax
+; SSE4-NEXT: xorl $3, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB6_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB6_3
; SSE4-NEXT: .LBB6_4: # %else2
; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB6_1: # %cond.store
+; SSE4-NEXT: movss %xmm3, (%rdi)
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: je .LBB6_4
+; SSE4-NEXT: .LBB6_3: # %cond.store1
+; SSE4-NEXT: extractps $2, %xmm3, 4(%rdi)
+; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v2i64_v2i32:
; AVX1: # %bb.0:
@@ -2280,93 +2206,95 @@ define void @truncstore_v2i64_v2i32(<2 x
define void @truncstore_v2i64_v2i16(<2 x i64> %x, <2 x i16>* %p, <2 x i64> %mask) {
; SSE2-LABEL: truncstore_v2i64_v2i16:
; SSE2: # %bb.0:
-; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,0,3,2]
-; SSE2-NEXT: pand %xmm2, %xmm1
+; SSE2-NEXT: pxor %xmm3, %xmm3
; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
; SSE2-NEXT: pxor %xmm0, %xmm2
-; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [9223372039002324991,9223372039002324991]
-; SSE2-NEXT: movdqa %xmm3, %xmm4
-; SSE2-NEXT: pcmpgtd %xmm2, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; SSE2-NEXT: pand %xmm5, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
-; SSE2-NEXT: por %xmm3, %xmm2
+; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [9223372039002324991,9223372039002324991]
+; SSE2-NEXT: movdqa %xmm4, %xmm5
+; SSE2-NEXT: pcmpgtd %xmm2, %xmm5
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
+; SSE2-NEXT: pcmpeqd %xmm4, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
+; SSE2-NEXT: pand %xmm6, %xmm4
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm5[1,1,3,3]
+; SSE2-NEXT: por %xmm4, %xmm2
; SSE2-NEXT: pand %xmm2, %xmm0
; SSE2-NEXT: pandn {{.*}}(%rip), %xmm2
; SSE2-NEXT: por %xmm0, %xmm2
-; SSE2-NEXT: movd %xmm1, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB7_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movd %xmm2, %eax
-; SSE2-NEXT: movw %ax, (%rdi)
-; SSE2-NEXT: .LBB7_2: # %else
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm1
-; SSE2-NEXT: pextrw $4, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: pcmpeqd %xmm1, %xmm3
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,0,3,2]
+; SSE2-NEXT: pand %xmm3, %xmm0
+; SSE2-NEXT: movmskpd %xmm0, %eax
+; SSE2-NEXT: xorl $3, %eax
+; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: jne .LBB7_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB7_3
+; SSE2-NEXT: .LBB7_4: # %else2
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB7_1: # %cond.store
+; SSE2-NEXT: movd %xmm2, %ecx
+; SSE2-NEXT: movw %cx, (%rdi)
+; SSE2-NEXT: testb $2, %al
; SSE2-NEXT: je .LBB7_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
+; SSE2-NEXT: .LBB7_3: # %cond.store1
; SSE2-NEXT: pextrw $4, %xmm2, %eax
; SSE2-NEXT: movw %ax, 2(%rdi)
-; SSE2-NEXT: .LBB7_4: # %else2
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v2i64_v2i16:
; SSE4: # %bb.0:
; SSE4-NEXT: movdqa %xmm0, %xmm2
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqq %xmm1, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm3, %xmm3
-; SSE4-NEXT: pxor %xmm0, %xmm3
-; SSE4-NEXT: movapd {{.*#+}} xmm1 = [65535,65535]
-; SSE4-NEXT: movdqa {{.*#+}} xmm4 = [9223372036854775808,9223372036854775808]
-; SSE4-NEXT: pxor %xmm2, %xmm4
+; SSE4-NEXT: pxor %xmm4, %xmm4
+; SSE4-NEXT: movapd {{.*#+}} xmm3 = [65535,65535]
+; SSE4-NEXT: movdqa {{.*#+}} xmm5 = [9223372036854775808,9223372036854775808]
+; SSE4-NEXT: pxor %xmm0, %xmm5
; SSE4-NEXT: movdqa {{.*#+}} xmm0 = [9223372036854841343,9223372036854841343]
-; SSE4-NEXT: pcmpgtq %xmm4, %xmm0
-; SSE4-NEXT: blendvpd %xmm0, %xmm2, %xmm1
-; SSE4-NEXT: pextrb $0, %xmm3, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB7_2
-; SSE4-NEXT: # %bb.1: # %cond.store
-; SSE4-NEXT: pextrw $0, %xmm1, (%rdi)
-; SSE4-NEXT: .LBB7_2: # %else
-; SSE4-NEXT: pextrb $8, %xmm3, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB7_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
-; SSE4-NEXT: pextrw $4, %xmm1, 2(%rdi)
+; SSE4-NEXT: pcmpgtq %xmm5, %xmm0
+; SSE4-NEXT: blendvpd %xmm0, %xmm2, %xmm3
+; SSE4-NEXT: pcmpeqq %xmm1, %xmm4
+; SSE4-NEXT: movmskpd %xmm4, %eax
+; SSE4-NEXT: xorl $3, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB7_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB7_3
; SSE4-NEXT: .LBB7_4: # %else2
; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB7_1: # %cond.store
+; SSE4-NEXT: pextrw $0, %xmm3, (%rdi)
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: je .LBB7_4
+; SSE4-NEXT: .LBB7_3: # %cond.store1
+; SSE4-NEXT: pextrw $4, %xmm3, 2(%rdi)
+; SSE4-NEXT: retq
;
; AVX-LABEL: truncstore_v2i64_v2i16:
; AVX: # %bb.0:
; AVX-NEXT: vpxor %xmm2, %xmm2, %xmm2
+; AVX-NEXT: vmovapd {{.*#+}} xmm3 = [65535,65535]
+; AVX-NEXT: vpxor {{.*}}(%rip), %xmm0, %xmm4
+; AVX-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372036854841343,9223372036854841343]
+; AVX-NEXT: vpcmpgtq %xmm4, %xmm5, %xmm4
+; AVX-NEXT: vblendvpd %xmm4, %xmm0, %xmm3, %xmm0
; AVX-NEXT: vpcmpeqq %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vmovapd {{.*#+}} xmm2 = [65535,65535]
-; AVX-NEXT: vpxor {{.*}}(%rip), %xmm0, %xmm3
-; AVX-NEXT: vmovdqa {{.*#+}} xmm4 = [9223372036854841343,9223372036854841343]
-; AVX-NEXT: vpcmpgtq %xmm3, %xmm4, %xmm3
-; AVX-NEXT: vblendvpd %xmm3, %xmm0, %xmm2, %xmm0
-; AVX-NEXT: vpextrb $0, %xmm1, %eax
+; AVX-NEXT: vmovmskpd %xmm1, %eax
+; AVX-NEXT: xorl $3, %eax
; AVX-NEXT: testb $1, %al
-; AVX-NEXT: je .LBB7_2
-; AVX-NEXT: # %bb.1: # %cond.store
+; AVX-NEXT: jne .LBB7_1
+; AVX-NEXT: # %bb.2: # %else
+; AVX-NEXT: testb $2, %al
+; AVX-NEXT: jne .LBB7_3
+; AVX-NEXT: .LBB7_4: # %else2
+; AVX-NEXT: retq
+; AVX-NEXT: .LBB7_1: # %cond.store
; AVX-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX-NEXT: .LBB7_2: # %else
-; AVX-NEXT: vpextrb $8, %xmm1, %eax
-; AVX-NEXT: testb $1, %al
+; AVX-NEXT: testb $2, %al
; AVX-NEXT: je .LBB7_4
-; AVX-NEXT: # %bb.3: # %cond.store1
+; AVX-NEXT: .LBB7_3: # %cond.store1
; AVX-NEXT: vpextrw $4, %xmm0, 2(%rdi)
-; AVX-NEXT: .LBB7_4: # %else2
; AVX-NEXT: retq
;
; AVX512F-LABEL: truncstore_v2i64_v2i16:
@@ -2374,22 +2302,23 @@ define void @truncstore_v2i64_v2i16(<2 x
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vptestmq %zmm1, %zmm1, %k0
-; AVX512F-NEXT: vmovdqa {{.*#+}} xmm2 = [65535,65535]
-; AVX512F-NEXT: vpminuq %zmm2, %zmm0, %zmm0
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm1 = [65535,65535]
+; AVX512F-NEXT: vpminuq %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB7_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB7_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB7_3
+; AVX512F-NEXT: .LBB7_4: # %else2
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB7_1: # %cond.store
; AVX512F-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB7_2: # %else
-; AVX512F-NEXT: vptestmq %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB7_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB7_3: # %cond.store1
; AVX512F-NEXT: vpextrw $4, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB7_4: # %else2
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -2425,93 +2354,95 @@ define void @truncstore_v2i64_v2i16(<2 x
define void @truncstore_v2i64_v2i8(<2 x i64> %x, <2 x i8>* %p, <2 x i64> %mask) {
; SSE2-LABEL: truncstore_v2i64_v2i8:
; SSE2: # %bb.0:
-; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm2[1,0,3,2]
-; SSE2-NEXT: pand %xmm2, %xmm1
+; SSE2-NEXT: pxor %xmm3, %xmm3
; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [9223372039002259456,9223372039002259456]
; SSE2-NEXT: pxor %xmm0, %xmm2
-; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [9223372039002259711,9223372039002259711]
-; SSE2-NEXT: movdqa %xmm3, %xmm4
-; SSE2-NEXT: pcmpgtd %xmm2, %xmm4
-; SSE2-NEXT: pshufd {{.*#+}} xmm5 = xmm4[0,0,2,2]
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm2
-; SSE2-NEXT: pshufd {{.*#+}} xmm3 = xmm2[1,1,3,3]
-; SSE2-NEXT: pand %xmm5, %xmm3
-; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm4[1,1,3,3]
-; SSE2-NEXT: por %xmm3, %xmm2
+; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [9223372039002259711,9223372039002259711]
+; SSE2-NEXT: movdqa %xmm4, %xmm5
+; SSE2-NEXT: pcmpgtd %xmm2, %xmm5
+; SSE2-NEXT: pshufd {{.*#+}} xmm6 = xmm5[0,0,2,2]
+; SSE2-NEXT: pcmpeqd %xmm4, %xmm2
+; SSE2-NEXT: pshufd {{.*#+}} xmm4 = xmm2[1,1,3,3]
+; SSE2-NEXT: pand %xmm6, %xmm4
+; SSE2-NEXT: pshufd {{.*#+}} xmm2 = xmm5[1,1,3,3]
+; SSE2-NEXT: por %xmm4, %xmm2
; SSE2-NEXT: pand %xmm2, %xmm0
; SSE2-NEXT: pandn {{.*}}(%rip), %xmm2
; SSE2-NEXT: por %xmm0, %xmm2
-; SSE2-NEXT: movd %xmm1, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB8_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movd %xmm2, %eax
-; SSE2-NEXT: movb %al, (%rdi)
-; SSE2-NEXT: .LBB8_2: # %else
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm1
-; SSE2-NEXT: pextrw $4, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: pcmpeqd %xmm1, %xmm3
+; SSE2-NEXT: pshufd {{.*#+}} xmm0 = xmm3[1,0,3,2]
+; SSE2-NEXT: pand %xmm3, %xmm0
+; SSE2-NEXT: movmskpd %xmm0, %eax
+; SSE2-NEXT: xorl $3, %eax
+; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: jne .LBB8_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB8_3
+; SSE2-NEXT: .LBB8_4: # %else2
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB8_1: # %cond.store
+; SSE2-NEXT: movd %xmm2, %ecx
+; SSE2-NEXT: movb %cl, (%rdi)
+; SSE2-NEXT: testb $2, %al
; SSE2-NEXT: je .LBB8_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
+; SSE2-NEXT: .LBB8_3: # %cond.store1
; SSE2-NEXT: pextrw $4, %xmm2, %eax
; SSE2-NEXT: movb %al, 1(%rdi)
-; SSE2-NEXT: .LBB8_4: # %else2
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v2i64_v2i8:
; SSE4: # %bb.0:
; SSE4-NEXT: movdqa %xmm0, %xmm2
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqq %xmm1, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm3, %xmm3
-; SSE4-NEXT: pxor %xmm0, %xmm3
-; SSE4-NEXT: movapd {{.*#+}} xmm1 = [255,255]
-; SSE4-NEXT: movdqa {{.*#+}} xmm4 = [9223372036854775808,9223372036854775808]
-; SSE4-NEXT: pxor %xmm2, %xmm4
+; SSE4-NEXT: pxor %xmm4, %xmm4
+; SSE4-NEXT: movapd {{.*#+}} xmm3 = [255,255]
+; SSE4-NEXT: movdqa {{.*#+}} xmm5 = [9223372036854775808,9223372036854775808]
+; SSE4-NEXT: pxor %xmm0, %xmm5
; SSE4-NEXT: movdqa {{.*#+}} xmm0 = [9223372036854776063,9223372036854776063]
-; SSE4-NEXT: pcmpgtq %xmm4, %xmm0
-; SSE4-NEXT: blendvpd %xmm0, %xmm2, %xmm1
-; SSE4-NEXT: pextrb $0, %xmm3, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB8_2
-; SSE4-NEXT: # %bb.1: # %cond.store
-; SSE4-NEXT: pextrb $0, %xmm1, (%rdi)
-; SSE4-NEXT: .LBB8_2: # %else
-; SSE4-NEXT: pextrb $8, %xmm3, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB8_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
-; SSE4-NEXT: pextrb $8, %xmm1, 1(%rdi)
+; SSE4-NEXT: pcmpgtq %xmm5, %xmm0
+; SSE4-NEXT: blendvpd %xmm0, %xmm2, %xmm3
+; SSE4-NEXT: pcmpeqq %xmm1, %xmm4
+; SSE4-NEXT: movmskpd %xmm4, %eax
+; SSE4-NEXT: xorl $3, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB8_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB8_3
; SSE4-NEXT: .LBB8_4: # %else2
; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB8_1: # %cond.store
+; SSE4-NEXT: pextrb $0, %xmm3, (%rdi)
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: je .LBB8_4
+; SSE4-NEXT: .LBB8_3: # %cond.store1
+; SSE4-NEXT: pextrb $8, %xmm3, 1(%rdi)
+; SSE4-NEXT: retq
;
; AVX-LABEL: truncstore_v2i64_v2i8:
; AVX: # %bb.0:
; AVX-NEXT: vpxor %xmm2, %xmm2, %xmm2
+; AVX-NEXT: vmovapd {{.*#+}} xmm3 = [255,255]
+; AVX-NEXT: vpxor {{.*}}(%rip), %xmm0, %xmm4
+; AVX-NEXT: vmovdqa {{.*#+}} xmm5 = [9223372036854776063,9223372036854776063]
+; AVX-NEXT: vpcmpgtq %xmm4, %xmm5, %xmm4
+; AVX-NEXT: vblendvpd %xmm4, %xmm0, %xmm3, %xmm0
; AVX-NEXT: vpcmpeqq %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vmovapd {{.*#+}} xmm2 = [255,255]
-; AVX-NEXT: vpxor {{.*}}(%rip), %xmm0, %xmm3
-; AVX-NEXT: vmovdqa {{.*#+}} xmm4 = [9223372036854776063,9223372036854776063]
-; AVX-NEXT: vpcmpgtq %xmm3, %xmm4, %xmm3
-; AVX-NEXT: vblendvpd %xmm3, %xmm0, %xmm2, %xmm0
-; AVX-NEXT: vpextrb $0, %xmm1, %eax
+; AVX-NEXT: vmovmskpd %xmm1, %eax
+; AVX-NEXT: xorl $3, %eax
; AVX-NEXT: testb $1, %al
-; AVX-NEXT: je .LBB8_2
-; AVX-NEXT: # %bb.1: # %cond.store
+; AVX-NEXT: jne .LBB8_1
+; AVX-NEXT: # %bb.2: # %else
+; AVX-NEXT: testb $2, %al
+; AVX-NEXT: jne .LBB8_3
+; AVX-NEXT: .LBB8_4: # %else2
+; AVX-NEXT: retq
+; AVX-NEXT: .LBB8_1: # %cond.store
; AVX-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX-NEXT: .LBB8_2: # %else
-; AVX-NEXT: vpextrb $8, %xmm1, %eax
-; AVX-NEXT: testb $1, %al
+; AVX-NEXT: testb $2, %al
; AVX-NEXT: je .LBB8_4
-; AVX-NEXT: # %bb.3: # %cond.store1
+; AVX-NEXT: .LBB8_3: # %cond.store1
; AVX-NEXT: vpextrb $8, %xmm0, 1(%rdi)
-; AVX-NEXT: .LBB8_4: # %else2
; AVX-NEXT: retq
;
; AVX512F-LABEL: truncstore_v2i64_v2i8:
@@ -2519,22 +2450,23 @@ define void @truncstore_v2i64_v2i8(<2 x
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: # kill: def $xmm0 killed $xmm0 def $zmm0
; AVX512F-NEXT: vptestmq %zmm1, %zmm1, %k0
-; AVX512F-NEXT: vmovdqa {{.*#+}} xmm2 = [255,255]
-; AVX512F-NEXT: vpminuq %zmm2, %zmm0, %zmm0
+; AVX512F-NEXT: vmovdqa {{.*#+}} xmm1 = [255,255]
+; AVX512F-NEXT: vpminuq %zmm1, %zmm0, %zmm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB8_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB8_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB8_3
+; AVX512F-NEXT: .LBB8_4: # %else2
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB8_1: # %cond.store
; AVX512F-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB8_2: # %else
-; AVX512F-NEXT: vptestmq %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB8_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB8_3: # %cond.store1
; AVX512F-NEXT: vpextrb $8, %xmm0, 1(%rdi)
-; AVX512F-NEXT: .LBB8_4: # %else2
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -2570,14 +2502,13 @@ define void @truncstore_v16i32_v16i16(<1
; SSE2-LABEL: truncstore_v16i32_v16i16:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm12, %xmm12
-; SSE2-NEXT: pcmpeqd %xmm4, %xmm12
-; SSE2-NEXT: pcmpeqd %xmm9, %xmm9
; SSE2-NEXT: movdqa {{.*#+}} xmm11 = [2147483648,2147483648,2147483648,2147483648]
-; SSE2-NEXT: movdqa %xmm0, %xmm13
-; SSE2-NEXT: pxor %xmm11, %xmm13
+; SSE2-NEXT: movdqa %xmm0, %xmm9
+; SSE2-NEXT: pxor %xmm11, %xmm9
; SSE2-NEXT: movdqa {{.*#+}} xmm10 = [2147549183,2147549183,2147549183,2147549183]
; SSE2-NEXT: movdqa %xmm10, %xmm8
-; SSE2-NEXT: pcmpgtd %xmm13, %xmm8
+; SSE2-NEXT: pcmpgtd %xmm9, %xmm8
+; SSE2-NEXT: pcmpeqd %xmm9, %xmm9
; SSE2-NEXT: pand %xmm8, %xmm0
; SSE2-NEXT: pxor %xmm9, %xmm8
; SSE2-NEXT: por %xmm0, %xmm8
@@ -2593,676 +2524,554 @@ define void @truncstore_v16i32_v16i16(<1
; SSE2-NEXT: pslld $16, %xmm8
; SSE2-NEXT: psrad $16, %xmm8
; SSE2-NEXT: packssdw %xmm0, %xmm8
-; SSE2-NEXT: movd %xmm12, %eax
-; SSE2-NEXT: notl %eax
+; SSE2-NEXT: pcmpeqd %xmm12, %xmm7
+; SSE2-NEXT: pxor %xmm9, %xmm7
+; SSE2-NEXT: pcmpeqd %xmm12, %xmm6
+; SSE2-NEXT: pxor %xmm9, %xmm6
+; SSE2-NEXT: packssdw %xmm7, %xmm6
+; SSE2-NEXT: pcmpeqd %xmm12, %xmm5
+; SSE2-NEXT: pxor %xmm9, %xmm5
+; SSE2-NEXT: pcmpeqd %xmm12, %xmm4
+; SSE2-NEXT: pxor %xmm9, %xmm4
+; SSE2-NEXT: packssdw %xmm5, %xmm4
+; SSE2-NEXT: packsswb %xmm6, %xmm4
+; SSE2-NEXT: pmovmskb %xmm4, %eax
; SSE2-NEXT: testb $1, %al
; SSE2-NEXT: je .LBB9_2
; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movd %xmm8, %eax
-; SSE2-NEXT: movw %ax, (%rdi)
+; SSE2-NEXT: movd %xmm8, %ecx
+; SSE2-NEXT: movw %cx, (%rdi)
; SSE2-NEXT: .LBB9_2: # %else
-; SSE2-NEXT: pxor %xmm9, %xmm12
-; SSE2-NEXT: pextrw $2, %xmm12, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $2, %al
; SSE2-NEXT: je .LBB9_4
; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: pextrw $1, %xmm8, %eax
-; SSE2-NEXT: movw %ax, 2(%rdi)
+; SSE2-NEXT: pextrw $1, %xmm8, %ecx
+; SSE2-NEXT: movw %cx, 2(%rdi)
; SSE2-NEXT: .LBB9_4: # %else2
-; SSE2-NEXT: movdqa %xmm2, %xmm12
-; SSE2-NEXT: pxor %xmm11, %xmm12
+; SSE2-NEXT: movdqa %xmm2, %xmm1
+; SSE2-NEXT: pxor %xmm11, %xmm1
; SSE2-NEXT: pxor %xmm3, %xmm11
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm4
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE2-NEXT: pxor %xmm4, %xmm1
-; SSE2-NEXT: pextrw $4, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $4, %al
; SSE2-NEXT: je .LBB9_6
; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: pextrw $2, %xmm8, %eax
-; SSE2-NEXT: movw %ax, 4(%rdi)
+; SSE2-NEXT: pextrw $2, %xmm8, %ecx
+; SSE2-NEXT: movw %cx, 4(%rdi)
; SSE2-NEXT: .LBB9_6: # %else4
; SSE2-NEXT: movdqa %xmm10, %xmm0
-; SSE2-NEXT: pcmpgtd %xmm12, %xmm0
+; SSE2-NEXT: pcmpgtd %xmm1, %xmm0
; SSE2-NEXT: pcmpgtd %xmm11, %xmm10
-; SSE2-NEXT: pextrw $6, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $8, %al
; SSE2-NEXT: je .LBB9_8
; SSE2-NEXT: # %bb.7: # %cond.store5
-; SSE2-NEXT: pextrw $3, %xmm8, %eax
-; SSE2-NEXT: movw %ax, 6(%rdi)
+; SSE2-NEXT: pextrw $3, %xmm8, %ecx
+; SSE2-NEXT: movw %cx, 6(%rdi)
; SSE2-NEXT: .LBB9_8: # %else6
; SSE2-NEXT: pand %xmm0, %xmm2
; SSE2-NEXT: pxor %xmm9, %xmm0
; SSE2-NEXT: pxor %xmm10, %xmm9
; SSE2-NEXT: pand %xmm10, %xmm3
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm5, %xmm1
-; SSE2-NEXT: movd %xmm1, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $16, %al
; SSE2-NEXT: je .LBB9_10
; SSE2-NEXT: # %bb.9: # %cond.store7
-; SSE2-NEXT: pextrw $4, %xmm8, %eax
-; SSE2-NEXT: movw %ax, 8(%rdi)
+; SSE2-NEXT: pextrw $4, %xmm8, %ecx
+; SSE2-NEXT: movw %cx, 8(%rdi)
; SSE2-NEXT: .LBB9_10: # %else8
; SSE2-NEXT: por %xmm0, %xmm2
; SSE2-NEXT: por %xmm9, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm1
-; SSE2-NEXT: pextrw $2, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $32, %al
; SSE2-NEXT: je .LBB9_12
; SSE2-NEXT: # %bb.11: # %cond.store9
-; SSE2-NEXT: pextrw $5, %xmm8, %eax
-; SSE2-NEXT: movw %ax, 10(%rdi)
+; SSE2-NEXT: pextrw $5, %xmm8, %ecx
+; SSE2-NEXT: movw %cx, 10(%rdi)
; SSE2-NEXT: .LBB9_12: # %else10
; SSE2-NEXT: pslld $16, %xmm3
; SSE2-NEXT: pslld $16, %xmm2
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm5
-; SSE2-NEXT: pxor %xmm5, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $64, %al
; SSE2-NEXT: je .LBB9_14
; SSE2-NEXT: # %bb.13: # %cond.store11
-; SSE2-NEXT: pextrw $6, %xmm8, %eax
-; SSE2-NEXT: movw %ax, 12(%rdi)
+; SSE2-NEXT: pextrw $6, %xmm8, %ecx
+; SSE2-NEXT: movw %cx, 12(%rdi)
; SSE2-NEXT: .LBB9_14: # %else12
; SSE2-NEXT: psrad $16, %xmm3
; SSE2-NEXT: psrad $16, %xmm2
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm5
-; SSE2-NEXT: pextrw $6, %xmm5, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $-128, %al
; SSE2-NEXT: je .LBB9_16
; SSE2-NEXT: # %bb.15: # %cond.store13
-; SSE2-NEXT: pextrw $7, %xmm8, %eax
-; SSE2-NEXT: movw %ax, 14(%rdi)
+; SSE2-NEXT: pextrw $7, %xmm8, %ecx
+; SSE2-NEXT: movw %cx, 14(%rdi)
; SSE2-NEXT: .LBB9_16: # %else14
; SSE2-NEXT: packssdw %xmm3, %xmm2
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm6, %xmm0
-; SSE2-NEXT: movd %xmm0, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB9_18
-; SSE2-NEXT: # %bb.17: # %cond.store15
-; SSE2-NEXT: movd %xmm2, %eax
-; SSE2-NEXT: movw %ax, 16(%rdi)
-; SSE2-NEXT: .LBB9_18: # %else16
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE2-NEXT: pxor %xmm1, %xmm0
-; SSE2-NEXT: pextrw $2, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB9_20
-; SSE2-NEXT: # %bb.19: # %cond.store17
-; SSE2-NEXT: pextrw $1, %xmm2, %eax
-; SSE2-NEXT: movw %ax, 18(%rdi)
+; SSE2-NEXT: testl $256, %eax # imm = 0x100
+; SSE2-NEXT: jne .LBB9_17
+; SSE2-NEXT: # %bb.18: # %else16
+; SSE2-NEXT: testl $512, %eax # imm = 0x200
+; SSE2-NEXT: jne .LBB9_19
; SSE2-NEXT: .LBB9_20: # %else18
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm6
-; SSE2-NEXT: pxor %xmm6, %xmm1
-; SSE2-NEXT: pextrw $4, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB9_22
-; SSE2-NEXT: # %bb.21: # %cond.store19
-; SSE2-NEXT: pextrw $2, %xmm2, %eax
-; SSE2-NEXT: movw %ax, 20(%rdi)
+; SSE2-NEXT: testl $1024, %eax # imm = 0x400
+; SSE2-NEXT: jne .LBB9_21
; SSE2-NEXT: .LBB9_22: # %else20
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm6
-; SSE2-NEXT: pextrw $6, %xmm6, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB9_24
-; SSE2-NEXT: # %bb.23: # %cond.store21
-; SSE2-NEXT: pextrw $3, %xmm2, %eax
-; SSE2-NEXT: movw %ax, 22(%rdi)
+; SSE2-NEXT: testl $2048, %eax # imm = 0x800
+; SSE2-NEXT: jne .LBB9_23
; SSE2-NEXT: .LBB9_24: # %else22
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm7, %xmm0
-; SSE2-NEXT: movd %xmm0, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB9_26
-; SSE2-NEXT: # %bb.25: # %cond.store23
-; SSE2-NEXT: pextrw $4, %xmm2, %eax
-; SSE2-NEXT: movw %ax, 24(%rdi)
+; SSE2-NEXT: testl $4096, %eax # imm = 0x1000
+; SSE2-NEXT: jne .LBB9_25
; SSE2-NEXT: .LBB9_26: # %else24
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE2-NEXT: pxor %xmm1, %xmm0
-; SSE2-NEXT: pextrw $2, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB9_28
-; SSE2-NEXT: # %bb.27: # %cond.store25
-; SSE2-NEXT: pextrw $5, %xmm2, %eax
-; SSE2-NEXT: movw %ax, 26(%rdi)
+; SSE2-NEXT: testl $8192, %eax # imm = 0x2000
+; SSE2-NEXT: jne .LBB9_27
; SSE2-NEXT: .LBB9_28: # %else26
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm7
-; SSE2-NEXT: pxor %xmm7, %xmm1
-; SSE2-NEXT: pextrw $4, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB9_30
-; SSE2-NEXT: # %bb.29: # %cond.store27
-; SSE2-NEXT: pextrw $6, %xmm2, %eax
-; SSE2-NEXT: movw %ax, 28(%rdi)
+; SSE2-NEXT: testl $16384, %eax # imm = 0x4000
+; SSE2-NEXT: jne .LBB9_29
; SSE2-NEXT: .LBB9_30: # %else28
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm7
-; SSE2-NEXT: pextrw $6, %xmm7, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $32768, %eax # imm = 0x8000
+; SSE2-NEXT: jne .LBB9_31
+; SSE2-NEXT: .LBB9_32: # %else30
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB9_17: # %cond.store15
+; SSE2-NEXT: movd %xmm2, %ecx
+; SSE2-NEXT: movw %cx, 16(%rdi)
+; SSE2-NEXT: testl $512, %eax # imm = 0x200
+; SSE2-NEXT: je .LBB9_20
+; SSE2-NEXT: .LBB9_19: # %cond.store17
+; SSE2-NEXT: pextrw $1, %xmm2, %ecx
+; SSE2-NEXT: movw %cx, 18(%rdi)
+; SSE2-NEXT: testl $1024, %eax # imm = 0x400
+; SSE2-NEXT: je .LBB9_22
+; SSE2-NEXT: .LBB9_21: # %cond.store19
+; SSE2-NEXT: pextrw $2, %xmm2, %ecx
+; SSE2-NEXT: movw %cx, 20(%rdi)
+; SSE2-NEXT: testl $2048, %eax # imm = 0x800
+; SSE2-NEXT: je .LBB9_24
+; SSE2-NEXT: .LBB9_23: # %cond.store21
+; SSE2-NEXT: pextrw $3, %xmm2, %ecx
+; SSE2-NEXT: movw %cx, 22(%rdi)
+; SSE2-NEXT: testl $4096, %eax # imm = 0x1000
+; SSE2-NEXT: je .LBB9_26
+; SSE2-NEXT: .LBB9_25: # %cond.store23
+; SSE2-NEXT: pextrw $4, %xmm2, %ecx
+; SSE2-NEXT: movw %cx, 24(%rdi)
+; SSE2-NEXT: testl $8192, %eax # imm = 0x2000
+; SSE2-NEXT: je .LBB9_28
+; SSE2-NEXT: .LBB9_27: # %cond.store25
+; SSE2-NEXT: pextrw $5, %xmm2, %ecx
+; SSE2-NEXT: movw %cx, 26(%rdi)
+; SSE2-NEXT: testl $16384, %eax # imm = 0x4000
+; SSE2-NEXT: je .LBB9_30
+; SSE2-NEXT: .LBB9_29: # %cond.store27
+; SSE2-NEXT: pextrw $6, %xmm2, %ecx
+; SSE2-NEXT: movw %cx, 28(%rdi)
+; SSE2-NEXT: testl $32768, %eax # imm = 0x8000
; SSE2-NEXT: je .LBB9_32
-; SSE2-NEXT: # %bb.31: # %cond.store29
+; SSE2-NEXT: .LBB9_31: # %cond.store29
; SSE2-NEXT: pextrw $7, %xmm2, %eax
; SSE2-NEXT: movw %ax, 30(%rdi)
-; SSE2-NEXT: .LBB9_32: # %else30
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v16i32_v16i16:
; SSE4: # %bb.0:
-; SSE4-NEXT: pxor %xmm8, %xmm8
-; SSE4-NEXT: pcmpeqd %xmm4, %xmm8
-; SSE4-NEXT: pcmpeqd %xmm9, %xmm9
-; SSE4-NEXT: pxor %xmm8, %xmm9
+; SSE4-NEXT: pxor %xmm9, %xmm9
; SSE4-NEXT: movdqa {{.*#+}} xmm8 = [65535,65535,65535,65535]
; SSE4-NEXT: pminud %xmm8, %xmm1
; SSE4-NEXT: pminud %xmm8, %xmm0
; SSE4-NEXT: packusdw %xmm1, %xmm0
-; SSE4-NEXT: pextrb $0, %xmm9, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB9_2
-; SSE4-NEXT: # %bb.1: # %cond.store
-; SSE4-NEXT: pextrw $0, %xmm0, (%rdi)
-; SSE4-NEXT: .LBB9_2: # %else
-; SSE4-NEXT: pextrb $4, %xmm9, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB9_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
-; SSE4-NEXT: pextrw $1, %xmm0, 2(%rdi)
-; SSE4-NEXT: .LBB9_4: # %else2
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm4
+; SSE4-NEXT: pcmpeqd %xmm9, %xmm7
; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm4, %xmm1
-; SSE4-NEXT: pextrb $8, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB9_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
-; SSE4-NEXT: pextrw $2, %xmm0, 4(%rdi)
+; SSE4-NEXT: pxor %xmm1, %xmm7
+; SSE4-NEXT: pcmpeqd %xmm9, %xmm6
+; SSE4-NEXT: pxor %xmm1, %xmm6
+; SSE4-NEXT: packssdw %xmm7, %xmm6
+; SSE4-NEXT: pcmpeqd %xmm9, %xmm5
+; SSE4-NEXT: pxor %xmm1, %xmm5
+; SSE4-NEXT: pcmpeqd %xmm9, %xmm4
+; SSE4-NEXT: pxor %xmm1, %xmm4
+; SSE4-NEXT: packssdw %xmm5, %xmm4
+; SSE4-NEXT: packsswb %xmm6, %xmm4
+; SSE4-NEXT: pmovmskb %xmm4, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB9_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB9_3
+; SSE4-NEXT: .LBB9_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB9_5
; SSE4-NEXT: .LBB9_6: # %else4
-; SSE4-NEXT: pextrb $12, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB9_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
-; SSE4-NEXT: pextrw $3, %xmm0, 6(%rdi)
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB9_7
; SSE4-NEXT: .LBB9_8: # %else6
-; SSE4-NEXT: pxor %xmm4, %xmm4
-; SSE4-NEXT: pcmpeqd %xmm5, %xmm4
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm4, %xmm1
-; SSE4-NEXT: pextrb $0, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB9_10
-; SSE4-NEXT: # %bb.9: # %cond.store7
-; SSE4-NEXT: pextrw $4, %xmm0, 8(%rdi)
+; SSE4-NEXT: testb $16, %al
+; SSE4-NEXT: jne .LBB9_9
; SSE4-NEXT: .LBB9_10: # %else8
-; SSE4-NEXT: pextrb $4, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB9_12
-; SSE4-NEXT: # %bb.11: # %cond.store9
-; SSE4-NEXT: pextrw $5, %xmm0, 10(%rdi)
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: jne .LBB9_11
; SSE4-NEXT: .LBB9_12: # %else10
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm5
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm5, %xmm1
-; SSE4-NEXT: pextrb $8, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $64, %al
; SSE4-NEXT: je .LBB9_14
-; SSE4-NEXT: # %bb.13: # %cond.store11
+; SSE4-NEXT: .LBB9_13: # %cond.store11
; SSE4-NEXT: pextrw $6, %xmm0, 12(%rdi)
; SSE4-NEXT: .LBB9_14: # %else12
; SSE4-NEXT: pminud %xmm8, %xmm3
; SSE4-NEXT: pminud %xmm8, %xmm2
-; SSE4-NEXT: pextrb $12, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $-128, %al
; SSE4-NEXT: je .LBB9_16
; SSE4-NEXT: # %bb.15: # %cond.store13
; SSE4-NEXT: pextrw $7, %xmm0, 14(%rdi)
; SSE4-NEXT: .LBB9_16: # %else14
; SSE4-NEXT: packusdw %xmm3, %xmm2
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm6, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm1, %xmm0
-; SSE4-NEXT: pextrb $0, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB9_18
-; SSE4-NEXT: # %bb.17: # %cond.store15
+; SSE4-NEXT: testl $256, %eax # imm = 0x100
+; SSE4-NEXT: jne .LBB9_17
+; SSE4-NEXT: # %bb.18: # %else16
+; SSE4-NEXT: testl $512, %eax # imm = 0x200
+; SSE4-NEXT: jne .LBB9_19
+; SSE4-NEXT: .LBB9_20: # %else18
+; SSE4-NEXT: testl $1024, %eax # imm = 0x400
+; SSE4-NEXT: jne .LBB9_21
+; SSE4-NEXT: .LBB9_22: # %else20
+; SSE4-NEXT: testl $2048, %eax # imm = 0x800
+; SSE4-NEXT: jne .LBB9_23
+; SSE4-NEXT: .LBB9_24: # %else22
+; SSE4-NEXT: testl $4096, %eax # imm = 0x1000
+; SSE4-NEXT: jne .LBB9_25
+; SSE4-NEXT: .LBB9_26: # %else24
+; SSE4-NEXT: testl $8192, %eax # imm = 0x2000
+; SSE4-NEXT: jne .LBB9_27
+; SSE4-NEXT: .LBB9_28: # %else26
+; SSE4-NEXT: testl $16384, %eax # imm = 0x4000
+; SSE4-NEXT: jne .LBB9_29
+; SSE4-NEXT: .LBB9_30: # %else28
+; SSE4-NEXT: testl $32768, %eax # imm = 0x8000
+; SSE4-NEXT: jne .LBB9_31
+; SSE4-NEXT: .LBB9_32: # %else30
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB9_1: # %cond.store
+; SSE4-NEXT: pextrw $0, %xmm0, (%rdi)
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: je .LBB9_4
+; SSE4-NEXT: .LBB9_3: # %cond.store1
+; SSE4-NEXT: pextrw $1, %xmm0, 2(%rdi)
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: je .LBB9_6
+; SSE4-NEXT: .LBB9_5: # %cond.store3
+; SSE4-NEXT: pextrw $2, %xmm0, 4(%rdi)
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: je .LBB9_8
+; SSE4-NEXT: .LBB9_7: # %cond.store5
+; SSE4-NEXT: pextrw $3, %xmm0, 6(%rdi)
+; SSE4-NEXT: testb $16, %al
+; SSE4-NEXT: je .LBB9_10
+; SSE4-NEXT: .LBB9_9: # %cond.store7
+; SSE4-NEXT: pextrw $4, %xmm0, 8(%rdi)
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: je .LBB9_12
+; SSE4-NEXT: .LBB9_11: # %cond.store9
+; SSE4-NEXT: pextrw $5, %xmm0, 10(%rdi)
+; SSE4-NEXT: testb $64, %al
+; SSE4-NEXT: jne .LBB9_13
+; SSE4-NEXT: jmp .LBB9_14
+; SSE4-NEXT: .LBB9_17: # %cond.store15
; SSE4-NEXT: pextrw $0, %xmm2, 16(%rdi)
-; SSE4-NEXT: .LBB9_18: # %else16
-; SSE4-NEXT: pextrb $4, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $512, %eax # imm = 0x200
; SSE4-NEXT: je .LBB9_20
-; SSE4-NEXT: # %bb.19: # %cond.store17
+; SSE4-NEXT: .LBB9_19: # %cond.store17
; SSE4-NEXT: pextrw $1, %xmm2, 18(%rdi)
-; SSE4-NEXT: .LBB9_20: # %else18
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm6
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm6, %xmm0
-; SSE4-NEXT: pextrb $8, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $1024, %eax # imm = 0x400
; SSE4-NEXT: je .LBB9_22
-; SSE4-NEXT: # %bb.21: # %cond.store19
+; SSE4-NEXT: .LBB9_21: # %cond.store19
; SSE4-NEXT: pextrw $2, %xmm2, 20(%rdi)
-; SSE4-NEXT: .LBB9_22: # %else20
-; SSE4-NEXT: pextrb $12, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $2048, %eax # imm = 0x800
; SSE4-NEXT: je .LBB9_24
-; SSE4-NEXT: # %bb.23: # %cond.store21
+; SSE4-NEXT: .LBB9_23: # %cond.store21
; SSE4-NEXT: pextrw $3, %xmm2, 22(%rdi)
-; SSE4-NEXT: .LBB9_24: # %else22
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm7, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm1, %xmm0
-; SSE4-NEXT: pextrb $0, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $4096, %eax # imm = 0x1000
; SSE4-NEXT: je .LBB9_26
-; SSE4-NEXT: # %bb.25: # %cond.store23
+; SSE4-NEXT: .LBB9_25: # %cond.store23
; SSE4-NEXT: pextrw $4, %xmm2, 24(%rdi)
-; SSE4-NEXT: .LBB9_26: # %else24
-; SSE4-NEXT: pextrb $4, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $8192, %eax # imm = 0x2000
; SSE4-NEXT: je .LBB9_28
-; SSE4-NEXT: # %bb.27: # %cond.store25
+; SSE4-NEXT: .LBB9_27: # %cond.store25
; SSE4-NEXT: pextrw $5, %xmm2, 26(%rdi)
-; SSE4-NEXT: .LBB9_28: # %else26
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm7
-; SSE4-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE4-NEXT: pxor %xmm7, %xmm0
-; SSE4-NEXT: pextrb $8, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $16384, %eax # imm = 0x4000
; SSE4-NEXT: je .LBB9_30
-; SSE4-NEXT: # %bb.29: # %cond.store27
+; SSE4-NEXT: .LBB9_29: # %cond.store27
; SSE4-NEXT: pextrw $6, %xmm2, 28(%rdi)
-; SSE4-NEXT: .LBB9_30: # %else28
-; SSE4-NEXT: pextrb $12, %xmm0, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $32768, %eax # imm = 0x8000
; SSE4-NEXT: je .LBB9_32
-; SSE4-NEXT: # %bb.31: # %cond.store29
+; SSE4-NEXT: .LBB9_31: # %cond.store29
; SSE4-NEXT: pextrw $7, %xmm2, 30(%rdi)
-; SSE4-NEXT: .LBB9_32: # %else30
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v16i32_v16i16:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpxor %xmm8, %xmm8, %xmm8
-; AVX1-NEXT: vpcmpeqd %xmm8, %xmm2, %xmm6
-; AVX1-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
-; AVX1-NEXT: vpxor %xmm5, %xmm6, %xmm6
-; AVX1-NEXT: vpackssdw %xmm0, %xmm6, %xmm6
-; AVX1-NEXT: vpacksswb %xmm0, %xmm6, %xmm6
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm7
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [65535,65535,65535,65535]
-; AVX1-NEXT: vpminud %xmm4, %xmm7, %xmm7
-; AVX1-NEXT: vpminud %xmm4, %xmm0, %xmm0
-; AVX1-NEXT: vpackusdw %xmm7, %xmm0, %xmm0
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm7
-; AVX1-NEXT: vpminud %xmm4, %xmm7, %xmm7
-; AVX1-NEXT: vpminud %xmm4, %xmm1, %xmm1
-; AVX1-NEXT: vpackusdw %xmm7, %xmm1, %xmm1
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm4
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [65535,65535,65535,65535]
+; AVX1-NEXT: vpminud %xmm5, %xmm4, %xmm4
+; AVX1-NEXT: vpminud %xmm5, %xmm0, %xmm0
+; AVX1-NEXT: vpackusdw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
+; AVX1-NEXT: vpminud %xmm5, %xmm4, %xmm4
+; AVX1-NEXT: vpminud %xmm5, %xmm1, %xmm1
+; AVX1-NEXT: vpackusdw %xmm4, %xmm1, %xmm1
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT: vpextrb $0, %xmm6, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB9_2
-; AVX1-NEXT: # %bb.1: # %cond.store
-; AVX1-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX1-NEXT: .LBB9_2: # %else
-; AVX1-NEXT: vpcmpeqd %xmm8, %xmm2, %xmm1
+; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm1
+; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT: vpcmpeqd %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
; AVX1-NEXT: vpxor %xmm5, %xmm1, %xmm1
-; AVX1-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $1, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB9_4
-; AVX1-NEXT: # %bb.3: # %cond.store1
-; AVX1-NEXT: vpextrw $1, %xmm0, 2(%rdi)
+; AVX1-NEXT: vpcmpeqd %xmm4, %xmm3, %xmm3
+; AVX1-NEXT: vpxor %xmm5, %xmm3, %xmm3
+; AVX1-NEXT: vpackssdw %xmm1, %xmm3, %xmm1
+; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT: vpcmpeqd %xmm4, %xmm3, %xmm3
+; AVX1-NEXT: vpxor %xmm5, %xmm3, %xmm3
+; AVX1-NEXT: vpcmpeqd %xmm4, %xmm2, %xmm2
+; AVX1-NEXT: vpxor %xmm5, %xmm2, %xmm2
+; AVX1-NEXT: vpackssdw %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vpacksswb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT: vpmovmskb %xmm1, %eax
+; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: jne .LBB9_1
+; AVX1-NEXT: # %bb.2: # %else
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: jne .LBB9_3
; AVX1-NEXT: .LBB9_4: # %else2
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm5
-; AVX1-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpxor %xmm4, %xmm5, %xmm5
-; AVX1-NEXT: vpackssdw %xmm0, %xmm5, %xmm5
-; AVX1-NEXT: vpacksswb %xmm0, %xmm5, %xmm5
-; AVX1-NEXT: vpextrb $2, %xmm5, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB9_6
-; AVX1-NEXT: # %bb.5: # %cond.store3
-; AVX1-NEXT: vpextrw $2, %xmm0, 4(%rdi)
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: jne .LBB9_5
; AVX1-NEXT: .LBB9_6: # %else4
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm4, %xmm1, %xmm1
-; AVX1-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $3, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB9_8
-; AVX1-NEXT: # %bb.7: # %cond.store5
-; AVX1-NEXT: vpextrw $3, %xmm0, 6(%rdi)
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: jne .LBB9_7
; AVX1-NEXT: .LBB9_8: # %else6
-; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpxor %xmm4, %xmm2, %xmm2
-; AVX1-NEXT: vpackssdw %xmm2, %xmm0, %xmm4
-; AVX1-NEXT: vpacksswb %xmm0, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $4, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB9_10
-; AVX1-NEXT: # %bb.9: # %cond.store7
-; AVX1-NEXT: vpextrw $4, %xmm0, 8(%rdi)
+; AVX1-NEXT: testb $16, %al
+; AVX1-NEXT: jne .LBB9_9
; AVX1-NEXT: .LBB9_10: # %else8
-; AVX1-NEXT: vpackssdw %xmm2, %xmm0, %xmm2
-; AVX1-NEXT: vpacksswb %xmm0, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $5, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB9_12
-; AVX1-NEXT: # %bb.11: # %cond.store9
-; AVX1-NEXT: vpextrw $5, %xmm0, 10(%rdi)
+; AVX1-NEXT: testb $32, %al
+; AVX1-NEXT: jne .LBB9_11
; AVX1-NEXT: .LBB9_12: # %else10
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpackssdw %xmm1, %xmm0, %xmm2
-; AVX1-NEXT: vpacksswb %xmm0, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $6, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB9_14
-; AVX1-NEXT: # %bb.13: # %cond.store11
-; AVX1-NEXT: vpextrw $6, %xmm0, 12(%rdi)
+; AVX1-NEXT: testb $64, %al
+; AVX1-NEXT: jne .LBB9_13
; AVX1-NEXT: .LBB9_14: # %else12
-; AVX1-NEXT: vpackssdw %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $7, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $-128, %al
; AVX1-NEXT: je .LBB9_16
-; AVX1-NEXT: # %bb.15: # %cond.store13
+; AVX1-NEXT: .LBB9_15: # %cond.store13
; AVX1-NEXT: vpextrw $7, %xmm0, 14(%rdi)
; AVX1-NEXT: .LBB9_16: # %else14
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm4, %xmm4
-; AVX1-NEXT: vpackssdw %xmm0, %xmm4, %xmm4
-; AVX1-NEXT: vpacksswb %xmm4, %xmm0, %xmm4
-; AVX1-NEXT: vpextrb $8, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $256, %eax # imm = 0x100
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT: je .LBB9_18
-; AVX1-NEXT: # %bb.17: # %cond.store15
-; AVX1-NEXT: vpextrw $0, %xmm0, 16(%rdi)
-; AVX1-NEXT: .LBB9_18: # %else16
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpacksswb %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vpextrb $9, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB9_20
-; AVX1-NEXT: # %bb.19: # %cond.store17
-; AVX1-NEXT: vpextrw $1, %xmm0, 18(%rdi)
+; AVX1-NEXT: jne .LBB9_17
+; AVX1-NEXT: # %bb.18: # %else16
+; AVX1-NEXT: testl $512, %eax # imm = 0x200
+; AVX1-NEXT: jne .LBB9_19
; AVX1-NEXT: .LBB9_20: # %else18
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm4, %xmm4
-; AVX1-NEXT: vpackssdw %xmm0, %xmm4, %xmm4
-; AVX1-NEXT: vpacksswb %xmm4, %xmm0, %xmm4
-; AVX1-NEXT: vpextrb $10, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB9_22
-; AVX1-NEXT: # %bb.21: # %cond.store19
-; AVX1-NEXT: vpextrw $2, %xmm0, 20(%rdi)
+; AVX1-NEXT: testl $1024, %eax # imm = 0x400
+; AVX1-NEXT: jne .LBB9_21
; AVX1-NEXT: .LBB9_22: # %else20
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpacksswb %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vpextrb $11, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $2048, %eax # imm = 0x800
+; AVX1-NEXT: jne .LBB9_23
+; AVX1-NEXT: .LBB9_24: # %else22
+; AVX1-NEXT: testl $4096, %eax # imm = 0x1000
+; AVX1-NEXT: jne .LBB9_25
+; AVX1-NEXT: .LBB9_26: # %else24
+; AVX1-NEXT: testl $8192, %eax # imm = 0x2000
+; AVX1-NEXT: jne .LBB9_27
+; AVX1-NEXT: .LBB9_28: # %else26
+; AVX1-NEXT: testl $16384, %eax # imm = 0x4000
+; AVX1-NEXT: jne .LBB9_29
+; AVX1-NEXT: .LBB9_30: # %else28
+; AVX1-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX1-NEXT: jne .LBB9_31
+; AVX1-NEXT: .LBB9_32: # %else30
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+; AVX1-NEXT: .LBB9_1: # %cond.store
+; AVX1-NEXT: vpextrw $0, %xmm0, (%rdi)
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: je .LBB9_4
+; AVX1-NEXT: .LBB9_3: # %cond.store1
+; AVX1-NEXT: vpextrw $1, %xmm0, 2(%rdi)
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: je .LBB9_6
+; AVX1-NEXT: .LBB9_5: # %cond.store3
+; AVX1-NEXT: vpextrw $2, %xmm0, 4(%rdi)
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: je .LBB9_8
+; AVX1-NEXT: .LBB9_7: # %cond.store5
+; AVX1-NEXT: vpextrw $3, %xmm0, 6(%rdi)
+; AVX1-NEXT: testb $16, %al
+; AVX1-NEXT: je .LBB9_10
+; AVX1-NEXT: .LBB9_9: # %cond.store7
+; AVX1-NEXT: vpextrw $4, %xmm0, 8(%rdi)
+; AVX1-NEXT: testb $32, %al
+; AVX1-NEXT: je .LBB9_12
+; AVX1-NEXT: .LBB9_11: # %cond.store9
+; AVX1-NEXT: vpextrw $5, %xmm0, 10(%rdi)
+; AVX1-NEXT: testb $64, %al
+; AVX1-NEXT: je .LBB9_14
+; AVX1-NEXT: .LBB9_13: # %cond.store11
+; AVX1-NEXT: vpextrw $6, %xmm0, 12(%rdi)
+; AVX1-NEXT: testb $-128, %al
+; AVX1-NEXT: jne .LBB9_15
+; AVX1-NEXT: jmp .LBB9_16
+; AVX1-NEXT: .LBB9_17: # %cond.store15
+; AVX1-NEXT: vpextrw $0, %xmm0, 16(%rdi)
+; AVX1-NEXT: testl $512, %eax # imm = 0x200
+; AVX1-NEXT: je .LBB9_20
+; AVX1-NEXT: .LBB9_19: # %cond.store17
+; AVX1-NEXT: vpextrw $1, %xmm0, 18(%rdi)
+; AVX1-NEXT: testl $1024, %eax # imm = 0x400
+; AVX1-NEXT: je .LBB9_22
+; AVX1-NEXT: .LBB9_21: # %cond.store19
+; AVX1-NEXT: vpextrw $2, %xmm0, 20(%rdi)
+; AVX1-NEXT: testl $2048, %eax # imm = 0x800
; AVX1-NEXT: je .LBB9_24
-; AVX1-NEXT: # %bb.23: # %cond.store21
+; AVX1-NEXT: .LBB9_23: # %cond.store21
; AVX1-NEXT: vpextrw $3, %xmm0, 22(%rdi)
-; AVX1-NEXT: .LBB9_24: # %else22
-; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpackssdw %xmm2, %xmm0, %xmm3
-; AVX1-NEXT: vpacksswb %xmm3, %xmm0, %xmm3
-; AVX1-NEXT: vpextrb $12, %xmm3, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $4096, %eax # imm = 0x1000
; AVX1-NEXT: je .LBB9_26
-; AVX1-NEXT: # %bb.25: # %cond.store23
+; AVX1-NEXT: .LBB9_25: # %cond.store23
; AVX1-NEXT: vpextrw $4, %xmm0, 24(%rdi)
-; AVX1-NEXT: .LBB9_26: # %else24
-; AVX1-NEXT: vpackssdw %xmm2, %xmm0, %xmm2
-; AVX1-NEXT: vpacksswb %xmm2, %xmm0, %xmm2
-; AVX1-NEXT: vpextrb $13, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $8192, %eax # imm = 0x2000
; AVX1-NEXT: je .LBB9_28
-; AVX1-NEXT: # %bb.27: # %cond.store25
+; AVX1-NEXT: .LBB9_27: # %cond.store25
; AVX1-NEXT: vpextrw $5, %xmm0, 26(%rdi)
-; AVX1-NEXT: .LBB9_28: # %else26
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpackssdw %xmm1, %xmm0, %xmm2
-; AVX1-NEXT: vpacksswb %xmm2, %xmm0, %xmm2
-; AVX1-NEXT: vpextrb $14, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $16384, %eax # imm = 0x4000
; AVX1-NEXT: je .LBB9_30
-; AVX1-NEXT: # %bb.29: # %cond.store27
+; AVX1-NEXT: .LBB9_29: # %cond.store27
; AVX1-NEXT: vpextrw $6, %xmm0, 28(%rdi)
-; AVX1-NEXT: .LBB9_30: # %else28
-; AVX1-NEXT: vpackssdw %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vpacksswb %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vpextrb $15, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $32768, %eax # imm = 0x8000
; AVX1-NEXT: je .LBB9_32
-; AVX1-NEXT: # %bb.31: # %cond.store29
+; AVX1-NEXT: .LBB9_31: # %cond.store29
; AVX1-NEXT: vpextrw $7, %xmm0, 30(%rdi)
-; AVX1-NEXT: .LBB9_32: # %else30
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: truncstore_v16i32_v16i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm4, %xmm4, %xmm4
-; AVX2-NEXT: vpcmpeqd %xmm4, %xmm2, %xmm6
-; AVX2-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
-; AVX2-NEXT: vpxor %xmm5, %xmm6, %xmm6
-; AVX2-NEXT: vpackssdw %xmm0, %xmm6, %xmm6
-; AVX2-NEXT: vpacksswb %xmm0, %xmm6, %xmm6
-; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm7 = [65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX2-NEXT: vpminud %ymm7, %ymm1, %ymm1
-; AVX2-NEXT: vpminud %ymm7, %ymm0, %ymm0
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm5 = [65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX2-NEXT: vpminud %ymm5, %ymm1, %ymm1
+; AVX2-NEXT: vpminud %ymm5, %ymm0, %ymm0
; AVX2-NEXT: vpackusdw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-NEXT: vpextrb $0, %xmm6, %eax
-; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB9_2
-; AVX2-NEXT: # %bb.1: # %cond.store
+; AVX2-NEXT: vpcmpeqd %ymm4, %ymm3, %ymm1
+; AVX2-NEXT: vpcmpeqd %ymm3, %ymm3, %ymm3
+; AVX2-NEXT: vpxor %ymm3, %ymm1, %ymm1
+; AVX2-NEXT: vpcmpeqd %ymm4, %ymm2, %ymm2
+; AVX2-NEXT: vpxor %ymm3, %ymm2, %ymm2
+; AVX2-NEXT: vpackssdw %ymm1, %ymm2, %ymm1
+; AVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,2,1,3]
+; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT: vpacksswb %xmm2, %xmm1, %xmm1
+; AVX2-NEXT: vpmovmskb %xmm1, %eax
+; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: jne .LBB9_1
+; AVX2-NEXT: # %bb.2: # %else
+; AVX2-NEXT: testb $2, %al
+; AVX2-NEXT: jne .LBB9_3
+; AVX2-NEXT: .LBB9_4: # %else2
+; AVX2-NEXT: testb $4, %al
+; AVX2-NEXT: jne .LBB9_5
+; AVX2-NEXT: .LBB9_6: # %else4
+; AVX2-NEXT: testb $8, %al
+; AVX2-NEXT: jne .LBB9_7
+; AVX2-NEXT: .LBB9_8: # %else6
+; AVX2-NEXT: testb $16, %al
+; AVX2-NEXT: jne .LBB9_9
+; AVX2-NEXT: .LBB9_10: # %else8
+; AVX2-NEXT: testb $32, %al
+; AVX2-NEXT: jne .LBB9_11
+; AVX2-NEXT: .LBB9_12: # %else10
+; AVX2-NEXT: testb $64, %al
+; AVX2-NEXT: jne .LBB9_13
+; AVX2-NEXT: .LBB9_14: # %else12
+; AVX2-NEXT: testb $-128, %al
+; AVX2-NEXT: je .LBB9_16
+; AVX2-NEXT: .LBB9_15: # %cond.store13
+; AVX2-NEXT: vpextrw $7, %xmm0, 14(%rdi)
+; AVX2-NEXT: .LBB9_16: # %else14
+; AVX2-NEXT: testl $256, %eax # imm = 0x100
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX2-NEXT: jne .LBB9_17
+; AVX2-NEXT: # %bb.18: # %else16
+; AVX2-NEXT: testl $512, %eax # imm = 0x200
+; AVX2-NEXT: jne .LBB9_19
+; AVX2-NEXT: .LBB9_20: # %else18
+; AVX2-NEXT: testl $1024, %eax # imm = 0x400
+; AVX2-NEXT: jne .LBB9_21
+; AVX2-NEXT: .LBB9_22: # %else20
+; AVX2-NEXT: testl $2048, %eax # imm = 0x800
+; AVX2-NEXT: jne .LBB9_23
+; AVX2-NEXT: .LBB9_24: # %else22
+; AVX2-NEXT: testl $4096, %eax # imm = 0x1000
+; AVX2-NEXT: jne .LBB9_25
+; AVX2-NEXT: .LBB9_26: # %else24
+; AVX2-NEXT: testl $8192, %eax # imm = 0x2000
+; AVX2-NEXT: jne .LBB9_27
+; AVX2-NEXT: .LBB9_28: # %else26
+; AVX2-NEXT: testl $16384, %eax # imm = 0x4000
+; AVX2-NEXT: jne .LBB9_29
+; AVX2-NEXT: .LBB9_30: # %else28
+; AVX2-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX2-NEXT: jne .LBB9_31
+; AVX2-NEXT: .LBB9_32: # %else30
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+; AVX2-NEXT: .LBB9_1: # %cond.store
; AVX2-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX2-NEXT: .LBB9_2: # %else
-; AVX2-NEXT: vpcmpeqd %xmm4, %xmm2, %xmm1
-; AVX2-NEXT: vpxor %xmm5, %xmm1, %xmm1
-; AVX2-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $1, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $2, %al
; AVX2-NEXT: je .LBB9_4
-; AVX2-NEXT: # %bb.3: # %cond.store1
+; AVX2-NEXT: .LBB9_3: # %cond.store1
; AVX2-NEXT: vpextrw $1, %xmm0, 2(%rdi)
-; AVX2-NEXT: .LBB9_4: # %else2
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm5
-; AVX2-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX2-NEXT: vpxor %xmm4, %xmm5, %xmm5
-; AVX2-NEXT: vpackssdw %xmm0, %xmm5, %xmm5
-; AVX2-NEXT: vpacksswb %xmm0, %xmm5, %xmm5
-; AVX2-NEXT: vpextrb $2, %xmm5, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $4, %al
; AVX2-NEXT: je .LBB9_6
-; AVX2-NEXT: # %bb.5: # %cond.store3
+; AVX2-NEXT: .LBB9_5: # %cond.store3
; AVX2-NEXT: vpextrw $2, %xmm0, 4(%rdi)
-; AVX2-NEXT: .LBB9_6: # %else4
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm1
-; AVX2-NEXT: vpxor %xmm4, %xmm1, %xmm1
-; AVX2-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $3, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $8, %al
; AVX2-NEXT: je .LBB9_8
-; AVX2-NEXT: # %bb.7: # %cond.store5
+; AVX2-NEXT: .LBB9_7: # %cond.store5
; AVX2-NEXT: vpextrw $3, %xmm0, 6(%rdi)
-; AVX2-NEXT: .LBB9_8: # %else6
-; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX2-NEXT: vpxor %xmm4, %xmm2, %xmm2
-; AVX2-NEXT: vpackssdw %xmm2, %xmm0, %xmm4
-; AVX2-NEXT: vpacksswb %xmm0, %xmm4, %xmm4
-; AVX2-NEXT: vpextrb $4, %xmm4, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $16, %al
; AVX2-NEXT: je .LBB9_10
-; AVX2-NEXT: # %bb.9: # %cond.store7
+; AVX2-NEXT: .LBB9_9: # %cond.store7
; AVX2-NEXT: vpextrw $4, %xmm0, 8(%rdi)
-; AVX2-NEXT: .LBB9_10: # %else8
-; AVX2-NEXT: vpackssdw %xmm2, %xmm0, %xmm2
-; AVX2-NEXT: vpacksswb %xmm0, %xmm2, %xmm2
-; AVX2-NEXT: vpextrb $5, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $32, %al
; AVX2-NEXT: je .LBB9_12
-; AVX2-NEXT: # %bb.11: # %cond.store9
+; AVX2-NEXT: .LBB9_11: # %cond.store9
; AVX2-NEXT: vpextrw $5, %xmm0, 10(%rdi)
-; AVX2-NEXT: .LBB9_12: # %else10
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpackssdw %xmm1, %xmm0, %xmm2
-; AVX2-NEXT: vpacksswb %xmm0, %xmm2, %xmm2
-; AVX2-NEXT: vpextrb $6, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $64, %al
; AVX2-NEXT: je .LBB9_14
-; AVX2-NEXT: # %bb.13: # %cond.store11
+; AVX2-NEXT: .LBB9_13: # %cond.store11
; AVX2-NEXT: vpextrw $6, %xmm0, 12(%rdi)
-; AVX2-NEXT: .LBB9_14: # %else12
-; AVX2-NEXT: vpackssdw %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $7, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB9_16
-; AVX2-NEXT: # %bb.15: # %cond.store13
-; AVX2-NEXT: vpextrw $7, %xmm0, 14(%rdi)
-; AVX2-NEXT: .LBB9_16: # %else14
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm4
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm4, %xmm4
-; AVX2-NEXT: vpackssdw %xmm0, %xmm4, %xmm4
-; AVX2-NEXT: vpacksswb %xmm4, %xmm0, %xmm4
-; AVX2-NEXT: vpextrb $8, %xmm4, %eax
-; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX2-NEXT: je .LBB9_18
-; AVX2-NEXT: # %bb.17: # %cond.store15
+; AVX2-NEXT: testb $-128, %al
+; AVX2-NEXT: jne .LBB9_15
+; AVX2-NEXT: jmp .LBB9_16
+; AVX2-NEXT: .LBB9_17: # %cond.store15
; AVX2-NEXT: vpextrw $0, %xmm0, 16(%rdi)
-; AVX2-NEXT: .LBB9_18: # %else16
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpacksswb %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vpextrb $9, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $512, %eax # imm = 0x200
; AVX2-NEXT: je .LBB9_20
-; AVX2-NEXT: # %bb.19: # %cond.store17
+; AVX2-NEXT: .LBB9_19: # %cond.store17
; AVX2-NEXT: vpextrw $1, %xmm0, 18(%rdi)
-; AVX2-NEXT: .LBB9_20: # %else18
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm4
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm4, %xmm4
-; AVX2-NEXT: vpackssdw %xmm0, %xmm4, %xmm4
-; AVX2-NEXT: vpacksswb %xmm4, %xmm0, %xmm4
-; AVX2-NEXT: vpextrb $10, %xmm4, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $1024, %eax # imm = 0x400
; AVX2-NEXT: je .LBB9_22
-; AVX2-NEXT: # %bb.21: # %cond.store19
+; AVX2-NEXT: .LBB9_21: # %cond.store19
; AVX2-NEXT: vpextrw $2, %xmm0, 20(%rdi)
-; AVX2-NEXT: .LBB9_22: # %else20
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpacksswb %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vpextrb $11, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $2048, %eax # imm = 0x800
; AVX2-NEXT: je .LBB9_24
-; AVX2-NEXT: # %bb.23: # %cond.store21
+; AVX2-NEXT: .LBB9_23: # %cond.store21
; AVX2-NEXT: vpextrw $3, %xmm0, 22(%rdi)
-; AVX2-NEXT: .LBB9_24: # %else22
-; AVX2-NEXT: vextracti128 $1, %ymm3, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX2-NEXT: vpackssdw %xmm2, %xmm0, %xmm3
-; AVX2-NEXT: vpacksswb %xmm3, %xmm0, %xmm3
-; AVX2-NEXT: vpextrb $12, %xmm3, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $4096, %eax # imm = 0x1000
; AVX2-NEXT: je .LBB9_26
-; AVX2-NEXT: # %bb.25: # %cond.store23
+; AVX2-NEXT: .LBB9_25: # %cond.store23
; AVX2-NEXT: vpextrw $4, %xmm0, 24(%rdi)
-; AVX2-NEXT: .LBB9_26: # %else24
-; AVX2-NEXT: vpackssdw %xmm2, %xmm0, %xmm2
-; AVX2-NEXT: vpacksswb %xmm2, %xmm0, %xmm2
-; AVX2-NEXT: vpextrb $13, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $8192, %eax # imm = 0x2000
; AVX2-NEXT: je .LBB9_28
-; AVX2-NEXT: # %bb.27: # %cond.store25
+; AVX2-NEXT: .LBB9_27: # %cond.store25
; AVX2-NEXT: vpextrw $5, %xmm0, 26(%rdi)
-; AVX2-NEXT: .LBB9_28: # %else26
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpackssdw %xmm1, %xmm0, %xmm2
-; AVX2-NEXT: vpacksswb %xmm2, %xmm0, %xmm2
-; AVX2-NEXT: vpextrb $14, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $16384, %eax # imm = 0x4000
; AVX2-NEXT: je .LBB9_30
-; AVX2-NEXT: # %bb.29: # %cond.store27
+; AVX2-NEXT: .LBB9_29: # %cond.store27
; AVX2-NEXT: vpextrw $6, %xmm0, 28(%rdi)
-; AVX2-NEXT: .LBB9_30: # %else28
-; AVX2-NEXT: vpackssdw %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vpacksswb %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vpextrb $15, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $32768, %eax # imm = 0x8000
; AVX2-NEXT: je .LBB9_32
-; AVX2-NEXT: # %bb.31: # %cond.store29
+; AVX2-NEXT: .LBB9_31: # %cond.store29
; AVX2-NEXT: vpextrw $7, %xmm0, 30(%rdi)
-; AVX2-NEXT: .LBB9_32: # %else30
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
@@ -3272,116 +3081,117 @@ define void @truncstore_v16i32_v16i16(<1
; AVX512F-NEXT: vpmovusdw %zmm0, %ymm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB9_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB9_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB9_3
+; AVX512F-NEXT: .LBB9_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB9_5
+; AVX512F-NEXT: .LBB9_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB9_7
+; AVX512F-NEXT: .LBB9_8: # %else6
+; AVX512F-NEXT: testb $16, %al
+; AVX512F-NEXT: jne .LBB9_9
+; AVX512F-NEXT: .LBB9_10: # %else8
+; AVX512F-NEXT: testb $32, %al
+; AVX512F-NEXT: jne .LBB9_11
+; AVX512F-NEXT: .LBB9_12: # %else10
+; AVX512F-NEXT: testb $64, %al
+; AVX512F-NEXT: jne .LBB9_13
+; AVX512F-NEXT: .LBB9_14: # %else12
+; AVX512F-NEXT: testb $-128, %al
+; AVX512F-NEXT: je .LBB9_16
+; AVX512F-NEXT: .LBB9_15: # %cond.store13
+; AVX512F-NEXT: vpextrw $7, %xmm0, 14(%rdi)
+; AVX512F-NEXT: .LBB9_16: # %else14
+; AVX512F-NEXT: testl $256, %eax # imm = 0x100
+; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX512F-NEXT: jne .LBB9_17
+; AVX512F-NEXT: # %bb.18: # %else16
+; AVX512F-NEXT: testl $512, %eax # imm = 0x200
+; AVX512F-NEXT: jne .LBB9_19
+; AVX512F-NEXT: .LBB9_20: # %else18
+; AVX512F-NEXT: testl $1024, %eax # imm = 0x400
+; AVX512F-NEXT: jne .LBB9_21
+; AVX512F-NEXT: .LBB9_22: # %else20
+; AVX512F-NEXT: testl $2048, %eax # imm = 0x800
+; AVX512F-NEXT: jne .LBB9_23
+; AVX512F-NEXT: .LBB9_24: # %else22
+; AVX512F-NEXT: testl $4096, %eax # imm = 0x1000
+; AVX512F-NEXT: jne .LBB9_25
+; AVX512F-NEXT: .LBB9_26: # %else24
+; AVX512F-NEXT: testl $8192, %eax # imm = 0x2000
+; AVX512F-NEXT: jne .LBB9_27
+; AVX512F-NEXT: .LBB9_28: # %else26
+; AVX512F-NEXT: testl $16384, %eax # imm = 0x4000
+; AVX512F-NEXT: jne .LBB9_29
+; AVX512F-NEXT: .LBB9_30: # %else28
+; AVX512F-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX512F-NEXT: jne .LBB9_31
+; AVX512F-NEXT: .LBB9_32: # %else30
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB9_1: # %cond.store
; AVX512F-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB9_2: # %else
-; AVX512F-NEXT: kshiftrw $1, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB9_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB9_3: # %cond.store1
; AVX512F-NEXT: vpextrw $1, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB9_4: # %else2
-; AVX512F-NEXT: kshiftrw $2, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB9_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB9_5: # %cond.store3
; AVX512F-NEXT: vpextrw $2, %xmm0, 4(%rdi)
-; AVX512F-NEXT: .LBB9_6: # %else4
-; AVX512F-NEXT: kshiftrw $3, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB9_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB9_7: # %cond.store5
; AVX512F-NEXT: vpextrw $3, %xmm0, 6(%rdi)
-; AVX512F-NEXT: .LBB9_8: # %else6
-; AVX512F-NEXT: kshiftrw $4, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $16, %al
; AVX512F-NEXT: je .LBB9_10
-; AVX512F-NEXT: # %bb.9: # %cond.store7
+; AVX512F-NEXT: .LBB9_9: # %cond.store7
; AVX512F-NEXT: vpextrw $4, %xmm0, 8(%rdi)
-; AVX512F-NEXT: .LBB9_10: # %else8
-; AVX512F-NEXT: kshiftrw $5, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $32, %al
; AVX512F-NEXT: je .LBB9_12
-; AVX512F-NEXT: # %bb.11: # %cond.store9
+; AVX512F-NEXT: .LBB9_11: # %cond.store9
; AVX512F-NEXT: vpextrw $5, %xmm0, 10(%rdi)
-; AVX512F-NEXT: .LBB9_12: # %else10
-; AVX512F-NEXT: kshiftrw $6, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $64, %al
; AVX512F-NEXT: je .LBB9_14
-; AVX512F-NEXT: # %bb.13: # %cond.store11
+; AVX512F-NEXT: .LBB9_13: # %cond.store11
; AVX512F-NEXT: vpextrw $6, %xmm0, 12(%rdi)
-; AVX512F-NEXT: .LBB9_14: # %else12
-; AVX512F-NEXT: kshiftrw $7, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB9_16
-; AVX512F-NEXT: # %bb.15: # %cond.store13
-; AVX512F-NEXT: vpextrw $7, %xmm0, 14(%rdi)
-; AVX512F-NEXT: .LBB9_16: # %else14
-; AVX512F-NEXT: kshiftrw $8, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX512F-NEXT: je .LBB9_18
-; AVX512F-NEXT: # %bb.17: # %cond.store15
+; AVX512F-NEXT: testb $-128, %al
+; AVX512F-NEXT: jne .LBB9_15
+; AVX512F-NEXT: jmp .LBB9_16
+; AVX512F-NEXT: .LBB9_17: # %cond.store15
; AVX512F-NEXT: vpextrw $0, %xmm0, 16(%rdi)
-; AVX512F-NEXT: .LBB9_18: # %else16
-; AVX512F-NEXT: kshiftrw $9, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $512, %eax # imm = 0x200
; AVX512F-NEXT: je .LBB9_20
-; AVX512F-NEXT: # %bb.19: # %cond.store17
+; AVX512F-NEXT: .LBB9_19: # %cond.store17
; AVX512F-NEXT: vpextrw $1, %xmm0, 18(%rdi)
-; AVX512F-NEXT: .LBB9_20: # %else18
-; AVX512F-NEXT: kshiftrw $10, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $1024, %eax # imm = 0x400
; AVX512F-NEXT: je .LBB9_22
-; AVX512F-NEXT: # %bb.21: # %cond.store19
+; AVX512F-NEXT: .LBB9_21: # %cond.store19
; AVX512F-NEXT: vpextrw $2, %xmm0, 20(%rdi)
-; AVX512F-NEXT: .LBB9_22: # %else20
-; AVX512F-NEXT: kshiftrw $11, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $2048, %eax # imm = 0x800
; AVX512F-NEXT: je .LBB9_24
-; AVX512F-NEXT: # %bb.23: # %cond.store21
+; AVX512F-NEXT: .LBB9_23: # %cond.store21
; AVX512F-NEXT: vpextrw $3, %xmm0, 22(%rdi)
-; AVX512F-NEXT: .LBB9_24: # %else22
-; AVX512F-NEXT: kshiftrw $12, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $4096, %eax # imm = 0x1000
; AVX512F-NEXT: je .LBB9_26
-; AVX512F-NEXT: # %bb.25: # %cond.store23
+; AVX512F-NEXT: .LBB9_25: # %cond.store23
; AVX512F-NEXT: vpextrw $4, %xmm0, 24(%rdi)
-; AVX512F-NEXT: .LBB9_26: # %else24
-; AVX512F-NEXT: kshiftrw $13, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $8192, %eax # imm = 0x2000
; AVX512F-NEXT: je .LBB9_28
-; AVX512F-NEXT: # %bb.27: # %cond.store25
+; AVX512F-NEXT: .LBB9_27: # %cond.store25
; AVX512F-NEXT: vpextrw $5, %xmm0, 26(%rdi)
-; AVX512F-NEXT: .LBB9_28: # %else26
-; AVX512F-NEXT: kshiftrw $14, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $16384, %eax # imm = 0x4000
; AVX512F-NEXT: je .LBB9_30
-; AVX512F-NEXT: # %bb.29: # %cond.store27
+; AVX512F-NEXT: .LBB9_29: # %cond.store27
; AVX512F-NEXT: vpextrw $6, %xmm0, 28(%rdi)
-; AVX512F-NEXT: .LBB9_30: # %else28
-; AVX512F-NEXT: kshiftrw $15, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $32768, %eax # imm = 0x8000
; AVX512F-NEXT: je .LBB9_32
-; AVX512F-NEXT: # %bb.31: # %cond.store29
+; AVX512F-NEXT: .LBB9_31: # %cond.store29
; AVX512F-NEXT: vpextrw $7, %xmm0, 30(%rdi)
-; AVX512F-NEXT: .LBB9_32: # %else30
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -3412,7 +3222,6 @@ define void @truncstore_v16i32_v16i8(<16
; SSE2-LABEL: truncstore_v16i32_v16i8:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm8, %xmm8
-; SSE2-NEXT: pcmpeqd %xmm4, %xmm8
; SSE2-NEXT: movdqa {{.*#+}} xmm10 = [255,255,255,255]
; SSE2-NEXT: movdqa {{.*#+}} xmm11 = [2147483648,2147483648,2147483648,2147483648]
; SSE2-NEXT: movdqa %xmm1, %xmm12
@@ -3445,158 +3254,124 @@ define void @truncstore_v16i32_v16i8(<16
; SSE2-NEXT: por %xmm2, %xmm9
; SSE2-NEXT: packuswb %xmm1, %xmm9
; SSE2-NEXT: packuswb %xmm9, %xmm12
-; SSE2-NEXT: movd %xmm8, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: movd %xmm12, %eax
-; SSE2-NEXT: je .LBB10_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movb %al, (%rdi)
-; SSE2-NEXT: .LBB10_2: # %else
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm7
; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm8
-; SSE2-NEXT: pextrw $2, %xmm8, %ecx
-; SSE2-NEXT: testb $1, %cl
-; SSE2-NEXT: je .LBB10_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: movb %ah, 1(%rdi)
+; SSE2-NEXT: pxor %xmm0, %xmm7
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm6
+; SSE2-NEXT: pxor %xmm0, %xmm6
+; SSE2-NEXT: packssdw %xmm7, %xmm6
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm5
+; SSE2-NEXT: pxor %xmm0, %xmm5
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm4
+; SSE2-NEXT: pxor %xmm0, %xmm4
+; SSE2-NEXT: packssdw %xmm5, %xmm4
+; SSE2-NEXT: packsswb %xmm6, %xmm4
+; SSE2-NEXT: pmovmskb %xmm4, %eax
+; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: movd %xmm12, %ecx
+; SSE2-NEXT: jne .LBB10_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB10_3
; SSE2-NEXT: .LBB10_4: # %else2
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm4
-; SSE2-NEXT: pxor %xmm4, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %ecx
-; SSE2-NEXT: testb $1, %cl
-; SSE2-NEXT: je .LBB10_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: shrl $16, %ecx
-; SSE2-NEXT: movb %cl, 2(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB10_5
; SSE2-NEXT: .LBB10_6: # %else4
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm4
-; SSE2-NEXT: pextrw $6, %xmm4, %ecx
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testb $8, %al
; SSE2-NEXT: je .LBB10_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
-; SSE2-NEXT: shrl $24, %eax
-; SSE2-NEXT: movb %al, 3(%rdi)
+; SSE2-NEXT: .LBB10_7: # %cond.store5
+; SSE2-NEXT: shrl $24, %ecx
+; SSE2-NEXT: movb %cl, 3(%rdi)
; SSE2-NEXT: .LBB10_8: # %else6
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm5, %xmm0
-; SSE2-NEXT: movd %xmm0, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: pextrw $2, %xmm12, %eax
+; SSE2-NEXT: testb $16, %al
+; SSE2-NEXT: pextrw $2, %xmm12, %ecx
; SSE2-NEXT: je .LBB10_10
; SSE2-NEXT: # %bb.9: # %cond.store7
-; SSE2-NEXT: movb %al, 4(%rdi)
+; SSE2-NEXT: movb %cl, 4(%rdi)
; SSE2-NEXT: .LBB10_10: # %else8
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE2-NEXT: pxor %xmm2, %xmm0
-; SSE2-NEXT: pextrw $2, %xmm0, %ecx
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testb $32, %al
; SSE2-NEXT: je .LBB10_12
; SSE2-NEXT: # %bb.11: # %cond.store9
-; SSE2-NEXT: movb %ah, 5(%rdi)
+; SSE2-NEXT: movb %ch, 5(%rdi)
; SSE2-NEXT: .LBB10_12: # %else10
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm5
-; SSE2-NEXT: pxor %xmm5, %xmm2
-; SSE2-NEXT: pextrw $4, %xmm2, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: pextrw $3, %xmm12, %eax
+; SSE2-NEXT: testb $64, %al
+; SSE2-NEXT: pextrw $3, %xmm12, %ecx
; SSE2-NEXT: je .LBB10_14
; SSE2-NEXT: # %bb.13: # %cond.store11
-; SSE2-NEXT: movb %al, 6(%rdi)
+; SSE2-NEXT: movb %cl, 6(%rdi)
; SSE2-NEXT: .LBB10_14: # %else12
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm5
-; SSE2-NEXT: pextrw $6, %xmm5, %ecx
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testb $-128, %al
; SSE2-NEXT: je .LBB10_16
; SSE2-NEXT: # %bb.15: # %cond.store13
-; SSE2-NEXT: movb %ah, 7(%rdi)
+; SSE2-NEXT: movb %ch, 7(%rdi)
; SSE2-NEXT: .LBB10_16: # %else14
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm6, %xmm0
-; SSE2-NEXT: movd %xmm0, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: pextrw $4, %xmm12, %eax
+; SSE2-NEXT: testl $256, %eax # imm = 0x100
+; SSE2-NEXT: pextrw $4, %xmm12, %ecx
; SSE2-NEXT: je .LBB10_18
; SSE2-NEXT: # %bb.17: # %cond.store15
-; SSE2-NEXT: movb %al, 8(%rdi)
+; SSE2-NEXT: movb %cl, 8(%rdi)
; SSE2-NEXT: .LBB10_18: # %else16
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE2-NEXT: pxor %xmm2, %xmm0
-; SSE2-NEXT: pextrw $2, %xmm0, %ecx
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $512, %eax # imm = 0x200
; SSE2-NEXT: je .LBB10_20
; SSE2-NEXT: # %bb.19: # %cond.store17
-; SSE2-NEXT: movb %ah, 9(%rdi)
+; SSE2-NEXT: movb %ch, 9(%rdi)
; SSE2-NEXT: .LBB10_20: # %else18
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm6
-; SSE2-NEXT: pxor %xmm6, %xmm2
-; SSE2-NEXT: pextrw $4, %xmm2, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: pextrw $5, %xmm12, %eax
+; SSE2-NEXT: testl $1024, %eax # imm = 0x400
+; SSE2-NEXT: pextrw $5, %xmm12, %ecx
; SSE2-NEXT: je .LBB10_22
; SSE2-NEXT: # %bb.21: # %cond.store19
-; SSE2-NEXT: movb %al, 10(%rdi)
+; SSE2-NEXT: movb %cl, 10(%rdi)
; SSE2-NEXT: .LBB10_22: # %else20
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm6
-; SSE2-NEXT: pextrw $6, %xmm6, %ecx
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $2048, %eax # imm = 0x800
; SSE2-NEXT: je .LBB10_24
; SSE2-NEXT: # %bb.23: # %cond.store21
-; SSE2-NEXT: movb %ah, 11(%rdi)
+; SSE2-NEXT: movb %ch, 11(%rdi)
; SSE2-NEXT: .LBB10_24: # %else22
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm7, %xmm0
-; SSE2-NEXT: movd %xmm0, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: pextrw $6, %xmm12, %eax
+; SSE2-NEXT: testl $4096, %eax # imm = 0x1000
+; SSE2-NEXT: pextrw $6, %xmm12, %ecx
; SSE2-NEXT: je .LBB10_26
; SSE2-NEXT: # %bb.25: # %cond.store23
-; SSE2-NEXT: movb %al, 12(%rdi)
+; SSE2-NEXT: movb %cl, 12(%rdi)
; SSE2-NEXT: .LBB10_26: # %else24
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE2-NEXT: pxor %xmm2, %xmm0
-; SSE2-NEXT: pextrw $2, %xmm0, %ecx
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $8192, %eax # imm = 0x2000
; SSE2-NEXT: je .LBB10_28
; SSE2-NEXT: # %bb.27: # %cond.store25
-; SSE2-NEXT: movb %ah, 13(%rdi)
+; SSE2-NEXT: movb %ch, 13(%rdi)
; SSE2-NEXT: .LBB10_28: # %else26
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm7
-; SSE2-NEXT: pxor %xmm7, %xmm2
-; SSE2-NEXT: pextrw $4, %xmm2, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: pextrw $7, %xmm12, %eax
-; SSE2-NEXT: je .LBB10_30
-; SSE2-NEXT: # %bb.29: # %cond.store27
-; SSE2-NEXT: movb %al, 14(%rdi)
-; SSE2-NEXT: .LBB10_30: # %else28
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm7
-; SSE2-NEXT: pextrw $6, %xmm7, %ecx
-; SSE2-NEXT: testb $1, %cl
-; SSE2-NEXT: je .LBB10_32
-; SSE2-NEXT: # %bb.31: # %cond.store29
-; SSE2-NEXT: movb %ah, 15(%rdi)
+; SSE2-NEXT: testl $16384, %eax # imm = 0x4000
+; SSE2-NEXT: pextrw $7, %xmm12, %ecx
+; SSE2-NEXT: jne .LBB10_29
+; SSE2-NEXT: # %bb.30: # %else28
+; SSE2-NEXT: testl $32768, %eax # imm = 0x8000
+; SSE2-NEXT: jne .LBB10_31
; SSE2-NEXT: .LBB10_32: # %else30
; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB10_1: # %cond.store
+; SSE2-NEXT: movb %cl, (%rdi)
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: je .LBB10_4
+; SSE2-NEXT: .LBB10_3: # %cond.store1
+; SSE2-NEXT: movb %ch, 1(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: je .LBB10_6
+; SSE2-NEXT: .LBB10_5: # %cond.store3
+; SSE2-NEXT: movl %ecx, %edx
+; SSE2-NEXT: shrl $16, %edx
+; SSE2-NEXT: movb %dl, 2(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB10_7
+; SSE2-NEXT: jmp .LBB10_8
+; SSE2-NEXT: .LBB10_29: # %cond.store27
+; SSE2-NEXT: movb %cl, 14(%rdi)
+; SSE2-NEXT: testl $32768, %eax # imm = 0x8000
+; SSE2-NEXT: je .LBB10_32
+; SSE2-NEXT: .LBB10_31: # %cond.store29
+; SSE2-NEXT: movb %ch, 15(%rdi)
+; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v16i32_v16i8:
; SSE4: # %bb.0:
-; SSE4-NEXT: pxor %xmm9, %xmm9
-; SSE4-NEXT: pcmpeqd %xmm4, %xmm9
-; SSE4-NEXT: pcmpeqd %xmm8, %xmm8
-; SSE4-NEXT: pxor %xmm9, %xmm8
+; SSE4-NEXT: pxor %xmm8, %xmm8
; SSE4-NEXT: movdqa {{.*#+}} xmm9 = [255,255,255,255]
; SSE4-NEXT: pminud %xmm9, %xmm1
; SSE4-NEXT: pminud %xmm9, %xmm0
@@ -3605,496 +3380,407 @@ define void @truncstore_v16i32_v16i8(<16
; SSE4-NEXT: pminud %xmm9, %xmm2
; SSE4-NEXT: packusdw %xmm3, %xmm2
; SSE4-NEXT: packuswb %xmm2, %xmm0
-; SSE4-NEXT: pextrb $0, %xmm8, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB10_2
-; SSE4-NEXT: # %bb.1: # %cond.store
-; SSE4-NEXT: pextrb $0, %xmm0, (%rdi)
-; SSE4-NEXT: .LBB10_2: # %else
-; SSE4-NEXT: pextrb $4, %xmm8, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB10_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
-; SSE4-NEXT: pextrb $1, %xmm0, 1(%rdi)
-; SSE4-NEXT: .LBB10_4: # %else2
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm4
+; SSE4-NEXT: pcmpeqd %xmm8, %xmm7
; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm4, %xmm1
-; SSE4-NEXT: pextrb $8, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB10_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
-; SSE4-NEXT: pextrb $2, %xmm0, 2(%rdi)
+; SSE4-NEXT: pxor %xmm1, %xmm7
+; SSE4-NEXT: pcmpeqd %xmm8, %xmm6
+; SSE4-NEXT: pxor %xmm1, %xmm6
+; SSE4-NEXT: packssdw %xmm7, %xmm6
+; SSE4-NEXT: pcmpeqd %xmm8, %xmm5
+; SSE4-NEXT: pxor %xmm1, %xmm5
+; SSE4-NEXT: pcmpeqd %xmm8, %xmm4
+; SSE4-NEXT: pxor %xmm1, %xmm4
+; SSE4-NEXT: packssdw %xmm5, %xmm4
+; SSE4-NEXT: packsswb %xmm6, %xmm4
+; SSE4-NEXT: pmovmskb %xmm4, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB10_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB10_3
+; SSE4-NEXT: .LBB10_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB10_5
; SSE4-NEXT: .LBB10_6: # %else4
-; SSE4-NEXT: pextrb $12, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB10_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
-; SSE4-NEXT: pextrb $3, %xmm0, 3(%rdi)
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB10_7
; SSE4-NEXT: .LBB10_8: # %else6
-; SSE4-NEXT: pxor %xmm2, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm5, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm2, %xmm1
-; SSE4-NEXT: pextrb $0, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB10_10
-; SSE4-NEXT: # %bb.9: # %cond.store7
-; SSE4-NEXT: pextrb $4, %xmm0, 4(%rdi)
+; SSE4-NEXT: testb $16, %al
+; SSE4-NEXT: jne .LBB10_9
; SSE4-NEXT: .LBB10_10: # %else8
-; SSE4-NEXT: pextrb $4, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB10_12
-; SSE4-NEXT: # %bb.11: # %cond.store9
-; SSE4-NEXT: pextrb $5, %xmm0, 5(%rdi)
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: jne .LBB10_11
; SSE4-NEXT: .LBB10_12: # %else10
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm5
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm5, %xmm1
-; SSE4-NEXT: pextrb $8, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $64, %al
+; SSE4-NEXT: jne .LBB10_13
+; SSE4-NEXT: .LBB10_14: # %else12
+; SSE4-NEXT: testb $-128, %al
+; SSE4-NEXT: jne .LBB10_15
+; SSE4-NEXT: .LBB10_16: # %else14
+; SSE4-NEXT: testl $256, %eax # imm = 0x100
+; SSE4-NEXT: jne .LBB10_17
+; SSE4-NEXT: .LBB10_18: # %else16
+; SSE4-NEXT: testl $512, %eax # imm = 0x200
+; SSE4-NEXT: jne .LBB10_19
+; SSE4-NEXT: .LBB10_20: # %else18
+; SSE4-NEXT: testl $1024, %eax # imm = 0x400
+; SSE4-NEXT: jne .LBB10_21
+; SSE4-NEXT: .LBB10_22: # %else20
+; SSE4-NEXT: testl $2048, %eax # imm = 0x800
+; SSE4-NEXT: jne .LBB10_23
+; SSE4-NEXT: .LBB10_24: # %else22
+; SSE4-NEXT: testl $4096, %eax # imm = 0x1000
+; SSE4-NEXT: jne .LBB10_25
+; SSE4-NEXT: .LBB10_26: # %else24
+; SSE4-NEXT: testl $8192, %eax # imm = 0x2000
+; SSE4-NEXT: jne .LBB10_27
+; SSE4-NEXT: .LBB10_28: # %else26
+; SSE4-NEXT: testl $16384, %eax # imm = 0x4000
+; SSE4-NEXT: jne .LBB10_29
+; SSE4-NEXT: .LBB10_30: # %else28
+; SSE4-NEXT: testl $32768, %eax # imm = 0x8000
+; SSE4-NEXT: jne .LBB10_31
+; SSE4-NEXT: .LBB10_32: # %else30
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB10_1: # %cond.store
+; SSE4-NEXT: pextrb $0, %xmm0, (%rdi)
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: je .LBB10_4
+; SSE4-NEXT: .LBB10_3: # %cond.store1
+; SSE4-NEXT: pextrb $1, %xmm0, 1(%rdi)
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: je .LBB10_6
+; SSE4-NEXT: .LBB10_5: # %cond.store3
+; SSE4-NEXT: pextrb $2, %xmm0, 2(%rdi)
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: je .LBB10_8
+; SSE4-NEXT: .LBB10_7: # %cond.store5
+; SSE4-NEXT: pextrb $3, %xmm0, 3(%rdi)
+; SSE4-NEXT: testb $16, %al
+; SSE4-NEXT: je .LBB10_10
+; SSE4-NEXT: .LBB10_9: # %cond.store7
+; SSE4-NEXT: pextrb $4, %xmm0, 4(%rdi)
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: je .LBB10_12
+; SSE4-NEXT: .LBB10_11: # %cond.store9
+; SSE4-NEXT: pextrb $5, %xmm0, 5(%rdi)
+; SSE4-NEXT: testb $64, %al
; SSE4-NEXT: je .LBB10_14
-; SSE4-NEXT: # %bb.13: # %cond.store11
+; SSE4-NEXT: .LBB10_13: # %cond.store11
; SSE4-NEXT: pextrb $6, %xmm0, 6(%rdi)
-; SSE4-NEXT: .LBB10_14: # %else12
-; SSE4-NEXT: pextrb $12, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $-128, %al
; SSE4-NEXT: je .LBB10_16
-; SSE4-NEXT: # %bb.15: # %cond.store13
+; SSE4-NEXT: .LBB10_15: # %cond.store13
; SSE4-NEXT: pextrb $7, %xmm0, 7(%rdi)
-; SSE4-NEXT: .LBB10_16: # %else14
-; SSE4-NEXT: pxor %xmm2, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm6, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm2, %xmm1
-; SSE4-NEXT: pextrb $0, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $256, %eax # imm = 0x100
; SSE4-NEXT: je .LBB10_18
-; SSE4-NEXT: # %bb.17: # %cond.store15
+; SSE4-NEXT: .LBB10_17: # %cond.store15
; SSE4-NEXT: pextrb $8, %xmm0, 8(%rdi)
-; SSE4-NEXT: .LBB10_18: # %else16
-; SSE4-NEXT: pextrb $4, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $512, %eax # imm = 0x200
; SSE4-NEXT: je .LBB10_20
-; SSE4-NEXT: # %bb.19: # %cond.store17
+; SSE4-NEXT: .LBB10_19: # %cond.store17
; SSE4-NEXT: pextrb $9, %xmm0, 9(%rdi)
-; SSE4-NEXT: .LBB10_20: # %else18
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm6
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm6, %xmm1
-; SSE4-NEXT: pextrb $8, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $1024, %eax # imm = 0x400
; SSE4-NEXT: je .LBB10_22
-; SSE4-NEXT: # %bb.21: # %cond.store19
+; SSE4-NEXT: .LBB10_21: # %cond.store19
; SSE4-NEXT: pextrb $10, %xmm0, 10(%rdi)
-; SSE4-NEXT: .LBB10_22: # %else20
-; SSE4-NEXT: pextrb $12, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $2048, %eax # imm = 0x800
; SSE4-NEXT: je .LBB10_24
-; SSE4-NEXT: # %bb.23: # %cond.store21
+; SSE4-NEXT: .LBB10_23: # %cond.store21
; SSE4-NEXT: pextrb $11, %xmm0, 11(%rdi)
-; SSE4-NEXT: .LBB10_24: # %else22
-; SSE4-NEXT: pxor %xmm2, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm7, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm2, %xmm1
-; SSE4-NEXT: pextrb $0, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $4096, %eax # imm = 0x1000
; SSE4-NEXT: je .LBB10_26
-; SSE4-NEXT: # %bb.25: # %cond.store23
+; SSE4-NEXT: .LBB10_25: # %cond.store23
; SSE4-NEXT: pextrb $12, %xmm0, 12(%rdi)
-; SSE4-NEXT: .LBB10_26: # %else24
-; SSE4-NEXT: pextrb $4, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $8192, %eax # imm = 0x2000
; SSE4-NEXT: je .LBB10_28
-; SSE4-NEXT: # %bb.27: # %cond.store25
+; SSE4-NEXT: .LBB10_27: # %cond.store25
; SSE4-NEXT: pextrb $13, %xmm0, 13(%rdi)
-; SSE4-NEXT: .LBB10_28: # %else26
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm7
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm7, %xmm1
-; SSE4-NEXT: pextrb $8, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $16384, %eax # imm = 0x4000
; SSE4-NEXT: je .LBB10_30
-; SSE4-NEXT: # %bb.29: # %cond.store27
+; SSE4-NEXT: .LBB10_29: # %cond.store27
; SSE4-NEXT: pextrb $14, %xmm0, 14(%rdi)
-; SSE4-NEXT: .LBB10_30: # %else28
-; SSE4-NEXT: pextrb $12, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $32768, %eax # imm = 0x8000
; SSE4-NEXT: je .LBB10_32
-; SSE4-NEXT: # %bb.31: # %cond.store29
+; SSE4-NEXT: .LBB10_31: # %cond.store29
; SSE4-NEXT: pextrb $15, %xmm0, 15(%rdi)
-; SSE4-NEXT: .LBB10_32: # %else30
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v16i32_v16i8:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpxor %xmm8, %xmm8, %xmm8
-; AVX1-NEXT: vpcmpeqd %xmm8, %xmm2, %xmm6
-; AVX1-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
-; AVX1-NEXT: vpxor %xmm5, %xmm6, %xmm6
-; AVX1-NEXT: vpackssdw %xmm0, %xmm6, %xmm6
-; AVX1-NEXT: vpacksswb %xmm0, %xmm6, %xmm6
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm7
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255]
-; AVX1-NEXT: vpminud %xmm4, %xmm7, %xmm7
-; AVX1-NEXT: vpminud %xmm4, %xmm0, %xmm0
-; AVX1-NEXT: vpackusdw %xmm7, %xmm0, %xmm0
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm7
-; AVX1-NEXT: vpminud %xmm4, %xmm7, %xmm7
-; AVX1-NEXT: vpminud %xmm4, %xmm1, %xmm1
-; AVX1-NEXT: vpackusdw %xmm7, %xmm1, %xmm1
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm4
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm5 = [255,255,255,255]
+; AVX1-NEXT: vpminud %xmm5, %xmm4, %xmm4
+; AVX1-NEXT: vpminud %xmm5, %xmm0, %xmm0
+; AVX1-NEXT: vpackusdw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm4
+; AVX1-NEXT: vpminud %xmm5, %xmm4, %xmm4
+; AVX1-NEXT: vpminud %xmm5, %xmm1, %xmm1
+; AVX1-NEXT: vpackusdw %xmm4, %xmm1, %xmm1
; AVX1-NEXT: vpackuswb %xmm1, %xmm0, %xmm0
-; AVX1-NEXT: vpextrb $0, %xmm6, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB10_2
-; AVX1-NEXT: # %bb.1: # %cond.store
-; AVX1-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX1-NEXT: .LBB10_2: # %else
-; AVX1-NEXT: vpcmpeqd %xmm8, %xmm2, %xmm1
+; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm1
+; AVX1-NEXT: vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT: vpcmpeqd %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
; AVX1-NEXT: vpxor %xmm5, %xmm1, %xmm1
-; AVX1-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $1, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: vpcmpeqd %xmm4, %xmm3, %xmm3
+; AVX1-NEXT: vpxor %xmm5, %xmm3, %xmm3
+; AVX1-NEXT: vpackssdw %xmm1, %xmm3, %xmm1
+; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm3
+; AVX1-NEXT: vpcmpeqd %xmm4, %xmm3, %xmm3
+; AVX1-NEXT: vpxor %xmm5, %xmm3, %xmm3
+; AVX1-NEXT: vpcmpeqd %xmm4, %xmm2, %xmm2
+; AVX1-NEXT: vpxor %xmm5, %xmm2, %xmm2
+; AVX1-NEXT: vpackssdw %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vpacksswb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT: vpmovmskb %xmm1, %eax
+; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: jne .LBB10_1
+; AVX1-NEXT: # %bb.2: # %else
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: jne .LBB10_3
+; AVX1-NEXT: .LBB10_4: # %else2
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: jne .LBB10_5
+; AVX1-NEXT: .LBB10_6: # %else4
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: jne .LBB10_7
+; AVX1-NEXT: .LBB10_8: # %else6
+; AVX1-NEXT: testb $16, %al
+; AVX1-NEXT: jne .LBB10_9
+; AVX1-NEXT: .LBB10_10: # %else8
+; AVX1-NEXT: testb $32, %al
+; AVX1-NEXT: jne .LBB10_11
+; AVX1-NEXT: .LBB10_12: # %else10
+; AVX1-NEXT: testb $64, %al
+; AVX1-NEXT: jne .LBB10_13
+; AVX1-NEXT: .LBB10_14: # %else12
+; AVX1-NEXT: testb $-128, %al
+; AVX1-NEXT: jne .LBB10_15
+; AVX1-NEXT: .LBB10_16: # %else14
+; AVX1-NEXT: testl $256, %eax # imm = 0x100
+; AVX1-NEXT: jne .LBB10_17
+; AVX1-NEXT: .LBB10_18: # %else16
+; AVX1-NEXT: testl $512, %eax # imm = 0x200
+; AVX1-NEXT: jne .LBB10_19
+; AVX1-NEXT: .LBB10_20: # %else18
+; AVX1-NEXT: testl $1024, %eax # imm = 0x400
+; AVX1-NEXT: jne .LBB10_21
+; AVX1-NEXT: .LBB10_22: # %else20
+; AVX1-NEXT: testl $2048, %eax # imm = 0x800
+; AVX1-NEXT: jne .LBB10_23
+; AVX1-NEXT: .LBB10_24: # %else22
+; AVX1-NEXT: testl $4096, %eax # imm = 0x1000
+; AVX1-NEXT: jne .LBB10_25
+; AVX1-NEXT: .LBB10_26: # %else24
+; AVX1-NEXT: testl $8192, %eax # imm = 0x2000
+; AVX1-NEXT: jne .LBB10_27
+; AVX1-NEXT: .LBB10_28: # %else26
+; AVX1-NEXT: testl $16384, %eax # imm = 0x4000
+; AVX1-NEXT: jne .LBB10_29
+; AVX1-NEXT: .LBB10_30: # %else28
+; AVX1-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX1-NEXT: jne .LBB10_31
+; AVX1-NEXT: .LBB10_32: # %else30
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+; AVX1-NEXT: .LBB10_1: # %cond.store
+; AVX1-NEXT: vpextrb $0, %xmm0, (%rdi)
+; AVX1-NEXT: testb $2, %al
; AVX1-NEXT: je .LBB10_4
-; AVX1-NEXT: # %bb.3: # %cond.store1
+; AVX1-NEXT: .LBB10_3: # %cond.store1
; AVX1-NEXT: vpextrb $1, %xmm0, 1(%rdi)
-; AVX1-NEXT: .LBB10_4: # %else2
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm5
-; AVX1-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpxor %xmm4, %xmm5, %xmm5
-; AVX1-NEXT: vpackssdw %xmm0, %xmm5, %xmm5
-; AVX1-NEXT: vpacksswb %xmm0, %xmm5, %xmm5
-; AVX1-NEXT: vpextrb $2, %xmm5, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $4, %al
; AVX1-NEXT: je .LBB10_6
-; AVX1-NEXT: # %bb.5: # %cond.store3
+; AVX1-NEXT: .LBB10_5: # %cond.store3
; AVX1-NEXT: vpextrb $2, %xmm0, 2(%rdi)
-; AVX1-NEXT: .LBB10_6: # %else4
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm4, %xmm1, %xmm1
-; AVX1-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $3, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $8, %al
; AVX1-NEXT: je .LBB10_8
-; AVX1-NEXT: # %bb.7: # %cond.store5
+; AVX1-NEXT: .LBB10_7: # %cond.store5
; AVX1-NEXT: vpextrb $3, %xmm0, 3(%rdi)
-; AVX1-NEXT: .LBB10_8: # %else6
-; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpxor %xmm4, %xmm2, %xmm2
-; AVX1-NEXT: vpackssdw %xmm2, %xmm0, %xmm4
-; AVX1-NEXT: vpacksswb %xmm0, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $4, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $16, %al
; AVX1-NEXT: je .LBB10_10
-; AVX1-NEXT: # %bb.9: # %cond.store7
+; AVX1-NEXT: .LBB10_9: # %cond.store7
; AVX1-NEXT: vpextrb $4, %xmm0, 4(%rdi)
-; AVX1-NEXT: .LBB10_10: # %else8
-; AVX1-NEXT: vpackssdw %xmm2, %xmm0, %xmm2
-; AVX1-NEXT: vpacksswb %xmm0, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $5, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $32, %al
; AVX1-NEXT: je .LBB10_12
-; AVX1-NEXT: # %bb.11: # %cond.store9
+; AVX1-NEXT: .LBB10_11: # %cond.store9
; AVX1-NEXT: vpextrb $5, %xmm0, 5(%rdi)
-; AVX1-NEXT: .LBB10_12: # %else10
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpackssdw %xmm1, %xmm0, %xmm2
-; AVX1-NEXT: vpacksswb %xmm0, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $6, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $64, %al
; AVX1-NEXT: je .LBB10_14
-; AVX1-NEXT: # %bb.13: # %cond.store11
+; AVX1-NEXT: .LBB10_13: # %cond.store11
; AVX1-NEXT: vpextrb $6, %xmm0, 6(%rdi)
-; AVX1-NEXT: .LBB10_14: # %else12
-; AVX1-NEXT: vpackssdw %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $7, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $-128, %al
; AVX1-NEXT: je .LBB10_16
-; AVX1-NEXT: # %bb.15: # %cond.store13
+; AVX1-NEXT: .LBB10_15: # %cond.store13
; AVX1-NEXT: vpextrb $7, %xmm0, 7(%rdi)
-; AVX1-NEXT: .LBB10_16: # %else14
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm4, %xmm4
-; AVX1-NEXT: vpackssdw %xmm0, %xmm4, %xmm4
-; AVX1-NEXT: vpacksswb %xmm4, %xmm0, %xmm4
-; AVX1-NEXT: vpextrb $8, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $256, %eax # imm = 0x100
; AVX1-NEXT: je .LBB10_18
-; AVX1-NEXT: # %bb.17: # %cond.store15
+; AVX1-NEXT: .LBB10_17: # %cond.store15
; AVX1-NEXT: vpextrb $8, %xmm0, 8(%rdi)
-; AVX1-NEXT: .LBB10_18: # %else16
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpacksswb %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vpextrb $9, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $512, %eax # imm = 0x200
; AVX1-NEXT: je .LBB10_20
-; AVX1-NEXT: # %bb.19: # %cond.store17
+; AVX1-NEXT: .LBB10_19: # %cond.store17
; AVX1-NEXT: vpextrb $9, %xmm0, 9(%rdi)
-; AVX1-NEXT: .LBB10_20: # %else18
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm4, %xmm4
-; AVX1-NEXT: vpackssdw %xmm0, %xmm4, %xmm4
-; AVX1-NEXT: vpacksswb %xmm4, %xmm0, %xmm4
-; AVX1-NEXT: vpextrb $10, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $1024, %eax # imm = 0x400
; AVX1-NEXT: je .LBB10_22
-; AVX1-NEXT: # %bb.21: # %cond.store19
+; AVX1-NEXT: .LBB10_21: # %cond.store19
; AVX1-NEXT: vpextrb $10, %xmm0, 10(%rdi)
-; AVX1-NEXT: .LBB10_22: # %else20
-; AVX1-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX1-NEXT: vpacksswb %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vpextrb $11, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $2048, %eax # imm = 0x800
; AVX1-NEXT: je .LBB10_24
-; AVX1-NEXT: # %bb.23: # %cond.store21
+; AVX1-NEXT: .LBB10_23: # %cond.store21
; AVX1-NEXT: vpextrb $11, %xmm0, 11(%rdi)
-; AVX1-NEXT: .LBB10_24: # %else22
-; AVX1-NEXT: vextractf128 $1, %ymm3, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpackssdw %xmm2, %xmm0, %xmm3
-; AVX1-NEXT: vpacksswb %xmm3, %xmm0, %xmm3
-; AVX1-NEXT: vpextrb $12, %xmm3, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $4096, %eax # imm = 0x1000
; AVX1-NEXT: je .LBB10_26
-; AVX1-NEXT: # %bb.25: # %cond.store23
+; AVX1-NEXT: .LBB10_25: # %cond.store23
; AVX1-NEXT: vpextrb $12, %xmm0, 12(%rdi)
-; AVX1-NEXT: .LBB10_26: # %else24
-; AVX1-NEXT: vpackssdw %xmm2, %xmm0, %xmm2
-; AVX1-NEXT: vpacksswb %xmm2, %xmm0, %xmm2
-; AVX1-NEXT: vpextrb $13, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $8192, %eax # imm = 0x2000
; AVX1-NEXT: je .LBB10_28
-; AVX1-NEXT: # %bb.27: # %cond.store25
+; AVX1-NEXT: .LBB10_27: # %cond.store25
; AVX1-NEXT: vpextrb $13, %xmm0, 13(%rdi)
-; AVX1-NEXT: .LBB10_28: # %else26
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpackssdw %xmm1, %xmm0, %xmm2
-; AVX1-NEXT: vpacksswb %xmm2, %xmm0, %xmm2
-; AVX1-NEXT: vpextrb $14, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $16384, %eax # imm = 0x4000
; AVX1-NEXT: je .LBB10_30
-; AVX1-NEXT: # %bb.29: # %cond.store27
+; AVX1-NEXT: .LBB10_29: # %cond.store27
; AVX1-NEXT: vpextrb $14, %xmm0, 14(%rdi)
-; AVX1-NEXT: .LBB10_30: # %else28
-; AVX1-NEXT: vpackssdw %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vpacksswb %xmm1, %xmm0, %xmm1
-; AVX1-NEXT: vpextrb $15, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $32768, %eax # imm = 0x8000
; AVX1-NEXT: je .LBB10_32
-; AVX1-NEXT: # %bb.31: # %cond.store29
+; AVX1-NEXT: .LBB10_31: # %cond.store29
; AVX1-NEXT: vpextrb $15, %xmm0, 15(%rdi)
-; AVX1-NEXT: .LBB10_32: # %else30
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: truncstore_v16i32_v16i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm4, %xmm4, %xmm4
-; AVX2-NEXT: vpcmpeqd %xmm4, %xmm2, %xmm6
-; AVX2-NEXT: vpcmpeqd %xmm5, %xmm5, %xmm5
-; AVX2-NEXT: vpxor %xmm5, %xmm6, %xmm6
-; AVX2-NEXT: vpackssdw %xmm0, %xmm6, %xmm6
-; AVX2-NEXT: vpacksswb %xmm0, %xmm6, %xmm6
-; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm7 = [255,255,255,255,255,255,255,255]
-; AVX2-NEXT: vpminud %ymm7, %ymm1, %ymm1
-; AVX2-NEXT: vpminud %ymm7, %ymm0, %ymm0
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255]
+; AVX2-NEXT: vpminud %ymm5, %ymm1, %ymm1
+; AVX2-NEXT: vpminud %ymm5, %ymm0, %ymm0
; AVX2-NEXT: vpackusdw %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm1
; AVX2-NEXT: vpackuswb %xmm1, %xmm0, %xmm0
-; AVX2-NEXT: vpextrb $0, %xmm6, %eax
-; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB10_2
-; AVX2-NEXT: # %bb.1: # %cond.store
+; AVX2-NEXT: vpcmpeqd %ymm4, %ymm3, %ymm1
+; AVX2-NEXT: vpcmpeqd %ymm3, %ymm3, %ymm3
+; AVX2-NEXT: vpxor %ymm3, %ymm1, %ymm1
+; AVX2-NEXT: vpcmpeqd %ymm4, %ymm2, %ymm2
+; AVX2-NEXT: vpxor %ymm3, %ymm2, %ymm2
+; AVX2-NEXT: vpackssdw %ymm1, %ymm2, %ymm1
+; AVX2-NEXT: vpermq {{.*#+}} ymm1 = ymm1[0,2,1,3]
+; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT: vpacksswb %xmm2, %xmm1, %xmm1
+; AVX2-NEXT: vpmovmskb %xmm1, %eax
+; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: jne .LBB10_1
+; AVX2-NEXT: # %bb.2: # %else
+; AVX2-NEXT: testb $2, %al
+; AVX2-NEXT: jne .LBB10_3
+; AVX2-NEXT: .LBB10_4: # %else2
+; AVX2-NEXT: testb $4, %al
+; AVX2-NEXT: jne .LBB10_5
+; AVX2-NEXT: .LBB10_6: # %else4
+; AVX2-NEXT: testb $8, %al
+; AVX2-NEXT: jne .LBB10_7
+; AVX2-NEXT: .LBB10_8: # %else6
+; AVX2-NEXT: testb $16, %al
+; AVX2-NEXT: jne .LBB10_9
+; AVX2-NEXT: .LBB10_10: # %else8
+; AVX2-NEXT: testb $32, %al
+; AVX2-NEXT: jne .LBB10_11
+; AVX2-NEXT: .LBB10_12: # %else10
+; AVX2-NEXT: testb $64, %al
+; AVX2-NEXT: jne .LBB10_13
+; AVX2-NEXT: .LBB10_14: # %else12
+; AVX2-NEXT: testb $-128, %al
+; AVX2-NEXT: jne .LBB10_15
+; AVX2-NEXT: .LBB10_16: # %else14
+; AVX2-NEXT: testl $256, %eax # imm = 0x100
+; AVX2-NEXT: jne .LBB10_17
+; AVX2-NEXT: .LBB10_18: # %else16
+; AVX2-NEXT: testl $512, %eax # imm = 0x200
+; AVX2-NEXT: jne .LBB10_19
+; AVX2-NEXT: .LBB10_20: # %else18
+; AVX2-NEXT: testl $1024, %eax # imm = 0x400
+; AVX2-NEXT: jne .LBB10_21
+; AVX2-NEXT: .LBB10_22: # %else20
+; AVX2-NEXT: testl $2048, %eax # imm = 0x800
+; AVX2-NEXT: jne .LBB10_23
+; AVX2-NEXT: .LBB10_24: # %else22
+; AVX2-NEXT: testl $4096, %eax # imm = 0x1000
+; AVX2-NEXT: jne .LBB10_25
+; AVX2-NEXT: .LBB10_26: # %else24
+; AVX2-NEXT: testl $8192, %eax # imm = 0x2000
+; AVX2-NEXT: jne .LBB10_27
+; AVX2-NEXT: .LBB10_28: # %else26
+; AVX2-NEXT: testl $16384, %eax # imm = 0x4000
+; AVX2-NEXT: jne .LBB10_29
+; AVX2-NEXT: .LBB10_30: # %else28
+; AVX2-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX2-NEXT: jne .LBB10_31
+; AVX2-NEXT: .LBB10_32: # %else30
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+; AVX2-NEXT: .LBB10_1: # %cond.store
; AVX2-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX2-NEXT: .LBB10_2: # %else
-; AVX2-NEXT: vpcmpeqd %xmm4, %xmm2, %xmm1
-; AVX2-NEXT: vpxor %xmm5, %xmm1, %xmm1
-; AVX2-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $1, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $2, %al
; AVX2-NEXT: je .LBB10_4
-; AVX2-NEXT: # %bb.3: # %cond.store1
+; AVX2-NEXT: .LBB10_3: # %cond.store1
; AVX2-NEXT: vpextrb $1, %xmm0, 1(%rdi)
-; AVX2-NEXT: .LBB10_4: # %else2
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm5
-; AVX2-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX2-NEXT: vpxor %xmm4, %xmm5, %xmm5
-; AVX2-NEXT: vpackssdw %xmm0, %xmm5, %xmm5
-; AVX2-NEXT: vpacksswb %xmm0, %xmm5, %xmm5
-; AVX2-NEXT: vpextrb $2, %xmm5, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $4, %al
; AVX2-NEXT: je .LBB10_6
-; AVX2-NEXT: # %bb.5: # %cond.store3
+; AVX2-NEXT: .LBB10_5: # %cond.store3
; AVX2-NEXT: vpextrb $2, %xmm0, 2(%rdi)
-; AVX2-NEXT: .LBB10_6: # %else4
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm2, %xmm1
-; AVX2-NEXT: vpxor %xmm4, %xmm1, %xmm1
-; AVX2-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $3, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $8, %al
; AVX2-NEXT: je .LBB10_8
-; AVX2-NEXT: # %bb.7: # %cond.store5
+; AVX2-NEXT: .LBB10_7: # %cond.store5
; AVX2-NEXT: vpextrb $3, %xmm0, 3(%rdi)
-; AVX2-NEXT: .LBB10_8: # %else6
-; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX2-NEXT: vpxor %xmm4, %xmm2, %xmm2
-; AVX2-NEXT: vpackssdw %xmm2, %xmm0, %xmm4
-; AVX2-NEXT: vpacksswb %xmm0, %xmm4, %xmm4
-; AVX2-NEXT: vpextrb $4, %xmm4, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $16, %al
; AVX2-NEXT: je .LBB10_10
-; AVX2-NEXT: # %bb.9: # %cond.store7
+; AVX2-NEXT: .LBB10_9: # %cond.store7
; AVX2-NEXT: vpextrb $4, %xmm0, 4(%rdi)
-; AVX2-NEXT: .LBB10_10: # %else8
-; AVX2-NEXT: vpackssdw %xmm2, %xmm0, %xmm2
-; AVX2-NEXT: vpacksswb %xmm0, %xmm2, %xmm2
-; AVX2-NEXT: vpextrb $5, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $32, %al
; AVX2-NEXT: je .LBB10_12
-; AVX2-NEXT: # %bb.11: # %cond.store9
+; AVX2-NEXT: .LBB10_11: # %cond.store9
; AVX2-NEXT: vpextrb $5, %xmm0, 5(%rdi)
-; AVX2-NEXT: .LBB10_12: # %else10
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpackssdw %xmm1, %xmm0, %xmm2
-; AVX2-NEXT: vpacksswb %xmm0, %xmm2, %xmm2
-; AVX2-NEXT: vpextrb $6, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $64, %al
; AVX2-NEXT: je .LBB10_14
-; AVX2-NEXT: # %bb.13: # %cond.store11
+; AVX2-NEXT: .LBB10_13: # %cond.store11
; AVX2-NEXT: vpextrb $6, %xmm0, 6(%rdi)
-; AVX2-NEXT: .LBB10_14: # %else12
-; AVX2-NEXT: vpackssdw %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $7, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $-128, %al
; AVX2-NEXT: je .LBB10_16
-; AVX2-NEXT: # %bb.15: # %cond.store13
+; AVX2-NEXT: .LBB10_15: # %cond.store13
; AVX2-NEXT: vpextrb $7, %xmm0, 7(%rdi)
-; AVX2-NEXT: .LBB10_16: # %else14
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm4
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm4, %xmm4
-; AVX2-NEXT: vpackssdw %xmm0, %xmm4, %xmm4
-; AVX2-NEXT: vpacksswb %xmm4, %xmm0, %xmm4
-; AVX2-NEXT: vpextrb $8, %xmm4, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $256, %eax # imm = 0x100
; AVX2-NEXT: je .LBB10_18
-; AVX2-NEXT: # %bb.17: # %cond.store15
+; AVX2-NEXT: .LBB10_17: # %cond.store15
; AVX2-NEXT: vpextrb $8, %xmm0, 8(%rdi)
-; AVX2-NEXT: .LBB10_18: # %else16
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpacksswb %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vpextrb $9, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $512, %eax # imm = 0x200
; AVX2-NEXT: je .LBB10_20
-; AVX2-NEXT: # %bb.19: # %cond.store17
+; AVX2-NEXT: .LBB10_19: # %cond.store17
; AVX2-NEXT: vpextrb $9, %xmm0, 9(%rdi)
-; AVX2-NEXT: .LBB10_20: # %else18
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm4
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm4, %xmm4
-; AVX2-NEXT: vpackssdw %xmm0, %xmm4, %xmm4
-; AVX2-NEXT: vpacksswb %xmm4, %xmm0, %xmm4
-; AVX2-NEXT: vpextrb $10, %xmm4, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $1024, %eax # imm = 0x400
; AVX2-NEXT: je .LBB10_22
-; AVX2-NEXT: # %bb.21: # %cond.store19
+; AVX2-NEXT: .LBB10_21: # %cond.store19
; AVX2-NEXT: vpextrb $10, %xmm0, 10(%rdi)
-; AVX2-NEXT: .LBB10_22: # %else20
-; AVX2-NEXT: vpcmpeqd %xmm1, %xmm3, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpackssdw %xmm0, %xmm1, %xmm1
-; AVX2-NEXT: vpacksswb %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vpextrb $11, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $2048, %eax # imm = 0x800
; AVX2-NEXT: je .LBB10_24
-; AVX2-NEXT: # %bb.23: # %cond.store21
+; AVX2-NEXT: .LBB10_23: # %cond.store21
; AVX2-NEXT: vpextrb $11, %xmm0, 11(%rdi)
-; AVX2-NEXT: .LBB10_24: # %else22
-; AVX2-NEXT: vextracti128 $1, %ymm3, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX2-NEXT: vpackssdw %xmm2, %xmm0, %xmm3
-; AVX2-NEXT: vpacksswb %xmm3, %xmm0, %xmm3
-; AVX2-NEXT: vpextrb $12, %xmm3, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $4096, %eax # imm = 0x1000
; AVX2-NEXT: je .LBB10_26
-; AVX2-NEXT: # %bb.25: # %cond.store23
+; AVX2-NEXT: .LBB10_25: # %cond.store23
; AVX2-NEXT: vpextrb $12, %xmm0, 12(%rdi)
-; AVX2-NEXT: .LBB10_26: # %else24
-; AVX2-NEXT: vpackssdw %xmm2, %xmm0, %xmm2
-; AVX2-NEXT: vpacksswb %xmm2, %xmm0, %xmm2
-; AVX2-NEXT: vpextrb $13, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $8192, %eax # imm = 0x2000
; AVX2-NEXT: je .LBB10_28
-; AVX2-NEXT: # %bb.27: # %cond.store25
+; AVX2-NEXT: .LBB10_27: # %cond.store25
; AVX2-NEXT: vpextrb $13, %xmm0, 13(%rdi)
-; AVX2-NEXT: .LBB10_28: # %else26
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpackssdw %xmm1, %xmm0, %xmm2
-; AVX2-NEXT: vpacksswb %xmm2, %xmm0, %xmm2
-; AVX2-NEXT: vpextrb $14, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $16384, %eax # imm = 0x4000
; AVX2-NEXT: je .LBB10_30
-; AVX2-NEXT: # %bb.29: # %cond.store27
+; AVX2-NEXT: .LBB10_29: # %cond.store27
; AVX2-NEXT: vpextrb $14, %xmm0, 14(%rdi)
-; AVX2-NEXT: .LBB10_30: # %else28
-; AVX2-NEXT: vpackssdw %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vpacksswb %xmm1, %xmm0, %xmm1
-; AVX2-NEXT: vpextrb $15, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $32768, %eax # imm = 0x8000
; AVX2-NEXT: je .LBB10_32
-; AVX2-NEXT: # %bb.31: # %cond.store29
+; AVX2-NEXT: .LBB10_31: # %cond.store29
; AVX2-NEXT: vpextrb $15, %xmm0, 15(%rdi)
-; AVX2-NEXT: .LBB10_32: # %else30
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
@@ -4104,115 +3790,117 @@ define void @truncstore_v16i32_v16i8(<16
; AVX512F-NEXT: vpmovusdb %zmm0, %xmm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB10_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB10_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB10_3
+; AVX512F-NEXT: .LBB10_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB10_5
+; AVX512F-NEXT: .LBB10_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB10_7
+; AVX512F-NEXT: .LBB10_8: # %else6
+; AVX512F-NEXT: testb $16, %al
+; AVX512F-NEXT: jne .LBB10_9
+; AVX512F-NEXT: .LBB10_10: # %else8
+; AVX512F-NEXT: testb $32, %al
+; AVX512F-NEXT: jne .LBB10_11
+; AVX512F-NEXT: .LBB10_12: # %else10
+; AVX512F-NEXT: testb $64, %al
+; AVX512F-NEXT: jne .LBB10_13
+; AVX512F-NEXT: .LBB10_14: # %else12
+; AVX512F-NEXT: testb $-128, %al
+; AVX512F-NEXT: jne .LBB10_15
+; AVX512F-NEXT: .LBB10_16: # %else14
+; AVX512F-NEXT: testl $256, %eax # imm = 0x100
+; AVX512F-NEXT: jne .LBB10_17
+; AVX512F-NEXT: .LBB10_18: # %else16
+; AVX512F-NEXT: testl $512, %eax # imm = 0x200
+; AVX512F-NEXT: jne .LBB10_19
+; AVX512F-NEXT: .LBB10_20: # %else18
+; AVX512F-NEXT: testl $1024, %eax # imm = 0x400
+; AVX512F-NEXT: jne .LBB10_21
+; AVX512F-NEXT: .LBB10_22: # %else20
+; AVX512F-NEXT: testl $2048, %eax # imm = 0x800
+; AVX512F-NEXT: jne .LBB10_23
+; AVX512F-NEXT: .LBB10_24: # %else22
+; AVX512F-NEXT: testl $4096, %eax # imm = 0x1000
+; AVX512F-NEXT: jne .LBB10_25
+; AVX512F-NEXT: .LBB10_26: # %else24
+; AVX512F-NEXT: testl $8192, %eax # imm = 0x2000
+; AVX512F-NEXT: jne .LBB10_27
+; AVX512F-NEXT: .LBB10_28: # %else26
+; AVX512F-NEXT: testl $16384, %eax # imm = 0x4000
+; AVX512F-NEXT: jne .LBB10_29
+; AVX512F-NEXT: .LBB10_30: # %else28
+; AVX512F-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX512F-NEXT: jne .LBB10_31
+; AVX512F-NEXT: .LBB10_32: # %else30
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB10_1: # %cond.store
; AVX512F-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB10_2: # %else
-; AVX512F-NEXT: kshiftrw $1, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB10_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB10_3: # %cond.store1
; AVX512F-NEXT: vpextrb $1, %xmm0, 1(%rdi)
-; AVX512F-NEXT: .LBB10_4: # %else2
-; AVX512F-NEXT: kshiftrw $2, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB10_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB10_5: # %cond.store3
; AVX512F-NEXT: vpextrb $2, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB10_6: # %else4
-; AVX512F-NEXT: kshiftrw $3, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB10_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB10_7: # %cond.store5
; AVX512F-NEXT: vpextrb $3, %xmm0, 3(%rdi)
-; AVX512F-NEXT: .LBB10_8: # %else6
-; AVX512F-NEXT: kshiftrw $4, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $16, %al
; AVX512F-NEXT: je .LBB10_10
-; AVX512F-NEXT: # %bb.9: # %cond.store7
+; AVX512F-NEXT: .LBB10_9: # %cond.store7
; AVX512F-NEXT: vpextrb $4, %xmm0, 4(%rdi)
-; AVX512F-NEXT: .LBB10_10: # %else8
-; AVX512F-NEXT: kshiftrw $5, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $32, %al
; AVX512F-NEXT: je .LBB10_12
-; AVX512F-NEXT: # %bb.11: # %cond.store9
+; AVX512F-NEXT: .LBB10_11: # %cond.store9
; AVX512F-NEXT: vpextrb $5, %xmm0, 5(%rdi)
-; AVX512F-NEXT: .LBB10_12: # %else10
-; AVX512F-NEXT: kshiftrw $6, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $64, %al
; AVX512F-NEXT: je .LBB10_14
-; AVX512F-NEXT: # %bb.13: # %cond.store11
+; AVX512F-NEXT: .LBB10_13: # %cond.store11
; AVX512F-NEXT: vpextrb $6, %xmm0, 6(%rdi)
-; AVX512F-NEXT: .LBB10_14: # %else12
-; AVX512F-NEXT: kshiftrw $7, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $-128, %al
; AVX512F-NEXT: je .LBB10_16
-; AVX512F-NEXT: # %bb.15: # %cond.store13
+; AVX512F-NEXT: .LBB10_15: # %cond.store13
; AVX512F-NEXT: vpextrb $7, %xmm0, 7(%rdi)
-; AVX512F-NEXT: .LBB10_16: # %else14
-; AVX512F-NEXT: kshiftrw $8, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $256, %eax # imm = 0x100
; AVX512F-NEXT: je .LBB10_18
-; AVX512F-NEXT: # %bb.17: # %cond.store15
+; AVX512F-NEXT: .LBB10_17: # %cond.store15
; AVX512F-NEXT: vpextrb $8, %xmm0, 8(%rdi)
-; AVX512F-NEXT: .LBB10_18: # %else16
-; AVX512F-NEXT: kshiftrw $9, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $512, %eax # imm = 0x200
; AVX512F-NEXT: je .LBB10_20
-; AVX512F-NEXT: # %bb.19: # %cond.store17
+; AVX512F-NEXT: .LBB10_19: # %cond.store17
; AVX512F-NEXT: vpextrb $9, %xmm0, 9(%rdi)
-; AVX512F-NEXT: .LBB10_20: # %else18
-; AVX512F-NEXT: kshiftrw $10, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $1024, %eax # imm = 0x400
; AVX512F-NEXT: je .LBB10_22
-; AVX512F-NEXT: # %bb.21: # %cond.store19
+; AVX512F-NEXT: .LBB10_21: # %cond.store19
; AVX512F-NEXT: vpextrb $10, %xmm0, 10(%rdi)
-; AVX512F-NEXT: .LBB10_22: # %else20
-; AVX512F-NEXT: kshiftrw $11, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $2048, %eax # imm = 0x800
; AVX512F-NEXT: je .LBB10_24
-; AVX512F-NEXT: # %bb.23: # %cond.store21
+; AVX512F-NEXT: .LBB10_23: # %cond.store21
; AVX512F-NEXT: vpextrb $11, %xmm0, 11(%rdi)
-; AVX512F-NEXT: .LBB10_24: # %else22
-; AVX512F-NEXT: kshiftrw $12, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $4096, %eax # imm = 0x1000
; AVX512F-NEXT: je .LBB10_26
-; AVX512F-NEXT: # %bb.25: # %cond.store23
+; AVX512F-NEXT: .LBB10_25: # %cond.store23
; AVX512F-NEXT: vpextrb $12, %xmm0, 12(%rdi)
-; AVX512F-NEXT: .LBB10_26: # %else24
-; AVX512F-NEXT: kshiftrw $13, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $8192, %eax # imm = 0x2000
; AVX512F-NEXT: je .LBB10_28
-; AVX512F-NEXT: # %bb.27: # %cond.store25
+; AVX512F-NEXT: .LBB10_27: # %cond.store25
; AVX512F-NEXT: vpextrb $13, %xmm0, 13(%rdi)
-; AVX512F-NEXT: .LBB10_28: # %else26
-; AVX512F-NEXT: kshiftrw $14, %k0, %k1
-; AVX512F-NEXT: kmovw %k1, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $16384, %eax # imm = 0x4000
; AVX512F-NEXT: je .LBB10_30
-; AVX512F-NEXT: # %bb.29: # %cond.store27
+; AVX512F-NEXT: .LBB10_29: # %cond.store27
; AVX512F-NEXT: vpextrb $14, %xmm0, 14(%rdi)
-; AVX512F-NEXT: .LBB10_30: # %else28
-; AVX512F-NEXT: kshiftrw $15, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $32768, %eax # imm = 0x8000
; AVX512F-NEXT: je .LBB10_32
-; AVX512F-NEXT: # %bb.31: # %cond.store29
+; AVX512F-NEXT: .LBB10_31: # %cond.store29
; AVX512F-NEXT: vpextrb $15, %xmm0, 15(%rdi)
-; AVX512F-NEXT: .LBB10_32: # %else30
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -4242,329 +3930,310 @@ define void @truncstore_v16i32_v16i8(<16
define void @truncstore_v8i32_v8i16(<8 x i32> %x, <8 x i16>* %p, <8 x i32> %mask) {
; SSE2-LABEL: truncstore_v8i32_v8i16:
; SSE2: # %bb.0:
-; SSE2-NEXT: pxor %xmm10, %xmm10
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm10
-; SSE2-NEXT: pcmpeqd %xmm9, %xmm9
-; SSE2-NEXT: pxor %xmm9, %xmm10
-; SSE2-NEXT: movdqa %xmm10, %xmm8
-; SSE2-NEXT: packssdw %xmm0, %xmm8
-; SSE2-NEXT: movdqa {{.*#+}} xmm6 = [2147483648,2147483648,2147483648,2147483648]
-; SSE2-NEXT: movdqa %xmm0, %xmm7
-; SSE2-NEXT: pxor %xmm6, %xmm7
+; SSE2-NEXT: pxor %xmm8, %xmm8
+; SSE2-NEXT: movdqa {{.*#+}} xmm7 = [2147483648,2147483648,2147483648,2147483648]
+; SSE2-NEXT: movdqa %xmm0, %xmm6
+; SSE2-NEXT: pxor %xmm7, %xmm6
; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [2147549183,2147549183,2147549183,2147549183]
; SSE2-NEXT: movdqa %xmm4, %xmm5
-; SSE2-NEXT: pcmpgtd %xmm7, %xmm5
+; SSE2-NEXT: pcmpgtd %xmm6, %xmm5
+; SSE2-NEXT: pcmpeqd %xmm6, %xmm6
; SSE2-NEXT: pand %xmm5, %xmm0
-; SSE2-NEXT: pxor %xmm9, %xmm5
+; SSE2-NEXT: pxor %xmm6, %xmm5
; SSE2-NEXT: por %xmm0, %xmm5
-; SSE2-NEXT: pxor %xmm1, %xmm6
-; SSE2-NEXT: pcmpgtd %xmm6, %xmm4
-; SSE2-NEXT: pxor %xmm4, %xmm9
-; SSE2-NEXT: pand %xmm1, %xmm4
-; SSE2-NEXT: por %xmm9, %xmm4
+; SSE2-NEXT: pxor %xmm1, %xmm7
+; SSE2-NEXT: pcmpgtd %xmm7, %xmm4
+; SSE2-NEXT: pand %xmm4, %xmm1
+; SSE2-NEXT: pxor %xmm6, %xmm4
+; SSE2-NEXT: por %xmm1, %xmm4
; SSE2-NEXT: pslld $16, %xmm4
; SSE2-NEXT: psrad $16, %xmm4
; SSE2-NEXT: pslld $16, %xmm5
; SSE2-NEXT: psrad $16, %xmm5
; SSE2-NEXT: packssdw %xmm4, %xmm5
-; SSE2-NEXT: movd %xmm8, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB11_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movd %xmm5, %eax
-; SSE2-NEXT: movw %ax, (%rdi)
-; SSE2-NEXT: .LBB11_2: # %else
-; SSE2-NEXT: psrlq $16, %xmm10
-; SSE2-NEXT: movd %xmm10, %eax
-; SSE2-NEXT: shrl $16, %eax
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm3
+; SSE2-NEXT: pxor %xmm6, %xmm3
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm2
+; SSE2-NEXT: pxor %xmm6, %xmm2
+; SSE2-NEXT: packssdw %xmm3, %xmm2
+; SSE2-NEXT: packsswb %xmm0, %xmm2
+; SSE2-NEXT: pmovmskb %xmm2, %eax
; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB11_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: pextrw $1, %xmm5, %eax
-; SSE2-NEXT: movw %ax, 2(%rdi)
+; SSE2-NEXT: jne .LBB11_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB11_3
; SSE2-NEXT: .LBB11_4: # %else2
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm2
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm2, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB11_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: pextrw $2, %xmm5, %eax
-; SSE2-NEXT: movw %ax, 4(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB11_5
; SSE2-NEXT: .LBB11_6: # %else4
-; SSE2-NEXT: pextrw $6, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB11_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
-; SSE2-NEXT: pextrw $3, %xmm5, %eax
-; SSE2-NEXT: movw %ax, 6(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB11_7
; SSE2-NEXT: .LBB11_8: # %else6
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm1, %xmm0
-; SSE2-NEXT: pextrw $0, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB11_10
-; SSE2-NEXT: # %bb.9: # %cond.store7
-; SSE2-NEXT: pextrw $4, %xmm5, %eax
-; SSE2-NEXT: movw %ax, 8(%rdi)
+; SSE2-NEXT: testb $16, %al
+; SSE2-NEXT: jne .LBB11_9
; SSE2-NEXT: .LBB11_10: # %else8
-; SSE2-NEXT: pextrw $2, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB11_12
-; SSE2-NEXT: # %bb.11: # %cond.store9
-; SSE2-NEXT: pextrw $5, %xmm5, %eax
-; SSE2-NEXT: movw %ax, 10(%rdi)
+; SSE2-NEXT: testb $32, %al
+; SSE2-NEXT: jne .LBB11_11
; SSE2-NEXT: .LBB11_12: # %else10
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm3, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB11_14
-; SSE2-NEXT: # %bb.13: # %cond.store11
-; SSE2-NEXT: pextrw $6, %xmm5, %eax
-; SSE2-NEXT: movw %ax, 12(%rdi)
+; SSE2-NEXT: testb $64, %al
+; SSE2-NEXT: jne .LBB11_13
; SSE2-NEXT: .LBB11_14: # %else12
-; SSE2-NEXT: pextrw $6, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $-128, %al
+; SSE2-NEXT: jne .LBB11_15
+; SSE2-NEXT: .LBB11_16: # %else14
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB11_1: # %cond.store
+; SSE2-NEXT: movd %xmm5, %ecx
+; SSE2-NEXT: movw %cx, (%rdi)
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: je .LBB11_4
+; SSE2-NEXT: .LBB11_3: # %cond.store1
+; SSE2-NEXT: pextrw $1, %xmm5, %ecx
+; SSE2-NEXT: movw %cx, 2(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: je .LBB11_6
+; SSE2-NEXT: .LBB11_5: # %cond.store3
+; SSE2-NEXT: pextrw $2, %xmm5, %ecx
+; SSE2-NEXT: movw %cx, 4(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: je .LBB11_8
+; SSE2-NEXT: .LBB11_7: # %cond.store5
+; SSE2-NEXT: pextrw $3, %xmm5, %ecx
+; SSE2-NEXT: movw %cx, 6(%rdi)
+; SSE2-NEXT: testb $16, %al
+; SSE2-NEXT: je .LBB11_10
+; SSE2-NEXT: .LBB11_9: # %cond.store7
+; SSE2-NEXT: pextrw $4, %xmm5, %ecx
+; SSE2-NEXT: movw %cx, 8(%rdi)
+; SSE2-NEXT: testb $32, %al
+; SSE2-NEXT: je .LBB11_12
+; SSE2-NEXT: .LBB11_11: # %cond.store9
+; SSE2-NEXT: pextrw $5, %xmm5, %ecx
+; SSE2-NEXT: movw %cx, 10(%rdi)
+; SSE2-NEXT: testb $64, %al
+; SSE2-NEXT: je .LBB11_14
+; SSE2-NEXT: .LBB11_13: # %cond.store11
+; SSE2-NEXT: pextrw $6, %xmm5, %ecx
+; SSE2-NEXT: movw %cx, 12(%rdi)
+; SSE2-NEXT: testb $-128, %al
; SSE2-NEXT: je .LBB11_16
-; SSE2-NEXT: # %bb.15: # %cond.store13
+; SSE2-NEXT: .LBB11_15: # %cond.store13
; SSE2-NEXT: pextrw $7, %xmm5, %eax
; SSE2-NEXT: movw %ax, 14(%rdi)
-; SSE2-NEXT: .LBB11_16: # %else14
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v8i32_v8i16:
; SSE4: # %bb.0:
-; SSE4-NEXT: pxor %xmm5, %xmm5
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm5
-; SSE4-NEXT: pcmpeqd %xmm4, %xmm4
-; SSE4-NEXT: pxor %xmm5, %xmm4
+; SSE4-NEXT: pxor %xmm4, %xmm4
; SSE4-NEXT: movdqa {{.*#+}} xmm5 = [65535,65535,65535,65535]
; SSE4-NEXT: pminud %xmm5, %xmm1
; SSE4-NEXT: pminud %xmm5, %xmm0
; SSE4-NEXT: packusdw %xmm1, %xmm0
-; SSE4-NEXT: pextrb $0, %xmm4, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB11_2
-; SSE4-NEXT: # %bb.1: # %cond.store
+; SSE4-NEXT: pcmpeqd %xmm4, %xmm3
+; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
+; SSE4-NEXT: pxor %xmm1, %xmm3
+; SSE4-NEXT: pcmpeqd %xmm4, %xmm2
+; SSE4-NEXT: pxor %xmm1, %xmm2
+; SSE4-NEXT: packssdw %xmm3, %xmm2
+; SSE4-NEXT: packsswb %xmm0, %xmm2
+; SSE4-NEXT: pmovmskb %xmm2, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB11_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB11_3
+; SSE4-NEXT: .LBB11_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB11_5
+; SSE4-NEXT: .LBB11_6: # %else4
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB11_7
+; SSE4-NEXT: .LBB11_8: # %else6
+; SSE4-NEXT: testb $16, %al
+; SSE4-NEXT: jne .LBB11_9
+; SSE4-NEXT: .LBB11_10: # %else8
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: jne .LBB11_11
+; SSE4-NEXT: .LBB11_12: # %else10
+; SSE4-NEXT: testb $64, %al
+; SSE4-NEXT: jne .LBB11_13
+; SSE4-NEXT: .LBB11_14: # %else12
+; SSE4-NEXT: testb $-128, %al
+; SSE4-NEXT: jne .LBB11_15
+; SSE4-NEXT: .LBB11_16: # %else14
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB11_1: # %cond.store
; SSE4-NEXT: pextrw $0, %xmm0, (%rdi)
-; SSE4-NEXT: .LBB11_2: # %else
-; SSE4-NEXT: pextrb $4, %xmm4, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $2, %al
; SSE4-NEXT: je .LBB11_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
+; SSE4-NEXT: .LBB11_3: # %cond.store1
; SSE4-NEXT: pextrw $1, %xmm0, 2(%rdi)
-; SSE4-NEXT: .LBB11_4: # %else2
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm2, %xmm1
-; SSE4-NEXT: pextrb $8, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $4, %al
; SSE4-NEXT: je .LBB11_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
+; SSE4-NEXT: .LBB11_5: # %cond.store3
; SSE4-NEXT: pextrw $2, %xmm0, 4(%rdi)
-; SSE4-NEXT: .LBB11_6: # %else4
-; SSE4-NEXT: pextrb $12, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $8, %al
; SSE4-NEXT: je .LBB11_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
+; SSE4-NEXT: .LBB11_7: # %cond.store5
; SSE4-NEXT: pextrw $3, %xmm0, 6(%rdi)
-; SSE4-NEXT: .LBB11_8: # %else6
-; SSE4-NEXT: pxor %xmm2, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm3, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm2, %xmm1
-; SSE4-NEXT: pextrb $0, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $16, %al
; SSE4-NEXT: je .LBB11_10
-; SSE4-NEXT: # %bb.9: # %cond.store7
+; SSE4-NEXT: .LBB11_9: # %cond.store7
; SSE4-NEXT: pextrw $4, %xmm0, 8(%rdi)
-; SSE4-NEXT: .LBB11_10: # %else8
-; SSE4-NEXT: pextrb $4, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $32, %al
; SSE4-NEXT: je .LBB11_12
-; SSE4-NEXT: # %bb.11: # %cond.store9
+; SSE4-NEXT: .LBB11_11: # %cond.store9
; SSE4-NEXT: pextrw $5, %xmm0, 10(%rdi)
-; SSE4-NEXT: .LBB11_12: # %else10
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm3
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm3, %xmm1
-; SSE4-NEXT: pextrb $8, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $64, %al
; SSE4-NEXT: je .LBB11_14
-; SSE4-NEXT: # %bb.13: # %cond.store11
+; SSE4-NEXT: .LBB11_13: # %cond.store11
; SSE4-NEXT: pextrw $6, %xmm0, 12(%rdi)
-; SSE4-NEXT: .LBB11_14: # %else12
-; SSE4-NEXT: pextrb $12, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $-128, %al
; SSE4-NEXT: je .LBB11_16
-; SSE4-NEXT: # %bb.15: # %cond.store13
+; SSE4-NEXT: .LBB11_15: # %cond.store13
; SSE4-NEXT: pextrw $7, %xmm0, 14(%rdi)
-; SSE4-NEXT: .LBB11_16: # %else14
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v8i32_v8i16:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [65535,65535,65535,65535]
-; AVX1-NEXT: vpminud %xmm6, %xmm5, %xmm5
-; AVX1-NEXT: vpminud %xmm6, %xmm0, %xmm0
-; AVX1-NEXT: vpackusdw %xmm5, %xmm0, %xmm0
-; AVX1-NEXT: vpextrb $0, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB11_2
-; AVX1-NEXT: # %bb.1: # %cond.store
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [65535,65535,65535,65535]
+; AVX1-NEXT: vpminud %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vpminud %xmm3, %xmm0, %xmm0
+; AVX1-NEXT: vpackusdw %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT: vpcmpeqd %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vpcmpeqd %xmm3, %xmm1, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT: vmovmskps %ymm1, %eax
+; AVX1-NEXT: notl %eax
+; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: jne .LBB11_1
+; AVX1-NEXT: # %bb.2: # %else
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: jne .LBB11_3
+; AVX1-NEXT: .LBB11_4: # %else2
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: jne .LBB11_5
+; AVX1-NEXT: .LBB11_6: # %else4
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: jne .LBB11_7
+; AVX1-NEXT: .LBB11_8: # %else6
+; AVX1-NEXT: testb $16, %al
+; AVX1-NEXT: jne .LBB11_9
+; AVX1-NEXT: .LBB11_10: # %else8
+; AVX1-NEXT: testb $32, %al
+; AVX1-NEXT: jne .LBB11_11
+; AVX1-NEXT: .LBB11_12: # %else10
+; AVX1-NEXT: testb $64, %al
+; AVX1-NEXT: jne .LBB11_13
+; AVX1-NEXT: .LBB11_14: # %else12
+; AVX1-NEXT: testb $-128, %al
+; AVX1-NEXT: jne .LBB11_15
+; AVX1-NEXT: .LBB11_16: # %else14
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+; AVX1-NEXT: .LBB11_1: # %cond.store
; AVX1-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX1-NEXT: .LBB11_2: # %else
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $4, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $2, %al
; AVX1-NEXT: je .LBB11_4
-; AVX1-NEXT: # %bb.3: # %cond.store1
+; AVX1-NEXT: .LBB11_3: # %cond.store1
; AVX1-NEXT: vpextrw $1, %xmm0, 2(%rdi)
-; AVX1-NEXT: .LBB11_4: # %else2
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $8, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $4, %al
; AVX1-NEXT: je .LBB11_6
-; AVX1-NEXT: # %bb.5: # %cond.store3
+; AVX1-NEXT: .LBB11_5: # %cond.store3
; AVX1-NEXT: vpextrw $2, %xmm0, 4(%rdi)
-; AVX1-NEXT: .LBB11_6: # %else4
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $12, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $8, %al
; AVX1-NEXT: je .LBB11_8
-; AVX1-NEXT: # %bb.7: # %cond.store5
+; AVX1-NEXT: .LBB11_7: # %cond.store5
; AVX1-NEXT: vpextrw $3, %xmm0, 6(%rdi)
-; AVX1-NEXT: .LBB11_8: # %else6
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $0, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $16, %al
; AVX1-NEXT: je .LBB11_10
-; AVX1-NEXT: # %bb.9: # %cond.store7
+; AVX1-NEXT: .LBB11_9: # %cond.store7
; AVX1-NEXT: vpextrw $4, %xmm0, 8(%rdi)
-; AVX1-NEXT: .LBB11_10: # %else8
-; AVX1-NEXT: vpextrb $4, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $32, %al
; AVX1-NEXT: je .LBB11_12
-; AVX1-NEXT: # %bb.11: # %cond.store9
+; AVX1-NEXT: .LBB11_11: # %cond.store9
; AVX1-NEXT: vpextrw $5, %xmm0, 10(%rdi)
-; AVX1-NEXT: .LBB11_12: # %else10
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $8, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $64, %al
; AVX1-NEXT: je .LBB11_14
-; AVX1-NEXT: # %bb.13: # %cond.store11
+; AVX1-NEXT: .LBB11_13: # %cond.store11
; AVX1-NEXT: vpextrw $6, %xmm0, 12(%rdi)
-; AVX1-NEXT: .LBB11_14: # %else12
-; AVX1-NEXT: vpextrb $12, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $-128, %al
; AVX1-NEXT: je .LBB11_16
-; AVX1-NEXT: # %bb.15: # %cond.store13
+; AVX1-NEXT: .LBB11_15: # %cond.store13
; AVX1-NEXT: vpextrw $7, %xmm0, 14(%rdi)
-; AVX1-NEXT: .LBB11_16: # %else14
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: truncstore_v8i32_v8i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm4
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm5 = [65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX2-NEXT: vpminud %ymm5, %ymm0, %ymm0
-; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm5
-; AVX2-NEXT: vpackusdw %xmm5, %xmm0, %xmm0
-; AVX2-NEXT: vpextrb $0, %xmm4, %eax
-; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB11_2
-; AVX2-NEXT: # %bb.1: # %cond.store
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm3 = [65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX2-NEXT: vpminud %ymm3, %ymm0, %ymm0
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX2-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
+; AVX2-NEXT: vpcmpeqd %ymm2, %ymm1, %ymm1
+; AVX2-NEXT: vmovmskps %ymm1, %eax
+; AVX2-NEXT: notl %eax
+; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: jne .LBB11_1
+; AVX2-NEXT: # %bb.2: # %else
+; AVX2-NEXT: testb $2, %al
+; AVX2-NEXT: jne .LBB11_3
+; AVX2-NEXT: .LBB11_4: # %else2
+; AVX2-NEXT: testb $4, %al
+; AVX2-NEXT: jne .LBB11_5
+; AVX2-NEXT: .LBB11_6: # %else4
+; AVX2-NEXT: testb $8, %al
+; AVX2-NEXT: jne .LBB11_7
+; AVX2-NEXT: .LBB11_8: # %else6
+; AVX2-NEXT: testb $16, %al
+; AVX2-NEXT: jne .LBB11_9
+; AVX2-NEXT: .LBB11_10: # %else8
+; AVX2-NEXT: testb $32, %al
+; AVX2-NEXT: jne .LBB11_11
+; AVX2-NEXT: .LBB11_12: # %else10
+; AVX2-NEXT: testb $64, %al
+; AVX2-NEXT: jne .LBB11_13
+; AVX2-NEXT: .LBB11_14: # %else12
+; AVX2-NEXT: testb $-128, %al
+; AVX2-NEXT: jne .LBB11_15
+; AVX2-NEXT: .LBB11_16: # %else14
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+; AVX2-NEXT: .LBB11_1: # %cond.store
; AVX2-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX2-NEXT: .LBB11_2: # %else
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX2-NEXT: vpextrb $4, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $2, %al
; AVX2-NEXT: je .LBB11_4
-; AVX2-NEXT: # %bb.3: # %cond.store1
+; AVX2-NEXT: .LBB11_3: # %cond.store1
; AVX2-NEXT: vpextrw $1, %xmm0, 2(%rdi)
-; AVX2-NEXT: .LBB11_4: # %else2
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm4
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX2-NEXT: vpextrb $8, %xmm4, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $4, %al
; AVX2-NEXT: je .LBB11_6
-; AVX2-NEXT: # %bb.5: # %cond.store3
+; AVX2-NEXT: .LBB11_5: # %cond.store3
; AVX2-NEXT: vpextrw $2, %xmm0, 4(%rdi)
-; AVX2-NEXT: .LBB11_6: # %else4
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX2-NEXT: vpextrb $12, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $8, %al
; AVX2-NEXT: je .LBB11_8
-; AVX2-NEXT: # %bb.7: # %cond.store5
+; AVX2-NEXT: .LBB11_7: # %cond.store5
; AVX2-NEXT: vpextrw $3, %xmm0, 6(%rdi)
-; AVX2-NEXT: .LBB11_8: # %else6
-; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX2-NEXT: vpextrb $0, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $16, %al
; AVX2-NEXT: je .LBB11_10
-; AVX2-NEXT: # %bb.9: # %cond.store7
+; AVX2-NEXT: .LBB11_9: # %cond.store7
; AVX2-NEXT: vpextrw $4, %xmm0, 8(%rdi)
-; AVX2-NEXT: .LBB11_10: # %else8
-; AVX2-NEXT: vpextrb $4, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $32, %al
; AVX2-NEXT: je .LBB11_12
-; AVX2-NEXT: # %bb.11: # %cond.store9
+; AVX2-NEXT: .LBB11_11: # %cond.store9
; AVX2-NEXT: vpextrw $5, %xmm0, 10(%rdi)
-; AVX2-NEXT: .LBB11_12: # %else10
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $8, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $64, %al
; AVX2-NEXT: je .LBB11_14
-; AVX2-NEXT: # %bb.13: # %cond.store11
+; AVX2-NEXT: .LBB11_13: # %cond.store11
; AVX2-NEXT: vpextrw $6, %xmm0, 12(%rdi)
-; AVX2-NEXT: .LBB11_14: # %else12
-; AVX2-NEXT: vpextrb $12, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $-128, %al
; AVX2-NEXT: je .LBB11_16
-; AVX2-NEXT: # %bb.15: # %cond.store13
+; AVX2-NEXT: .LBB11_15: # %cond.store13
; AVX2-NEXT: vpextrw $7, %xmm0, 14(%rdi)
-; AVX2-NEXT: .LBB11_16: # %else14
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
@@ -4572,71 +4241,66 @@ define void @truncstore_v8i32_v8i16(<8 x
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: vpbroadcastd {{.*#+}} ymm2 = [65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX512F-NEXT: vpminud %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT: vpbroadcastd {{.*#+}} ymm1 = [65535,65535,65535,65535,65535,65535,65535,65535]
+; AVX512F-NEXT: vpminud %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpmovdw %zmm0, %ymm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB11_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB11_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB11_3
+; AVX512F-NEXT: .LBB11_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB11_5
+; AVX512F-NEXT: .LBB11_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB11_7
+; AVX512F-NEXT: .LBB11_8: # %else6
+; AVX512F-NEXT: testb $16, %al
+; AVX512F-NEXT: jne .LBB11_9
+; AVX512F-NEXT: .LBB11_10: # %else8
+; AVX512F-NEXT: testb $32, %al
+; AVX512F-NEXT: jne .LBB11_11
+; AVX512F-NEXT: .LBB11_12: # %else10
+; AVX512F-NEXT: testb $64, %al
+; AVX512F-NEXT: jne .LBB11_13
+; AVX512F-NEXT: .LBB11_14: # %else12
+; AVX512F-NEXT: testb $-128, %al
+; AVX512F-NEXT: jne .LBB11_15
+; AVX512F-NEXT: .LBB11_16: # %else14
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB11_1: # %cond.store
; AVX512F-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB11_2: # %else
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB11_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB11_3: # %cond.store1
; AVX512F-NEXT: vpextrw $1, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB11_4: # %else2
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $2, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB11_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB11_5: # %cond.store3
; AVX512F-NEXT: vpextrw $2, %xmm0, 4(%rdi)
-; AVX512F-NEXT: .LBB11_6: # %else4
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $3, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB11_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB11_7: # %cond.store5
; AVX512F-NEXT: vpextrw $3, %xmm0, 6(%rdi)
-; AVX512F-NEXT: .LBB11_8: # %else6
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $4, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $16, %al
; AVX512F-NEXT: je .LBB11_10
-; AVX512F-NEXT: # %bb.9: # %cond.store7
+; AVX512F-NEXT: .LBB11_9: # %cond.store7
; AVX512F-NEXT: vpextrw $4, %xmm0, 8(%rdi)
-; AVX512F-NEXT: .LBB11_10: # %else8
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $5, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $32, %al
; AVX512F-NEXT: je .LBB11_12
-; AVX512F-NEXT: # %bb.11: # %cond.store9
+; AVX512F-NEXT: .LBB11_11: # %cond.store9
; AVX512F-NEXT: vpextrw $5, %xmm0, 10(%rdi)
-; AVX512F-NEXT: .LBB11_12: # %else10
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $6, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $64, %al
; AVX512F-NEXT: je .LBB11_14
-; AVX512F-NEXT: # %bb.13: # %cond.store11
+; AVX512F-NEXT: .LBB11_13: # %cond.store11
; AVX512F-NEXT: vpextrw $6, %xmm0, 12(%rdi)
-; AVX512F-NEXT: .LBB11_14: # %else12
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $7, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $-128, %al
; AVX512F-NEXT: je .LBB11_16
-; AVX512F-NEXT: # %bb.15: # %cond.store13
+; AVX512F-NEXT: .LBB11_15: # %cond.store13
; AVX512F-NEXT: vpextrw $7, %xmm0, 14(%rdi)
-; AVX512F-NEXT: .LBB11_16: # %else14
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -4671,326 +4335,307 @@ define void @truncstore_v8i32_v8i16(<8 x
define void @truncstore_v8i32_v8i8(<8 x i32> %x, <8 x i8>* %p, <8 x i32> %mask) {
; SSE2-LABEL: truncstore_v8i32_v8i8:
; SSE2: # %bb.0:
-; SSE2-NEXT: pxor %xmm4, %xmm4
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm4
-; SSE2-NEXT: pcmpeqd %xmm5, %xmm5
-; SSE2-NEXT: pxor %xmm4, %xmm5
-; SSE2-NEXT: movdqa %xmm5, %xmm9
-; SSE2-NEXT: packssdw %xmm0, %xmm9
-; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [255,255,255,255]
+; SSE2-NEXT: pxor %xmm8, %xmm8
+; SSE2-NEXT: movdqa {{.*#+}} xmm9 = [255,255,255,255]
; SSE2-NEXT: movdqa {{.*#+}} xmm7 = [2147483648,2147483648,2147483648,2147483648]
-; SSE2-NEXT: movdqa %xmm1, %xmm6
-; SSE2-NEXT: pxor %xmm7, %xmm6
-; SSE2-NEXT: movdqa {{.*#+}} xmm10 = [2147483903,2147483903,2147483903,2147483903]
-; SSE2-NEXT: movdqa %xmm10, %xmm4
-; SSE2-NEXT: pcmpgtd %xmm6, %xmm4
+; SSE2-NEXT: movdqa %xmm1, %xmm5
+; SSE2-NEXT: pxor %xmm7, %xmm5
+; SSE2-NEXT: movdqa {{.*#+}} xmm6 = [2147483903,2147483903,2147483903,2147483903]
+; SSE2-NEXT: movdqa %xmm6, %xmm4
+; SSE2-NEXT: pcmpgtd %xmm5, %xmm4
; SSE2-NEXT: pand %xmm4, %xmm1
-; SSE2-NEXT: pandn %xmm8, %xmm4
+; SSE2-NEXT: pandn %xmm9, %xmm4
; SSE2-NEXT: por %xmm1, %xmm4
; SSE2-NEXT: pxor %xmm0, %xmm7
-; SSE2-NEXT: pcmpgtd %xmm7, %xmm10
-; SSE2-NEXT: pand %xmm10, %xmm0
-; SSE2-NEXT: pandn %xmm8, %xmm10
-; SSE2-NEXT: por %xmm0, %xmm10
-; SSE2-NEXT: packuswb %xmm4, %xmm10
-; SSE2-NEXT: movd %xmm9, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: movd %xmm10, %eax
-; SSE2-NEXT: je .LBB12_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movb %al, (%rdi)
-; SSE2-NEXT: .LBB12_2: # %else
-; SSE2-NEXT: psrlq $16, %xmm5
-; SSE2-NEXT: movd %xmm5, %ecx
-; SSE2-NEXT: shrl $16, %ecx
-; SSE2-NEXT: testb $1, %cl
-; SSE2-NEXT: je .LBB12_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: shrl $16, %eax
-; SSE2-NEXT: movb %al, 1(%rdi)
-; SSE2-NEXT: .LBB12_4: # %else2
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm2
+; SSE2-NEXT: pcmpgtd %xmm7, %xmm6
+; SSE2-NEXT: pand %xmm6, %xmm0
+; SSE2-NEXT: pandn %xmm9, %xmm6
+; SSE2-NEXT: por %xmm0, %xmm6
+; SSE2-NEXT: packuswb %xmm4, %xmm6
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm3
; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm2, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
+; SSE2-NEXT: pxor %xmm0, %xmm3
+; SSE2-NEXT: pcmpeqd %xmm8, %xmm2
+; SSE2-NEXT: pxor %xmm0, %xmm2
+; SSE2-NEXT: packssdw %xmm3, %xmm2
+; SSE2-NEXT: packsswb %xmm0, %xmm2
+; SSE2-NEXT: pmovmskb %xmm2, %eax
; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB12_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: pextrw $2, %xmm10, %eax
-; SSE2-NEXT: movb %al, 2(%rdi)
+; SSE2-NEXT: movd %xmm6, %ecx
+; SSE2-NEXT: jne .LBB12_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB12_3
+; SSE2-NEXT: .LBB12_4: # %else2
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB12_5
; SSE2-NEXT: .LBB12_6: # %else4
-; SSE2-NEXT: pextrw $6, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB12_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
-; SSE2-NEXT: pextrw $3, %xmm10, %eax
-; SSE2-NEXT: movb %al, 3(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB12_7
; SSE2-NEXT: .LBB12_8: # %else6
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm1, %xmm0
-; SSE2-NEXT: pextrw $0, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB12_10
-; SSE2-NEXT: # %bb.9: # %cond.store7
-; SSE2-NEXT: pextrw $4, %xmm10, %eax
-; SSE2-NEXT: movb %al, 4(%rdi)
+; SSE2-NEXT: testb $16, %al
+; SSE2-NEXT: jne .LBB12_9
; SSE2-NEXT: .LBB12_10: # %else8
-; SSE2-NEXT: pextrw $2, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB12_12
-; SSE2-NEXT: # %bb.11: # %cond.store9
-; SSE2-NEXT: pextrw $5, %xmm10, %eax
-; SSE2-NEXT: movb %al, 5(%rdi)
+; SSE2-NEXT: testb $32, %al
+; SSE2-NEXT: jne .LBB12_11
; SSE2-NEXT: .LBB12_12: # %else10
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm3, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB12_14
-; SSE2-NEXT: # %bb.13: # %cond.store11
-; SSE2-NEXT: pextrw $6, %xmm10, %eax
-; SSE2-NEXT: movb %al, 6(%rdi)
+; SSE2-NEXT: testb $64, %al
+; SSE2-NEXT: jne .LBB12_13
; SSE2-NEXT: .LBB12_14: # %else12
-; SSE2-NEXT: pextrw $6, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $-128, %al
+; SSE2-NEXT: jne .LBB12_15
+; SSE2-NEXT: .LBB12_16: # %else14
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB12_1: # %cond.store
+; SSE2-NEXT: movb %cl, (%rdi)
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: je .LBB12_4
+; SSE2-NEXT: .LBB12_3: # %cond.store1
+; SSE2-NEXT: shrl $16, %ecx
+; SSE2-NEXT: movb %cl, 1(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: je .LBB12_6
+; SSE2-NEXT: .LBB12_5: # %cond.store3
+; SSE2-NEXT: pextrw $2, %xmm6, %ecx
+; SSE2-NEXT: movb %cl, 2(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: je .LBB12_8
+; SSE2-NEXT: .LBB12_7: # %cond.store5
+; SSE2-NEXT: pextrw $3, %xmm6, %ecx
+; SSE2-NEXT: movb %cl, 3(%rdi)
+; SSE2-NEXT: testb $16, %al
+; SSE2-NEXT: je .LBB12_10
+; SSE2-NEXT: .LBB12_9: # %cond.store7
+; SSE2-NEXT: pextrw $4, %xmm6, %ecx
+; SSE2-NEXT: movb %cl, 4(%rdi)
+; SSE2-NEXT: testb $32, %al
+; SSE2-NEXT: je .LBB12_12
+; SSE2-NEXT: .LBB12_11: # %cond.store9
+; SSE2-NEXT: pextrw $5, %xmm6, %ecx
+; SSE2-NEXT: movb %cl, 5(%rdi)
+; SSE2-NEXT: testb $64, %al
+; SSE2-NEXT: je .LBB12_14
+; SSE2-NEXT: .LBB12_13: # %cond.store11
+; SSE2-NEXT: pextrw $6, %xmm6, %ecx
+; SSE2-NEXT: movb %cl, 6(%rdi)
+; SSE2-NEXT: testb $-128, %al
; SSE2-NEXT: je .LBB12_16
-; SSE2-NEXT: # %bb.15: # %cond.store13
-; SSE2-NEXT: pextrw $7, %xmm10, %eax
+; SSE2-NEXT: .LBB12_15: # %cond.store13
+; SSE2-NEXT: pextrw $7, %xmm6, %eax
; SSE2-NEXT: movb %al, 7(%rdi)
-; SSE2-NEXT: .LBB12_16: # %else14
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v8i32_v8i8:
; SSE4: # %bb.0:
-; SSE4-NEXT: pxor %xmm5, %xmm5
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm5
-; SSE4-NEXT: pcmpeqd %xmm4, %xmm4
-; SSE4-NEXT: pxor %xmm5, %xmm4
+; SSE4-NEXT: pxor %xmm4, %xmm4
; SSE4-NEXT: movdqa {{.*#+}} xmm5 = [255,255,255,255]
; SSE4-NEXT: pminud %xmm5, %xmm1
; SSE4-NEXT: pminud %xmm5, %xmm0
; SSE4-NEXT: packusdw %xmm1, %xmm0
-; SSE4-NEXT: pextrb $0, %xmm4, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB12_2
-; SSE4-NEXT: # %bb.1: # %cond.store
-; SSE4-NEXT: pextrb $0, %xmm0, (%rdi)
-; SSE4-NEXT: .LBB12_2: # %else
-; SSE4-NEXT: pextrb $4, %xmm4, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB12_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
-; SSE4-NEXT: pextrb $2, %xmm0, 1(%rdi)
-; SSE4-NEXT: .LBB12_4: # %else2
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm2
+; SSE4-NEXT: pcmpeqd %xmm4, %xmm3
; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm2, %xmm1
-; SSE4-NEXT: pextrb $8, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB12_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
-; SSE4-NEXT: pextrb $4, %xmm0, 2(%rdi)
+; SSE4-NEXT: pxor %xmm1, %xmm3
+; SSE4-NEXT: pcmpeqd %xmm4, %xmm2
+; SSE4-NEXT: pxor %xmm1, %xmm2
+; SSE4-NEXT: packssdw %xmm3, %xmm2
+; SSE4-NEXT: packsswb %xmm0, %xmm2
+; SSE4-NEXT: pmovmskb %xmm2, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB12_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB12_3
+; SSE4-NEXT: .LBB12_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB12_5
; SSE4-NEXT: .LBB12_6: # %else4
-; SSE4-NEXT: pextrb $12, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB12_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
-; SSE4-NEXT: pextrb $6, %xmm0, 3(%rdi)
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB12_7
; SSE4-NEXT: .LBB12_8: # %else6
-; SSE4-NEXT: pxor %xmm2, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm3, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm2, %xmm1
-; SSE4-NEXT: pextrb $0, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $16, %al
+; SSE4-NEXT: jne .LBB12_9
+; SSE4-NEXT: .LBB12_10: # %else8
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: jne .LBB12_11
+; SSE4-NEXT: .LBB12_12: # %else10
+; SSE4-NEXT: testb $64, %al
+; SSE4-NEXT: jne .LBB12_13
+; SSE4-NEXT: .LBB12_14: # %else12
+; SSE4-NEXT: testb $-128, %al
+; SSE4-NEXT: jne .LBB12_15
+; SSE4-NEXT: .LBB12_16: # %else14
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB12_1: # %cond.store
+; SSE4-NEXT: pextrb $0, %xmm0, (%rdi)
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: je .LBB12_4
+; SSE4-NEXT: .LBB12_3: # %cond.store1
+; SSE4-NEXT: pextrb $2, %xmm0, 1(%rdi)
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: je .LBB12_6
+; SSE4-NEXT: .LBB12_5: # %cond.store3
+; SSE4-NEXT: pextrb $4, %xmm0, 2(%rdi)
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: je .LBB12_8
+; SSE4-NEXT: .LBB12_7: # %cond.store5
+; SSE4-NEXT: pextrb $6, %xmm0, 3(%rdi)
+; SSE4-NEXT: testb $16, %al
; SSE4-NEXT: je .LBB12_10
-; SSE4-NEXT: # %bb.9: # %cond.store7
+; SSE4-NEXT: .LBB12_9: # %cond.store7
; SSE4-NEXT: pextrb $8, %xmm0, 4(%rdi)
-; SSE4-NEXT: .LBB12_10: # %else8
-; SSE4-NEXT: pextrb $4, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $32, %al
; SSE4-NEXT: je .LBB12_12
-; SSE4-NEXT: # %bb.11: # %cond.store9
+; SSE4-NEXT: .LBB12_11: # %cond.store9
; SSE4-NEXT: pextrb $10, %xmm0, 5(%rdi)
-; SSE4-NEXT: .LBB12_12: # %else10
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm3
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
-; SSE4-NEXT: pxor %xmm3, %xmm1
-; SSE4-NEXT: pextrb $8, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $64, %al
; SSE4-NEXT: je .LBB12_14
-; SSE4-NEXT: # %bb.13: # %cond.store11
+; SSE4-NEXT: .LBB12_13: # %cond.store11
; SSE4-NEXT: pextrb $12, %xmm0, 6(%rdi)
-; SSE4-NEXT: .LBB12_14: # %else12
-; SSE4-NEXT: pextrb $12, %xmm1, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $-128, %al
; SSE4-NEXT: je .LBB12_16
-; SSE4-NEXT: # %bb.15: # %cond.store13
+; SSE4-NEXT: .LBB12_15: # %cond.store13
; SSE4-NEXT: pextrb $14, %xmm0, 7(%rdi)
-; SSE4-NEXT: .LBB12_16: # %else14
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v8i32_v8i8:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm5
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm6 = [255,255,255,255]
-; AVX1-NEXT: vpminud %xmm6, %xmm5, %xmm5
-; AVX1-NEXT: vpminud %xmm6, %xmm0, %xmm0
-; AVX1-NEXT: vpackusdw %xmm5, %xmm0, %xmm0
-; AVX1-NEXT: vpextrb $0, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB12_2
-; AVX1-NEXT: # %bb.1: # %cond.store
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm3 = [255,255,255,255]
+; AVX1-NEXT: vpminud %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vpminud %xmm3, %xmm0, %xmm0
+; AVX1-NEXT: vpackusdw %xmm2, %xmm0, %xmm0
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT: vpcmpeqd %xmm3, %xmm2, %xmm2
+; AVX1-NEXT: vpcmpeqd %xmm3, %xmm1, %xmm1
+; AVX1-NEXT: vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT: vmovmskps %ymm1, %eax
+; AVX1-NEXT: notl %eax
+; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: jne .LBB12_1
+; AVX1-NEXT: # %bb.2: # %else
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: jne .LBB12_3
+; AVX1-NEXT: .LBB12_4: # %else2
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: jne .LBB12_5
+; AVX1-NEXT: .LBB12_6: # %else4
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: jne .LBB12_7
+; AVX1-NEXT: .LBB12_8: # %else6
+; AVX1-NEXT: testb $16, %al
+; AVX1-NEXT: jne .LBB12_9
+; AVX1-NEXT: .LBB12_10: # %else8
+; AVX1-NEXT: testb $32, %al
+; AVX1-NEXT: jne .LBB12_11
+; AVX1-NEXT: .LBB12_12: # %else10
+; AVX1-NEXT: testb $64, %al
+; AVX1-NEXT: jne .LBB12_13
+; AVX1-NEXT: .LBB12_14: # %else12
+; AVX1-NEXT: testb $-128, %al
+; AVX1-NEXT: jne .LBB12_15
+; AVX1-NEXT: .LBB12_16: # %else14
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+; AVX1-NEXT: .LBB12_1: # %cond.store
; AVX1-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX1-NEXT: .LBB12_2: # %else
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $4, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $2, %al
; AVX1-NEXT: je .LBB12_4
-; AVX1-NEXT: # %bb.3: # %cond.store1
+; AVX1-NEXT: .LBB12_3: # %cond.store1
; AVX1-NEXT: vpextrb $2, %xmm0, 1(%rdi)
-; AVX1-NEXT: .LBB12_4: # %else2
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $8, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $4, %al
; AVX1-NEXT: je .LBB12_6
-; AVX1-NEXT: # %bb.5: # %cond.store3
+; AVX1-NEXT: .LBB12_5: # %cond.store3
; AVX1-NEXT: vpextrb $4, %xmm0, 2(%rdi)
-; AVX1-NEXT: .LBB12_6: # %else4
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $12, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $8, %al
; AVX1-NEXT: je .LBB12_8
-; AVX1-NEXT: # %bb.7: # %cond.store5
+; AVX1-NEXT: .LBB12_7: # %cond.store5
; AVX1-NEXT: vpextrb $6, %xmm0, 3(%rdi)
-; AVX1-NEXT: .LBB12_8: # %else6
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $0, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $16, %al
; AVX1-NEXT: je .LBB12_10
-; AVX1-NEXT: # %bb.9: # %cond.store7
+; AVX1-NEXT: .LBB12_9: # %cond.store7
; AVX1-NEXT: vpextrb $8, %xmm0, 4(%rdi)
-; AVX1-NEXT: .LBB12_10: # %else8
-; AVX1-NEXT: vpextrb $4, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $32, %al
; AVX1-NEXT: je .LBB12_12
-; AVX1-NEXT: # %bb.11: # %cond.store9
+; AVX1-NEXT: .LBB12_11: # %cond.store9
; AVX1-NEXT: vpextrb $10, %xmm0, 5(%rdi)
-; AVX1-NEXT: .LBB12_12: # %else10
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $8, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $64, %al
; AVX1-NEXT: je .LBB12_14
-; AVX1-NEXT: # %bb.13: # %cond.store11
+; AVX1-NEXT: .LBB12_13: # %cond.store11
; AVX1-NEXT: vpextrb $12, %xmm0, 6(%rdi)
-; AVX1-NEXT: .LBB12_14: # %else12
-; AVX1-NEXT: vpextrb $12, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $-128, %al
; AVX1-NEXT: je .LBB12_16
-; AVX1-NEXT: # %bb.15: # %cond.store13
+; AVX1-NEXT: .LBB12_15: # %cond.store13
; AVX1-NEXT: vpextrb $14, %xmm0, 7(%rdi)
-; AVX1-NEXT: .LBB12_16: # %else14
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: truncstore_v8i32_v8i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm4
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm5 = [255,255,255,255,255,255,255,255]
-; AVX2-NEXT: vpminud %ymm5, %ymm0, %ymm0
-; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm5
-; AVX2-NEXT: vpackusdw %xmm5, %xmm0, %xmm0
-; AVX2-NEXT: vpextrb $0, %xmm4, %eax
-; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB12_2
-; AVX2-NEXT: # %bb.1: # %cond.store
+; AVX2-NEXT: vpbroadcastd {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255]
+; AVX2-NEXT: vpminud %ymm3, %ymm0, %ymm0
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm3
+; AVX2-NEXT: vpackusdw %xmm3, %xmm0, %xmm0
+; AVX2-NEXT: vpcmpeqd %ymm2, %ymm1, %ymm1
+; AVX2-NEXT: vmovmskps %ymm1, %eax
+; AVX2-NEXT: notl %eax
+; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: jne .LBB12_1
+; AVX2-NEXT: # %bb.2: # %else
+; AVX2-NEXT: testb $2, %al
+; AVX2-NEXT: jne .LBB12_3
+; AVX2-NEXT: .LBB12_4: # %else2
+; AVX2-NEXT: testb $4, %al
+; AVX2-NEXT: jne .LBB12_5
+; AVX2-NEXT: .LBB12_6: # %else4
+; AVX2-NEXT: testb $8, %al
+; AVX2-NEXT: jne .LBB12_7
+; AVX2-NEXT: .LBB12_8: # %else6
+; AVX2-NEXT: testb $16, %al
+; AVX2-NEXT: jne .LBB12_9
+; AVX2-NEXT: .LBB12_10: # %else8
+; AVX2-NEXT: testb $32, %al
+; AVX2-NEXT: jne .LBB12_11
+; AVX2-NEXT: .LBB12_12: # %else10
+; AVX2-NEXT: testb $64, %al
+; AVX2-NEXT: jne .LBB12_13
+; AVX2-NEXT: .LBB12_14: # %else12
+; AVX2-NEXT: testb $-128, %al
+; AVX2-NEXT: jne .LBB12_15
+; AVX2-NEXT: .LBB12_16: # %else14
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+; AVX2-NEXT: .LBB12_1: # %cond.store
; AVX2-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX2-NEXT: .LBB12_2: # %else
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX2-NEXT: vpextrb $4, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $2, %al
; AVX2-NEXT: je .LBB12_4
-; AVX2-NEXT: # %bb.3: # %cond.store1
+; AVX2-NEXT: .LBB12_3: # %cond.store1
; AVX2-NEXT: vpextrb $2, %xmm0, 1(%rdi)
-; AVX2-NEXT: .LBB12_4: # %else2
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm4
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX2-NEXT: vpextrb $8, %xmm4, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $4, %al
; AVX2-NEXT: je .LBB12_6
-; AVX2-NEXT: # %bb.5: # %cond.store3
+; AVX2-NEXT: .LBB12_5: # %cond.store3
; AVX2-NEXT: vpextrb $4, %xmm0, 2(%rdi)
-; AVX2-NEXT: .LBB12_6: # %else4
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX2-NEXT: vpextrb $12, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $8, %al
; AVX2-NEXT: je .LBB12_8
-; AVX2-NEXT: # %bb.7: # %cond.store5
+; AVX2-NEXT: .LBB12_7: # %cond.store5
; AVX2-NEXT: vpextrb $6, %xmm0, 3(%rdi)
-; AVX2-NEXT: .LBB12_8: # %else6
-; AVX2-NEXT: vextracti128 $1, %ymm1, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX2-NEXT: vpextrb $0, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $16, %al
; AVX2-NEXT: je .LBB12_10
-; AVX2-NEXT: # %bb.9: # %cond.store7
+; AVX2-NEXT: .LBB12_9: # %cond.store7
; AVX2-NEXT: vpextrb $8, %xmm0, 4(%rdi)
-; AVX2-NEXT: .LBB12_10: # %else8
-; AVX2-NEXT: vpextrb $4, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $32, %al
; AVX2-NEXT: je .LBB12_12
-; AVX2-NEXT: # %bb.11: # %cond.store9
+; AVX2-NEXT: .LBB12_11: # %cond.store9
; AVX2-NEXT: vpextrb $10, %xmm0, 5(%rdi)
-; AVX2-NEXT: .LBB12_12: # %else10
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $8, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $64, %al
; AVX2-NEXT: je .LBB12_14
-; AVX2-NEXT: # %bb.13: # %cond.store11
+; AVX2-NEXT: .LBB12_13: # %cond.store11
; AVX2-NEXT: vpextrb $12, %xmm0, 6(%rdi)
-; AVX2-NEXT: .LBB12_14: # %else12
-; AVX2-NEXT: vpextrb $12, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $-128, %al
; AVX2-NEXT: je .LBB12_16
-; AVX2-NEXT: # %bb.15: # %cond.store13
+; AVX2-NEXT: .LBB12_15: # %cond.store13
; AVX2-NEXT: vpextrb $14, %xmm0, 7(%rdi)
-; AVX2-NEXT: .LBB12_16: # %else14
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
@@ -4998,71 +4643,66 @@ define void @truncstore_v8i32_v8i8(<8 x
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $ymm1 killed $ymm1 def $zmm1
; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: vpbroadcastd {{.*#+}} ymm2 = [255,255,255,255,255,255,255,255]
-; AVX512F-NEXT: vpminud %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT: vpbroadcastd {{.*#+}} ymm1 = [255,255,255,255,255,255,255,255]
+; AVX512F-NEXT: vpminud %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpmovdw %zmm0, %ymm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB12_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB12_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB12_3
+; AVX512F-NEXT: .LBB12_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB12_5
+; AVX512F-NEXT: .LBB12_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB12_7
+; AVX512F-NEXT: .LBB12_8: # %else6
+; AVX512F-NEXT: testb $16, %al
+; AVX512F-NEXT: jne .LBB12_9
+; AVX512F-NEXT: .LBB12_10: # %else8
+; AVX512F-NEXT: testb $32, %al
+; AVX512F-NEXT: jne .LBB12_11
+; AVX512F-NEXT: .LBB12_12: # %else10
+; AVX512F-NEXT: testb $64, %al
+; AVX512F-NEXT: jne .LBB12_13
+; AVX512F-NEXT: .LBB12_14: # %else12
+; AVX512F-NEXT: testb $-128, %al
+; AVX512F-NEXT: jne .LBB12_15
+; AVX512F-NEXT: .LBB12_16: # %else14
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB12_1: # %cond.store
; AVX512F-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB12_2: # %else
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB12_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB12_3: # %cond.store1
; AVX512F-NEXT: vpextrb $2, %xmm0, 1(%rdi)
-; AVX512F-NEXT: .LBB12_4: # %else2
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $2, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB12_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB12_5: # %cond.store3
; AVX512F-NEXT: vpextrb $4, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB12_6: # %else4
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $3, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB12_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB12_7: # %cond.store5
; AVX512F-NEXT: vpextrb $6, %xmm0, 3(%rdi)
-; AVX512F-NEXT: .LBB12_8: # %else6
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $4, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $16, %al
; AVX512F-NEXT: je .LBB12_10
-; AVX512F-NEXT: # %bb.9: # %cond.store7
+; AVX512F-NEXT: .LBB12_9: # %cond.store7
; AVX512F-NEXT: vpextrb $8, %xmm0, 4(%rdi)
-; AVX512F-NEXT: .LBB12_10: # %else8
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $5, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $32, %al
; AVX512F-NEXT: je .LBB12_12
-; AVX512F-NEXT: # %bb.11: # %cond.store9
+; AVX512F-NEXT: .LBB12_11: # %cond.store9
; AVX512F-NEXT: vpextrb $10, %xmm0, 5(%rdi)
-; AVX512F-NEXT: .LBB12_12: # %else10
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $6, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $64, %al
; AVX512F-NEXT: je .LBB12_14
-; AVX512F-NEXT: # %bb.13: # %cond.store11
+; AVX512F-NEXT: .LBB12_13: # %cond.store11
; AVX512F-NEXT: vpextrb $12, %xmm0, 6(%rdi)
-; AVX512F-NEXT: .LBB12_14: # %else12
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $7, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $-128, %al
; AVX512F-NEXT: je .LBB12_16
-; AVX512F-NEXT: # %bb.15: # %cond.store13
+; AVX512F-NEXT: .LBB12_15: # %cond.store13
; AVX512F-NEXT: vpextrb $14, %xmm0, 7(%rdi)
-; AVX512F-NEXT: .LBB12_16: # %else14
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -5099,7 +4739,6 @@ define void @truncstore_v4i32_v4i16(<4 x
; SSE2-LABEL: truncstore_v4i32_v4i16:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm3
; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [2147483648,2147483648,2147483648,2147483648]
; SSE2-NEXT: pxor %xmm0, %xmm4
; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [2147549183,2147549183,2147549183,2147549183]
@@ -5107,192 +4746,186 @@ define void @truncstore_v4i32_v4i16(<4 x
; SSE2-NEXT: pand %xmm2, %xmm0
; SSE2-NEXT: pandn {{.*}}(%rip), %xmm2
; SSE2-NEXT: por %xmm0, %xmm2
-; SSE2-NEXT: movd %xmm3, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB13_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movd %xmm2, %eax
-; SSE2-NEXT: movw %ax, (%rdi)
-; SSE2-NEXT: .LBB13_2: # %else
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm3
-; SSE2-NEXT: pextrw $2, %xmm3, %eax
+; SSE2-NEXT: pcmpeqd %xmm1, %xmm3
+; SSE2-NEXT: movmskps %xmm3, %eax
+; SSE2-NEXT: xorl $15, %eax
; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB13_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: pextrw $2, %xmm2, %eax
-; SSE2-NEXT: movw %ax, 2(%rdi)
+; SSE2-NEXT: jne .LBB13_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB13_3
; SSE2-NEXT: .LBB13_4: # %else2
-; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm1
-; SSE2-NEXT: pxor %xmm1, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB13_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: pextrw $4, %xmm2, %eax
-; SSE2-NEXT: movw %ax, 4(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB13_5
; SSE2-NEXT: .LBB13_6: # %else4
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm1
-; SSE2-NEXT: pextrw $6, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB13_7
+; SSE2-NEXT: .LBB13_8: # %else6
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB13_1: # %cond.store
+; SSE2-NEXT: movd %xmm2, %ecx
+; SSE2-NEXT: movw %cx, (%rdi)
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: je .LBB13_4
+; SSE2-NEXT: .LBB13_3: # %cond.store1
+; SSE2-NEXT: pextrw $2, %xmm2, %ecx
+; SSE2-NEXT: movw %cx, 2(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: je .LBB13_6
+; SSE2-NEXT: .LBB13_5: # %cond.store3
+; SSE2-NEXT: pextrw $4, %xmm2, %ecx
+; SSE2-NEXT: movw %cx, 4(%rdi)
+; SSE2-NEXT: testb $8, %al
; SSE2-NEXT: je .LBB13_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
+; SSE2-NEXT: .LBB13_7: # %cond.store5
; SSE2-NEXT: pextrw $6, %xmm2, %eax
; SSE2-NEXT: movw %ax, 6(%rdi)
-; SSE2-NEXT: .LBB13_8: # %else6
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v4i32_v4i16:
; SSE4: # %bb.0:
-; SSE4-NEXT: pxor %xmm3, %xmm3
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm3
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE4-NEXT: pxor %xmm3, %xmm2
+; SSE4-NEXT: pxor %xmm2, %xmm2
; SSE4-NEXT: pminud {{.*}}(%rip), %xmm0
-; SSE4-NEXT: pextrb $0, %xmm2, %eax
+; SSE4-NEXT: pcmpeqd %xmm1, %xmm2
+; SSE4-NEXT: movmskps %xmm2, %eax
+; SSE4-NEXT: xorl $15, %eax
; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB13_2
-; SSE4-NEXT: # %bb.1: # %cond.store
+; SSE4-NEXT: jne .LBB13_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB13_3
+; SSE4-NEXT: .LBB13_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB13_5
+; SSE4-NEXT: .LBB13_6: # %else4
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB13_7
+; SSE4-NEXT: .LBB13_8: # %else6
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB13_1: # %cond.store
; SSE4-NEXT: pextrw $0, %xmm0, (%rdi)
-; SSE4-NEXT: .LBB13_2: # %else
-; SSE4-NEXT: pextrb $4, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $2, %al
; SSE4-NEXT: je .LBB13_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
+; SSE4-NEXT: .LBB13_3: # %cond.store1
; SSE4-NEXT: pextrw $2, %xmm0, 2(%rdi)
-; SSE4-NEXT: .LBB13_4: # %else2
-; SSE4-NEXT: pxor %xmm2, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE4-NEXT: pxor %xmm1, %xmm2
-; SSE4-NEXT: pextrb $8, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $4, %al
; SSE4-NEXT: je .LBB13_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
+; SSE4-NEXT: .LBB13_5: # %cond.store3
; SSE4-NEXT: pextrw $4, %xmm0, 4(%rdi)
-; SSE4-NEXT: .LBB13_6: # %else4
-; SSE4-NEXT: pextrb $12, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $8, %al
; SSE4-NEXT: je .LBB13_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
+; SSE4-NEXT: .LBB13_7: # %cond.store5
; SSE4-NEXT: pextrw $6, %xmm0, 6(%rdi)
-; SSE4-NEXT: .LBB13_8: # %else6
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v4i32_v4i16:
; AVX1: # %bb.0:
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpminud {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT: vpextrb $0, %xmm2, %eax
+; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT: vmovmskps %xmm1, %eax
+; AVX1-NEXT: xorl $15, %eax
; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB13_2
-; AVX1-NEXT: # %bb.1: # %cond.store
+; AVX1-NEXT: jne .LBB13_1
+; AVX1-NEXT: # %bb.2: # %else
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: jne .LBB13_3
+; AVX1-NEXT: .LBB13_4: # %else2
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: jne .LBB13_5
+; AVX1-NEXT: .LBB13_6: # %else4
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: jne .LBB13_7
+; AVX1-NEXT: .LBB13_8: # %else6
+; AVX1-NEXT: retq
+; AVX1-NEXT: .LBB13_1: # %cond.store
; AVX1-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX1-NEXT: .LBB13_2: # %else
-; AVX1-NEXT: vpextrb $4, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $2, %al
; AVX1-NEXT: je .LBB13_4
-; AVX1-NEXT: # %bb.3: # %cond.store1
+; AVX1-NEXT: .LBB13_3: # %cond.store1
; AVX1-NEXT: vpextrw $2, %xmm0, 2(%rdi)
-; AVX1-NEXT: .LBB13_4: # %else2
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $8, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $4, %al
; AVX1-NEXT: je .LBB13_6
-; AVX1-NEXT: # %bb.5: # %cond.store3
+; AVX1-NEXT: .LBB13_5: # %cond.store3
; AVX1-NEXT: vpextrw $4, %xmm0, 4(%rdi)
-; AVX1-NEXT: .LBB13_6: # %else4
-; AVX1-NEXT: vpextrb $12, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $8, %al
; AVX1-NEXT: je .LBB13_8
-; AVX1-NEXT: # %bb.7: # %cond.store5
+; AVX1-NEXT: .LBB13_7: # %cond.store5
; AVX1-NEXT: vpextrw $6, %xmm0, 6(%rdi)
-; AVX1-NEXT: .LBB13_8: # %else6
; AVX1-NEXT: retq
;
; AVX2-LABEL: truncstore_v4i32_v4i16:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [65535,65535,65535,65535]
; AVX2-NEXT: vpminud %xmm3, %xmm0, %xmm0
-; AVX2-NEXT: vpextrb $0, %xmm2, %eax
+; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
+; AVX2-NEXT: vmovmskps %xmm1, %eax
+; AVX2-NEXT: xorl $15, %eax
; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB13_2
-; AVX2-NEXT: # %bb.1: # %cond.store
+; AVX2-NEXT: jne .LBB13_1
+; AVX2-NEXT: # %bb.2: # %else
+; AVX2-NEXT: testb $2, %al
+; AVX2-NEXT: jne .LBB13_3
+; AVX2-NEXT: .LBB13_4: # %else2
+; AVX2-NEXT: testb $4, %al
+; AVX2-NEXT: jne .LBB13_5
+; AVX2-NEXT: .LBB13_6: # %else4
+; AVX2-NEXT: testb $8, %al
+; AVX2-NEXT: jne .LBB13_7
+; AVX2-NEXT: .LBB13_8: # %else6
+; AVX2-NEXT: retq
+; AVX2-NEXT: .LBB13_1: # %cond.store
; AVX2-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX2-NEXT: .LBB13_2: # %else
-; AVX2-NEXT: vpextrb $4, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $2, %al
; AVX2-NEXT: je .LBB13_4
-; AVX2-NEXT: # %bb.3: # %cond.store1
+; AVX2-NEXT: .LBB13_3: # %cond.store1
; AVX2-NEXT: vpextrw $2, %xmm0, 2(%rdi)
-; AVX2-NEXT: .LBB13_4: # %else2
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $8, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $4, %al
; AVX2-NEXT: je .LBB13_6
-; AVX2-NEXT: # %bb.5: # %cond.store3
+; AVX2-NEXT: .LBB13_5: # %cond.store3
; AVX2-NEXT: vpextrw $4, %xmm0, 4(%rdi)
-; AVX2-NEXT: .LBB13_6: # %else4
-; AVX2-NEXT: vpextrb $12, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $8, %al
; AVX2-NEXT: je .LBB13_8
-; AVX2-NEXT: # %bb.7: # %cond.store5
+; AVX2-NEXT: .LBB13_7: # %cond.store5
; AVX2-NEXT: vpextrw $6, %xmm0, 6(%rdi)
-; AVX2-NEXT: .LBB13_8: # %else6
; AVX2-NEXT: retq
;
; AVX512F-LABEL: truncstore_v4i32_v4i16:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm2 = [65535,65535,65535,65535]
-; AVX512F-NEXT: vpminud %xmm2, %xmm0, %xmm0
+; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm1 = [65535,65535,65535,65535]
+; AVX512F-NEXT: vpminud %xmm1, %xmm0, %xmm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB13_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB13_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB13_3
+; AVX512F-NEXT: .LBB13_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB13_5
+; AVX512F-NEXT: .LBB13_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB13_7
+; AVX512F-NEXT: .LBB13_8: # %else6
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB13_1: # %cond.store
; AVX512F-NEXT: vpextrw $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB13_2: # %else
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB13_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB13_3: # %cond.store1
; AVX512F-NEXT: vpextrw $2, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB13_4: # %else2
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $2, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB13_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB13_5: # %cond.store3
; AVX512F-NEXT: vpextrw $4, %xmm0, 4(%rdi)
-; AVX512F-NEXT: .LBB13_6: # %else4
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $3, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB13_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB13_7: # %cond.store5
; AVX512F-NEXT: vpextrw $6, %xmm0, 6(%rdi)
-; AVX512F-NEXT: .LBB13_8: # %else6
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -5327,7 +4960,6 @@ define void @truncstore_v4i32_v4i8(<4 x
; SSE2-LABEL: truncstore_v4i32_v4i8:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm1, %xmm3
; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [2147483648,2147483648,2147483648,2147483648]
; SSE2-NEXT: pxor %xmm0, %xmm4
; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [2147483903,2147483903,2147483903,2147483903]
@@ -5335,192 +4967,186 @@ define void @truncstore_v4i32_v4i8(<4 x
; SSE2-NEXT: pand %xmm2, %xmm0
; SSE2-NEXT: pandn {{.*}}(%rip), %xmm2
; SSE2-NEXT: por %xmm0, %xmm2
-; SSE2-NEXT: movd %xmm3, %eax
-; SSE2-NEXT: notl %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB14_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movd %xmm2, %eax
-; SSE2-NEXT: movb %al, (%rdi)
-; SSE2-NEXT: .LBB14_2: # %else
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm3
-; SSE2-NEXT: pextrw $2, %xmm3, %eax
+; SSE2-NEXT: pcmpeqd %xmm1, %xmm3
+; SSE2-NEXT: movmskps %xmm3, %eax
+; SSE2-NEXT: xorl $15, %eax
; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB14_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: pextrw $2, %xmm2, %eax
-; SSE2-NEXT: movb %al, 1(%rdi)
+; SSE2-NEXT: jne .LBB14_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB14_3
; SSE2-NEXT: .LBB14_4: # %else2
-; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm1
-; SSE2-NEXT: pxor %xmm1, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB14_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: pextrw $4, %xmm2, %eax
-; SSE2-NEXT: movb %al, 2(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB14_5
; SSE2-NEXT: .LBB14_6: # %else4
-; SSE2-NEXT: pcmpeqd %xmm0, %xmm0
-; SSE2-NEXT: pxor %xmm0, %xmm1
-; SSE2-NEXT: pextrw $6, %xmm1, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB14_7
+; SSE2-NEXT: .LBB14_8: # %else6
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB14_1: # %cond.store
+; SSE2-NEXT: movd %xmm2, %ecx
+; SSE2-NEXT: movb %cl, (%rdi)
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: je .LBB14_4
+; SSE2-NEXT: .LBB14_3: # %cond.store1
+; SSE2-NEXT: pextrw $2, %xmm2, %ecx
+; SSE2-NEXT: movb %cl, 1(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: je .LBB14_6
+; SSE2-NEXT: .LBB14_5: # %cond.store3
+; SSE2-NEXT: pextrw $4, %xmm2, %ecx
+; SSE2-NEXT: movb %cl, 2(%rdi)
+; SSE2-NEXT: testb $8, %al
; SSE2-NEXT: je .LBB14_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
+; SSE2-NEXT: .LBB14_7: # %cond.store5
; SSE2-NEXT: pextrw $6, %xmm2, %eax
; SSE2-NEXT: movb %al, 3(%rdi)
-; SSE2-NEXT: .LBB14_8: # %else6
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v4i32_v4i8:
; SSE4: # %bb.0:
-; SSE4-NEXT: pxor %xmm3, %xmm3
-; SSE4-NEXT: pcmpeqd %xmm1, %xmm3
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE4-NEXT: pxor %xmm3, %xmm2
+; SSE4-NEXT: pxor %xmm2, %xmm2
; SSE4-NEXT: pminud {{.*}}(%rip), %xmm0
-; SSE4-NEXT: pextrb $0, %xmm2, %eax
+; SSE4-NEXT: pcmpeqd %xmm1, %xmm2
+; SSE4-NEXT: movmskps %xmm2, %eax
+; SSE4-NEXT: xorl $15, %eax
; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB14_2
-; SSE4-NEXT: # %bb.1: # %cond.store
+; SSE4-NEXT: jne .LBB14_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB14_3
+; SSE4-NEXT: .LBB14_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB14_5
+; SSE4-NEXT: .LBB14_6: # %else4
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB14_7
+; SSE4-NEXT: .LBB14_8: # %else6
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB14_1: # %cond.store
; SSE4-NEXT: pextrb $0, %xmm0, (%rdi)
-; SSE4-NEXT: .LBB14_2: # %else
-; SSE4-NEXT: pextrb $4, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $2, %al
; SSE4-NEXT: je .LBB14_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
+; SSE4-NEXT: .LBB14_3: # %cond.store1
; SSE4-NEXT: pextrb $4, %xmm0, 1(%rdi)
-; SSE4-NEXT: .LBB14_4: # %else2
-; SSE4-NEXT: pxor %xmm2, %xmm2
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE4-NEXT: pxor %xmm1, %xmm2
-; SSE4-NEXT: pextrb $8, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $4, %al
; SSE4-NEXT: je .LBB14_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
+; SSE4-NEXT: .LBB14_5: # %cond.store3
; SSE4-NEXT: pextrb $8, %xmm0, 2(%rdi)
-; SSE4-NEXT: .LBB14_6: # %else4
-; SSE4-NEXT: pextrb $12, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $8, %al
; SSE4-NEXT: je .LBB14_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
+; SSE4-NEXT: .LBB14_7: # %cond.store5
; SSE4-NEXT: pextrb $12, %xmm0, 3(%rdi)
-; SSE4-NEXT: .LBB14_8: # %else6
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v4i32_v4i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
; AVX1-NEXT: vpminud {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT: vpextrb $0, %xmm2, %eax
+; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT: vmovmskps %xmm1, %eax
+; AVX1-NEXT: xorl $15, %eax
; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB14_2
-; AVX1-NEXT: # %bb.1: # %cond.store
+; AVX1-NEXT: jne .LBB14_1
+; AVX1-NEXT: # %bb.2: # %else
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: jne .LBB14_3
+; AVX1-NEXT: .LBB14_4: # %else2
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: jne .LBB14_5
+; AVX1-NEXT: .LBB14_6: # %else4
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: jne .LBB14_7
+; AVX1-NEXT: .LBB14_8: # %else6
+; AVX1-NEXT: retq
+; AVX1-NEXT: .LBB14_1: # %cond.store
; AVX1-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX1-NEXT: .LBB14_2: # %else
-; AVX1-NEXT: vpextrb $4, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $2, %al
; AVX1-NEXT: je .LBB14_4
-; AVX1-NEXT: # %bb.3: # %cond.store1
+; AVX1-NEXT: .LBB14_3: # %cond.store1
; AVX1-NEXT: vpextrb $4, %xmm0, 1(%rdi)
-; AVX1-NEXT: .LBB14_4: # %else2
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $8, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $4, %al
; AVX1-NEXT: je .LBB14_6
-; AVX1-NEXT: # %bb.5: # %cond.store3
+; AVX1-NEXT: .LBB14_5: # %cond.store3
; AVX1-NEXT: vpextrb $8, %xmm0, 2(%rdi)
-; AVX1-NEXT: .LBB14_6: # %else4
-; AVX1-NEXT: vpextrb $12, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $8, %al
; AVX1-NEXT: je .LBB14_8
-; AVX1-NEXT: # %bb.7: # %cond.store5
+; AVX1-NEXT: .LBB14_7: # %cond.store5
; AVX1-NEXT: vpextrb $12, %xmm0, 3(%rdi)
-; AVX1-NEXT: .LBB14_8: # %else6
; AVX1-NEXT: retq
;
; AVX2-LABEL: truncstore_v4i32_v4i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX2-NEXT: vpxor %xmm3, %xmm2, %xmm2
; AVX2-NEXT: vpbroadcastd {{.*#+}} xmm3 = [255,255,255,255]
; AVX2-NEXT: vpminud %xmm3, %xmm0, %xmm0
-; AVX2-NEXT: vpextrb $0, %xmm2, %eax
+; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
+; AVX2-NEXT: vmovmskps %xmm1, %eax
+; AVX2-NEXT: xorl $15, %eax
; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB14_2
-; AVX2-NEXT: # %bb.1: # %cond.store
+; AVX2-NEXT: jne .LBB14_1
+; AVX2-NEXT: # %bb.2: # %else
+; AVX2-NEXT: testb $2, %al
+; AVX2-NEXT: jne .LBB14_3
+; AVX2-NEXT: .LBB14_4: # %else2
+; AVX2-NEXT: testb $4, %al
+; AVX2-NEXT: jne .LBB14_5
+; AVX2-NEXT: .LBB14_6: # %else4
+; AVX2-NEXT: testb $8, %al
+; AVX2-NEXT: jne .LBB14_7
+; AVX2-NEXT: .LBB14_8: # %else6
+; AVX2-NEXT: retq
+; AVX2-NEXT: .LBB14_1: # %cond.store
; AVX2-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX2-NEXT: .LBB14_2: # %else
-; AVX2-NEXT: vpextrb $4, %xmm2, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $2, %al
; AVX2-NEXT: je .LBB14_4
-; AVX2-NEXT: # %bb.3: # %cond.store1
+; AVX2-NEXT: .LBB14_3: # %cond.store1
; AVX2-NEXT: vpextrb $4, %xmm0, 1(%rdi)
-; AVX2-NEXT: .LBB14_4: # %else2
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $8, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $4, %al
; AVX2-NEXT: je .LBB14_6
-; AVX2-NEXT: # %bb.5: # %cond.store3
+; AVX2-NEXT: .LBB14_5: # %cond.store3
; AVX2-NEXT: vpextrb $8, %xmm0, 2(%rdi)
-; AVX2-NEXT: .LBB14_6: # %else4
-; AVX2-NEXT: vpextrb $12, %xmm1, %eax
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $8, %al
; AVX2-NEXT: je .LBB14_8
-; AVX2-NEXT: # %bb.7: # %cond.store5
+; AVX2-NEXT: .LBB14_7: # %cond.store5
; AVX2-NEXT: vpextrb $12, %xmm0, 3(%rdi)
-; AVX2-NEXT: .LBB14_8: # %else6
; AVX2-NEXT: retq
;
; AVX512F-LABEL: truncstore_v4i32_v4i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: # kill: def $xmm1 killed $xmm1 def $zmm1
; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm2 = [255,255,255,255]
-; AVX512F-NEXT: vpminud %xmm2, %xmm0, %xmm0
+; AVX512F-NEXT: vpbroadcastd {{.*#+}} xmm1 = [255,255,255,255]
+; AVX512F-NEXT: vpminud %xmm1, %xmm0, %xmm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB14_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB14_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB14_3
+; AVX512F-NEXT: .LBB14_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB14_5
+; AVX512F-NEXT: .LBB14_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB14_7
+; AVX512F-NEXT: .LBB14_8: # %else6
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB14_1: # %cond.store
; AVX512F-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB14_2: # %else
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB14_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB14_3: # %cond.store1
; AVX512F-NEXT: vpextrb $4, %xmm0, 1(%rdi)
-; AVX512F-NEXT: .LBB14_4: # %else2
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $2, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB14_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB14_5: # %cond.store3
; AVX512F-NEXT: vpextrb $8, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB14_6: # %else4
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $3, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB14_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB14_7: # %cond.store5
; AVX512F-NEXT: vpextrb $12, %xmm0, 3(%rdi)
-; AVX512F-NEXT: .LBB14_8: # %else6
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -5555,7 +5181,6 @@ define void @truncstore_v32i16_v32i8(<32
; SSE2-LABEL: truncstore_v32i16_v32i8:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm7, %xmm7
-; SSE2-NEXT: pcmpeqb %xmm4, %xmm7
; SSE2-NEXT: movdqa {{.*#+}} xmm6 = [32768,32768,32768,32768,32768,32768,32768,32768]
; SSE2-NEXT: pxor %xmm6, %xmm1
; SSE2-NEXT: movdqa {{.*#+}} xmm8 = [33023,33023,33023,33023,33023,33023,33023,33023]
@@ -5565,123 +5190,77 @@ define void @truncstore_v32i16_v32i8(<32
; SSE2-NEXT: pminsw %xmm8, %xmm0
; SSE2-NEXT: pxor %xmm6, %xmm0
; SSE2-NEXT: packuswb %xmm1, %xmm0
-; SSE2-NEXT: movd %xmm7, %ecx
-; SSE2-NEXT: movl %ecx, %eax
-; SSE2-NEXT: notb %al
+; SSE2-NEXT: pcmpeqb %xmm7, %xmm4
+; SSE2-NEXT: pmovmskb %xmm4, %ecx
+; SSE2-NEXT: xorl $65535, %ecx # imm = 0xFFFF
+; SSE2-NEXT: pcmpeqb %xmm7, %xmm5
+; SSE2-NEXT: pmovmskb %xmm5, %eax
+; SSE2-NEXT: notl %eax
+; SSE2-NEXT: shll $16, %eax
+; SSE2-NEXT: orl %ecx, %eax
; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: movd %xmm0, %eax
-; SSE2-NEXT: je .LBB15_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movb %al, (%rdi)
-; SSE2-NEXT: .LBB15_2: # %else
-; SSE2-NEXT: shrl $8, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
-; SSE2-NEXT: je .LBB15_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: movb %ah, 1(%rdi)
+; SSE2-NEXT: movd %xmm0, %ecx
+; SSE2-NEXT: jne .LBB15_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB15_3
; SSE2-NEXT: .LBB15_4: # %else2
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm4, %xmm1
-; SSE2-NEXT: movd %xmm1, %ecx
-; SSE2-NEXT: movl %ecx, %edx
-; SSE2-NEXT: shrl $16, %edx
-; SSE2-NEXT: notb %dl
-; SSE2-NEXT: testb $1, %dl
-; SSE2-NEXT: je .LBB15_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: movl %eax, %edx
-; SSE2-NEXT: shrl $16, %edx
-; SSE2-NEXT: movb %dl, 2(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB15_5
; SSE2-NEXT: .LBB15_6: # %else4
-; SSE2-NEXT: shrl $24, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testb $8, %al
; SSE2-NEXT: je .LBB15_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
-; SSE2-NEXT: shrl $24, %eax
-; SSE2-NEXT: movb %al, 3(%rdi)
+; SSE2-NEXT: .LBB15_7: # %cond.store5
+; SSE2-NEXT: shrl $24, %ecx
+; SSE2-NEXT: movb %cl, 3(%rdi)
; SSE2-NEXT: .LBB15_8: # %else6
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm4, %xmm1
-; SSE2-NEXT: pextrw $2, %xmm1, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testb $16, %al
; SSE2-NEXT: pextrw $2, %xmm0, %ecx
; SSE2-NEXT: je .LBB15_10
; SSE2-NEXT: # %bb.9: # %cond.store7
; SSE2-NEXT: movb %cl, 4(%rdi)
; SSE2-NEXT: .LBB15_10: # %else8
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $32, %al
; SSE2-NEXT: je .LBB15_12
; SSE2-NEXT: # %bb.11: # %cond.store9
; SSE2-NEXT: movb %ch, 5(%rdi)
; SSE2-NEXT: .LBB15_12: # %else10
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm4, %xmm1
-; SSE2-NEXT: pextrw $3, %xmm1, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testb $64, %al
; SSE2-NEXT: pextrw $3, %xmm0, %ecx
; SSE2-NEXT: je .LBB15_14
; SSE2-NEXT: # %bb.13: # %cond.store11
; SSE2-NEXT: movb %cl, 6(%rdi)
; SSE2-NEXT: .LBB15_14: # %else12
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $-128, %al
; SSE2-NEXT: je .LBB15_16
; SSE2-NEXT: # %bb.15: # %cond.store13
; SSE2-NEXT: movb %ch, 7(%rdi)
; SSE2-NEXT: .LBB15_16: # %else14
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm4, %xmm1
-; SSE2-NEXT: pextrw $4, %xmm1, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $256, %eax # imm = 0x100
; SSE2-NEXT: pextrw $4, %xmm0, %ecx
; SSE2-NEXT: je .LBB15_18
; SSE2-NEXT: # %bb.17: # %cond.store15
; SSE2-NEXT: movb %cl, 8(%rdi)
; SSE2-NEXT: .LBB15_18: # %else16
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $512, %eax # imm = 0x200
; SSE2-NEXT: je .LBB15_20
; SSE2-NEXT: # %bb.19: # %cond.store17
; SSE2-NEXT: movb %ch, 9(%rdi)
; SSE2-NEXT: .LBB15_20: # %else18
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm4, %xmm1
-; SSE2-NEXT: pextrw $5, %xmm1, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $1024, %eax # imm = 0x400
; SSE2-NEXT: pextrw $5, %xmm0, %ecx
; SSE2-NEXT: je .LBB15_22
; SSE2-NEXT: # %bb.21: # %cond.store19
; SSE2-NEXT: movb %cl, 10(%rdi)
; SSE2-NEXT: .LBB15_22: # %else20
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $2048, %eax # imm = 0x800
; SSE2-NEXT: je .LBB15_24
; SSE2-NEXT: # %bb.23: # %cond.store21
; SSE2-NEXT: movb %ch, 11(%rdi)
; SSE2-NEXT: .LBB15_24: # %else22
; SSE2-NEXT: pxor %xmm6, %xmm3
; SSE2-NEXT: pxor %xmm6, %xmm2
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm4, %xmm1
-; SSE2-NEXT: pextrw $6, %xmm1, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $4096, %eax # imm = 0x1000
; SSE2-NEXT: pextrw $6, %xmm0, %ecx
; SSE2-NEXT: je .LBB15_26
; SSE2-NEXT: # %bb.25: # %cond.store23
@@ -5689,1393 +5268,1107 @@ define void @truncstore_v32i16_v32i8(<32
; SSE2-NEXT: .LBB15_26: # %else24
; SSE2-NEXT: pminsw %xmm8, %xmm3
; SSE2-NEXT: pminsw %xmm8, %xmm2
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $8192, %eax # imm = 0x2000
; SSE2-NEXT: je .LBB15_28
; SSE2-NEXT: # %bb.27: # %cond.store25
; SSE2-NEXT: movb %ch, 13(%rdi)
; SSE2-NEXT: .LBB15_28: # %else26
; SSE2-NEXT: pxor %xmm6, %xmm3
; SSE2-NEXT: pxor %xmm6, %xmm2
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm1, %xmm4
-; SSE2-NEXT: pextrw $7, %xmm4, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $16384, %eax # imm = 0x4000
; SSE2-NEXT: pextrw $7, %xmm0, %ecx
; SSE2-NEXT: je .LBB15_30
; SSE2-NEXT: # %bb.29: # %cond.store27
; SSE2-NEXT: movb %cl, 14(%rdi)
; SSE2-NEXT: .LBB15_30: # %else28
; SSE2-NEXT: packuswb %xmm3, %xmm2
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $32768, %eax # imm = 0x8000
; SSE2-NEXT: je .LBB15_32
; SSE2-NEXT: # %bb.31: # %cond.store29
; SSE2-NEXT: movb %ch, 15(%rdi)
; SSE2-NEXT: .LBB15_32: # %else30
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE2-NEXT: movd %xmm0, %ecx
-; SSE2-NEXT: movl %ecx, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: movd %xmm2, %eax
-; SSE2-NEXT: je .LBB15_34
-; SSE2-NEXT: # %bb.33: # %cond.store31
-; SSE2-NEXT: movb %al, 16(%rdi)
-; SSE2-NEXT: .LBB15_34: # %else32
-; SSE2-NEXT: shrl $8, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
-; SSE2-NEXT: je .LBB15_36
-; SSE2-NEXT: # %bb.35: # %cond.store33
-; SSE2-NEXT: movb %ah, 17(%rdi)
+; SSE2-NEXT: testl $65536, %eax # imm = 0x10000
+; SSE2-NEXT: movd %xmm2, %ecx
+; SSE2-NEXT: jne .LBB15_33
+; SSE2-NEXT: # %bb.34: # %else32
+; SSE2-NEXT: testl $131072, %eax # imm = 0x20000
+; SSE2-NEXT: jne .LBB15_35
; SSE2-NEXT: .LBB15_36: # %else34
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE2-NEXT: movd %xmm0, %ecx
-; SSE2-NEXT: movl %ecx, %edx
-; SSE2-NEXT: shrl $16, %edx
-; SSE2-NEXT: notb %dl
-; SSE2-NEXT: testb $1, %dl
-; SSE2-NEXT: je .LBB15_38
-; SSE2-NEXT: # %bb.37: # %cond.store35
-; SSE2-NEXT: movl %eax, %edx
-; SSE2-NEXT: shrl $16, %edx
-; SSE2-NEXT: movb %dl, 18(%rdi)
+; SSE2-NEXT: testl $262144, %eax # imm = 0x40000
+; SSE2-NEXT: jne .LBB15_37
; SSE2-NEXT: .LBB15_38: # %else36
-; SSE2-NEXT: shrl $24, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $524288, %eax # imm = 0x80000
; SSE2-NEXT: je .LBB15_40
-; SSE2-NEXT: # %bb.39: # %cond.store37
-; SSE2-NEXT: shrl $24, %eax
-; SSE2-NEXT: movb %al, 19(%rdi)
+; SSE2-NEXT: .LBB15_39: # %cond.store37
+; SSE2-NEXT: shrl $24, %ecx
+; SSE2-NEXT: movb %cl, 19(%rdi)
; SSE2-NEXT: .LBB15_40: # %else38
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE2-NEXT: pextrw $2, %xmm0, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $1048576, %eax # imm = 0x100000
; SSE2-NEXT: pextrw $2, %xmm2, %ecx
; SSE2-NEXT: je .LBB15_42
; SSE2-NEXT: # %bb.41: # %cond.store39
; SSE2-NEXT: movb %cl, 20(%rdi)
; SSE2-NEXT: .LBB15_42: # %else40
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $2097152, %eax # imm = 0x200000
; SSE2-NEXT: je .LBB15_44
; SSE2-NEXT: # %bb.43: # %cond.store41
; SSE2-NEXT: movb %ch, 21(%rdi)
; SSE2-NEXT: .LBB15_44: # %else42
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE2-NEXT: pextrw $3, %xmm0, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $4194304, %eax # imm = 0x400000
; SSE2-NEXT: pextrw $3, %xmm2, %ecx
; SSE2-NEXT: je .LBB15_46
; SSE2-NEXT: # %bb.45: # %cond.store43
; SSE2-NEXT: movb %cl, 22(%rdi)
; SSE2-NEXT: .LBB15_46: # %else44
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $8388608, %eax # imm = 0x800000
; SSE2-NEXT: je .LBB15_48
; SSE2-NEXT: # %bb.47: # %cond.store45
; SSE2-NEXT: movb %ch, 23(%rdi)
; SSE2-NEXT: .LBB15_48: # %else46
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $16777216, %eax # imm = 0x1000000
; SSE2-NEXT: pextrw $4, %xmm2, %ecx
; SSE2-NEXT: je .LBB15_50
; SSE2-NEXT: # %bb.49: # %cond.store47
; SSE2-NEXT: movb %cl, 24(%rdi)
; SSE2-NEXT: .LBB15_50: # %else48
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $33554432, %eax # imm = 0x2000000
; SSE2-NEXT: je .LBB15_52
; SSE2-NEXT: # %bb.51: # %cond.store49
; SSE2-NEXT: movb %ch, 25(%rdi)
; SSE2-NEXT: .LBB15_52: # %else50
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE2-NEXT: pextrw $5, %xmm0, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $67108864, %eax # imm = 0x4000000
; SSE2-NEXT: pextrw $5, %xmm2, %ecx
; SSE2-NEXT: je .LBB15_54
; SSE2-NEXT: # %bb.53: # %cond.store51
; SSE2-NEXT: movb %cl, 26(%rdi)
; SSE2-NEXT: .LBB15_54: # %else52
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $134217728, %eax # imm = 0x8000000
; SSE2-NEXT: je .LBB15_56
; SSE2-NEXT: # %bb.55: # %cond.store53
; SSE2-NEXT: movb %ch, 27(%rdi)
; SSE2-NEXT: .LBB15_56: # %else54
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE2-NEXT: pextrw $6, %xmm0, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $268435456, %eax # imm = 0x10000000
; SSE2-NEXT: pextrw $6, %xmm2, %ecx
; SSE2-NEXT: je .LBB15_58
; SSE2-NEXT: # %bb.57: # %cond.store55
; SSE2-NEXT: movb %cl, 28(%rdi)
; SSE2-NEXT: .LBB15_58: # %else56
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $536870912, %eax # imm = 0x20000000
; SSE2-NEXT: je .LBB15_60
; SSE2-NEXT: # %bb.59: # %cond.store57
; SSE2-NEXT: movb %ch, 29(%rdi)
; SSE2-NEXT: .LBB15_60: # %else58
-; SSE2-NEXT: pxor %xmm0, %xmm0
-; SSE2-NEXT: pcmpeqb %xmm0, %xmm5
-; SSE2-NEXT: pextrw $7, %xmm5, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $1073741824, %eax # imm = 0x40000000
; SSE2-NEXT: pextrw $7, %xmm2, %ecx
-; SSE2-NEXT: je .LBB15_62
-; SSE2-NEXT: # %bb.61: # %cond.store59
+; SSE2-NEXT: jne .LBB15_61
+; SSE2-NEXT: # %bb.62: # %else60
+; SSE2-NEXT: testl $-2147483648, %eax # imm = 0x80000000
+; SSE2-NEXT: jne .LBB15_63
+; SSE2-NEXT: .LBB15_64: # %else62
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB15_1: # %cond.store
+; SSE2-NEXT: movb %cl, (%rdi)
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: je .LBB15_4
+; SSE2-NEXT: .LBB15_3: # %cond.store1
+; SSE2-NEXT: movb %ch, 1(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: je .LBB15_6
+; SSE2-NEXT: .LBB15_5: # %cond.store3
+; SSE2-NEXT: movl %ecx, %edx
+; SSE2-NEXT: shrl $16, %edx
+; SSE2-NEXT: movb %dl, 2(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB15_7
+; SSE2-NEXT: jmp .LBB15_8
+; SSE2-NEXT: .LBB15_33: # %cond.store31
+; SSE2-NEXT: movb %cl, 16(%rdi)
+; SSE2-NEXT: testl $131072, %eax # imm = 0x20000
+; SSE2-NEXT: je .LBB15_36
+; SSE2-NEXT: .LBB15_35: # %cond.store33
+; SSE2-NEXT: movb %ch, 17(%rdi)
+; SSE2-NEXT: testl $262144, %eax # imm = 0x40000
+; SSE2-NEXT: je .LBB15_38
+; SSE2-NEXT: .LBB15_37: # %cond.store35
+; SSE2-NEXT: movl %ecx, %edx
+; SSE2-NEXT: shrl $16, %edx
+; SSE2-NEXT: movb %dl, 18(%rdi)
+; SSE2-NEXT: testl $524288, %eax # imm = 0x80000
+; SSE2-NEXT: jne .LBB15_39
+; SSE2-NEXT: jmp .LBB15_40
+; SSE2-NEXT: .LBB15_61: # %cond.store59
; SSE2-NEXT: movb %cl, 30(%rdi)
-; SSE2-NEXT: .LBB15_62: # %else60
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $-2147483648, %eax # imm = 0x80000000
; SSE2-NEXT: je .LBB15_64
-; SSE2-NEXT: # %bb.63: # %cond.store61
+; SSE2-NEXT: .LBB15_63: # %cond.store61
; SSE2-NEXT: movb %ch, 31(%rdi)
-; SSE2-NEXT: .LBB15_64: # %else62
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v32i16_v32i8:
; SSE4: # %bb.0:
; SSE4-NEXT: pxor %xmm7, %xmm7
-; SSE4-NEXT: pcmpeqb %xmm4, %xmm7
; SSE4-NEXT: movdqa {{.*#+}} xmm6 = [255,255,255,255,255,255,255,255]
; SSE4-NEXT: pminuw %xmm6, %xmm1
; SSE4-NEXT: pminuw %xmm6, %xmm0
; SSE4-NEXT: packuswb %xmm1, %xmm0
-; SSE4-NEXT: pextrb $0, %xmm7, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB15_2
-; SSE4-NEXT: # %bb.1: # %cond.store
+; SSE4-NEXT: pcmpeqb %xmm7, %xmm4
+; SSE4-NEXT: pmovmskb %xmm4, %ecx
+; SSE4-NEXT: xorl $65535, %ecx # imm = 0xFFFF
+; SSE4-NEXT: pcmpeqb %xmm7, %xmm5
+; SSE4-NEXT: pmovmskb %xmm5, %eax
+; SSE4-NEXT: notl %eax
+; SSE4-NEXT: shll $16, %eax
+; SSE4-NEXT: orl %ecx, %eax
+; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: jne .LBB15_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB15_3
+; SSE4-NEXT: .LBB15_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB15_5
+; SSE4-NEXT: .LBB15_6: # %else4
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB15_7
+; SSE4-NEXT: .LBB15_8: # %else6
+; SSE4-NEXT: testb $16, %al
+; SSE4-NEXT: jne .LBB15_9
+; SSE4-NEXT: .LBB15_10: # %else8
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: jne .LBB15_11
+; SSE4-NEXT: .LBB15_12: # %else10
+; SSE4-NEXT: testb $64, %al
+; SSE4-NEXT: jne .LBB15_13
+; SSE4-NEXT: .LBB15_14: # %else12
+; SSE4-NEXT: testb $-128, %al
+; SSE4-NEXT: jne .LBB15_15
+; SSE4-NEXT: .LBB15_16: # %else14
+; SSE4-NEXT: testl $256, %eax # imm = 0x100
+; SSE4-NEXT: jne .LBB15_17
+; SSE4-NEXT: .LBB15_18: # %else16
+; SSE4-NEXT: testl $512, %eax # imm = 0x200
+; SSE4-NEXT: jne .LBB15_19
+; SSE4-NEXT: .LBB15_20: # %else18
+; SSE4-NEXT: testl $1024, %eax # imm = 0x400
+; SSE4-NEXT: jne .LBB15_21
+; SSE4-NEXT: .LBB15_22: # %else20
+; SSE4-NEXT: testl $2048, %eax # imm = 0x800
+; SSE4-NEXT: jne .LBB15_23
+; SSE4-NEXT: .LBB15_24: # %else22
+; SSE4-NEXT: testl $4096, %eax # imm = 0x1000
+; SSE4-NEXT: jne .LBB15_25
+; SSE4-NEXT: .LBB15_26: # %else24
+; SSE4-NEXT: testl $8192, %eax # imm = 0x2000
+; SSE4-NEXT: jne .LBB15_27
+; SSE4-NEXT: .LBB15_28: # %else26
+; SSE4-NEXT: testl $16384, %eax # imm = 0x4000
+; SSE4-NEXT: je .LBB15_30
+; SSE4-NEXT: .LBB15_29: # %cond.store27
+; SSE4-NEXT: pextrb $14, %xmm0, 14(%rdi)
+; SSE4-NEXT: .LBB15_30: # %else28
+; SSE4-NEXT: pminuw %xmm6, %xmm3
+; SSE4-NEXT: pminuw %xmm6, %xmm2
+; SSE4-NEXT: testl $32768, %eax # imm = 0x8000
+; SSE4-NEXT: je .LBB15_32
+; SSE4-NEXT: # %bb.31: # %cond.store29
+; SSE4-NEXT: pextrb $15, %xmm0, 15(%rdi)
+; SSE4-NEXT: .LBB15_32: # %else30
+; SSE4-NEXT: packuswb %xmm3, %xmm2
+; SSE4-NEXT: testl $65536, %eax # imm = 0x10000
+; SSE4-NEXT: jne .LBB15_33
+; SSE4-NEXT: # %bb.34: # %else32
+; SSE4-NEXT: testl $131072, %eax # imm = 0x20000
+; SSE4-NEXT: jne .LBB15_35
+; SSE4-NEXT: .LBB15_36: # %else34
+; SSE4-NEXT: testl $262144, %eax # imm = 0x40000
+; SSE4-NEXT: jne .LBB15_37
+; SSE4-NEXT: .LBB15_38: # %else36
+; SSE4-NEXT: testl $524288, %eax # imm = 0x80000
+; SSE4-NEXT: jne .LBB15_39
+; SSE4-NEXT: .LBB15_40: # %else38
+; SSE4-NEXT: testl $1048576, %eax # imm = 0x100000
+; SSE4-NEXT: jne .LBB15_41
+; SSE4-NEXT: .LBB15_42: # %else40
+; SSE4-NEXT: testl $2097152, %eax # imm = 0x200000
+; SSE4-NEXT: jne .LBB15_43
+; SSE4-NEXT: .LBB15_44: # %else42
+; SSE4-NEXT: testl $4194304, %eax # imm = 0x400000
+; SSE4-NEXT: jne .LBB15_45
+; SSE4-NEXT: .LBB15_46: # %else44
+; SSE4-NEXT: testl $8388608, %eax # imm = 0x800000
+; SSE4-NEXT: jne .LBB15_47
+; SSE4-NEXT: .LBB15_48: # %else46
+; SSE4-NEXT: testl $16777216, %eax # imm = 0x1000000
+; SSE4-NEXT: jne .LBB15_49
+; SSE4-NEXT: .LBB15_50: # %else48
+; SSE4-NEXT: testl $33554432, %eax # imm = 0x2000000
+; SSE4-NEXT: jne .LBB15_51
+; SSE4-NEXT: .LBB15_52: # %else50
+; SSE4-NEXT: testl $67108864, %eax # imm = 0x4000000
+; SSE4-NEXT: jne .LBB15_53
+; SSE4-NEXT: .LBB15_54: # %else52
+; SSE4-NEXT: testl $134217728, %eax # imm = 0x8000000
+; SSE4-NEXT: jne .LBB15_55
+; SSE4-NEXT: .LBB15_56: # %else54
+; SSE4-NEXT: testl $268435456, %eax # imm = 0x10000000
+; SSE4-NEXT: jne .LBB15_57
+; SSE4-NEXT: .LBB15_58: # %else56
+; SSE4-NEXT: testl $536870912, %eax # imm = 0x20000000
+; SSE4-NEXT: jne .LBB15_59
+; SSE4-NEXT: .LBB15_60: # %else58
+; SSE4-NEXT: testl $1073741824, %eax # imm = 0x40000000
+; SSE4-NEXT: jne .LBB15_61
+; SSE4-NEXT: .LBB15_62: # %else60
+; SSE4-NEXT: testl $-2147483648, %eax # imm = 0x80000000
+; SSE4-NEXT: jne .LBB15_63
+; SSE4-NEXT: .LBB15_64: # %else62
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB15_1: # %cond.store
; SSE4-NEXT: pextrb $0, %xmm0, (%rdi)
-; SSE4-NEXT: .LBB15_2: # %else
-; SSE4-NEXT: pextrb $1, %xmm7, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $2, %al
; SSE4-NEXT: je .LBB15_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
+; SSE4-NEXT: .LBB15_3: # %cond.store1
; SSE4-NEXT: pextrb $1, %xmm0, 1(%rdi)
-; SSE4-NEXT: .LBB15_4: # %else2
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm4, %xmm1
-; SSE4-NEXT: pextrb $2, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $4, %al
; SSE4-NEXT: je .LBB15_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
+; SSE4-NEXT: .LBB15_5: # %cond.store3
; SSE4-NEXT: pextrb $2, %xmm0, 2(%rdi)
-; SSE4-NEXT: .LBB15_6: # %else4
-; SSE4-NEXT: pextrb $3, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $8, %al
; SSE4-NEXT: je .LBB15_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
+; SSE4-NEXT: .LBB15_7: # %cond.store5
; SSE4-NEXT: pextrb $3, %xmm0, 3(%rdi)
-; SSE4-NEXT: .LBB15_8: # %else6
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm4, %xmm1
-; SSE4-NEXT: pextrb $4, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $16, %al
; SSE4-NEXT: je .LBB15_10
-; SSE4-NEXT: # %bb.9: # %cond.store7
+; SSE4-NEXT: .LBB15_9: # %cond.store7
; SSE4-NEXT: pextrb $4, %xmm0, 4(%rdi)
-; SSE4-NEXT: .LBB15_10: # %else8
-; SSE4-NEXT: pextrb $5, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $32, %al
; SSE4-NEXT: je .LBB15_12
-; SSE4-NEXT: # %bb.11: # %cond.store9
+; SSE4-NEXT: .LBB15_11: # %cond.store9
; SSE4-NEXT: pextrb $5, %xmm0, 5(%rdi)
-; SSE4-NEXT: .LBB15_12: # %else10
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm4, %xmm1
-; SSE4-NEXT: pextrb $6, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $64, %al
; SSE4-NEXT: je .LBB15_14
-; SSE4-NEXT: # %bb.13: # %cond.store11
+; SSE4-NEXT: .LBB15_13: # %cond.store11
; SSE4-NEXT: pextrb $6, %xmm0, 6(%rdi)
-; SSE4-NEXT: .LBB15_14: # %else12
-; SSE4-NEXT: pextrb $7, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $-128, %al
; SSE4-NEXT: je .LBB15_16
-; SSE4-NEXT: # %bb.15: # %cond.store13
+; SSE4-NEXT: .LBB15_15: # %cond.store13
; SSE4-NEXT: pextrb $7, %xmm0, 7(%rdi)
-; SSE4-NEXT: .LBB15_16: # %else14
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm4, %xmm1
-; SSE4-NEXT: pextrb $8, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $256, %eax # imm = 0x100
; SSE4-NEXT: je .LBB15_18
-; SSE4-NEXT: # %bb.17: # %cond.store15
+; SSE4-NEXT: .LBB15_17: # %cond.store15
; SSE4-NEXT: pextrb $8, %xmm0, 8(%rdi)
-; SSE4-NEXT: .LBB15_18: # %else16
-; SSE4-NEXT: pextrb $9, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $512, %eax # imm = 0x200
; SSE4-NEXT: je .LBB15_20
-; SSE4-NEXT: # %bb.19: # %cond.store17
+; SSE4-NEXT: .LBB15_19: # %cond.store17
; SSE4-NEXT: pextrb $9, %xmm0, 9(%rdi)
-; SSE4-NEXT: .LBB15_20: # %else18
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm4, %xmm1
-; SSE4-NEXT: pextrb $10, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $1024, %eax # imm = 0x400
; SSE4-NEXT: je .LBB15_22
-; SSE4-NEXT: # %bb.21: # %cond.store19
+; SSE4-NEXT: .LBB15_21: # %cond.store19
; SSE4-NEXT: pextrb $10, %xmm0, 10(%rdi)
-; SSE4-NEXT: .LBB15_22: # %else20
-; SSE4-NEXT: pextrb $11, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $2048, %eax # imm = 0x800
; SSE4-NEXT: je .LBB15_24
-; SSE4-NEXT: # %bb.23: # %cond.store21
+; SSE4-NEXT: .LBB15_23: # %cond.store21
; SSE4-NEXT: pextrb $11, %xmm0, 11(%rdi)
-; SSE4-NEXT: .LBB15_24: # %else22
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm4, %xmm1
-; SSE4-NEXT: pextrb $12, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $4096, %eax # imm = 0x1000
; SSE4-NEXT: je .LBB15_26
-; SSE4-NEXT: # %bb.25: # %cond.store23
+; SSE4-NEXT: .LBB15_25: # %cond.store23
; SSE4-NEXT: pextrb $12, %xmm0, 12(%rdi)
-; SSE4-NEXT: .LBB15_26: # %else24
-; SSE4-NEXT: pextrb $13, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $8192, %eax # imm = 0x2000
; SSE4-NEXT: je .LBB15_28
-; SSE4-NEXT: # %bb.27: # %cond.store25
+; SSE4-NEXT: .LBB15_27: # %cond.store25
; SSE4-NEXT: pextrb $13, %xmm0, 13(%rdi)
-; SSE4-NEXT: .LBB15_28: # %else26
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm1, %xmm4
-; SSE4-NEXT: pextrb $14, %xmm4, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB15_30
-; SSE4-NEXT: # %bb.29: # %cond.store27
-; SSE4-NEXT: pextrb $14, %xmm0, 14(%rdi)
-; SSE4-NEXT: .LBB15_30: # %else28
-; SSE4-NEXT: pminuw %xmm6, %xmm3
-; SSE4-NEXT: pminuw %xmm6, %xmm2
-; SSE4-NEXT: pextrb $15, %xmm4, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB15_32
-; SSE4-NEXT: # %bb.31: # %cond.store29
-; SSE4-NEXT: pextrb $15, %xmm0, 15(%rdi)
-; SSE4-NEXT: .LBB15_32: # %else30
-; SSE4-NEXT: packuswb %xmm3, %xmm2
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE4-NEXT: pextrb $0, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB15_34
-; SSE4-NEXT: # %bb.33: # %cond.store31
+; SSE4-NEXT: testl $16384, %eax # imm = 0x4000
+; SSE4-NEXT: jne .LBB15_29
+; SSE4-NEXT: jmp .LBB15_30
+; SSE4-NEXT: .LBB15_33: # %cond.store31
; SSE4-NEXT: pextrb $0, %xmm2, 16(%rdi)
-; SSE4-NEXT: .LBB15_34: # %else32
-; SSE4-NEXT: pextrb $1, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $131072, %eax # imm = 0x20000
; SSE4-NEXT: je .LBB15_36
-; SSE4-NEXT: # %bb.35: # %cond.store33
+; SSE4-NEXT: .LBB15_35: # %cond.store33
; SSE4-NEXT: pextrb $1, %xmm2, 17(%rdi)
-; SSE4-NEXT: .LBB15_36: # %else34
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE4-NEXT: pextrb $2, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $262144, %eax # imm = 0x40000
; SSE4-NEXT: je .LBB15_38
-; SSE4-NEXT: # %bb.37: # %cond.store35
+; SSE4-NEXT: .LBB15_37: # %cond.store35
; SSE4-NEXT: pextrb $2, %xmm2, 18(%rdi)
-; SSE4-NEXT: .LBB15_38: # %else36
-; SSE4-NEXT: pextrb $3, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $524288, %eax # imm = 0x80000
; SSE4-NEXT: je .LBB15_40
-; SSE4-NEXT: # %bb.39: # %cond.store37
+; SSE4-NEXT: .LBB15_39: # %cond.store37
; SSE4-NEXT: pextrb $3, %xmm2, 19(%rdi)
-; SSE4-NEXT: .LBB15_40: # %else38
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE4-NEXT: pextrb $4, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $1048576, %eax # imm = 0x100000
; SSE4-NEXT: je .LBB15_42
-; SSE4-NEXT: # %bb.41: # %cond.store39
+; SSE4-NEXT: .LBB15_41: # %cond.store39
; SSE4-NEXT: pextrb $4, %xmm2, 20(%rdi)
-; SSE4-NEXT: .LBB15_42: # %else40
-; SSE4-NEXT: pextrb $5, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $2097152, %eax # imm = 0x200000
; SSE4-NEXT: je .LBB15_44
-; SSE4-NEXT: # %bb.43: # %cond.store41
+; SSE4-NEXT: .LBB15_43: # %cond.store41
; SSE4-NEXT: pextrb $5, %xmm2, 21(%rdi)
-; SSE4-NEXT: .LBB15_44: # %else42
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE4-NEXT: pextrb $6, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $4194304, %eax # imm = 0x400000
; SSE4-NEXT: je .LBB15_46
-; SSE4-NEXT: # %bb.45: # %cond.store43
+; SSE4-NEXT: .LBB15_45: # %cond.store43
; SSE4-NEXT: pextrb $6, %xmm2, 22(%rdi)
-; SSE4-NEXT: .LBB15_46: # %else44
-; SSE4-NEXT: pextrb $7, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $8388608, %eax # imm = 0x800000
; SSE4-NEXT: je .LBB15_48
-; SSE4-NEXT: # %bb.47: # %cond.store45
+; SSE4-NEXT: .LBB15_47: # %cond.store45
; SSE4-NEXT: pextrb $7, %xmm2, 23(%rdi)
-; SSE4-NEXT: .LBB15_48: # %else46
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE4-NEXT: pextrb $8, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $16777216, %eax # imm = 0x1000000
; SSE4-NEXT: je .LBB15_50
-; SSE4-NEXT: # %bb.49: # %cond.store47
+; SSE4-NEXT: .LBB15_49: # %cond.store47
; SSE4-NEXT: pextrb $8, %xmm2, 24(%rdi)
-; SSE4-NEXT: .LBB15_50: # %else48
-; SSE4-NEXT: pextrb $9, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $33554432, %eax # imm = 0x2000000
; SSE4-NEXT: je .LBB15_52
-; SSE4-NEXT: # %bb.51: # %cond.store49
+; SSE4-NEXT: .LBB15_51: # %cond.store49
; SSE4-NEXT: pextrb $9, %xmm2, 25(%rdi)
-; SSE4-NEXT: .LBB15_52: # %else50
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE4-NEXT: pextrb $10, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $67108864, %eax # imm = 0x4000000
; SSE4-NEXT: je .LBB15_54
-; SSE4-NEXT: # %bb.53: # %cond.store51
+; SSE4-NEXT: .LBB15_53: # %cond.store51
; SSE4-NEXT: pextrb $10, %xmm2, 26(%rdi)
-; SSE4-NEXT: .LBB15_54: # %else52
-; SSE4-NEXT: pextrb $11, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $134217728, %eax # imm = 0x8000000
; SSE4-NEXT: je .LBB15_56
-; SSE4-NEXT: # %bb.55: # %cond.store53
+; SSE4-NEXT: .LBB15_55: # %cond.store53
; SSE4-NEXT: pextrb $11, %xmm2, 27(%rdi)
-; SSE4-NEXT: .LBB15_56: # %else54
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqb %xmm5, %xmm0
-; SSE4-NEXT: pextrb $12, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $268435456, %eax # imm = 0x10000000
; SSE4-NEXT: je .LBB15_58
-; SSE4-NEXT: # %bb.57: # %cond.store55
+; SSE4-NEXT: .LBB15_57: # %cond.store55
; SSE4-NEXT: pextrb $12, %xmm2, 28(%rdi)
-; SSE4-NEXT: .LBB15_58: # %else56
-; SSE4-NEXT: pextrb $13, %xmm0, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $536870912, %eax # imm = 0x20000000
; SSE4-NEXT: je .LBB15_60
-; SSE4-NEXT: # %bb.59: # %cond.store57
+; SSE4-NEXT: .LBB15_59: # %cond.store57
; SSE4-NEXT: pextrb $13, %xmm2, 29(%rdi)
-; SSE4-NEXT: .LBB15_60: # %else58
-; SSE4-NEXT: pxor %xmm0, %xmm0
-; SSE4-NEXT: pcmpeqb %xmm0, %xmm5
-; SSE4-NEXT: pextrb $14, %xmm5, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $1073741824, %eax # imm = 0x40000000
; SSE4-NEXT: je .LBB15_62
-; SSE4-NEXT: # %bb.61: # %cond.store59
+; SSE4-NEXT: .LBB15_61: # %cond.store59
; SSE4-NEXT: pextrb $14, %xmm2, 30(%rdi)
-; SSE4-NEXT: .LBB15_62: # %else60
-; SSE4-NEXT: pextrb $15, %xmm5, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $-2147483648, %eax # imm = 0x80000000
; SSE4-NEXT: je .LBB15_64
-; SSE4-NEXT: # %bb.63: # %cond.store61
+; SSE4-NEXT: .LBB15_63: # %cond.store61
; SSE4-NEXT: pextrb $15, %xmm2, 31(%rdi)
-; SSE4-NEXT: .LBB15_64: # %else62
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v32i16_v32i8:
; AVX1: # %bb.0:
-; AVX1-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpcmpeqb %xmm3, %xmm2, %xmm5
-; AVX1-NEXT: vpcmpeqd %xmm4, %xmm4, %xmm4
-; AVX1-NEXT: vpxor %xmm4, %xmm5, %xmm5
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm6
-; AVX1-NEXT: vmovdqa {{.*#+}} xmm7 = [255,255,255,255,255,255,255,255]
-; AVX1-NEXT: vpminuw %xmm7, %xmm6, %xmm6
-; AVX1-NEXT: vpminuw %xmm7, %xmm0, %xmm0
-; AVX1-NEXT: vpackuswb %xmm6, %xmm0, %xmm0
-; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm6
-; AVX1-NEXT: vpminuw %xmm7, %xmm6, %xmm6
-; AVX1-NEXT: vpminuw %xmm7, %xmm1, %xmm1
-; AVX1-NEXT: vpackuswb %xmm6, %xmm1, %xmm1
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
+; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
+; AVX1-NEXT: vpminuw %xmm4, %xmm3, %xmm3
+; AVX1-NEXT: vpminuw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT: vpackuswb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT: vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT: vpminuw %xmm4, %xmm3, %xmm3
+; AVX1-NEXT: vpminuw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT: vpackuswb %xmm3, %xmm1, %xmm1
; AVX1-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT: vpextrb $0, %xmm5, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB15_2
-; AVX1-NEXT: # %bb.1: # %cond.store
-; AVX1-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX1-NEXT: .LBB15_2: # %else
-; AVX1-NEXT: vpcmpeqb %xmm3, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm4, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $1, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB15_4
-; AVX1-NEXT: # %bb.3: # %cond.store1
-; AVX1-NEXT: vpextrb $1, %xmm0, 1(%rdi)
-; AVX1-NEXT: .LBB15_4: # %else2
; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $2, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB15_6
-; AVX1-NEXT: # %bb.5: # %cond.store3
-; AVX1-NEXT: vpextrb $2, %xmm0, 2(%rdi)
-; AVX1-NEXT: .LBB15_6: # %else4
+; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
+; AVX1-NEXT: vpmovmskb %xmm3, %ecx
+; AVX1-NEXT: xorl $65535, %ecx # imm = 0xFFFF
+; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm2
; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $3, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB15_8
-; AVX1-NEXT: # %bb.7: # %cond.store5
-; AVX1-NEXT: vpextrb $3, %xmm0, 3(%rdi)
+; AVX1-NEXT: vpmovmskb %xmm1, %eax
+; AVX1-NEXT: notl %eax
+; AVX1-NEXT: shll $16, %eax
+; AVX1-NEXT: orl %ecx, %eax
+; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: jne .LBB15_1
+; AVX1-NEXT: # %bb.2: # %else
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: jne .LBB15_3
+; AVX1-NEXT: .LBB15_4: # %else2
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: jne .LBB15_5
+; AVX1-NEXT: .LBB15_6: # %else4
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: jne .LBB15_7
; AVX1-NEXT: .LBB15_8: # %else6
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $4, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB15_10
-; AVX1-NEXT: # %bb.9: # %cond.store7
-; AVX1-NEXT: vpextrb $4, %xmm0, 4(%rdi)
+; AVX1-NEXT: testb $16, %al
+; AVX1-NEXT: jne .LBB15_9
; AVX1-NEXT: .LBB15_10: # %else8
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $5, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB15_12
-; AVX1-NEXT: # %bb.11: # %cond.store9
-; AVX1-NEXT: vpextrb $5, %xmm0, 5(%rdi)
+; AVX1-NEXT: testb $32, %al
+; AVX1-NEXT: jne .LBB15_11
; AVX1-NEXT: .LBB15_12: # %else10
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $6, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB15_14
-; AVX1-NEXT: # %bb.13: # %cond.store11
-; AVX1-NEXT: vpextrb $6, %xmm0, 6(%rdi)
+; AVX1-NEXT: testb $64, %al
+; AVX1-NEXT: jne .LBB15_13
; AVX1-NEXT: .LBB15_14: # %else12
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $7, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB15_16
-; AVX1-NEXT: # %bb.15: # %cond.store13
-; AVX1-NEXT: vpextrb $7, %xmm0, 7(%rdi)
+; AVX1-NEXT: testb $-128, %al
+; AVX1-NEXT: jne .LBB15_15
; AVX1-NEXT: .LBB15_16: # %else14
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $8, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $256, %eax # imm = 0x100
+; AVX1-NEXT: jne .LBB15_17
+; AVX1-NEXT: .LBB15_18: # %else16
+; AVX1-NEXT: testl $512, %eax # imm = 0x200
+; AVX1-NEXT: jne .LBB15_19
+; AVX1-NEXT: .LBB15_20: # %else18
+; AVX1-NEXT: testl $1024, %eax # imm = 0x400
+; AVX1-NEXT: jne .LBB15_21
+; AVX1-NEXT: .LBB15_22: # %else20
+; AVX1-NEXT: testl $2048, %eax # imm = 0x800
+; AVX1-NEXT: jne .LBB15_23
+; AVX1-NEXT: .LBB15_24: # %else22
+; AVX1-NEXT: testl $4096, %eax # imm = 0x1000
+; AVX1-NEXT: jne .LBB15_25
+; AVX1-NEXT: .LBB15_26: # %else24
+; AVX1-NEXT: testl $8192, %eax # imm = 0x2000
+; AVX1-NEXT: jne .LBB15_27
+; AVX1-NEXT: .LBB15_28: # %else26
+; AVX1-NEXT: testl $16384, %eax # imm = 0x4000
+; AVX1-NEXT: jne .LBB15_29
+; AVX1-NEXT: .LBB15_30: # %else28
+; AVX1-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX1-NEXT: je .LBB15_32
+; AVX1-NEXT: .LBB15_31: # %cond.store29
+; AVX1-NEXT: vpextrb $15, %xmm0, 15(%rdi)
+; AVX1-NEXT: .LBB15_32: # %else30
+; AVX1-NEXT: testl $65536, %eax # imm = 0x10000
+; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT: jne .LBB15_33
+; AVX1-NEXT: # %bb.34: # %else32
+; AVX1-NEXT: testl $131072, %eax # imm = 0x20000
+; AVX1-NEXT: jne .LBB15_35
+; AVX1-NEXT: .LBB15_36: # %else34
+; AVX1-NEXT: testl $262144, %eax # imm = 0x40000
+; AVX1-NEXT: jne .LBB15_37
+; AVX1-NEXT: .LBB15_38: # %else36
+; AVX1-NEXT: testl $524288, %eax # imm = 0x80000
+; AVX1-NEXT: jne .LBB15_39
+; AVX1-NEXT: .LBB15_40: # %else38
+; AVX1-NEXT: testl $1048576, %eax # imm = 0x100000
+; AVX1-NEXT: jne .LBB15_41
+; AVX1-NEXT: .LBB15_42: # %else40
+; AVX1-NEXT: testl $2097152, %eax # imm = 0x200000
+; AVX1-NEXT: jne .LBB15_43
+; AVX1-NEXT: .LBB15_44: # %else42
+; AVX1-NEXT: testl $4194304, %eax # imm = 0x400000
+; AVX1-NEXT: jne .LBB15_45
+; AVX1-NEXT: .LBB15_46: # %else44
+; AVX1-NEXT: testl $8388608, %eax # imm = 0x800000
+; AVX1-NEXT: jne .LBB15_47
+; AVX1-NEXT: .LBB15_48: # %else46
+; AVX1-NEXT: testl $16777216, %eax # imm = 0x1000000
+; AVX1-NEXT: jne .LBB15_49
+; AVX1-NEXT: .LBB15_50: # %else48
+; AVX1-NEXT: testl $33554432, %eax # imm = 0x2000000
+; AVX1-NEXT: jne .LBB15_51
+; AVX1-NEXT: .LBB15_52: # %else50
+; AVX1-NEXT: testl $67108864, %eax # imm = 0x4000000
+; AVX1-NEXT: jne .LBB15_53
+; AVX1-NEXT: .LBB15_54: # %else52
+; AVX1-NEXT: testl $134217728, %eax # imm = 0x8000000
+; AVX1-NEXT: jne .LBB15_55
+; AVX1-NEXT: .LBB15_56: # %else54
+; AVX1-NEXT: testl $268435456, %eax # imm = 0x10000000
+; AVX1-NEXT: jne .LBB15_57
+; AVX1-NEXT: .LBB15_58: # %else56
+; AVX1-NEXT: testl $536870912, %eax # imm = 0x20000000
+; AVX1-NEXT: jne .LBB15_59
+; AVX1-NEXT: .LBB15_60: # %else58
+; AVX1-NEXT: testl $1073741824, %eax # imm = 0x40000000
+; AVX1-NEXT: jne .LBB15_61
+; AVX1-NEXT: .LBB15_62: # %else60
+; AVX1-NEXT: testl $-2147483648, %eax # imm = 0x80000000
+; AVX1-NEXT: jne .LBB15_63
+; AVX1-NEXT: .LBB15_64: # %else62
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+; AVX1-NEXT: .LBB15_1: # %cond.store
+; AVX1-NEXT: vpextrb $0, %xmm0, (%rdi)
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: je .LBB15_4
+; AVX1-NEXT: .LBB15_3: # %cond.store1
+; AVX1-NEXT: vpextrb $1, %xmm0, 1(%rdi)
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: je .LBB15_6
+; AVX1-NEXT: .LBB15_5: # %cond.store3
+; AVX1-NEXT: vpextrb $2, %xmm0, 2(%rdi)
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: je .LBB15_8
+; AVX1-NEXT: .LBB15_7: # %cond.store5
+; AVX1-NEXT: vpextrb $3, %xmm0, 3(%rdi)
+; AVX1-NEXT: testb $16, %al
+; AVX1-NEXT: je .LBB15_10
+; AVX1-NEXT: .LBB15_9: # %cond.store7
+; AVX1-NEXT: vpextrb $4, %xmm0, 4(%rdi)
+; AVX1-NEXT: testb $32, %al
+; AVX1-NEXT: je .LBB15_12
+; AVX1-NEXT: .LBB15_11: # %cond.store9
+; AVX1-NEXT: vpextrb $5, %xmm0, 5(%rdi)
+; AVX1-NEXT: testb $64, %al
+; AVX1-NEXT: je .LBB15_14
+; AVX1-NEXT: .LBB15_13: # %cond.store11
+; AVX1-NEXT: vpextrb $6, %xmm0, 6(%rdi)
+; AVX1-NEXT: testb $-128, %al
+; AVX1-NEXT: je .LBB15_16
+; AVX1-NEXT: .LBB15_15: # %cond.store13
+; AVX1-NEXT: vpextrb $7, %xmm0, 7(%rdi)
+; AVX1-NEXT: testl $256, %eax # imm = 0x100
; AVX1-NEXT: je .LBB15_18
-; AVX1-NEXT: # %bb.17: # %cond.store15
+; AVX1-NEXT: .LBB15_17: # %cond.store15
; AVX1-NEXT: vpextrb $8, %xmm0, 8(%rdi)
-; AVX1-NEXT: .LBB15_18: # %else16
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $9, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $512, %eax # imm = 0x200
; AVX1-NEXT: je .LBB15_20
-; AVX1-NEXT: # %bb.19: # %cond.store17
+; AVX1-NEXT: .LBB15_19: # %cond.store17
; AVX1-NEXT: vpextrb $9, %xmm0, 9(%rdi)
-; AVX1-NEXT: .LBB15_20: # %else18
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $10, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $1024, %eax # imm = 0x400
; AVX1-NEXT: je .LBB15_22
-; AVX1-NEXT: # %bb.21: # %cond.store19
+; AVX1-NEXT: .LBB15_21: # %cond.store19
; AVX1-NEXT: vpextrb $10, %xmm0, 10(%rdi)
-; AVX1-NEXT: .LBB15_22: # %else20
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $11, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $2048, %eax # imm = 0x800
; AVX1-NEXT: je .LBB15_24
-; AVX1-NEXT: # %bb.23: # %cond.store21
+; AVX1-NEXT: .LBB15_23: # %cond.store21
; AVX1-NEXT: vpextrb $11, %xmm0, 11(%rdi)
-; AVX1-NEXT: .LBB15_24: # %else22
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $12, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $4096, %eax # imm = 0x1000
; AVX1-NEXT: je .LBB15_26
-; AVX1-NEXT: # %bb.25: # %cond.store23
+; AVX1-NEXT: .LBB15_25: # %cond.store23
; AVX1-NEXT: vpextrb $12, %xmm0, 12(%rdi)
-; AVX1-NEXT: .LBB15_26: # %else24
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $13, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $8192, %eax # imm = 0x2000
; AVX1-NEXT: je .LBB15_28
-; AVX1-NEXT: # %bb.27: # %cond.store25
+; AVX1-NEXT: .LBB15_27: # %cond.store25
; AVX1-NEXT: vpextrb $13, %xmm0, 13(%rdi)
-; AVX1-NEXT: .LBB15_28: # %else26
-; AVX1-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm4
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm4, %xmm4
-; AVX1-NEXT: vpextrb $14, %xmm4, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $16384, %eax # imm = 0x4000
; AVX1-NEXT: je .LBB15_30
-; AVX1-NEXT: # %bb.29: # %cond.store27
+; AVX1-NEXT: .LBB15_29: # %cond.store27
; AVX1-NEXT: vpextrb $14, %xmm0, 14(%rdi)
-; AVX1-NEXT: .LBB15_30: # %else28
-; AVX1-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX1-NEXT: vpxor %xmm3, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $15, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB15_32
-; AVX1-NEXT: # %bb.31: # %cond.store29
-; AVX1-NEXT: vpextrb $15, %xmm0, 15(%rdi)
-; AVX1-NEXT: .LBB15_32: # %else30
-; AVX1-NEXT: vextractf128 $1, %ymm2, %xmm1
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $0, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT: je .LBB15_34
-; AVX1-NEXT: # %bb.33: # %cond.store31
+; AVX1-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX1-NEXT: jne .LBB15_31
+; AVX1-NEXT: jmp .LBB15_32
+; AVX1-NEXT: .LBB15_33: # %cond.store31
; AVX1-NEXT: vpextrb $0, %xmm0, 16(%rdi)
-; AVX1-NEXT: .LBB15_34: # %else32
-; AVX1-NEXT: vpextrb $1, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $131072, %eax # imm = 0x20000
; AVX1-NEXT: je .LBB15_36
-; AVX1-NEXT: # %bb.35: # %cond.store33
+; AVX1-NEXT: .LBB15_35: # %cond.store33
; AVX1-NEXT: vpextrb $1, %xmm0, 17(%rdi)
-; AVX1-NEXT: .LBB15_36: # %else34
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $2, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $262144, %eax # imm = 0x40000
; AVX1-NEXT: je .LBB15_38
-; AVX1-NEXT: # %bb.37: # %cond.store35
+; AVX1-NEXT: .LBB15_37: # %cond.store35
; AVX1-NEXT: vpextrb $2, %xmm0, 18(%rdi)
-; AVX1-NEXT: .LBB15_38: # %else36
-; AVX1-NEXT: vpextrb $3, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $524288, %eax # imm = 0x80000
; AVX1-NEXT: je .LBB15_40
-; AVX1-NEXT: # %bb.39: # %cond.store37
+; AVX1-NEXT: .LBB15_39: # %cond.store37
; AVX1-NEXT: vpextrb $3, %xmm0, 19(%rdi)
-; AVX1-NEXT: .LBB15_40: # %else38
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $4, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $1048576, %eax # imm = 0x100000
; AVX1-NEXT: je .LBB15_42
-; AVX1-NEXT: # %bb.41: # %cond.store39
+; AVX1-NEXT: .LBB15_41: # %cond.store39
; AVX1-NEXT: vpextrb $4, %xmm0, 20(%rdi)
-; AVX1-NEXT: .LBB15_42: # %else40
-; AVX1-NEXT: vpextrb $5, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $2097152, %eax # imm = 0x200000
; AVX1-NEXT: je .LBB15_44
-; AVX1-NEXT: # %bb.43: # %cond.store41
+; AVX1-NEXT: .LBB15_43: # %cond.store41
; AVX1-NEXT: vpextrb $5, %xmm0, 21(%rdi)
-; AVX1-NEXT: .LBB15_44: # %else42
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $6, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $4194304, %eax # imm = 0x400000
; AVX1-NEXT: je .LBB15_46
-; AVX1-NEXT: # %bb.45: # %cond.store43
+; AVX1-NEXT: .LBB15_45: # %cond.store43
; AVX1-NEXT: vpextrb $6, %xmm0, 22(%rdi)
-; AVX1-NEXT: .LBB15_46: # %else44
-; AVX1-NEXT: vpextrb $7, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $8388608, %eax # imm = 0x800000
; AVX1-NEXT: je .LBB15_48
-; AVX1-NEXT: # %bb.47: # %cond.store45
+; AVX1-NEXT: .LBB15_47: # %cond.store45
; AVX1-NEXT: vpextrb $7, %xmm0, 23(%rdi)
-; AVX1-NEXT: .LBB15_48: # %else46
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $8, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $16777216, %eax # imm = 0x1000000
; AVX1-NEXT: je .LBB15_50
-; AVX1-NEXT: # %bb.49: # %cond.store47
+; AVX1-NEXT: .LBB15_49: # %cond.store47
; AVX1-NEXT: vpextrb $8, %xmm0, 24(%rdi)
-; AVX1-NEXT: .LBB15_50: # %else48
-; AVX1-NEXT: vpextrb $9, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $33554432, %eax # imm = 0x2000000
; AVX1-NEXT: je .LBB15_52
-; AVX1-NEXT: # %bb.51: # %cond.store49
+; AVX1-NEXT: .LBB15_51: # %cond.store49
; AVX1-NEXT: vpextrb $9, %xmm0, 25(%rdi)
-; AVX1-NEXT: .LBB15_52: # %else50
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $10, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $67108864, %eax # imm = 0x4000000
; AVX1-NEXT: je .LBB15_54
-; AVX1-NEXT: # %bb.53: # %cond.store51
+; AVX1-NEXT: .LBB15_53: # %cond.store51
; AVX1-NEXT: vpextrb $10, %xmm0, 26(%rdi)
-; AVX1-NEXT: .LBB15_54: # %else52
-; AVX1-NEXT: vpextrb $11, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $134217728, %eax # imm = 0x8000000
; AVX1-NEXT: je .LBB15_56
-; AVX1-NEXT: # %bb.55: # %cond.store53
+; AVX1-NEXT: .LBB15_55: # %cond.store53
; AVX1-NEXT: vpextrb $11, %xmm0, 27(%rdi)
-; AVX1-NEXT: .LBB15_56: # %else54
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX1-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX1-NEXT: vpextrb $12, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $268435456, %eax # imm = 0x10000000
; AVX1-NEXT: je .LBB15_58
-; AVX1-NEXT: # %bb.57: # %cond.store55
+; AVX1-NEXT: .LBB15_57: # %cond.store55
; AVX1-NEXT: vpextrb $12, %xmm0, 28(%rdi)
-; AVX1-NEXT: .LBB15_58: # %else56
-; AVX1-NEXT: vpextrb $13, %xmm2, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $536870912, %eax # imm = 0x20000000
; AVX1-NEXT: je .LBB15_60
-; AVX1-NEXT: # %bb.59: # %cond.store57
+; AVX1-NEXT: .LBB15_59: # %cond.store57
; AVX1-NEXT: vpextrb $13, %xmm0, 29(%rdi)
-; AVX1-NEXT: .LBB15_60: # %else58
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $14, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $1073741824, %eax # imm = 0x40000000
; AVX1-NEXT: je .LBB15_62
-; AVX1-NEXT: # %bb.61: # %cond.store59
+; AVX1-NEXT: .LBB15_61: # %cond.store59
; AVX1-NEXT: vpextrb $14, %xmm0, 30(%rdi)
-; AVX1-NEXT: .LBB15_62: # %else60
-; AVX1-NEXT: vpextrb $15, %xmm1, %eax
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $-2147483648, %eax # imm = 0x80000000
; AVX1-NEXT: je .LBB15_64
-; AVX1-NEXT: # %bb.63: # %cond.store61
+; AVX1-NEXT: .LBB15_63: # %cond.store61
; AVX1-NEXT: vpextrb $15, %xmm0, 31(%rdi)
-; AVX1-NEXT: .LBB15_64: # %else62
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: truncstore_v32i16_v32i8:
; AVX2: # %bb.0:
-; AVX2-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX2-NEXT: vpminuw %ymm3, %ymm1, %ymm1
-; AVX2-NEXT: vpminuw %ymm3, %ymm0, %ymm0
+; AVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
+; AVX2-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX2-NEXT: vpminuw %ymm4, %ymm1, %ymm1
+; AVX2-NEXT: vpminuw %ymm4, %ymm0, %ymm0
; AVX2-NEXT: vpackuswb %ymm1, %ymm0, %ymm0
; AVX2-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX2-NEXT: vpextrb $0, %xmm3, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB15_2
-; AVX2-NEXT: # %bb.1: # %cond.store
+; AVX2-NEXT: vpcmpeqb %ymm3, %ymm2, %ymm1
+; AVX2-NEXT: vpmovmskb %ymm1, %eax
+; AVX2-NEXT: notl %eax
+; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: jne .LBB15_1
+; AVX2-NEXT: # %bb.2: # %else
+; AVX2-NEXT: testb $2, %al
+; AVX2-NEXT: jne .LBB15_3
+; AVX2-NEXT: .LBB15_4: # %else2
+; AVX2-NEXT: testb $4, %al
+; AVX2-NEXT: jne .LBB15_5
+; AVX2-NEXT: .LBB15_6: # %else4
+; AVX2-NEXT: testb $8, %al
+; AVX2-NEXT: jne .LBB15_7
+; AVX2-NEXT: .LBB15_8: # %else6
+; AVX2-NEXT: testb $16, %al
+; AVX2-NEXT: jne .LBB15_9
+; AVX2-NEXT: .LBB15_10: # %else8
+; AVX2-NEXT: testb $32, %al
+; AVX2-NEXT: jne .LBB15_11
+; AVX2-NEXT: .LBB15_12: # %else10
+; AVX2-NEXT: testb $64, %al
+; AVX2-NEXT: jne .LBB15_13
+; AVX2-NEXT: .LBB15_14: # %else12
+; AVX2-NEXT: testb $-128, %al
+; AVX2-NEXT: jne .LBB15_15
+; AVX2-NEXT: .LBB15_16: # %else14
+; AVX2-NEXT: testl $256, %eax # imm = 0x100
+; AVX2-NEXT: jne .LBB15_17
+; AVX2-NEXT: .LBB15_18: # %else16
+; AVX2-NEXT: testl $512, %eax # imm = 0x200
+; AVX2-NEXT: jne .LBB15_19
+; AVX2-NEXT: .LBB15_20: # %else18
+; AVX2-NEXT: testl $1024, %eax # imm = 0x400
+; AVX2-NEXT: jne .LBB15_21
+; AVX2-NEXT: .LBB15_22: # %else20
+; AVX2-NEXT: testl $2048, %eax # imm = 0x800
+; AVX2-NEXT: jne .LBB15_23
+; AVX2-NEXT: .LBB15_24: # %else22
+; AVX2-NEXT: testl $4096, %eax # imm = 0x1000
+; AVX2-NEXT: jne .LBB15_25
+; AVX2-NEXT: .LBB15_26: # %else24
+; AVX2-NEXT: testl $8192, %eax # imm = 0x2000
+; AVX2-NEXT: jne .LBB15_27
+; AVX2-NEXT: .LBB15_28: # %else26
+; AVX2-NEXT: testl $16384, %eax # imm = 0x4000
+; AVX2-NEXT: jne .LBB15_29
+; AVX2-NEXT: .LBB15_30: # %else28
+; AVX2-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX2-NEXT: je .LBB15_32
+; AVX2-NEXT: .LBB15_31: # %cond.store29
+; AVX2-NEXT: vpextrb $15, %xmm0, 15(%rdi)
+; AVX2-NEXT: .LBB15_32: # %else30
+; AVX2-NEXT: testl $65536, %eax # imm = 0x10000
+; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX2-NEXT: jne .LBB15_33
+; AVX2-NEXT: # %bb.34: # %else32
+; AVX2-NEXT: testl $131072, %eax # imm = 0x20000
+; AVX2-NEXT: jne .LBB15_35
+; AVX2-NEXT: .LBB15_36: # %else34
+; AVX2-NEXT: testl $262144, %eax # imm = 0x40000
+; AVX2-NEXT: jne .LBB15_37
+; AVX2-NEXT: .LBB15_38: # %else36
+; AVX2-NEXT: testl $524288, %eax # imm = 0x80000
+; AVX2-NEXT: jne .LBB15_39
+; AVX2-NEXT: .LBB15_40: # %else38
+; AVX2-NEXT: testl $1048576, %eax # imm = 0x100000
+; AVX2-NEXT: jne .LBB15_41
+; AVX2-NEXT: .LBB15_42: # %else40
+; AVX2-NEXT: testl $2097152, %eax # imm = 0x200000
+; AVX2-NEXT: jne .LBB15_43
+; AVX2-NEXT: .LBB15_44: # %else42
+; AVX2-NEXT: testl $4194304, %eax # imm = 0x400000
+; AVX2-NEXT: jne .LBB15_45
+; AVX2-NEXT: .LBB15_46: # %else44
+; AVX2-NEXT: testl $8388608, %eax # imm = 0x800000
+; AVX2-NEXT: jne .LBB15_47
+; AVX2-NEXT: .LBB15_48: # %else46
+; AVX2-NEXT: testl $16777216, %eax # imm = 0x1000000
+; AVX2-NEXT: jne .LBB15_49
+; AVX2-NEXT: .LBB15_50: # %else48
+; AVX2-NEXT: testl $33554432, %eax # imm = 0x2000000
+; AVX2-NEXT: jne .LBB15_51
+; AVX2-NEXT: .LBB15_52: # %else50
+; AVX2-NEXT: testl $67108864, %eax # imm = 0x4000000
+; AVX2-NEXT: jne .LBB15_53
+; AVX2-NEXT: .LBB15_54: # %else52
+; AVX2-NEXT: testl $134217728, %eax # imm = 0x8000000
+; AVX2-NEXT: jne .LBB15_55
+; AVX2-NEXT: .LBB15_56: # %else54
+; AVX2-NEXT: testl $268435456, %eax # imm = 0x10000000
+; AVX2-NEXT: jne .LBB15_57
+; AVX2-NEXT: .LBB15_58: # %else56
+; AVX2-NEXT: testl $536870912, %eax # imm = 0x20000000
+; AVX2-NEXT: jne .LBB15_59
+; AVX2-NEXT: .LBB15_60: # %else58
+; AVX2-NEXT: testl $1073741824, %eax # imm = 0x40000000
+; AVX2-NEXT: jne .LBB15_61
+; AVX2-NEXT: .LBB15_62: # %else60
+; AVX2-NEXT: testl $-2147483648, %eax # imm = 0x80000000
+; AVX2-NEXT: jne .LBB15_63
+; AVX2-NEXT: .LBB15_64: # %else62
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+; AVX2-NEXT: .LBB15_1: # %cond.store
; AVX2-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX2-NEXT: .LBB15_2: # %else
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX2-NEXT: vpextrb $1, %xmm1, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $2, %al
; AVX2-NEXT: je .LBB15_4
-; AVX2-NEXT: # %bb.3: # %cond.store1
+; AVX2-NEXT: .LBB15_3: # %cond.store1
; AVX2-NEXT: vpextrb $1, %xmm0, 1(%rdi)
-; AVX2-NEXT: .LBB15_4: # %else2
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX2-NEXT: vpextrb $2, %xmm3, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $4, %al
; AVX2-NEXT: je .LBB15_6
-; AVX2-NEXT: # %bb.5: # %cond.store3
+; AVX2-NEXT: .LBB15_5: # %cond.store3
; AVX2-NEXT: vpextrb $2, %xmm0, 2(%rdi)
-; AVX2-NEXT: .LBB15_6: # %else4
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX2-NEXT: vpextrb $3, %xmm1, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $8, %al
; AVX2-NEXT: je .LBB15_8
-; AVX2-NEXT: # %bb.7: # %cond.store5
+; AVX2-NEXT: .LBB15_7: # %cond.store5
; AVX2-NEXT: vpextrb $3, %xmm0, 3(%rdi)
-; AVX2-NEXT: .LBB15_8: # %else6
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX2-NEXT: vpextrb $4, %xmm3, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $16, %al
; AVX2-NEXT: je .LBB15_10
-; AVX2-NEXT: # %bb.9: # %cond.store7
+; AVX2-NEXT: .LBB15_9: # %cond.store7
; AVX2-NEXT: vpextrb $4, %xmm0, 4(%rdi)
-; AVX2-NEXT: .LBB15_10: # %else8
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX2-NEXT: vpextrb $5, %xmm1, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $32, %al
; AVX2-NEXT: je .LBB15_12
-; AVX2-NEXT: # %bb.11: # %cond.store9
+; AVX2-NEXT: .LBB15_11: # %cond.store9
; AVX2-NEXT: vpextrb $5, %xmm0, 5(%rdi)
-; AVX2-NEXT: .LBB15_12: # %else10
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX2-NEXT: vpextrb $6, %xmm3, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $64, %al
; AVX2-NEXT: je .LBB15_14
-; AVX2-NEXT: # %bb.13: # %cond.store11
+; AVX2-NEXT: .LBB15_13: # %cond.store11
; AVX2-NEXT: vpextrb $6, %xmm0, 6(%rdi)
-; AVX2-NEXT: .LBB15_14: # %else12
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX2-NEXT: vpextrb $7, %xmm1, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $-128, %al
; AVX2-NEXT: je .LBB15_16
-; AVX2-NEXT: # %bb.15: # %cond.store13
+; AVX2-NEXT: .LBB15_15: # %cond.store13
; AVX2-NEXT: vpextrb $7, %xmm0, 7(%rdi)
-; AVX2-NEXT: .LBB15_16: # %else14
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX2-NEXT: vpextrb $8, %xmm3, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $256, %eax # imm = 0x100
; AVX2-NEXT: je .LBB15_18
-; AVX2-NEXT: # %bb.17: # %cond.store15
+; AVX2-NEXT: .LBB15_17: # %cond.store15
; AVX2-NEXT: vpextrb $8, %xmm0, 8(%rdi)
-; AVX2-NEXT: .LBB15_18: # %else16
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX2-NEXT: vpextrb $9, %xmm1, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $512, %eax # imm = 0x200
; AVX2-NEXT: je .LBB15_20
-; AVX2-NEXT: # %bb.19: # %cond.store17
+; AVX2-NEXT: .LBB15_19: # %cond.store17
; AVX2-NEXT: vpextrb $9, %xmm0, 9(%rdi)
-; AVX2-NEXT: .LBB15_20: # %else18
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX2-NEXT: vpextrb $10, %xmm3, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $1024, %eax # imm = 0x400
; AVX2-NEXT: je .LBB15_22
-; AVX2-NEXT: # %bb.21: # %cond.store19
+; AVX2-NEXT: .LBB15_21: # %cond.store19
; AVX2-NEXT: vpextrb $10, %xmm0, 10(%rdi)
-; AVX2-NEXT: .LBB15_22: # %else20
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX2-NEXT: vpextrb $11, %xmm1, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $2048, %eax # imm = 0x800
; AVX2-NEXT: je .LBB15_24
-; AVX2-NEXT: # %bb.23: # %cond.store21
+; AVX2-NEXT: .LBB15_23: # %cond.store21
; AVX2-NEXT: vpextrb $11, %xmm0, 11(%rdi)
-; AVX2-NEXT: .LBB15_24: # %else22
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX2-NEXT: vpextrb $12, %xmm3, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $4096, %eax # imm = 0x1000
; AVX2-NEXT: je .LBB15_26
-; AVX2-NEXT: # %bb.25: # %cond.store23
+; AVX2-NEXT: .LBB15_25: # %cond.store23
; AVX2-NEXT: vpextrb $12, %xmm0, 12(%rdi)
-; AVX2-NEXT: .LBB15_26: # %else24
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX2-NEXT: vpextrb $13, %xmm1, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $8192, %eax # imm = 0x2000
; AVX2-NEXT: je .LBB15_28
-; AVX2-NEXT: # %bb.27: # %cond.store25
+; AVX2-NEXT: .LBB15_27: # %cond.store25
; AVX2-NEXT: vpextrb $13, %xmm0, 13(%rdi)
-; AVX2-NEXT: .LBB15_28: # %else26
-; AVX2-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX2-NEXT: vpextrb $14, %xmm3, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $16384, %eax # imm = 0x4000
; AVX2-NEXT: je .LBB15_30
-; AVX2-NEXT: # %bb.29: # %cond.store27
+; AVX2-NEXT: .LBB15_29: # %cond.store27
; AVX2-NEXT: vpextrb $14, %xmm0, 14(%rdi)
-; AVX2-NEXT: .LBB15_30: # %else28
-; AVX2-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX2-NEXT: vpextrb $15, %xmm1, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB15_32
-; AVX2-NEXT: # %bb.31: # %cond.store29
-; AVX2-NEXT: vpextrb $15, %xmm0, 15(%rdi)
-; AVX2-NEXT: .LBB15_32: # %else30
-; AVX2-NEXT: vextracti128 $1, %ymm2, %xmm1
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $0, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX2-NEXT: je .LBB15_34
-; AVX2-NEXT: # %bb.33: # %cond.store31
+; AVX2-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX2-NEXT: jne .LBB15_31
+; AVX2-NEXT: jmp .LBB15_32
+; AVX2-NEXT: .LBB15_33: # %cond.store31
; AVX2-NEXT: vpextrb $0, %xmm0, 16(%rdi)
-; AVX2-NEXT: .LBB15_34: # %else32
-; AVX2-NEXT: vpextrb $1, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $131072, %eax # imm = 0x20000
; AVX2-NEXT: je .LBB15_36
-; AVX2-NEXT: # %bb.35: # %cond.store33
+; AVX2-NEXT: .LBB15_35: # %cond.store33
; AVX2-NEXT: vpextrb $1, %xmm0, 17(%rdi)
-; AVX2-NEXT: .LBB15_36: # %else34
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $2, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $262144, %eax # imm = 0x40000
; AVX2-NEXT: je .LBB15_38
-; AVX2-NEXT: # %bb.37: # %cond.store35
+; AVX2-NEXT: .LBB15_37: # %cond.store35
; AVX2-NEXT: vpextrb $2, %xmm0, 18(%rdi)
-; AVX2-NEXT: .LBB15_38: # %else36
-; AVX2-NEXT: vpextrb $3, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $524288, %eax # imm = 0x80000
; AVX2-NEXT: je .LBB15_40
-; AVX2-NEXT: # %bb.39: # %cond.store37
+; AVX2-NEXT: .LBB15_39: # %cond.store37
; AVX2-NEXT: vpextrb $3, %xmm0, 19(%rdi)
-; AVX2-NEXT: .LBB15_40: # %else38
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $4, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $1048576, %eax # imm = 0x100000
; AVX2-NEXT: je .LBB15_42
-; AVX2-NEXT: # %bb.41: # %cond.store39
+; AVX2-NEXT: .LBB15_41: # %cond.store39
; AVX2-NEXT: vpextrb $4, %xmm0, 20(%rdi)
-; AVX2-NEXT: .LBB15_42: # %else40
-; AVX2-NEXT: vpextrb $5, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $2097152, %eax # imm = 0x200000
; AVX2-NEXT: je .LBB15_44
-; AVX2-NEXT: # %bb.43: # %cond.store41
+; AVX2-NEXT: .LBB15_43: # %cond.store41
; AVX2-NEXT: vpextrb $5, %xmm0, 21(%rdi)
-; AVX2-NEXT: .LBB15_44: # %else42
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $6, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $4194304, %eax # imm = 0x400000
; AVX2-NEXT: je .LBB15_46
-; AVX2-NEXT: # %bb.45: # %cond.store43
+; AVX2-NEXT: .LBB15_45: # %cond.store43
; AVX2-NEXT: vpextrb $6, %xmm0, 22(%rdi)
-; AVX2-NEXT: .LBB15_46: # %else44
-; AVX2-NEXT: vpextrb $7, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $8388608, %eax # imm = 0x800000
; AVX2-NEXT: je .LBB15_48
-; AVX2-NEXT: # %bb.47: # %cond.store45
+; AVX2-NEXT: .LBB15_47: # %cond.store45
; AVX2-NEXT: vpextrb $7, %xmm0, 23(%rdi)
-; AVX2-NEXT: .LBB15_48: # %else46
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $8, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $16777216, %eax # imm = 0x1000000
; AVX2-NEXT: je .LBB15_50
-; AVX2-NEXT: # %bb.49: # %cond.store47
+; AVX2-NEXT: .LBB15_49: # %cond.store47
; AVX2-NEXT: vpextrb $8, %xmm0, 24(%rdi)
-; AVX2-NEXT: .LBB15_50: # %else48
-; AVX2-NEXT: vpextrb $9, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $33554432, %eax # imm = 0x2000000
; AVX2-NEXT: je .LBB15_52
-; AVX2-NEXT: # %bb.51: # %cond.store49
+; AVX2-NEXT: .LBB15_51: # %cond.store49
; AVX2-NEXT: vpextrb $9, %xmm0, 25(%rdi)
-; AVX2-NEXT: .LBB15_52: # %else50
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $10, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $67108864, %eax # imm = 0x4000000
; AVX2-NEXT: je .LBB15_54
-; AVX2-NEXT: # %bb.53: # %cond.store51
+; AVX2-NEXT: .LBB15_53: # %cond.store51
; AVX2-NEXT: vpextrb $10, %xmm0, 26(%rdi)
-; AVX2-NEXT: .LBB15_54: # %else52
-; AVX2-NEXT: vpextrb $11, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $134217728, %eax # imm = 0x8000000
; AVX2-NEXT: je .LBB15_56
-; AVX2-NEXT: # %bb.55: # %cond.store53
+; AVX2-NEXT: .LBB15_55: # %cond.store53
; AVX2-NEXT: vpextrb $11, %xmm0, 27(%rdi)
-; AVX2-NEXT: .LBB15_56: # %else54
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $12, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $268435456, %eax # imm = 0x10000000
; AVX2-NEXT: je .LBB15_58
-; AVX2-NEXT: # %bb.57: # %cond.store55
+; AVX2-NEXT: .LBB15_57: # %cond.store55
; AVX2-NEXT: vpextrb $12, %xmm0, 28(%rdi)
-; AVX2-NEXT: .LBB15_58: # %else56
-; AVX2-NEXT: vpextrb $13, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $536870912, %eax # imm = 0x20000000
; AVX2-NEXT: je .LBB15_60
-; AVX2-NEXT: # %bb.59: # %cond.store57
+; AVX2-NEXT: .LBB15_59: # %cond.store57
; AVX2-NEXT: vpextrb $13, %xmm0, 29(%rdi)
-; AVX2-NEXT: .LBB15_60: # %else58
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $14, %xmm1, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $1073741824, %eax # imm = 0x40000000
; AVX2-NEXT: je .LBB15_62
-; AVX2-NEXT: # %bb.61: # %cond.store59
+; AVX2-NEXT: .LBB15_61: # %cond.store59
; AVX2-NEXT: vpextrb $14, %xmm0, 30(%rdi)
-; AVX2-NEXT: .LBB15_62: # %else60
-; AVX2-NEXT: vpextrb $15, %xmm1, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $-2147483648, %eax # imm = 0x80000000
; AVX2-NEXT: je .LBB15_64
-; AVX2-NEXT: # %bb.63: # %cond.store61
+; AVX2-NEXT: .LBB15_63: # %cond.store61
; AVX2-NEXT: vpextrb $15, %xmm0, 31(%rdi)
-; AVX2-NEXT: .LBB15_64: # %else62
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: truncstore_v32i16_v32i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpxor %xmm3, %xmm3, %xmm3
-; AVX512F-NEXT: vpcmpeqb %xmm3, %xmm2, %xmm4
-; AVX512F-NEXT: vpternlogq $15, %zmm4, %zmm4, %zmm4
-; AVX512F-NEXT: vpmovsxbd %xmm4, %zmm4
-; AVX512F-NEXT: vptestmd %zmm4, %zmm4, %k0
-; AVX512F-NEXT: vmovdqa {{.*#+}} ymm4 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; AVX512F-NEXT: vpminuw %ymm4, %ymm1, %ymm1
-; AVX512F-NEXT: vpminuw %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT: vpcmpeqb %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT: vmovdqa {{.*#+}} ymm3 = [255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
+; AVX512F-NEXT: vpminuw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT: vpminuw %ymm3, %ymm0, %ymm0
; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512F-NEXT: vpmovdb %zmm0, %xmm0
; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm1 = ymm1[0],zero,ymm1[1],zero,ymm1[2],zero,ymm1[3],zero,ymm1[4],zero,ymm1[5],zero,ymm1[6],zero,ymm1[7],zero,ymm1[8],zero,ymm1[9],zero,ymm1[10],zero,ymm1[11],zero,ymm1[12],zero,ymm1[13],zero,ymm1[14],zero,ymm1[15],zero
; AVX512F-NEXT: vpmovdb %zmm1, %xmm1
; AVX512F-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
-; AVX512F-NEXT: kmovw %k0, %eax
+; AVX512F-NEXT: vpmovmskb %ymm2, %eax
+; AVX512F-NEXT: notl %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB15_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB15_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB15_3
+; AVX512F-NEXT: .LBB15_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB15_5
+; AVX512F-NEXT: .LBB15_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB15_7
+; AVX512F-NEXT: .LBB15_8: # %else6
+; AVX512F-NEXT: testb $16, %al
+; AVX512F-NEXT: jne .LBB15_9
+; AVX512F-NEXT: .LBB15_10: # %else8
+; AVX512F-NEXT: testb $32, %al
+; AVX512F-NEXT: jne .LBB15_11
+; AVX512F-NEXT: .LBB15_12: # %else10
+; AVX512F-NEXT: testb $64, %al
+; AVX512F-NEXT: jne .LBB15_13
+; AVX512F-NEXT: .LBB15_14: # %else12
+; AVX512F-NEXT: testb $-128, %al
+; AVX512F-NEXT: jne .LBB15_15
+; AVX512F-NEXT: .LBB15_16: # %else14
+; AVX512F-NEXT: testl $256, %eax # imm = 0x100
+; AVX512F-NEXT: jne .LBB15_17
+; AVX512F-NEXT: .LBB15_18: # %else16
+; AVX512F-NEXT: testl $512, %eax # imm = 0x200
+; AVX512F-NEXT: jne .LBB15_19
+; AVX512F-NEXT: .LBB15_20: # %else18
+; AVX512F-NEXT: testl $1024, %eax # imm = 0x400
+; AVX512F-NEXT: jne .LBB15_21
+; AVX512F-NEXT: .LBB15_22: # %else20
+; AVX512F-NEXT: testl $2048, %eax # imm = 0x800
+; AVX512F-NEXT: jne .LBB15_23
+; AVX512F-NEXT: .LBB15_24: # %else22
+; AVX512F-NEXT: testl $4096, %eax # imm = 0x1000
+; AVX512F-NEXT: jne .LBB15_25
+; AVX512F-NEXT: .LBB15_26: # %else24
+; AVX512F-NEXT: testl $8192, %eax # imm = 0x2000
+; AVX512F-NEXT: jne .LBB15_27
+; AVX512F-NEXT: .LBB15_28: # %else26
+; AVX512F-NEXT: testl $16384, %eax # imm = 0x4000
+; AVX512F-NEXT: jne .LBB15_29
+; AVX512F-NEXT: .LBB15_30: # %else28
+; AVX512F-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX512F-NEXT: je .LBB15_32
+; AVX512F-NEXT: .LBB15_31: # %cond.store29
+; AVX512F-NEXT: vpextrb $15, %xmm0, 15(%rdi)
+; AVX512F-NEXT: .LBB15_32: # %else30
+; AVX512F-NEXT: testl $65536, %eax # imm = 0x10000
+; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm0
+; AVX512F-NEXT: jne .LBB15_33
+; AVX512F-NEXT: # %bb.34: # %else32
+; AVX512F-NEXT: testl $131072, %eax # imm = 0x20000
+; AVX512F-NEXT: jne .LBB15_35
+; AVX512F-NEXT: .LBB15_36: # %else34
+; AVX512F-NEXT: testl $262144, %eax # imm = 0x40000
+; AVX512F-NEXT: jne .LBB15_37
+; AVX512F-NEXT: .LBB15_38: # %else36
+; AVX512F-NEXT: testl $524288, %eax # imm = 0x80000
+; AVX512F-NEXT: jne .LBB15_39
+; AVX512F-NEXT: .LBB15_40: # %else38
+; AVX512F-NEXT: testl $1048576, %eax # imm = 0x100000
+; AVX512F-NEXT: jne .LBB15_41
+; AVX512F-NEXT: .LBB15_42: # %else40
+; AVX512F-NEXT: testl $2097152, %eax # imm = 0x200000
+; AVX512F-NEXT: jne .LBB15_43
+; AVX512F-NEXT: .LBB15_44: # %else42
+; AVX512F-NEXT: testl $4194304, %eax # imm = 0x400000
+; AVX512F-NEXT: jne .LBB15_45
+; AVX512F-NEXT: .LBB15_46: # %else44
+; AVX512F-NEXT: testl $8388608, %eax # imm = 0x800000
+; AVX512F-NEXT: jne .LBB15_47
+; AVX512F-NEXT: .LBB15_48: # %else46
+; AVX512F-NEXT: testl $16777216, %eax # imm = 0x1000000
+; AVX512F-NEXT: jne .LBB15_49
+; AVX512F-NEXT: .LBB15_50: # %else48
+; AVX512F-NEXT: testl $33554432, %eax # imm = 0x2000000
+; AVX512F-NEXT: jne .LBB15_51
+; AVX512F-NEXT: .LBB15_52: # %else50
+; AVX512F-NEXT: testl $67108864, %eax # imm = 0x4000000
+; AVX512F-NEXT: jne .LBB15_53
+; AVX512F-NEXT: .LBB15_54: # %else52
+; AVX512F-NEXT: testl $134217728, %eax # imm = 0x8000000
+; AVX512F-NEXT: jne .LBB15_55
+; AVX512F-NEXT: .LBB15_56: # %else54
+; AVX512F-NEXT: testl $268435456, %eax # imm = 0x10000000
+; AVX512F-NEXT: jne .LBB15_57
+; AVX512F-NEXT: .LBB15_58: # %else56
+; AVX512F-NEXT: testl $536870912, %eax # imm = 0x20000000
+; AVX512F-NEXT: jne .LBB15_59
+; AVX512F-NEXT: .LBB15_60: # %else58
+; AVX512F-NEXT: testl $1073741824, %eax # imm = 0x40000000
+; AVX512F-NEXT: jne .LBB15_61
+; AVX512F-NEXT: .LBB15_62: # %else60
+; AVX512F-NEXT: testl $-2147483648, %eax # imm = 0x80000000
+; AVX512F-NEXT: jne .LBB15_63
+; AVX512F-NEXT: .LBB15_64: # %else62
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB15_1: # %cond.store
; AVX512F-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB15_2: # %else
-; AVX512F-NEXT: vpcmpeqb %xmm3, %xmm2, %xmm1
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovsxbd %xmm1, %zmm1
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB15_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB15_3: # %cond.store1
; AVX512F-NEXT: vpextrb $1, %xmm0, 1(%rdi)
-; AVX512F-NEXT: .LBB15_4: # %else2
-; AVX512F-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm3, %zmm3, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $2, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB15_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB15_5: # %cond.store3
; AVX512F-NEXT: vpextrb $2, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB15_6: # %else4
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovsxbd %xmm1, %zmm1
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $3, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB15_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB15_7: # %cond.store5
; AVX512F-NEXT: vpextrb $3, %xmm0, 3(%rdi)
-; AVX512F-NEXT: .LBB15_8: # %else6
-; AVX512F-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm3, %zmm3, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $4, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $16, %al
; AVX512F-NEXT: je .LBB15_10
-; AVX512F-NEXT: # %bb.9: # %cond.store7
+; AVX512F-NEXT: .LBB15_9: # %cond.store7
; AVX512F-NEXT: vpextrb $4, %xmm0, 4(%rdi)
-; AVX512F-NEXT: .LBB15_10: # %else8
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovsxbd %xmm1, %zmm1
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $5, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $32, %al
; AVX512F-NEXT: je .LBB15_12
-; AVX512F-NEXT: # %bb.11: # %cond.store9
+; AVX512F-NEXT: .LBB15_11: # %cond.store9
; AVX512F-NEXT: vpextrb $5, %xmm0, 5(%rdi)
-; AVX512F-NEXT: .LBB15_12: # %else10
-; AVX512F-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm3, %zmm3, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $6, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $64, %al
; AVX512F-NEXT: je .LBB15_14
-; AVX512F-NEXT: # %bb.13: # %cond.store11
+; AVX512F-NEXT: .LBB15_13: # %cond.store11
; AVX512F-NEXT: vpextrb $6, %xmm0, 6(%rdi)
-; AVX512F-NEXT: .LBB15_14: # %else12
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovsxbd %xmm1, %zmm1
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $7, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $-128, %al
; AVX512F-NEXT: je .LBB15_16
-; AVX512F-NEXT: # %bb.15: # %cond.store13
+; AVX512F-NEXT: .LBB15_15: # %cond.store13
; AVX512F-NEXT: vpextrb $7, %xmm0, 7(%rdi)
-; AVX512F-NEXT: .LBB15_16: # %else14
-; AVX512F-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm3, %zmm3, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $8, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $256, %eax # imm = 0x100
; AVX512F-NEXT: je .LBB15_18
-; AVX512F-NEXT: # %bb.17: # %cond.store15
+; AVX512F-NEXT: .LBB15_17: # %cond.store15
; AVX512F-NEXT: vpextrb $8, %xmm0, 8(%rdi)
-; AVX512F-NEXT: .LBB15_18: # %else16
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovsxbd %xmm1, %zmm1
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $9, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $512, %eax # imm = 0x200
; AVX512F-NEXT: je .LBB15_20
-; AVX512F-NEXT: # %bb.19: # %cond.store17
+; AVX512F-NEXT: .LBB15_19: # %cond.store17
; AVX512F-NEXT: vpextrb $9, %xmm0, 9(%rdi)
-; AVX512F-NEXT: .LBB15_20: # %else18
-; AVX512F-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm3, %zmm3, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $10, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $1024, %eax # imm = 0x400
; AVX512F-NEXT: je .LBB15_22
-; AVX512F-NEXT: # %bb.21: # %cond.store19
+; AVX512F-NEXT: .LBB15_21: # %cond.store19
; AVX512F-NEXT: vpextrb $10, %xmm0, 10(%rdi)
-; AVX512F-NEXT: .LBB15_22: # %else20
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovsxbd %xmm1, %zmm1
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $11, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $2048, %eax # imm = 0x800
; AVX512F-NEXT: je .LBB15_24
-; AVX512F-NEXT: # %bb.23: # %cond.store21
+; AVX512F-NEXT: .LBB15_23: # %cond.store21
; AVX512F-NEXT: vpextrb $11, %xmm0, 11(%rdi)
-; AVX512F-NEXT: .LBB15_24: # %else22
-; AVX512F-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm3, %zmm3, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $12, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $4096, %eax # imm = 0x1000
; AVX512F-NEXT: je .LBB15_26
-; AVX512F-NEXT: # %bb.25: # %cond.store23
+; AVX512F-NEXT: .LBB15_25: # %cond.store23
; AVX512F-NEXT: vpextrb $12, %xmm0, 12(%rdi)
-; AVX512F-NEXT: .LBB15_26: # %else24
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovsxbd %xmm1, %zmm1
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $13, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $8192, %eax # imm = 0x2000
; AVX512F-NEXT: je .LBB15_28
-; AVX512F-NEXT: # %bb.27: # %cond.store25
+; AVX512F-NEXT: .LBB15_27: # %cond.store25
; AVX512F-NEXT: vpextrb $13, %xmm0, 13(%rdi)
-; AVX512F-NEXT: .LBB15_28: # %else26
-; AVX512F-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm3, %zmm3, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $14, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $16384, %eax # imm = 0x4000
; AVX512F-NEXT: je .LBB15_30
-; AVX512F-NEXT: # %bb.29: # %cond.store27
+; AVX512F-NEXT: .LBB15_29: # %cond.store27
; AVX512F-NEXT: vpextrb $14, %xmm0, 14(%rdi)
-; AVX512F-NEXT: .LBB15_30: # %else28
-; AVX512F-NEXT: vpcmpeqb %xmm1, %xmm2, %xmm1
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovsxbd %xmm1, %zmm1
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $15, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB15_32
-; AVX512F-NEXT: # %bb.31: # %cond.store29
-; AVX512F-NEXT: vpextrb $15, %xmm0, 15(%rdi)
-; AVX512F-NEXT: .LBB15_32: # %else30
-; AVX512F-NEXT: vextracti128 $1, %ymm2, %xmm1
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: vextracti128 $1, %ymm0, %xmm0
-; AVX512F-NEXT: je .LBB15_34
-; AVX512F-NEXT: # %bb.33: # %cond.store31
+; AVX512F-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX512F-NEXT: jne .LBB15_31
+; AVX512F-NEXT: jmp .LBB15_32
+; AVX512F-NEXT: .LBB15_33: # %cond.store31
; AVX512F-NEXT: vpextrb $0, %xmm0, 16(%rdi)
-; AVX512F-NEXT: .LBB15_34: # %else32
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $131072, %eax # imm = 0x20000
; AVX512F-NEXT: je .LBB15_36
-; AVX512F-NEXT: # %bb.35: # %cond.store33
+; AVX512F-NEXT: .LBB15_35: # %cond.store33
; AVX512F-NEXT: vpextrb $1, %xmm0, 17(%rdi)
-; AVX512F-NEXT: .LBB15_36: # %else34
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $2, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $262144, %eax # imm = 0x40000
; AVX512F-NEXT: je .LBB15_38
-; AVX512F-NEXT: # %bb.37: # %cond.store35
+; AVX512F-NEXT: .LBB15_37: # %cond.store35
; AVX512F-NEXT: vpextrb $2, %xmm0, 18(%rdi)
-; AVX512F-NEXT: .LBB15_38: # %else36
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $3, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $524288, %eax # imm = 0x80000
; AVX512F-NEXT: je .LBB15_40
-; AVX512F-NEXT: # %bb.39: # %cond.store37
+; AVX512F-NEXT: .LBB15_39: # %cond.store37
; AVX512F-NEXT: vpextrb $3, %xmm0, 19(%rdi)
-; AVX512F-NEXT: .LBB15_40: # %else38
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $4, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $1048576, %eax # imm = 0x100000
; AVX512F-NEXT: je .LBB15_42
-; AVX512F-NEXT: # %bb.41: # %cond.store39
+; AVX512F-NEXT: .LBB15_41: # %cond.store39
; AVX512F-NEXT: vpextrb $4, %xmm0, 20(%rdi)
-; AVX512F-NEXT: .LBB15_42: # %else40
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $5, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $2097152, %eax # imm = 0x200000
; AVX512F-NEXT: je .LBB15_44
-; AVX512F-NEXT: # %bb.43: # %cond.store41
+; AVX512F-NEXT: .LBB15_43: # %cond.store41
; AVX512F-NEXT: vpextrb $5, %xmm0, 21(%rdi)
-; AVX512F-NEXT: .LBB15_44: # %else42
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $6, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $4194304, %eax # imm = 0x400000
; AVX512F-NEXT: je .LBB15_46
-; AVX512F-NEXT: # %bb.45: # %cond.store43
+; AVX512F-NEXT: .LBB15_45: # %cond.store43
; AVX512F-NEXT: vpextrb $6, %xmm0, 22(%rdi)
-; AVX512F-NEXT: .LBB15_46: # %else44
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $7, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $8388608, %eax # imm = 0x800000
; AVX512F-NEXT: je .LBB15_48
-; AVX512F-NEXT: # %bb.47: # %cond.store45
+; AVX512F-NEXT: .LBB15_47: # %cond.store45
; AVX512F-NEXT: vpextrb $7, %xmm0, 23(%rdi)
-; AVX512F-NEXT: .LBB15_48: # %else46
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $8, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $16777216, %eax # imm = 0x1000000
; AVX512F-NEXT: je .LBB15_50
-; AVX512F-NEXT: # %bb.49: # %cond.store47
+; AVX512F-NEXT: .LBB15_49: # %cond.store47
; AVX512F-NEXT: vpextrb $8, %xmm0, 24(%rdi)
-; AVX512F-NEXT: .LBB15_50: # %else48
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $9, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $33554432, %eax # imm = 0x2000000
; AVX512F-NEXT: je .LBB15_52
-; AVX512F-NEXT: # %bb.51: # %cond.store49
+; AVX512F-NEXT: .LBB15_51: # %cond.store49
; AVX512F-NEXT: vpextrb $9, %xmm0, 25(%rdi)
-; AVX512F-NEXT: .LBB15_52: # %else50
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $10, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $67108864, %eax # imm = 0x4000000
; AVX512F-NEXT: je .LBB15_54
-; AVX512F-NEXT: # %bb.53: # %cond.store51
+; AVX512F-NEXT: .LBB15_53: # %cond.store51
; AVX512F-NEXT: vpextrb $10, %xmm0, 26(%rdi)
-; AVX512F-NEXT: .LBB15_54: # %else52
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $11, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $134217728, %eax # imm = 0x8000000
; AVX512F-NEXT: je .LBB15_56
-; AVX512F-NEXT: # %bb.55: # %cond.store53
+; AVX512F-NEXT: .LBB15_55: # %cond.store53
; AVX512F-NEXT: vpextrb $11, %xmm0, 27(%rdi)
-; AVX512F-NEXT: .LBB15_56: # %else54
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $12, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $268435456, %eax # imm = 0x10000000
; AVX512F-NEXT: je .LBB15_58
-; AVX512F-NEXT: # %bb.57: # %cond.store55
+; AVX512F-NEXT: .LBB15_57: # %cond.store55
; AVX512F-NEXT: vpextrb $12, %xmm0, 28(%rdi)
-; AVX512F-NEXT: .LBB15_58: # %else56
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $13, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $536870912, %eax # imm = 0x20000000
; AVX512F-NEXT: je .LBB15_60
-; AVX512F-NEXT: # %bb.59: # %cond.store57
+; AVX512F-NEXT: .LBB15_59: # %cond.store57
; AVX512F-NEXT: vpextrb $13, %xmm0, 29(%rdi)
-; AVX512F-NEXT: .LBB15_60: # %else58
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
-; AVX512F-NEXT: vmovdqa64 %zmm1, %zmm2
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $14, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $1073741824, %eax # imm = 0x40000000
; AVX512F-NEXT: je .LBB15_62
-; AVX512F-NEXT: # %bb.61: # %cond.store59
+; AVX512F-NEXT: .LBB15_61: # %cond.store59
; AVX512F-NEXT: vpextrb $14, %xmm0, 30(%rdi)
-; AVX512F-NEXT: .LBB15_62: # %else60
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovsxbd %xmm1, %zmm1
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $15, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $-2147483648, %eax # imm = 0x80000000
; AVX512F-NEXT: je .LBB15_64
-; AVX512F-NEXT: # %bb.63: # %cond.store61
+; AVX512F-NEXT: .LBB15_63: # %cond.store61
; AVX512F-NEXT: vpextrb $15, %xmm0, 31(%rdi)
-; AVX512F-NEXT: .LBB15_64: # %else62
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -7107,7 +6400,6 @@ define void @truncstore_v16i16_v16i8(<16
; SSE2-LABEL: truncstore_v16i16_v16i8:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqb %xmm2, %xmm3
; SSE2-NEXT: movdqa {{.*#+}} xmm4 = [32768,32768,32768,32768,32768,32768,32768,32768]
; SSE2-NEXT: pxor %xmm4, %xmm1
; SSE2-NEXT: movdqa {{.*#+}} xmm5 = [33023,33023,33023,33023,33023,33023,33023,33023]
@@ -7117,751 +6409,605 @@ define void @truncstore_v16i16_v16i8(<16
; SSE2-NEXT: pminsw %xmm5, %xmm0
; SSE2-NEXT: pxor %xmm4, %xmm0
; SSE2-NEXT: packuswb %xmm1, %xmm0
-; SSE2-NEXT: movd %xmm3, %ecx
-; SSE2-NEXT: movl %ecx, %eax
-; SSE2-NEXT: notb %al
+; SSE2-NEXT: pcmpeqb %xmm2, %xmm3
+; SSE2-NEXT: pmovmskb %xmm3, %eax
+; SSE2-NEXT: xorl $65535, %eax # imm = 0xFFFF
; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: movd %xmm0, %eax
-; SSE2-NEXT: je .LBB16_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movb %al, (%rdi)
-; SSE2-NEXT: .LBB16_2: # %else
-; SSE2-NEXT: shrl $8, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
-; SSE2-NEXT: je .LBB16_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: movb %ah, 1(%rdi)
+; SSE2-NEXT: movd %xmm0, %ecx
+; SSE2-NEXT: jne .LBB16_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB16_3
; SSE2-NEXT: .LBB16_4: # %else2
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm2, %xmm1
-; SSE2-NEXT: movd %xmm1, %ecx
-; SSE2-NEXT: movl %ecx, %edx
-; SSE2-NEXT: shrl $16, %edx
-; SSE2-NEXT: notb %dl
-; SSE2-NEXT: testb $1, %dl
-; SSE2-NEXT: je .LBB16_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: movl %eax, %edx
-; SSE2-NEXT: shrl $16, %edx
-; SSE2-NEXT: movb %dl, 2(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB16_5
; SSE2-NEXT: .LBB16_6: # %else4
-; SSE2-NEXT: shrl $24, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testb $8, %al
; SSE2-NEXT: je .LBB16_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
-; SSE2-NEXT: shrl $24, %eax
-; SSE2-NEXT: movb %al, 3(%rdi)
+; SSE2-NEXT: .LBB16_7: # %cond.store5
+; SSE2-NEXT: shrl $24, %ecx
+; SSE2-NEXT: movb %cl, 3(%rdi)
; SSE2-NEXT: .LBB16_8: # %else6
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm2, %xmm1
-; SSE2-NEXT: pextrw $2, %xmm1, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testb $16, %al
; SSE2-NEXT: pextrw $2, %xmm0, %ecx
; SSE2-NEXT: je .LBB16_10
; SSE2-NEXT: # %bb.9: # %cond.store7
; SSE2-NEXT: movb %cl, 4(%rdi)
; SSE2-NEXT: .LBB16_10: # %else8
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $32, %al
; SSE2-NEXT: je .LBB16_12
; SSE2-NEXT: # %bb.11: # %cond.store9
; SSE2-NEXT: movb %ch, 5(%rdi)
; SSE2-NEXT: .LBB16_12: # %else10
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm2, %xmm1
-; SSE2-NEXT: pextrw $3, %xmm1, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testb $64, %al
; SSE2-NEXT: pextrw $3, %xmm0, %ecx
; SSE2-NEXT: je .LBB16_14
; SSE2-NEXT: # %bb.13: # %cond.store11
; SSE2-NEXT: movb %cl, 6(%rdi)
; SSE2-NEXT: .LBB16_14: # %else12
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $-128, %al
; SSE2-NEXT: je .LBB16_16
; SSE2-NEXT: # %bb.15: # %cond.store13
; SSE2-NEXT: movb %ch, 7(%rdi)
; SSE2-NEXT: .LBB16_16: # %else14
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm2, %xmm1
-; SSE2-NEXT: pextrw $4, %xmm1, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $256, %eax # imm = 0x100
; SSE2-NEXT: pextrw $4, %xmm0, %ecx
; SSE2-NEXT: je .LBB16_18
; SSE2-NEXT: # %bb.17: # %cond.store15
; SSE2-NEXT: movb %cl, 8(%rdi)
; SSE2-NEXT: .LBB16_18: # %else16
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $512, %eax # imm = 0x200
; SSE2-NEXT: je .LBB16_20
; SSE2-NEXT: # %bb.19: # %cond.store17
; SSE2-NEXT: movb %ch, 9(%rdi)
; SSE2-NEXT: .LBB16_20: # %else18
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm2, %xmm1
-; SSE2-NEXT: pextrw $5, %xmm1, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $1024, %eax # imm = 0x400
; SSE2-NEXT: pextrw $5, %xmm0, %ecx
; SSE2-NEXT: je .LBB16_22
; SSE2-NEXT: # %bb.21: # %cond.store19
; SSE2-NEXT: movb %cl, 10(%rdi)
; SSE2-NEXT: .LBB16_22: # %else20
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $2048, %eax # imm = 0x800
; SSE2-NEXT: je .LBB16_24
; SSE2-NEXT: # %bb.23: # %cond.store21
; SSE2-NEXT: movb %ch, 11(%rdi)
; SSE2-NEXT: .LBB16_24: # %else22
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm2, %xmm1
-; SSE2-NEXT: pextrw $6, %xmm1, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $4096, %eax # imm = 0x1000
; SSE2-NEXT: pextrw $6, %xmm0, %ecx
; SSE2-NEXT: je .LBB16_26
; SSE2-NEXT: # %bb.25: # %cond.store23
; SSE2-NEXT: movb %cl, 12(%rdi)
; SSE2-NEXT: .LBB16_26: # %else24
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $8192, %eax # imm = 0x2000
; SSE2-NEXT: je .LBB16_28
; SSE2-NEXT: # %bb.27: # %cond.store25
; SSE2-NEXT: movb %ch, 13(%rdi)
; SSE2-NEXT: .LBB16_28: # %else26
-; SSE2-NEXT: pxor %xmm1, %xmm1
-; SSE2-NEXT: pcmpeqb %xmm1, %xmm2
-; SSE2-NEXT: pextrw $7, %xmm2, %eax
-; SSE2-NEXT: movl %eax, %ecx
-; SSE2-NEXT: notb %cl
-; SSE2-NEXT: testb $1, %cl
+; SSE2-NEXT: testl $16384, %eax # imm = 0x4000
; SSE2-NEXT: pextrw $7, %xmm0, %ecx
-; SSE2-NEXT: je .LBB16_30
-; SSE2-NEXT: # %bb.29: # %cond.store27
+; SSE2-NEXT: jne .LBB16_29
+; SSE2-NEXT: # %bb.30: # %else28
+; SSE2-NEXT: testl $32768, %eax # imm = 0x8000
+; SSE2-NEXT: jne .LBB16_31
+; SSE2-NEXT: .LBB16_32: # %else30
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB16_1: # %cond.store
+; SSE2-NEXT: movb %cl, (%rdi)
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: je .LBB16_4
+; SSE2-NEXT: .LBB16_3: # %cond.store1
+; SSE2-NEXT: movb %ch, 1(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: je .LBB16_6
+; SSE2-NEXT: .LBB16_5: # %cond.store3
+; SSE2-NEXT: movl %ecx, %edx
+; SSE2-NEXT: shrl $16, %edx
+; SSE2-NEXT: movb %dl, 2(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB16_7
+; SSE2-NEXT: jmp .LBB16_8
+; SSE2-NEXT: .LBB16_29: # %cond.store27
; SSE2-NEXT: movb %cl, 14(%rdi)
-; SSE2-NEXT: .LBB16_30: # %else28
-; SSE2-NEXT: shrl $8, %eax
-; SSE2-NEXT: notb %al
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testl $32768, %eax # imm = 0x8000
; SSE2-NEXT: je .LBB16_32
-; SSE2-NEXT: # %bb.31: # %cond.store29
+; SSE2-NEXT: .LBB16_31: # %cond.store29
; SSE2-NEXT: movb %ch, 15(%rdi)
-; SSE2-NEXT: .LBB16_32: # %else30
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v16i16_v16i8:
; SSE4: # %bb.0:
; SSE4-NEXT: pxor %xmm3, %xmm3
-; SSE4-NEXT: pcmpeqb %xmm2, %xmm3
; SSE4-NEXT: movdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
; SSE4-NEXT: pminuw %xmm4, %xmm1
; SSE4-NEXT: pminuw %xmm4, %xmm0
; SSE4-NEXT: packuswb %xmm1, %xmm0
-; SSE4-NEXT: pextrb $0, %xmm3, %eax
-; SSE4-NEXT: notb %al
+; SSE4-NEXT: pcmpeqb %xmm2, %xmm3
+; SSE4-NEXT: pmovmskb %xmm3, %eax
+; SSE4-NEXT: xorl $65535, %eax # imm = 0xFFFF
; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB16_2
-; SSE4-NEXT: # %bb.1: # %cond.store
+; SSE4-NEXT: jne .LBB16_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB16_3
+; SSE4-NEXT: .LBB16_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB16_5
+; SSE4-NEXT: .LBB16_6: # %else4
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB16_7
+; SSE4-NEXT: .LBB16_8: # %else6
+; SSE4-NEXT: testb $16, %al
+; SSE4-NEXT: jne .LBB16_9
+; SSE4-NEXT: .LBB16_10: # %else8
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: jne .LBB16_11
+; SSE4-NEXT: .LBB16_12: # %else10
+; SSE4-NEXT: testb $64, %al
+; SSE4-NEXT: jne .LBB16_13
+; SSE4-NEXT: .LBB16_14: # %else12
+; SSE4-NEXT: testb $-128, %al
+; SSE4-NEXT: jne .LBB16_15
+; SSE4-NEXT: .LBB16_16: # %else14
+; SSE4-NEXT: testl $256, %eax # imm = 0x100
+; SSE4-NEXT: jne .LBB16_17
+; SSE4-NEXT: .LBB16_18: # %else16
+; SSE4-NEXT: testl $512, %eax # imm = 0x200
+; SSE4-NEXT: jne .LBB16_19
+; SSE4-NEXT: .LBB16_20: # %else18
+; SSE4-NEXT: testl $1024, %eax # imm = 0x400
+; SSE4-NEXT: jne .LBB16_21
+; SSE4-NEXT: .LBB16_22: # %else20
+; SSE4-NEXT: testl $2048, %eax # imm = 0x800
+; SSE4-NEXT: jne .LBB16_23
+; SSE4-NEXT: .LBB16_24: # %else22
+; SSE4-NEXT: testl $4096, %eax # imm = 0x1000
+; SSE4-NEXT: jne .LBB16_25
+; SSE4-NEXT: .LBB16_26: # %else24
+; SSE4-NEXT: testl $8192, %eax # imm = 0x2000
+; SSE4-NEXT: jne .LBB16_27
+; SSE4-NEXT: .LBB16_28: # %else26
+; SSE4-NEXT: testl $16384, %eax # imm = 0x4000
+; SSE4-NEXT: jne .LBB16_29
+; SSE4-NEXT: .LBB16_30: # %else28
+; SSE4-NEXT: testl $32768, %eax # imm = 0x8000
+; SSE4-NEXT: jne .LBB16_31
+; SSE4-NEXT: .LBB16_32: # %else30
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB16_1: # %cond.store
; SSE4-NEXT: pextrb $0, %xmm0, (%rdi)
-; SSE4-NEXT: .LBB16_2: # %else
-; SSE4-NEXT: pextrb $1, %xmm3, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $2, %al
; SSE4-NEXT: je .LBB16_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
+; SSE4-NEXT: .LBB16_3: # %cond.store1
; SSE4-NEXT: pextrb $1, %xmm0, 1(%rdi)
-; SSE4-NEXT: .LBB16_4: # %else2
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm2, %xmm1
-; SSE4-NEXT: pextrb $2, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $4, %al
; SSE4-NEXT: je .LBB16_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
+; SSE4-NEXT: .LBB16_5: # %cond.store3
; SSE4-NEXT: pextrb $2, %xmm0, 2(%rdi)
-; SSE4-NEXT: .LBB16_6: # %else4
-; SSE4-NEXT: pextrb $3, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $8, %al
; SSE4-NEXT: je .LBB16_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
+; SSE4-NEXT: .LBB16_7: # %cond.store5
; SSE4-NEXT: pextrb $3, %xmm0, 3(%rdi)
-; SSE4-NEXT: .LBB16_8: # %else6
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm2, %xmm1
-; SSE4-NEXT: pextrb $4, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $16, %al
; SSE4-NEXT: je .LBB16_10
-; SSE4-NEXT: # %bb.9: # %cond.store7
+; SSE4-NEXT: .LBB16_9: # %cond.store7
; SSE4-NEXT: pextrb $4, %xmm0, 4(%rdi)
-; SSE4-NEXT: .LBB16_10: # %else8
-; SSE4-NEXT: pextrb $5, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $32, %al
; SSE4-NEXT: je .LBB16_12
-; SSE4-NEXT: # %bb.11: # %cond.store9
+; SSE4-NEXT: .LBB16_11: # %cond.store9
; SSE4-NEXT: pextrb $5, %xmm0, 5(%rdi)
-; SSE4-NEXT: .LBB16_12: # %else10
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm2, %xmm1
-; SSE4-NEXT: pextrb $6, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $64, %al
; SSE4-NEXT: je .LBB16_14
-; SSE4-NEXT: # %bb.13: # %cond.store11
+; SSE4-NEXT: .LBB16_13: # %cond.store11
; SSE4-NEXT: pextrb $6, %xmm0, 6(%rdi)
-; SSE4-NEXT: .LBB16_14: # %else12
-; SSE4-NEXT: pextrb $7, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $-128, %al
; SSE4-NEXT: je .LBB16_16
-; SSE4-NEXT: # %bb.15: # %cond.store13
+; SSE4-NEXT: .LBB16_15: # %cond.store13
; SSE4-NEXT: pextrb $7, %xmm0, 7(%rdi)
-; SSE4-NEXT: .LBB16_16: # %else14
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm2, %xmm1
-; SSE4-NEXT: pextrb $8, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $256, %eax # imm = 0x100
; SSE4-NEXT: je .LBB16_18
-; SSE4-NEXT: # %bb.17: # %cond.store15
+; SSE4-NEXT: .LBB16_17: # %cond.store15
; SSE4-NEXT: pextrb $8, %xmm0, 8(%rdi)
-; SSE4-NEXT: .LBB16_18: # %else16
-; SSE4-NEXT: pextrb $9, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $512, %eax # imm = 0x200
; SSE4-NEXT: je .LBB16_20
-; SSE4-NEXT: # %bb.19: # %cond.store17
+; SSE4-NEXT: .LBB16_19: # %cond.store17
; SSE4-NEXT: pextrb $9, %xmm0, 9(%rdi)
-; SSE4-NEXT: .LBB16_20: # %else18
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm2, %xmm1
-; SSE4-NEXT: pextrb $10, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $1024, %eax # imm = 0x400
; SSE4-NEXT: je .LBB16_22
-; SSE4-NEXT: # %bb.21: # %cond.store19
+; SSE4-NEXT: .LBB16_21: # %cond.store19
; SSE4-NEXT: pextrb $10, %xmm0, 10(%rdi)
-; SSE4-NEXT: .LBB16_22: # %else20
-; SSE4-NEXT: pextrb $11, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $2048, %eax # imm = 0x800
; SSE4-NEXT: je .LBB16_24
-; SSE4-NEXT: # %bb.23: # %cond.store21
+; SSE4-NEXT: .LBB16_23: # %cond.store21
; SSE4-NEXT: pextrb $11, %xmm0, 11(%rdi)
-; SSE4-NEXT: .LBB16_24: # %else22
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm2, %xmm1
-; SSE4-NEXT: pextrb $12, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $4096, %eax # imm = 0x1000
; SSE4-NEXT: je .LBB16_26
-; SSE4-NEXT: # %bb.25: # %cond.store23
+; SSE4-NEXT: .LBB16_25: # %cond.store23
; SSE4-NEXT: pextrb $12, %xmm0, 12(%rdi)
-; SSE4-NEXT: .LBB16_26: # %else24
-; SSE4-NEXT: pextrb $13, %xmm1, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $8192, %eax # imm = 0x2000
; SSE4-NEXT: je .LBB16_28
-; SSE4-NEXT: # %bb.27: # %cond.store25
+; SSE4-NEXT: .LBB16_27: # %cond.store25
; SSE4-NEXT: pextrb $13, %xmm0, 13(%rdi)
-; SSE4-NEXT: .LBB16_28: # %else26
-; SSE4-NEXT: pxor %xmm1, %xmm1
-; SSE4-NEXT: pcmpeqb %xmm1, %xmm2
-; SSE4-NEXT: pextrb $14, %xmm2, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $16384, %eax # imm = 0x4000
; SSE4-NEXT: je .LBB16_30
-; SSE4-NEXT: # %bb.29: # %cond.store27
+; SSE4-NEXT: .LBB16_29: # %cond.store27
; SSE4-NEXT: pextrb $14, %xmm0, 14(%rdi)
-; SSE4-NEXT: .LBB16_30: # %else28
-; SSE4-NEXT: pextrb $15, %xmm2, %eax
-; SSE4-NEXT: notb %al
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testl $32768, %eax # imm = 0x8000
; SSE4-NEXT: je .LBB16_32
-; SSE4-NEXT: # %bb.31: # %cond.store29
+; SSE4-NEXT: .LBB16_31: # %cond.store29
; SSE4-NEXT: pextrb $15, %xmm0, 15(%rdi)
-; SSE4-NEXT: .LBB16_32: # %else30
; SSE4-NEXT: retq
;
; AVX1-LABEL: truncstore_v16i16_v16i8:
; AVX1: # %bb.0:
; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
; AVX1-NEXT: vextractf128 $1, %ymm0, %xmm3
; AVX1-NEXT: vmovdqa {{.*#+}} xmm4 = [255,255,255,255,255,255,255,255]
; AVX1-NEXT: vpminuw %xmm4, %xmm3, %xmm3
; AVX1-NEXT: vpminuw %xmm4, %xmm0, %xmm0
; AVX1-NEXT: vpackuswb %xmm3, %xmm0, %xmm0
-; AVX1-NEXT: vpextrb $0, %xmm2, %eax
-; AVX1-NEXT: notb %al
+; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
+; AVX1-NEXT: vpmovmskb %xmm1, %eax
+; AVX1-NEXT: xorl $65535, %eax # imm = 0xFFFF
; AVX1-NEXT: testb $1, %al
-; AVX1-NEXT: je .LBB16_2
-; AVX1-NEXT: # %bb.1: # %cond.store
+; AVX1-NEXT: jne .LBB16_1
+; AVX1-NEXT: # %bb.2: # %else
+; AVX1-NEXT: testb $2, %al
+; AVX1-NEXT: jne .LBB16_3
+; AVX1-NEXT: .LBB16_4: # %else2
+; AVX1-NEXT: testb $4, %al
+; AVX1-NEXT: jne .LBB16_5
+; AVX1-NEXT: .LBB16_6: # %else4
+; AVX1-NEXT: testb $8, %al
+; AVX1-NEXT: jne .LBB16_7
+; AVX1-NEXT: .LBB16_8: # %else6
+; AVX1-NEXT: testb $16, %al
+; AVX1-NEXT: jne .LBB16_9
+; AVX1-NEXT: .LBB16_10: # %else8
+; AVX1-NEXT: testb $32, %al
+; AVX1-NEXT: jne .LBB16_11
+; AVX1-NEXT: .LBB16_12: # %else10
+; AVX1-NEXT: testb $64, %al
+; AVX1-NEXT: jne .LBB16_13
+; AVX1-NEXT: .LBB16_14: # %else12
+; AVX1-NEXT: testb $-128, %al
+; AVX1-NEXT: jne .LBB16_15
+; AVX1-NEXT: .LBB16_16: # %else14
+; AVX1-NEXT: testl $256, %eax # imm = 0x100
+; AVX1-NEXT: jne .LBB16_17
+; AVX1-NEXT: .LBB16_18: # %else16
+; AVX1-NEXT: testl $512, %eax # imm = 0x200
+; AVX1-NEXT: jne .LBB16_19
+; AVX1-NEXT: .LBB16_20: # %else18
+; AVX1-NEXT: testl $1024, %eax # imm = 0x400
+; AVX1-NEXT: jne .LBB16_21
+; AVX1-NEXT: .LBB16_22: # %else20
+; AVX1-NEXT: testl $2048, %eax # imm = 0x800
+; AVX1-NEXT: jne .LBB16_23
+; AVX1-NEXT: .LBB16_24: # %else22
+; AVX1-NEXT: testl $4096, %eax # imm = 0x1000
+; AVX1-NEXT: jne .LBB16_25
+; AVX1-NEXT: .LBB16_26: # %else24
+; AVX1-NEXT: testl $8192, %eax # imm = 0x2000
+; AVX1-NEXT: jne .LBB16_27
+; AVX1-NEXT: .LBB16_28: # %else26
+; AVX1-NEXT: testl $16384, %eax # imm = 0x4000
+; AVX1-NEXT: jne .LBB16_29
+; AVX1-NEXT: .LBB16_30: # %else28
+; AVX1-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX1-NEXT: jne .LBB16_31
+; AVX1-NEXT: .LBB16_32: # %else30
+; AVX1-NEXT: vzeroupper
+; AVX1-NEXT: retq
+; AVX1-NEXT: .LBB16_1: # %cond.store
; AVX1-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX1-NEXT: .LBB16_2: # %else
-; AVX1-NEXT: vpextrb $1, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $2, %al
; AVX1-NEXT: je .LBB16_4
-; AVX1-NEXT: # %bb.3: # %cond.store1
+; AVX1-NEXT: .LBB16_3: # %cond.store1
; AVX1-NEXT: vpextrb $1, %xmm0, 1(%rdi)
-; AVX1-NEXT: .LBB16_4: # %else2
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpextrb $2, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $4, %al
; AVX1-NEXT: je .LBB16_6
-; AVX1-NEXT: # %bb.5: # %cond.store3
+; AVX1-NEXT: .LBB16_5: # %cond.store3
; AVX1-NEXT: vpextrb $2, %xmm0, 2(%rdi)
-; AVX1-NEXT: .LBB16_6: # %else4
-; AVX1-NEXT: vpextrb $3, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $8, %al
; AVX1-NEXT: je .LBB16_8
-; AVX1-NEXT: # %bb.7: # %cond.store5
+; AVX1-NEXT: .LBB16_7: # %cond.store5
; AVX1-NEXT: vpextrb $3, %xmm0, 3(%rdi)
-; AVX1-NEXT: .LBB16_8: # %else6
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpextrb $4, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $16, %al
; AVX1-NEXT: je .LBB16_10
-; AVX1-NEXT: # %bb.9: # %cond.store7
+; AVX1-NEXT: .LBB16_9: # %cond.store7
; AVX1-NEXT: vpextrb $4, %xmm0, 4(%rdi)
-; AVX1-NEXT: .LBB16_10: # %else8
-; AVX1-NEXT: vpextrb $5, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $32, %al
; AVX1-NEXT: je .LBB16_12
-; AVX1-NEXT: # %bb.11: # %cond.store9
+; AVX1-NEXT: .LBB16_11: # %cond.store9
; AVX1-NEXT: vpextrb $5, %xmm0, 5(%rdi)
-; AVX1-NEXT: .LBB16_12: # %else10
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpextrb $6, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $64, %al
; AVX1-NEXT: je .LBB16_14
-; AVX1-NEXT: # %bb.13: # %cond.store11
+; AVX1-NEXT: .LBB16_13: # %cond.store11
; AVX1-NEXT: vpextrb $6, %xmm0, 6(%rdi)
-; AVX1-NEXT: .LBB16_14: # %else12
-; AVX1-NEXT: vpextrb $7, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testb $-128, %al
; AVX1-NEXT: je .LBB16_16
-; AVX1-NEXT: # %bb.15: # %cond.store13
+; AVX1-NEXT: .LBB16_15: # %cond.store13
; AVX1-NEXT: vpextrb $7, %xmm0, 7(%rdi)
-; AVX1-NEXT: .LBB16_16: # %else14
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpextrb $8, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $256, %eax # imm = 0x100
; AVX1-NEXT: je .LBB16_18
-; AVX1-NEXT: # %bb.17: # %cond.store15
+; AVX1-NEXT: .LBB16_17: # %cond.store15
; AVX1-NEXT: vpextrb $8, %xmm0, 8(%rdi)
-; AVX1-NEXT: .LBB16_18: # %else16
-; AVX1-NEXT: vpextrb $9, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $512, %eax # imm = 0x200
; AVX1-NEXT: je .LBB16_20
-; AVX1-NEXT: # %bb.19: # %cond.store17
+; AVX1-NEXT: .LBB16_19: # %cond.store17
; AVX1-NEXT: vpextrb $9, %xmm0, 9(%rdi)
-; AVX1-NEXT: .LBB16_20: # %else18
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpextrb $10, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $1024, %eax # imm = 0x400
; AVX1-NEXT: je .LBB16_22
-; AVX1-NEXT: # %bb.21: # %cond.store19
+; AVX1-NEXT: .LBB16_21: # %cond.store19
; AVX1-NEXT: vpextrb $10, %xmm0, 10(%rdi)
-; AVX1-NEXT: .LBB16_22: # %else20
-; AVX1-NEXT: vpextrb $11, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $2048, %eax # imm = 0x800
; AVX1-NEXT: je .LBB16_24
-; AVX1-NEXT: # %bb.23: # %cond.store21
+; AVX1-NEXT: .LBB16_23: # %cond.store21
; AVX1-NEXT: vpextrb $11, %xmm0, 11(%rdi)
-; AVX1-NEXT: .LBB16_24: # %else22
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX1-NEXT: vpextrb $12, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $4096, %eax # imm = 0x1000
; AVX1-NEXT: je .LBB16_26
-; AVX1-NEXT: # %bb.25: # %cond.store23
+; AVX1-NEXT: .LBB16_25: # %cond.store23
; AVX1-NEXT: vpextrb $12, %xmm0, 12(%rdi)
-; AVX1-NEXT: .LBB16_26: # %else24
-; AVX1-NEXT: vpextrb $13, %xmm2, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $8192, %eax # imm = 0x2000
; AVX1-NEXT: je .LBB16_28
-; AVX1-NEXT: # %bb.27: # %cond.store25
+; AVX1-NEXT: .LBB16_27: # %cond.store25
; AVX1-NEXT: vpextrb $13, %xmm0, 13(%rdi)
-; AVX1-NEXT: .LBB16_28: # %else26
-; AVX1-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT: vpextrb $14, %xmm1, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $16384, %eax # imm = 0x4000
; AVX1-NEXT: je .LBB16_30
-; AVX1-NEXT: # %bb.29: # %cond.store27
+; AVX1-NEXT: .LBB16_29: # %cond.store27
; AVX1-NEXT: vpextrb $14, %xmm0, 14(%rdi)
-; AVX1-NEXT: .LBB16_30: # %else28
-; AVX1-NEXT: vpextrb $15, %xmm1, %eax
-; AVX1-NEXT: notb %al
-; AVX1-NEXT: testb $1, %al
+; AVX1-NEXT: testl $32768, %eax # imm = 0x8000
; AVX1-NEXT: je .LBB16_32
-; AVX1-NEXT: # %bb.31: # %cond.store29
+; AVX1-NEXT: .LBB16_31: # %cond.store29
; AVX1-NEXT: vpextrb $15, %xmm0, 15(%rdi)
-; AVX1-NEXT: .LBB16_32: # %else30
; AVX1-NEXT: vzeroupper
; AVX1-NEXT: retq
;
; AVX2-LABEL: truncstore_v16i16_v16i8:
; AVX2: # %bb.0:
; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
; AVX2-NEXT: vpminuw {{.*}}(%rip), %ymm0, %ymm0
; AVX2-NEXT: vextracti128 $1, %ymm0, %xmm3
; AVX2-NEXT: vpackuswb %xmm3, %xmm0, %xmm0
-; AVX2-NEXT: vpextrb $0, %xmm2, %eax
-; AVX2-NEXT: notb %al
+; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
+; AVX2-NEXT: vpmovmskb %xmm1, %eax
+; AVX2-NEXT: xorl $65535, %eax # imm = 0xFFFF
; AVX2-NEXT: testb $1, %al
-; AVX2-NEXT: je .LBB16_2
-; AVX2-NEXT: # %bb.1: # %cond.store
+; AVX2-NEXT: jne .LBB16_1
+; AVX2-NEXT: # %bb.2: # %else
+; AVX2-NEXT: testb $2, %al
+; AVX2-NEXT: jne .LBB16_3
+; AVX2-NEXT: .LBB16_4: # %else2
+; AVX2-NEXT: testb $4, %al
+; AVX2-NEXT: jne .LBB16_5
+; AVX2-NEXT: .LBB16_6: # %else4
+; AVX2-NEXT: testb $8, %al
+; AVX2-NEXT: jne .LBB16_7
+; AVX2-NEXT: .LBB16_8: # %else6
+; AVX2-NEXT: testb $16, %al
+; AVX2-NEXT: jne .LBB16_9
+; AVX2-NEXT: .LBB16_10: # %else8
+; AVX2-NEXT: testb $32, %al
+; AVX2-NEXT: jne .LBB16_11
+; AVX2-NEXT: .LBB16_12: # %else10
+; AVX2-NEXT: testb $64, %al
+; AVX2-NEXT: jne .LBB16_13
+; AVX2-NEXT: .LBB16_14: # %else12
+; AVX2-NEXT: testb $-128, %al
+; AVX2-NEXT: jne .LBB16_15
+; AVX2-NEXT: .LBB16_16: # %else14
+; AVX2-NEXT: testl $256, %eax # imm = 0x100
+; AVX2-NEXT: jne .LBB16_17
+; AVX2-NEXT: .LBB16_18: # %else16
+; AVX2-NEXT: testl $512, %eax # imm = 0x200
+; AVX2-NEXT: jne .LBB16_19
+; AVX2-NEXT: .LBB16_20: # %else18
+; AVX2-NEXT: testl $1024, %eax # imm = 0x400
+; AVX2-NEXT: jne .LBB16_21
+; AVX2-NEXT: .LBB16_22: # %else20
+; AVX2-NEXT: testl $2048, %eax # imm = 0x800
+; AVX2-NEXT: jne .LBB16_23
+; AVX2-NEXT: .LBB16_24: # %else22
+; AVX2-NEXT: testl $4096, %eax # imm = 0x1000
+; AVX2-NEXT: jne .LBB16_25
+; AVX2-NEXT: .LBB16_26: # %else24
+; AVX2-NEXT: testl $8192, %eax # imm = 0x2000
+; AVX2-NEXT: jne .LBB16_27
+; AVX2-NEXT: .LBB16_28: # %else26
+; AVX2-NEXT: testl $16384, %eax # imm = 0x4000
+; AVX2-NEXT: jne .LBB16_29
+; AVX2-NEXT: .LBB16_30: # %else28
+; AVX2-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX2-NEXT: jne .LBB16_31
+; AVX2-NEXT: .LBB16_32: # %else30
+; AVX2-NEXT: vzeroupper
+; AVX2-NEXT: retq
+; AVX2-NEXT: .LBB16_1: # %cond.store
; AVX2-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX2-NEXT: .LBB16_2: # %else
-; AVX2-NEXT: vpextrb $1, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $2, %al
; AVX2-NEXT: je .LBB16_4
-; AVX2-NEXT: # %bb.3: # %cond.store1
+; AVX2-NEXT: .LBB16_3: # %cond.store1
; AVX2-NEXT: vpextrb $1, %xmm0, 1(%rdi)
-; AVX2-NEXT: .LBB16_4: # %else2
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $2, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $4, %al
; AVX2-NEXT: je .LBB16_6
-; AVX2-NEXT: # %bb.5: # %cond.store3
+; AVX2-NEXT: .LBB16_5: # %cond.store3
; AVX2-NEXT: vpextrb $2, %xmm0, 2(%rdi)
-; AVX2-NEXT: .LBB16_6: # %else4
-; AVX2-NEXT: vpextrb $3, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $8, %al
; AVX2-NEXT: je .LBB16_8
-; AVX2-NEXT: # %bb.7: # %cond.store5
+; AVX2-NEXT: .LBB16_7: # %cond.store5
; AVX2-NEXT: vpextrb $3, %xmm0, 3(%rdi)
-; AVX2-NEXT: .LBB16_8: # %else6
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $4, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $16, %al
; AVX2-NEXT: je .LBB16_10
-; AVX2-NEXT: # %bb.9: # %cond.store7
+; AVX2-NEXT: .LBB16_9: # %cond.store7
; AVX2-NEXT: vpextrb $4, %xmm0, 4(%rdi)
-; AVX2-NEXT: .LBB16_10: # %else8
-; AVX2-NEXT: vpextrb $5, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $32, %al
; AVX2-NEXT: je .LBB16_12
-; AVX2-NEXT: # %bb.11: # %cond.store9
+; AVX2-NEXT: .LBB16_11: # %cond.store9
; AVX2-NEXT: vpextrb $5, %xmm0, 5(%rdi)
-; AVX2-NEXT: .LBB16_12: # %else10
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $6, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $64, %al
; AVX2-NEXT: je .LBB16_14
-; AVX2-NEXT: # %bb.13: # %cond.store11
+; AVX2-NEXT: .LBB16_13: # %cond.store11
; AVX2-NEXT: vpextrb $6, %xmm0, 6(%rdi)
-; AVX2-NEXT: .LBB16_14: # %else12
-; AVX2-NEXT: vpextrb $7, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testb $-128, %al
; AVX2-NEXT: je .LBB16_16
-; AVX2-NEXT: # %bb.15: # %cond.store13
+; AVX2-NEXT: .LBB16_15: # %cond.store13
; AVX2-NEXT: vpextrb $7, %xmm0, 7(%rdi)
-; AVX2-NEXT: .LBB16_16: # %else14
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $8, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $256, %eax # imm = 0x100
; AVX2-NEXT: je .LBB16_18
-; AVX2-NEXT: # %bb.17: # %cond.store15
+; AVX2-NEXT: .LBB16_17: # %cond.store15
; AVX2-NEXT: vpextrb $8, %xmm0, 8(%rdi)
-; AVX2-NEXT: .LBB16_18: # %else16
-; AVX2-NEXT: vpextrb $9, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $512, %eax # imm = 0x200
; AVX2-NEXT: je .LBB16_20
-; AVX2-NEXT: # %bb.19: # %cond.store17
+; AVX2-NEXT: .LBB16_19: # %cond.store17
; AVX2-NEXT: vpextrb $9, %xmm0, 9(%rdi)
-; AVX2-NEXT: .LBB16_20: # %else18
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $10, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $1024, %eax # imm = 0x400
; AVX2-NEXT: je .LBB16_22
-; AVX2-NEXT: # %bb.21: # %cond.store19
+; AVX2-NEXT: .LBB16_21: # %cond.store19
; AVX2-NEXT: vpextrb $10, %xmm0, 10(%rdi)
-; AVX2-NEXT: .LBB16_22: # %else20
-; AVX2-NEXT: vpextrb $11, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $2048, %eax # imm = 0x800
; AVX2-NEXT: je .LBB16_24
-; AVX2-NEXT: # %bb.23: # %cond.store21
+; AVX2-NEXT: .LBB16_23: # %cond.store21
; AVX2-NEXT: vpextrb $11, %xmm0, 11(%rdi)
-; AVX2-NEXT: .LBB16_24: # %else22
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX2-NEXT: vpextrb $12, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $4096, %eax # imm = 0x1000
; AVX2-NEXT: je .LBB16_26
-; AVX2-NEXT: # %bb.25: # %cond.store23
+; AVX2-NEXT: .LBB16_25: # %cond.store23
; AVX2-NEXT: vpextrb $12, %xmm0, 12(%rdi)
-; AVX2-NEXT: .LBB16_26: # %else24
-; AVX2-NEXT: vpextrb $13, %xmm2, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $8192, %eax # imm = 0x2000
; AVX2-NEXT: je .LBB16_28
-; AVX2-NEXT: # %bb.27: # %cond.store25
+; AVX2-NEXT: .LBB16_27: # %cond.store25
; AVX2-NEXT: vpextrb $13, %xmm0, 13(%rdi)
-; AVX2-NEXT: .LBB16_28: # %else26
-; AVX2-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
-; AVX2-NEXT: vpextrb $14, %xmm1, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $16384, %eax # imm = 0x4000
; AVX2-NEXT: je .LBB16_30
-; AVX2-NEXT: # %bb.29: # %cond.store27
+; AVX2-NEXT: .LBB16_29: # %cond.store27
; AVX2-NEXT: vpextrb $14, %xmm0, 14(%rdi)
-; AVX2-NEXT: .LBB16_30: # %else28
-; AVX2-NEXT: vpextrb $15, %xmm1, %eax
-; AVX2-NEXT: notb %al
-; AVX2-NEXT: testb $1, %al
+; AVX2-NEXT: testl $32768, %eax # imm = 0x8000
; AVX2-NEXT: je .LBB16_32
-; AVX2-NEXT: # %bb.31: # %cond.store29
+; AVX2-NEXT: .LBB16_31: # %cond.store29
; AVX2-NEXT: vpextrb $15, %xmm0, 15(%rdi)
-; AVX2-NEXT: .LBB16_32: # %else30
; AVX2-NEXT: vzeroupper
; AVX2-NEXT: retq
;
; AVX512F-LABEL: truncstore_v16i16_v16i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
+; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
; AVX512F-NEXT: vpminuw {{.*}}(%rip), %ymm0, %ymm0
; AVX512F-NEXT: vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
; AVX512F-NEXT: vpmovdb %zmm0, %xmm0
-; AVX512F-NEXT: kmovw %k0, %eax
+; AVX512F-NEXT: vpmovmskb %xmm1, %eax
+; AVX512F-NEXT: xorl $65535, %eax # imm = 0xFFFF
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB16_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB16_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB16_3
+; AVX512F-NEXT: .LBB16_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB16_5
+; AVX512F-NEXT: .LBB16_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB16_7
+; AVX512F-NEXT: .LBB16_8: # %else6
+; AVX512F-NEXT: testb $16, %al
+; AVX512F-NEXT: jne .LBB16_9
+; AVX512F-NEXT: .LBB16_10: # %else8
+; AVX512F-NEXT: testb $32, %al
+; AVX512F-NEXT: jne .LBB16_11
+; AVX512F-NEXT: .LBB16_12: # %else10
+; AVX512F-NEXT: testb $64, %al
+; AVX512F-NEXT: jne .LBB16_13
+; AVX512F-NEXT: .LBB16_14: # %else12
+; AVX512F-NEXT: testb $-128, %al
+; AVX512F-NEXT: jne .LBB16_15
+; AVX512F-NEXT: .LBB16_16: # %else14
+; AVX512F-NEXT: testl $256, %eax # imm = 0x100
+; AVX512F-NEXT: jne .LBB16_17
+; AVX512F-NEXT: .LBB16_18: # %else16
+; AVX512F-NEXT: testl $512, %eax # imm = 0x200
+; AVX512F-NEXT: jne .LBB16_19
+; AVX512F-NEXT: .LBB16_20: # %else18
+; AVX512F-NEXT: testl $1024, %eax # imm = 0x400
+; AVX512F-NEXT: jne .LBB16_21
+; AVX512F-NEXT: .LBB16_22: # %else20
+; AVX512F-NEXT: testl $2048, %eax # imm = 0x800
+; AVX512F-NEXT: jne .LBB16_23
+; AVX512F-NEXT: .LBB16_24: # %else22
+; AVX512F-NEXT: testl $4096, %eax # imm = 0x1000
+; AVX512F-NEXT: jne .LBB16_25
+; AVX512F-NEXT: .LBB16_26: # %else24
+; AVX512F-NEXT: testl $8192, %eax # imm = 0x2000
+; AVX512F-NEXT: jne .LBB16_27
+; AVX512F-NEXT: .LBB16_28: # %else26
+; AVX512F-NEXT: testl $16384, %eax # imm = 0x4000
+; AVX512F-NEXT: jne .LBB16_29
+; AVX512F-NEXT: .LBB16_30: # %else28
+; AVX512F-NEXT: testl $32768, %eax # imm = 0x8000
+; AVX512F-NEXT: jne .LBB16_31
+; AVX512F-NEXT: .LBB16_32: # %else30
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB16_1: # %cond.store
; AVX512F-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB16_2: # %else
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB16_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB16_3: # %cond.store1
; AVX512F-NEXT: vpextrb $1, %xmm0, 1(%rdi)
-; AVX512F-NEXT: .LBB16_4: # %else2
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $2, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB16_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB16_5: # %cond.store3
; AVX512F-NEXT: vpextrb $2, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB16_6: # %else4
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $3, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB16_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB16_7: # %cond.store5
; AVX512F-NEXT: vpextrb $3, %xmm0, 3(%rdi)
-; AVX512F-NEXT: .LBB16_8: # %else6
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $4, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $16, %al
; AVX512F-NEXT: je .LBB16_10
-; AVX512F-NEXT: # %bb.9: # %cond.store7
+; AVX512F-NEXT: .LBB16_9: # %cond.store7
; AVX512F-NEXT: vpextrb $4, %xmm0, 4(%rdi)
-; AVX512F-NEXT: .LBB16_10: # %else8
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $5, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $32, %al
; AVX512F-NEXT: je .LBB16_12
-; AVX512F-NEXT: # %bb.11: # %cond.store9
+; AVX512F-NEXT: .LBB16_11: # %cond.store9
; AVX512F-NEXT: vpextrb $5, %xmm0, 5(%rdi)
-; AVX512F-NEXT: .LBB16_12: # %else10
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $6, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $64, %al
; AVX512F-NEXT: je .LBB16_14
-; AVX512F-NEXT: # %bb.13: # %cond.store11
+; AVX512F-NEXT: .LBB16_13: # %cond.store11
; AVX512F-NEXT: vpextrb $6, %xmm0, 6(%rdi)
-; AVX512F-NEXT: .LBB16_14: # %else12
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $7, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $-128, %al
; AVX512F-NEXT: je .LBB16_16
-; AVX512F-NEXT: # %bb.15: # %cond.store13
+; AVX512F-NEXT: .LBB16_15: # %cond.store13
; AVX512F-NEXT: vpextrb $7, %xmm0, 7(%rdi)
-; AVX512F-NEXT: .LBB16_16: # %else14
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $8, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $256, %eax # imm = 0x100
; AVX512F-NEXT: je .LBB16_18
-; AVX512F-NEXT: # %bb.17: # %cond.store15
+; AVX512F-NEXT: .LBB16_17: # %cond.store15
; AVX512F-NEXT: vpextrb $8, %xmm0, 8(%rdi)
-; AVX512F-NEXT: .LBB16_18: # %else16
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $9, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $512, %eax # imm = 0x200
; AVX512F-NEXT: je .LBB16_20
-; AVX512F-NEXT: # %bb.19: # %cond.store17
+; AVX512F-NEXT: .LBB16_19: # %cond.store17
; AVX512F-NEXT: vpextrb $9, %xmm0, 9(%rdi)
-; AVX512F-NEXT: .LBB16_20: # %else18
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $10, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $1024, %eax # imm = 0x400
; AVX512F-NEXT: je .LBB16_22
-; AVX512F-NEXT: # %bb.21: # %cond.store19
+; AVX512F-NEXT: .LBB16_21: # %cond.store19
; AVX512F-NEXT: vpextrb $10, %xmm0, 10(%rdi)
-; AVX512F-NEXT: .LBB16_22: # %else20
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $11, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $2048, %eax # imm = 0x800
; AVX512F-NEXT: je .LBB16_24
-; AVX512F-NEXT: # %bb.23: # %cond.store21
+; AVX512F-NEXT: .LBB16_23: # %cond.store21
; AVX512F-NEXT: vpextrb $11, %xmm0, 11(%rdi)
-; AVX512F-NEXT: .LBB16_24: # %else22
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxbd %xmm3, %zmm3
-; AVX512F-NEXT: vptestmd %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $12, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $4096, %eax # imm = 0x1000
; AVX512F-NEXT: je .LBB16_26
-; AVX512F-NEXT: # %bb.25: # %cond.store23
+; AVX512F-NEXT: .LBB16_25: # %cond.store23
; AVX512F-NEXT: vpextrb $12, %xmm0, 12(%rdi)
-; AVX512F-NEXT: .LBB16_26: # %else24
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $13, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $8192, %eax # imm = 0x2000
; AVX512F-NEXT: je .LBB16_28
-; AVX512F-NEXT: # %bb.27: # %cond.store25
+; AVX512F-NEXT: .LBB16_27: # %cond.store25
; AVX512F-NEXT: vpextrb $13, %xmm0, 13(%rdi)
-; AVX512F-NEXT: .LBB16_28: # %else26
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqb %xmm2, %xmm1, %xmm1
-; AVX512F-NEXT: vmovdqa64 %zmm1, %zmm2
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm2
-; AVX512F-NEXT: vpmovsxbd %xmm2, %zmm2
-; AVX512F-NEXT: vptestmd %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $14, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $16384, %eax # imm = 0x4000
; AVX512F-NEXT: je .LBB16_30
-; AVX512F-NEXT: # %bb.29: # %cond.store27
+; AVX512F-NEXT: .LBB16_29: # %cond.store27
; AVX512F-NEXT: vpextrb $14, %xmm0, 14(%rdi)
-; AVX512F-NEXT: .LBB16_30: # %else28
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovsxbd %xmm1, %zmm1
-; AVX512F-NEXT: vptestmd %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $15, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testl $32768, %eax # imm = 0x8000
; AVX512F-NEXT: je .LBB16_32
-; AVX512F-NEXT: # %bb.31: # %cond.store29
+; AVX512F-NEXT: .LBB16_31: # %cond.store29
; AVX512F-NEXT: vpextrb $15, %xmm0, 15(%rdi)
-; AVX512F-NEXT: .LBB16_32: # %else30
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@@ -7895,315 +7041,277 @@ define void @truncstore_v8i16_v8i8(<8 x
; SSE2-LABEL: truncstore_v8i16_v8i8:
; SSE2: # %bb.0:
; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: pcmpeqw %xmm1, %xmm2
-; SSE2-NEXT: pcmpeqd %xmm3, %xmm3
-; SSE2-NEXT: pxor %xmm2, %xmm3
-; SSE2-NEXT: movdqa {{.*#+}} xmm2 = [32768,32768,32768,32768,32768,32768,32768,32768]
-; SSE2-NEXT: pxor %xmm2, %xmm0
+; SSE2-NEXT: movdqa {{.*#+}} xmm3 = [32768,32768,32768,32768,32768,32768,32768,32768]
+; SSE2-NEXT: pxor %xmm3, %xmm0
; SSE2-NEXT: pminsw {{.*}}(%rip), %xmm0
-; SSE2-NEXT: pxor %xmm2, %xmm0
-; SSE2-NEXT: movd %xmm3, %ecx
-; SSE2-NEXT: testb $1, %cl
-; SSE2-NEXT: movd %xmm0, %eax
-; SSE2-NEXT: je .LBB17_2
-; SSE2-NEXT: # %bb.1: # %cond.store
-; SSE2-NEXT: movb %al, (%rdi)
-; SSE2-NEXT: .LBB17_2: # %else
-; SSE2-NEXT: shrl $16, %ecx
-; SSE2-NEXT: testb $1, %cl
-; SSE2-NEXT: je .LBB17_4
-; SSE2-NEXT: # %bb.3: # %cond.store1
-; SSE2-NEXT: shrl $16, %eax
-; SSE2-NEXT: movb %al, 1(%rdi)
-; SSE2-NEXT: .LBB17_4: # %else2
-; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqw %xmm1, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE2-NEXT: pxor %xmm3, %xmm2
-; SSE2-NEXT: pextrw $2, %xmm2, %eax
+; SSE2-NEXT: pxor %xmm3, %xmm0
+; SSE2-NEXT: pcmpeqw %xmm1, %xmm2
+; SSE2-NEXT: pcmpeqd %xmm1, %xmm1
+; SSE2-NEXT: pxor %xmm2, %xmm1
+; SSE2-NEXT: packsswb %xmm0, %xmm1
+; SSE2-NEXT: pmovmskb %xmm1, %eax
; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB17_6
-; SSE2-NEXT: # %bb.5: # %cond.store3
-; SSE2-NEXT: pextrw $2, %xmm0, %eax
-; SSE2-NEXT: movb %al, 2(%rdi)
+; SSE2-NEXT: movd %xmm0, %ecx
+; SSE2-NEXT: jne .LBB17_1
+; SSE2-NEXT: # %bb.2: # %else
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: jne .LBB17_3
+; SSE2-NEXT: .LBB17_4: # %else2
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: jne .LBB17_5
; SSE2-NEXT: .LBB17_6: # %else4
-; SSE2-NEXT: pextrw $3, %xmm2, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB17_8
-; SSE2-NEXT: # %bb.7: # %cond.store5
-; SSE2-NEXT: pextrw $3, %xmm0, %eax
-; SSE2-NEXT: movb %al, 3(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: jne .LBB17_7
; SSE2-NEXT: .LBB17_8: # %else6
-; SSE2-NEXT: pxor %xmm3, %xmm3
-; SSE2-NEXT: pcmpeqw %xmm1, %xmm3
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE2-NEXT: pxor %xmm3, %xmm2
-; SSE2-NEXT: pextrw $4, %xmm2, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB17_10
-; SSE2-NEXT: # %bb.9: # %cond.store7
-; SSE2-NEXT: pextrw $4, %xmm0, %eax
-; SSE2-NEXT: movb %al, 4(%rdi)
+; SSE2-NEXT: testb $16, %al
+; SSE2-NEXT: jne .LBB17_9
; SSE2-NEXT: .LBB17_10: # %else8
-; SSE2-NEXT: pextrw $5, %xmm2, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB17_12
-; SSE2-NEXT: # %bb.11: # %cond.store9
-; SSE2-NEXT: pextrw $5, %xmm0, %eax
-; SSE2-NEXT: movb %al, 5(%rdi)
+; SSE2-NEXT: testb $32, %al
+; SSE2-NEXT: jne .LBB17_11
; SSE2-NEXT: .LBB17_12: # %else10
-; SSE2-NEXT: pxor %xmm2, %xmm2
-; SSE2-NEXT: pcmpeqw %xmm2, %xmm1
-; SSE2-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE2-NEXT: pxor %xmm1, %xmm2
-; SSE2-NEXT: pextrw $6, %xmm2, %eax
-; SSE2-NEXT: testb $1, %al
-; SSE2-NEXT: je .LBB17_14
-; SSE2-NEXT: # %bb.13: # %cond.store11
-; SSE2-NEXT: pextrw $6, %xmm0, %eax
-; SSE2-NEXT: movb %al, 6(%rdi)
+; SSE2-NEXT: testb $64, %al
+; SSE2-NEXT: jne .LBB17_13
; SSE2-NEXT: .LBB17_14: # %else12
-; SSE2-NEXT: pextrw $7, %xmm2, %eax
-; SSE2-NEXT: testb $1, %al
+; SSE2-NEXT: testb $-128, %al
+; SSE2-NEXT: jne .LBB17_15
+; SSE2-NEXT: .LBB17_16: # %else14
+; SSE2-NEXT: retq
+; SSE2-NEXT: .LBB17_1: # %cond.store
+; SSE2-NEXT: movb %cl, (%rdi)
+; SSE2-NEXT: testb $2, %al
+; SSE2-NEXT: je .LBB17_4
+; SSE2-NEXT: .LBB17_3: # %cond.store1
+; SSE2-NEXT: shrl $16, %ecx
+; SSE2-NEXT: movb %cl, 1(%rdi)
+; SSE2-NEXT: testb $4, %al
+; SSE2-NEXT: je .LBB17_6
+; SSE2-NEXT: .LBB17_5: # %cond.store3
+; SSE2-NEXT: pextrw $2, %xmm0, %ecx
+; SSE2-NEXT: movb %cl, 2(%rdi)
+; SSE2-NEXT: testb $8, %al
+; SSE2-NEXT: je .LBB17_8
+; SSE2-NEXT: .LBB17_7: # %cond.store5
+; SSE2-NEXT: pextrw $3, %xmm0, %ecx
+; SSE2-NEXT: movb %cl, 3(%rdi)
+; SSE2-NEXT: testb $16, %al
+; SSE2-NEXT: je .LBB17_10
+; SSE2-NEXT: .LBB17_9: # %cond.store7
+; SSE2-NEXT: pextrw $4, %xmm0, %ecx
+; SSE2-NEXT: movb %cl, 4(%rdi)
+; SSE2-NEXT: testb $32, %al
+; SSE2-NEXT: je .LBB17_12
+; SSE2-NEXT: .LBB17_11: # %cond.store9
+; SSE2-NEXT: pextrw $5, %xmm0, %ecx
+; SSE2-NEXT: movb %cl, 5(%rdi)
+; SSE2-NEXT: testb $64, %al
+; SSE2-NEXT: je .LBB17_14
+; SSE2-NEXT: .LBB17_13: # %cond.store11
+; SSE2-NEXT: pextrw $6, %xmm0, %ecx
+; SSE2-NEXT: movb %cl, 6(%rdi)
+; SSE2-NEXT: testb $-128, %al
; SSE2-NEXT: je .LBB17_16
-; SSE2-NEXT: # %bb.15: # %cond.store13
+; SSE2-NEXT: .LBB17_15: # %cond.store13
; SSE2-NEXT: pextrw $7, %xmm0, %eax
; SSE2-NEXT: movb %al, 7(%rdi)
-; SSE2-NEXT: .LBB17_16: # %else14
; SSE2-NEXT: retq
;
; SSE4-LABEL: truncstore_v8i16_v8i8:
; SSE4: # %bb.0:
-; SSE4-NEXT: pxor %xmm3, %xmm3
-; SSE4-NEXT: pcmpeqw %xmm1, %xmm3
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE4-NEXT: pxor %xmm3, %xmm2
+; SSE4-NEXT: pxor %xmm2, %xmm2
; SSE4-NEXT: pminuw {{.*}}(%rip), %xmm0
-; SSE4-NEXT: pextrb $0, %xmm2, %eax
+; SSE4-NEXT: pcmpeqw %xmm1, %xmm2
+; SSE4-NEXT: pcmpeqd %xmm1, %xmm1
+; SSE4-NEXT: pxor %xmm2, %xmm1
+; SSE4-NEXT: packsswb %xmm0, %xmm1
+; SSE4-NEXT: pmovmskb %xmm1, %eax
; SSE4-NEXT: testb $1, %al
-; SSE4-NEXT: je .LBB17_2
-; SSE4-NEXT: # %bb.1: # %cond.store
+; SSE4-NEXT: jne .LBB17_1
+; SSE4-NEXT: # %bb.2: # %else
+; SSE4-NEXT: testb $2, %al
+; SSE4-NEXT: jne .LBB17_3
+; SSE4-NEXT: .LBB17_4: # %else2
+; SSE4-NEXT: testb $4, %al
+; SSE4-NEXT: jne .LBB17_5
+; SSE4-NEXT: .LBB17_6: # %else4
+; SSE4-NEXT: testb $8, %al
+; SSE4-NEXT: jne .LBB17_7
+; SSE4-NEXT: .LBB17_8: # %else6
+; SSE4-NEXT: testb $16, %al
+; SSE4-NEXT: jne .LBB17_9
+; SSE4-NEXT: .LBB17_10: # %else8
+; SSE4-NEXT: testb $32, %al
+; SSE4-NEXT: jne .LBB17_11
+; SSE4-NEXT: .LBB17_12: # %else10
+; SSE4-NEXT: testb $64, %al
+; SSE4-NEXT: jne .LBB17_13
+; SSE4-NEXT: .LBB17_14: # %else12
+; SSE4-NEXT: testb $-128, %al
+; SSE4-NEXT: jne .LBB17_15
+; SSE4-NEXT: .LBB17_16: # %else14
+; SSE4-NEXT: retq
+; SSE4-NEXT: .LBB17_1: # %cond.store
; SSE4-NEXT: pextrb $0, %xmm0, (%rdi)
-; SSE4-NEXT: .LBB17_2: # %else
-; SSE4-NEXT: pextrb $2, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $2, %al
; SSE4-NEXT: je .LBB17_4
-; SSE4-NEXT: # %bb.3: # %cond.store1
+; SSE4-NEXT: .LBB17_3: # %cond.store1
; SSE4-NEXT: pextrb $2, %xmm0, 1(%rdi)
-; SSE4-NEXT: .LBB17_4: # %else2
-; SSE4-NEXT: pxor %xmm3, %xmm3
-; SSE4-NEXT: pcmpeqw %xmm1, %xmm3
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE4-NEXT: pxor %xmm3, %xmm2
-; SSE4-NEXT: pextrb $4, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $4, %al
; SSE4-NEXT: je .LBB17_6
-; SSE4-NEXT: # %bb.5: # %cond.store3
+; SSE4-NEXT: .LBB17_5: # %cond.store3
; SSE4-NEXT: pextrb $4, %xmm0, 2(%rdi)
-; SSE4-NEXT: .LBB17_6: # %else4
-; SSE4-NEXT: pextrb $6, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $8, %al
; SSE4-NEXT: je .LBB17_8
-; SSE4-NEXT: # %bb.7: # %cond.store5
+; SSE4-NEXT: .LBB17_7: # %cond.store5
; SSE4-NEXT: pextrb $6, %xmm0, 3(%rdi)
-; SSE4-NEXT: .LBB17_8: # %else6
-; SSE4-NEXT: pxor %xmm3, %xmm3
-; SSE4-NEXT: pcmpeqw %xmm1, %xmm3
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE4-NEXT: pxor %xmm3, %xmm2
-; SSE4-NEXT: pextrb $8, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $16, %al
; SSE4-NEXT: je .LBB17_10
-; SSE4-NEXT: # %bb.9: # %cond.store7
+; SSE4-NEXT: .LBB17_9: # %cond.store7
; SSE4-NEXT: pextrb $8, %xmm0, 4(%rdi)
-; SSE4-NEXT: .LBB17_10: # %else8
-; SSE4-NEXT: pextrb $10, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $32, %al
; SSE4-NEXT: je .LBB17_12
-; SSE4-NEXT: # %bb.11: # %cond.store9
+; SSE4-NEXT: .LBB17_11: # %cond.store9
; SSE4-NEXT: pextrb $10, %xmm0, 5(%rdi)
-; SSE4-NEXT: .LBB17_12: # %else10
-; SSE4-NEXT: pxor %xmm2, %xmm2
-; SSE4-NEXT: pcmpeqw %xmm2, %xmm1
-; SSE4-NEXT: pcmpeqd %xmm2, %xmm2
-; SSE4-NEXT: pxor %xmm1, %xmm2
-; SSE4-NEXT: pextrb $12, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $64, %al
; SSE4-NEXT: je .LBB17_14
-; SSE4-NEXT: # %bb.13: # %cond.store11
+; SSE4-NEXT: .LBB17_13: # %cond.store11
; SSE4-NEXT: pextrb $12, %xmm0, 6(%rdi)
-; SSE4-NEXT: .LBB17_14: # %else12
-; SSE4-NEXT: pextrb $14, %xmm2, %eax
-; SSE4-NEXT: testb $1, %al
+; SSE4-NEXT: testb $-128, %al
; SSE4-NEXT: je .LBB17_16
-; SSE4-NEXT: # %bb.15: # %cond.store13
+; SSE4-NEXT: .LBB17_15: # %cond.store13
; SSE4-NEXT: pextrb $14, %xmm0, 7(%rdi)
-; SSE4-NEXT: .LBB17_16: # %else14
; SSE4-NEXT: retq
;
; AVX-LABEL: truncstore_v8i16_v8i8:
; AVX: # %bb.0:
; AVX-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm2
-; AVX-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX-NEXT: vpxor %xmm3, %xmm2, %xmm2
; AVX-NEXT: vpminuw {{.*}}(%rip), %xmm0, %xmm0
-; AVX-NEXT: vpextrb $0, %xmm2, %eax
+; AVX-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm1
+; AVX-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
+; AVX-NEXT: vpxor %xmm2, %xmm1, %xmm1
+; AVX-NEXT: vpacksswb %xmm0, %xmm1, %xmm1
+; AVX-NEXT: vpmovmskb %xmm1, %eax
; AVX-NEXT: testb $1, %al
-; AVX-NEXT: je .LBB17_2
-; AVX-NEXT: # %bb.1: # %cond.store
+; AVX-NEXT: jne .LBB17_1
+; AVX-NEXT: # %bb.2: # %else
+; AVX-NEXT: testb $2, %al
+; AVX-NEXT: jne .LBB17_3
+; AVX-NEXT: .LBB17_4: # %else2
+; AVX-NEXT: testb $4, %al
+; AVX-NEXT: jne .LBB17_5
+; AVX-NEXT: .LBB17_6: # %else4
+; AVX-NEXT: testb $8, %al
+; AVX-NEXT: jne .LBB17_7
+; AVX-NEXT: .LBB17_8: # %else6
+; AVX-NEXT: testb $16, %al
+; AVX-NEXT: jne .LBB17_9
+; AVX-NEXT: .LBB17_10: # %else8
+; AVX-NEXT: testb $32, %al
+; AVX-NEXT: jne .LBB17_11
+; AVX-NEXT: .LBB17_12: # %else10
+; AVX-NEXT: testb $64, %al
+; AVX-NEXT: jne .LBB17_13
+; AVX-NEXT: .LBB17_14: # %else12
+; AVX-NEXT: testb $-128, %al
+; AVX-NEXT: jne .LBB17_15
+; AVX-NEXT: .LBB17_16: # %else14
+; AVX-NEXT: retq
+; AVX-NEXT: .LBB17_1: # %cond.store
; AVX-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX-NEXT: .LBB17_2: # %else
-; AVX-NEXT: vpextrb $2, %xmm2, %eax
-; AVX-NEXT: testb $1, %al
+; AVX-NEXT: testb $2, %al
; AVX-NEXT: je .LBB17_4
-; AVX-NEXT: # %bb.3: # %cond.store1
+; AVX-NEXT: .LBB17_3: # %cond.store1
; AVX-NEXT: vpextrb $2, %xmm0, 1(%rdi)
-; AVX-NEXT: .LBB17_4: # %else2
-; AVX-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm2
-; AVX-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX-NEXT: vpextrb $4, %xmm2, %eax
-; AVX-NEXT: testb $1, %al
+; AVX-NEXT: testb $4, %al
; AVX-NEXT: je .LBB17_6
-; AVX-NEXT: # %bb.5: # %cond.store3
+; AVX-NEXT: .LBB17_5: # %cond.store3
; AVX-NEXT: vpextrb $4, %xmm0, 2(%rdi)
-; AVX-NEXT: .LBB17_6: # %else4
-; AVX-NEXT: vpextrb $6, %xmm2, %eax
-; AVX-NEXT: testb $1, %al
+; AVX-NEXT: testb $8, %al
; AVX-NEXT: je .LBB17_8
-; AVX-NEXT: # %bb.7: # %cond.store5
+; AVX-NEXT: .LBB17_7: # %cond.store5
; AVX-NEXT: vpextrb $6, %xmm0, 3(%rdi)
-; AVX-NEXT: .LBB17_8: # %else6
-; AVX-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm2
-; AVX-NEXT: vpcmpeqd %xmm3, %xmm3, %xmm3
-; AVX-NEXT: vpxor %xmm3, %xmm2, %xmm2
-; AVX-NEXT: vpextrb $8, %xmm2, %eax
-; AVX-NEXT: testb $1, %al
+; AVX-NEXT: testb $16, %al
; AVX-NEXT: je .LBB17_10
-; AVX-NEXT: # %bb.9: # %cond.store7
+; AVX-NEXT: .LBB17_9: # %cond.store7
; AVX-NEXT: vpextrb $8, %xmm0, 4(%rdi)
-; AVX-NEXT: .LBB17_10: # %else8
-; AVX-NEXT: vpextrb $10, %xmm2, %eax
-; AVX-NEXT: testb $1, %al
+; AVX-NEXT: testb $32, %al
; AVX-NEXT: je .LBB17_12
-; AVX-NEXT: # %bb.11: # %cond.store9
+; AVX-NEXT: .LBB17_11: # %cond.store9
; AVX-NEXT: vpextrb $10, %xmm0, 5(%rdi)
-; AVX-NEXT: .LBB17_12: # %else10
-; AVX-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX-NEXT: vpxor %xmm2, %xmm1, %xmm1
-; AVX-NEXT: vpextrb $12, %xmm1, %eax
-; AVX-NEXT: testb $1, %al
+; AVX-NEXT: testb $64, %al
; AVX-NEXT: je .LBB17_14
-; AVX-NEXT: # %bb.13: # %cond.store11
+; AVX-NEXT: .LBB17_13: # %cond.store11
; AVX-NEXT: vpextrb $12, %xmm0, 6(%rdi)
-; AVX-NEXT: .LBB17_14: # %else12
-; AVX-NEXT: vpextrb $14, %xmm1, %eax
-; AVX-NEXT: testb $1, %al
+; AVX-NEXT: testb $-128, %al
; AVX-NEXT: je .LBB17_16
-; AVX-NEXT: # %bb.15: # %cond.store13
+; AVX-NEXT: .LBB17_15: # %cond.store13
; AVX-NEXT: vpextrb $14, %xmm0, 7(%rdi)
-; AVX-NEXT: .LBB17_16: # %else14
; AVX-NEXT: retq
;
; AVX512F-LABEL: truncstore_v8i16_v8i8:
; AVX512F: # %bb.0:
; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxwq %xmm3, %zmm3
-; AVX512F-NEXT: vptestmq %zmm3, %zmm3, %k0
+; AVX512F-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm1
+; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm1
+; AVX512F-NEXT: vpmovsxwq %xmm1, %zmm1
+; AVX512F-NEXT: vptestmq %zmm1, %zmm1, %k0
; AVX512F-NEXT: vpminuw {{.*}}(%rip), %xmm0, %xmm0
; AVX512F-NEXT: kmovw %k0, %eax
; AVX512F-NEXT: testb $1, %al
-; AVX512F-NEXT: je .LBB17_2
-; AVX512F-NEXT: # %bb.1: # %cond.store
+; AVX512F-NEXT: jne .LBB17_1
+; AVX512F-NEXT: # %bb.2: # %else
+; AVX512F-NEXT: testb $2, %al
+; AVX512F-NEXT: jne .LBB17_3
+; AVX512F-NEXT: .LBB17_4: # %else2
+; AVX512F-NEXT: testb $4, %al
+; AVX512F-NEXT: jne .LBB17_5
+; AVX512F-NEXT: .LBB17_6: # %else4
+; AVX512F-NEXT: testb $8, %al
+; AVX512F-NEXT: jne .LBB17_7
+; AVX512F-NEXT: .LBB17_8: # %else6
+; AVX512F-NEXT: testb $16, %al
+; AVX512F-NEXT: jne .LBB17_9
+; AVX512F-NEXT: .LBB17_10: # %else8
+; AVX512F-NEXT: testb $32, %al
+; AVX512F-NEXT: jne .LBB17_11
+; AVX512F-NEXT: .LBB17_12: # %else10
+; AVX512F-NEXT: testb $64, %al
+; AVX512F-NEXT: jne .LBB17_13
+; AVX512F-NEXT: .LBB17_14: # %else12
+; AVX512F-NEXT: testb $-128, %al
+; AVX512F-NEXT: jne .LBB17_15
+; AVX512F-NEXT: .LBB17_16: # %else14
+; AVX512F-NEXT: vzeroupper
+; AVX512F-NEXT: retq
+; AVX512F-NEXT: .LBB17_1: # %cond.store
; AVX512F-NEXT: vpextrb $0, %xmm0, (%rdi)
-; AVX512F-NEXT: .LBB17_2: # %else
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxwq %xmm2, %zmm2
-; AVX512F-NEXT: vptestmq %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $1, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $2, %al
; AVX512F-NEXT: je .LBB17_4
-; AVX512F-NEXT: # %bb.3: # %cond.store1
+; AVX512F-NEXT: .LBB17_3: # %cond.store1
; AVX512F-NEXT: vpextrb $2, %xmm0, 1(%rdi)
-; AVX512F-NEXT: .LBB17_4: # %else2
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxwq %xmm3, %zmm3
-; AVX512F-NEXT: vptestmq %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $2, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $4, %al
; AVX512F-NEXT: je .LBB17_6
-; AVX512F-NEXT: # %bb.5: # %cond.store3
+; AVX512F-NEXT: .LBB17_5: # %cond.store3
; AVX512F-NEXT: vpextrb $4, %xmm0, 2(%rdi)
-; AVX512F-NEXT: .LBB17_6: # %else4
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxwq %xmm2, %zmm2
-; AVX512F-NEXT: vptestmq %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $3, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $8, %al
; AVX512F-NEXT: je .LBB17_8
-; AVX512F-NEXT: # %bb.7: # %cond.store5
+; AVX512F-NEXT: .LBB17_7: # %cond.store5
; AVX512F-NEXT: vpextrb $6, %xmm0, 3(%rdi)
-; AVX512F-NEXT: .LBB17_8: # %else6
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm2
-; AVX512F-NEXT: vmovdqa64 %zmm2, %zmm3
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm3
-; AVX512F-NEXT: vpmovsxwq %xmm3, %zmm3
-; AVX512F-NEXT: vptestmq %zmm3, %zmm3, %k0
-; AVX512F-NEXT: kshiftrw $4, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $16, %al
; AVX512F-NEXT: je .LBB17_10
-; AVX512F-NEXT: # %bb.9: # %cond.store7
+; AVX512F-NEXT: .LBB17_9: # %cond.store7
; AVX512F-NEXT: vpextrb $8, %xmm0, 4(%rdi)
-; AVX512F-NEXT: .LBB17_10: # %else8
-; AVX512F-NEXT: vpternlogq $15, %zmm2, %zmm2, %zmm2
-; AVX512F-NEXT: vpmovsxwq %xmm2, %zmm2
-; AVX512F-NEXT: vptestmq %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $5, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $32, %al
; AVX512F-NEXT: je .LBB17_12
-; AVX512F-NEXT: # %bb.11: # %cond.store9
+; AVX512F-NEXT: .LBB17_11: # %cond.store9
; AVX512F-NEXT: vpextrb $10, %xmm0, 5(%rdi)
-; AVX512F-NEXT: .LBB17_12: # %else10
-; AVX512F-NEXT: vpxor %xmm2, %xmm2, %xmm2
-; AVX512F-NEXT: vpcmpeqw %xmm2, %xmm1, %xmm1
-; AVX512F-NEXT: vmovdqa64 %zmm1, %zmm2
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm2
-; AVX512F-NEXT: vpmovsxwq %xmm2, %zmm2
-; AVX512F-NEXT: vptestmq %zmm2, %zmm2, %k0
-; AVX512F-NEXT: kshiftrw $6, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $64, %al
; AVX512F-NEXT: je .LBB17_14
-; AVX512F-NEXT: # %bb.13: # %cond.store11
+; AVX512F-NEXT: .LBB17_13: # %cond.store11
; AVX512F-NEXT: vpextrb $12, %xmm0, 6(%rdi)
-; AVX512F-NEXT: .LBB17_14: # %else12
-; AVX512F-NEXT: vpternlogq $15, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT: vpmovsxwq %xmm1, %zmm1
-; AVX512F-NEXT: vptestmq %zmm1, %zmm1, %k0
-; AVX512F-NEXT: kshiftrw $7, %k0, %k0
-; AVX512F-NEXT: kmovw %k0, %eax
-; AVX512F-NEXT: testb $1, %al
+; AVX512F-NEXT: testb $-128, %al
; AVX512F-NEXT: je .LBB17_16
-; AVX512F-NEXT: # %bb.15: # %cond.store13
+; AVX512F-NEXT: .LBB17_15: # %cond.store13
; AVX512F-NEXT: vpextrb $14, %xmm0, 7(%rdi)
-; AVX512F-NEXT: .LBB17_16: # %else14
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
Modified: llvm/trunk/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-gather.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-gather.ll?rev=367489&r1=367488&r2=367489&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-gather.ll (original)
+++ llvm/trunk/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-gather.ll Wed Jul 31 15:58:15 2019
@@ -3,8 +3,10 @@
define <2 x i64> @scalarize_v2i64(<2 x i64*> %p, <2 x i1> %mask, <2 x i64> %passthru) {
; CHECK-LABEL: @scalarize_v2i64(
-; CHECK-NEXT: [[MASK0:%.*]] = extractelement <2 x i1> [[MASK:%.*]], i64 0
-; CHECK-NEXT: br i1 [[MASK0]], label [[COND_LOAD:%.*]], label [[ELSE:%.*]]
+; CHECK-NEXT: [[SCALAR_MASK:%.*]] = bitcast <2 x i1> [[MASK:%.*]] to i2
+; CHECK-NEXT: [[TMP1:%.*]] = and i2 [[SCALAR_MASK]], 1
+; CHECK-NEXT: [[TMP2:%.*]] = icmp ne i2 [[TMP1]], 0
+; CHECK-NEXT: br i1 [[TMP2]], label [[COND_LOAD:%.*]], label [[ELSE:%.*]]
; CHECK: cond.load:
; CHECK-NEXT: [[PTR0:%.*]] = extractelement <2 x i64*> [[P:%.*]], i64 0
; CHECK-NEXT: [[LOAD0:%.*]] = load i64, i64* [[PTR0]], align 8
@@ -12,8 +14,9 @@ define <2 x i64> @scalarize_v2i64(<2 x i
; CHECK-NEXT: br label [[ELSE]]
; CHECK: else:
; CHECK-NEXT: [[RES_PHI_ELSE:%.*]] = phi <2 x i64> [ [[RES0]], [[COND_LOAD]] ], [ [[PASSTHRU]], [[TMP0:%.*]] ]
-; CHECK-NEXT: [[MASK1:%.*]] = extractelement <2 x i1> [[MASK]], i64 1
-; CHECK-NEXT: br i1 [[MASK1]], label [[COND_LOAD1:%.*]], label [[ELSE2:%.*]]
+; CHECK-NEXT: [[TMP3:%.*]] = and i2 [[SCALAR_MASK]], -2
+; CHECK-NEXT: [[TMP4:%.*]] = icmp ne i2 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[TMP4]], label [[COND_LOAD1:%.*]], label [[ELSE2:%.*]]
; CHECK: cond.load1:
; CHECK-NEXT: [[PTR1:%.*]] = extractelement <2 x i64*> [[P]], i64 1
; CHECK-NEXT: [[LOAD1:%.*]] = load i64, i64* [[PTR1]], align 8
Modified: llvm/trunk/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-load.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-load.ll?rev=367489&r1=367488&r2=367489&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-load.ll (original)
+++ llvm/trunk/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-load.ll Wed Jul 31 15:58:15 2019
@@ -4,24 +4,27 @@
define <2 x i64> @scalarize_v2i64(<2 x i64>* %p, <2 x i1> %mask, <2 x i64> %passthru) {
; CHECK-LABEL: @scalarize_v2i64(
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <2 x i64>* [[P:%.*]] to i64*
-; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x i1> [[MASK:%.*]], i64 0
-; CHECK-NEXT: br i1 [[TMP2]], label [[COND_LOAD:%.*]], label [[ELSE:%.*]]
+; CHECK-NEXT: [[SCALAR_MASK:%.*]] = bitcast <2 x i1> [[MASK:%.*]] to i2
+; CHECK-NEXT: [[TMP2:%.*]] = and i2 [[SCALAR_MASK]], 1
+; CHECK-NEXT: [[TMP3:%.*]] = icmp ne i2 [[TMP2]], 0
+; CHECK-NEXT: br i1 [[TMP3]], label [[COND_LOAD:%.*]], label [[ELSE:%.*]]
; CHECK: cond.load:
-; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i64, i64* [[TMP1]], i32 0
-; CHECK-NEXT: [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8
-; CHECK-NEXT: [[TMP5:%.*]] = insertelement <2 x i64> [[PASSTHRU:%.*]], i64 [[TMP4]], i64 0
+; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i64, i64* [[TMP1]], i32 0
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, i64* [[TMP4]], align 8
+; CHECK-NEXT: [[TMP6:%.*]] = insertelement <2 x i64> [[PASSTHRU:%.*]], i64 [[TMP5]], i64 0
; CHECK-NEXT: br label [[ELSE]]
; CHECK: else:
-; CHECK-NEXT: [[RES_PHI_ELSE:%.*]] = phi <2 x i64> [ [[TMP5]], [[COND_LOAD]] ], [ [[PASSTHRU]], [[TMP0:%.*]] ]
-; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x i1> [[MASK]], i64 1
-; CHECK-NEXT: br i1 [[TMP6]], label [[COND_LOAD1:%.*]], label [[ELSE2:%.*]]
+; CHECK-NEXT: [[RES_PHI_ELSE:%.*]] = phi <2 x i64> [ [[TMP6]], [[COND_LOAD]] ], [ [[PASSTHRU]], [[TMP0:%.*]] ]
+; CHECK-NEXT: [[TMP7:%.*]] = and i2 [[SCALAR_MASK]], -2
+; CHECK-NEXT: [[TMP8:%.*]] = icmp ne i2 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[TMP8]], label [[COND_LOAD1:%.*]], label [[ELSE2:%.*]]
; CHECK: cond.load1:
-; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i64, i64* [[TMP1]], i32 1
-; CHECK-NEXT: [[TMP8:%.*]] = load i64, i64* [[TMP7]], align 8
-; CHECK-NEXT: [[TMP9:%.*]] = insertelement <2 x i64> [[RES_PHI_ELSE]], i64 [[TMP8]], i64 1
+; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i64, i64* [[TMP1]], i32 1
+; CHECK-NEXT: [[TMP10:%.*]] = load i64, i64* [[TMP9]], align 8
+; CHECK-NEXT: [[TMP11:%.*]] = insertelement <2 x i64> [[RES_PHI_ELSE]], i64 [[TMP10]], i64 1
; CHECK-NEXT: br label [[ELSE2]]
; CHECK: else2:
-; CHECK-NEXT: [[RES_PHI_ELSE3:%.*]] = phi <2 x i64> [ [[TMP9]], [[COND_LOAD1]] ], [ [[RES_PHI_ELSE]], [[ELSE]] ]
+; CHECK-NEXT: [[RES_PHI_ELSE3:%.*]] = phi <2 x i64> [ [[TMP11]], [[COND_LOAD1]] ], [ [[RES_PHI_ELSE]], [[ELSE]] ]
; CHECK-NEXT: ret <2 x i64> [[RES_PHI_ELSE3]]
;
%ret = call <2 x i64> @llvm.masked.load.v2i64.p0v2i64(<2 x i64>* %p, i32 128, <2 x i1> %mask, <2 x i64> %passthru)
@@ -62,24 +65,27 @@ define <2 x i64> @scalarize_v2i64_const_
define <2 x i24> @scalarize_v2i24(<2 x i24>* %p, <2 x i1> %mask, <2 x i24> %passthru) {
; CHECK-LABEL: @scalarize_v2i24(
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <2 x i24>* [[P:%.*]] to i24*
-; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x i1> [[MASK:%.*]], i64 0
-; CHECK-NEXT: br i1 [[TMP2]], label [[COND_LOAD:%.*]], label [[ELSE:%.*]]
+; CHECK-NEXT: [[SCALAR_MASK:%.*]] = bitcast <2 x i1> [[MASK:%.*]] to i2
+; CHECK-NEXT: [[TMP2:%.*]] = and i2 [[SCALAR_MASK]], 1
+; CHECK-NEXT: [[TMP3:%.*]] = icmp ne i2 [[TMP2]], 0
+; CHECK-NEXT: br i1 [[TMP3]], label [[COND_LOAD:%.*]], label [[ELSE:%.*]]
; CHECK: cond.load:
-; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i24, i24* [[TMP1]], i32 0
-; CHECK-NEXT: [[TMP4:%.*]] = load i24, i24* [[TMP3]], align 1
-; CHECK-NEXT: [[TMP5:%.*]] = insertelement <2 x i24> [[PASSTHRU:%.*]], i24 [[TMP4]], i64 0
+; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i24, i24* [[TMP1]], i32 0
+; CHECK-NEXT: [[TMP5:%.*]] = load i24, i24* [[TMP4]], align 1
+; CHECK-NEXT: [[TMP6:%.*]] = insertelement <2 x i24> [[PASSTHRU:%.*]], i24 [[TMP5]], i64 0
; CHECK-NEXT: br label [[ELSE]]
; CHECK: else:
-; CHECK-NEXT: [[RES_PHI_ELSE:%.*]] = phi <2 x i24> [ [[TMP5]], [[COND_LOAD]] ], [ [[PASSTHRU]], [[TMP0:%.*]] ]
-; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x i1> [[MASK]], i64 1
-; CHECK-NEXT: br i1 [[TMP6]], label [[COND_LOAD1:%.*]], label [[ELSE2:%.*]]
+; CHECK-NEXT: [[RES_PHI_ELSE:%.*]] = phi <2 x i24> [ [[TMP6]], [[COND_LOAD]] ], [ [[PASSTHRU]], [[TMP0:%.*]] ]
+; CHECK-NEXT: [[TMP7:%.*]] = and i2 [[SCALAR_MASK]], -2
+; CHECK-NEXT: [[TMP8:%.*]] = icmp ne i2 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[TMP8]], label [[COND_LOAD1:%.*]], label [[ELSE2:%.*]]
; CHECK: cond.load1:
-; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i24, i24* [[TMP1]], i32 1
-; CHECK-NEXT: [[TMP8:%.*]] = load i24, i24* [[TMP7]], align 1
-; CHECK-NEXT: [[TMP9:%.*]] = insertelement <2 x i24> [[RES_PHI_ELSE]], i24 [[TMP8]], i64 1
+; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i24, i24* [[TMP1]], i32 1
+; CHECK-NEXT: [[TMP10:%.*]] = load i24, i24* [[TMP9]], align 1
+; CHECK-NEXT: [[TMP11:%.*]] = insertelement <2 x i24> [[RES_PHI_ELSE]], i24 [[TMP10]], i64 1
; CHECK-NEXT: br label [[ELSE2]]
; CHECK: else2:
-; CHECK-NEXT: [[RES_PHI_ELSE3:%.*]] = phi <2 x i24> [ [[TMP9]], [[COND_LOAD1]] ], [ [[RES_PHI_ELSE]], [[ELSE]] ]
+; CHECK-NEXT: [[RES_PHI_ELSE3:%.*]] = phi <2 x i24> [ [[TMP11]], [[COND_LOAD1]] ], [ [[RES_PHI_ELSE]], [[ELSE]] ]
; CHECK-NEXT: ret <2 x i24> [[RES_PHI_ELSE3]]
;
%ret = call <2 x i24> @llvm.masked.load.v2i24.p0v2i24(<2 x i24>* %p, i32 8, <2 x i1> %mask, <2 x i24> %passthru)
@@ -90,24 +96,27 @@ define <2 x i24> @scalarize_v2i24(<2 x i
define <2 x i48> @scalarize_v2i48(<2 x i48>* %p, <2 x i1> %mask, <2 x i48> %passthru) {
; CHECK-LABEL: @scalarize_v2i48(
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <2 x i48>* [[P:%.*]] to i48*
-; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x i1> [[MASK:%.*]], i64 0
-; CHECK-NEXT: br i1 [[TMP2]], label [[COND_LOAD:%.*]], label [[ELSE:%.*]]
+; CHECK-NEXT: [[SCALAR_MASK:%.*]] = bitcast <2 x i1> [[MASK:%.*]] to i2
+; CHECK-NEXT: [[TMP2:%.*]] = and i2 [[SCALAR_MASK]], 1
+; CHECK-NEXT: [[TMP3:%.*]] = icmp ne i2 [[TMP2]], 0
+; CHECK-NEXT: br i1 [[TMP3]], label [[COND_LOAD:%.*]], label [[ELSE:%.*]]
; CHECK: cond.load:
-; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i48, i48* [[TMP1]], i32 0
-; CHECK-NEXT: [[TMP4:%.*]] = load i48, i48* [[TMP3]], align 2
-; CHECK-NEXT: [[TMP5:%.*]] = insertelement <2 x i48> [[PASSTHRU:%.*]], i48 [[TMP4]], i64 0
+; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i48, i48* [[TMP1]], i32 0
+; CHECK-NEXT: [[TMP5:%.*]] = load i48, i48* [[TMP4]], align 2
+; CHECK-NEXT: [[TMP6:%.*]] = insertelement <2 x i48> [[PASSTHRU:%.*]], i48 [[TMP5]], i64 0
; CHECK-NEXT: br label [[ELSE]]
; CHECK: else:
-; CHECK-NEXT: [[RES_PHI_ELSE:%.*]] = phi <2 x i48> [ [[TMP5]], [[COND_LOAD]] ], [ [[PASSTHRU]], [[TMP0:%.*]] ]
-; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x i1> [[MASK]], i64 1
-; CHECK-NEXT: br i1 [[TMP6]], label [[COND_LOAD1:%.*]], label [[ELSE2:%.*]]
+; CHECK-NEXT: [[RES_PHI_ELSE:%.*]] = phi <2 x i48> [ [[TMP6]], [[COND_LOAD]] ], [ [[PASSTHRU]], [[TMP0:%.*]] ]
+; CHECK-NEXT: [[TMP7:%.*]] = and i2 [[SCALAR_MASK]], -2
+; CHECK-NEXT: [[TMP8:%.*]] = icmp ne i2 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[TMP8]], label [[COND_LOAD1:%.*]], label [[ELSE2:%.*]]
; CHECK: cond.load1:
-; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i48, i48* [[TMP1]], i32 1
-; CHECK-NEXT: [[TMP8:%.*]] = load i48, i48* [[TMP7]], align 2
-; CHECK-NEXT: [[TMP9:%.*]] = insertelement <2 x i48> [[RES_PHI_ELSE]], i48 [[TMP8]], i64 1
+; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i48, i48* [[TMP1]], i32 1
+; CHECK-NEXT: [[TMP10:%.*]] = load i48, i48* [[TMP9]], align 2
+; CHECK-NEXT: [[TMP11:%.*]] = insertelement <2 x i48> [[RES_PHI_ELSE]], i48 [[TMP10]], i64 1
; CHECK-NEXT: br label [[ELSE2]]
; CHECK: else2:
-; CHECK-NEXT: [[RES_PHI_ELSE3:%.*]] = phi <2 x i48> [ [[TMP9]], [[COND_LOAD1]] ], [ [[RES_PHI_ELSE]], [[ELSE]] ]
+; CHECK-NEXT: [[RES_PHI_ELSE3:%.*]] = phi <2 x i48> [ [[TMP11]], [[COND_LOAD1]] ], [ [[RES_PHI_ELSE]], [[ELSE]] ]
; CHECK-NEXT: ret <2 x i48> [[RES_PHI_ELSE3]]
;
%ret = call <2 x i48> @llvm.masked.load.v2i48.p0v2i48(<2 x i48>* %p, i32 16, <2 x i1> %mask, <2 x i48> %passthru)
Modified: llvm/trunk/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-store.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-store.ll?rev=367489&r1=367488&r2=367489&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-store.ll (original)
+++ llvm/trunk/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-store.ll Wed Jul 31 15:58:15 2019
@@ -4,20 +4,23 @@
define void @scalarize_v2i64(<2 x i64>* %p, <2 x i1> %mask, <2 x i64> %data) {
; CHECK-LABEL: @scalarize_v2i64(
; CHECK-NEXT: [[TMP1:%.*]] = bitcast <2 x i64>* [[P:%.*]] to i64*
-; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x i1> [[MASK:%.*]], i64 0
-; CHECK-NEXT: br i1 [[TMP2]], label [[COND_STORE:%.*]], label [[ELSE:%.*]]
+; CHECK-NEXT: [[SCALAR_MASK:%.*]] = bitcast <2 x i1> [[MASK:%.*]] to i2
+; CHECK-NEXT: [[TMP2:%.*]] = and i2 [[SCALAR_MASK]], 1
+; CHECK-NEXT: [[TMP3:%.*]] = icmp ne i2 [[TMP2]], 0
+; CHECK-NEXT: br i1 [[TMP3]], label [[COND_STORE:%.*]], label [[ELSE:%.*]]
; CHECK: cond.store:
-; CHECK-NEXT: [[TMP3:%.*]] = extractelement <2 x i64> [[DATA:%.*]], i64 0
-; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i64, i64* [[TMP1]], i32 0
-; CHECK-NEXT: store i64 [[TMP3]], i64* [[TMP4]], align 8
+; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x i64> [[DATA:%.*]], i64 0
+; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i64, i64* [[TMP1]], i32 0
+; CHECK-NEXT: store i64 [[TMP4]], i64* [[TMP5]], align 8
; CHECK-NEXT: br label [[ELSE]]
; CHECK: else:
-; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x i1> [[MASK]], i64 1
-; CHECK-NEXT: br i1 [[TMP5]], label [[COND_STORE1:%.*]], label [[ELSE2:%.*]]
+; CHECK-NEXT: [[TMP6:%.*]] = and i2 [[SCALAR_MASK]], -2
+; CHECK-NEXT: [[TMP7:%.*]] = icmp ne i2 [[TMP6]], 0
+; CHECK-NEXT: br i1 [[TMP7]], label [[COND_STORE1:%.*]], label [[ELSE2:%.*]]
; CHECK: cond.store1:
-; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x i64> [[DATA]], i64 1
-; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i64, i64* [[TMP1]], i32 1
-; CHECK-NEXT: store i64 [[TMP6]], i64* [[TMP7]], align 8
+; CHECK-NEXT: [[TMP8:%.*]] = extractelement <2 x i64> [[DATA]], i64 1
+; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i64, i64* [[TMP1]], i32 1
+; CHECK-NEXT: store i64 [[TMP8]], i64* [[TMP9]], align 8
; CHECK-NEXT: br label [[ELSE2]]
; CHECK: else2:
; CHECK-NEXT: ret void
More information about the llvm-commits
mailing list