diff options
Diffstat (limited to 'llvm/test/CodeGen/X86/recip-fastmath2.ll')
-rw-r--r-- | llvm/test/CodeGen/X86/recip-fastmath2.ll | 96 |
1 files changed, 48 insertions, 48 deletions
diff --git a/llvm/test/CodeGen/X86/recip-fastmath2.ll b/llvm/test/CodeGen/X86/recip-fastmath2.ll index 7cc09dc96f2..fdd441581dc 100644 --- a/llvm/test/CodeGen/X86/recip-fastmath2.ll +++ b/llvm/test/CodeGen/X86/recip-fastmath2.ll @@ -398,7 +398,7 @@ define <4 x float> @v4f32_one_step2(<4 x float> %x) #1 { ; SSE: # %bb.0: ; SSE-NEXT: rcpps %xmm0, %xmm2 ; SSE-NEXT: mulps %xmm2, %xmm0 -; SSE-NEXT: movaps {{.*#+}} xmm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm1 = [1,1,1,1] ; SSE-NEXT: subps %xmm0, %xmm1 ; SSE-NEXT: mulps %xmm2, %xmm1 ; SSE-NEXT: addps %xmm2, %xmm1 @@ -410,7 +410,7 @@ define <4 x float> @v4f32_one_step2(<4 x float> %x) #1 { ; AVX-RECIP: # %bb.0: ; AVX-RECIP-NEXT: vrcpps %xmm0, %xmm1 ; AVX-RECIP-NEXT: vmulps %xmm1, %xmm0, %xmm0 -; AVX-RECIP-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; AVX-RECIP-NEXT: vmovaps {{.*#+}} xmm2 = [1,1,1,1] ; AVX-RECIP-NEXT: vsubps %xmm0, %xmm2, %xmm0 ; AVX-RECIP-NEXT: vmulps %xmm0, %xmm1, %xmm0 ; AVX-RECIP-NEXT: vaddps %xmm0, %xmm1, %xmm0 @@ -427,7 +427,7 @@ define <4 x float> @v4f32_one_step2(<4 x float> %x) #1 { ; ; BTVER2-LABEL: v4f32_one_step2: ; BTVER2: # %bb.0: -; BTVER2-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00] +; BTVER2-NEXT: vmovaps {{.*#+}} xmm2 = [1,1,1,1] sched: [5:1.00] ; BTVER2-NEXT: vrcpps %xmm0, %xmm1 # sched: [2:1.00] ; BTVER2-NEXT: vmulps %xmm1, %xmm0, %xmm0 # sched: [2:1.00] ; BTVER2-NEXT: vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00] @@ -440,7 +440,7 @@ define <4 x float> @v4f32_one_step2(<4 x float> %x) #1 { ; SANDY: # %bb.0: ; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] ; SANDY-NEXT: vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} xmm2 = [1,1,1,1] sched: [6:0.50] ; SANDY-NEXT: vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] @@ -492,11 +492,11 @@ define <4 x float> @v4f32_one_step_2_divs(<4 x float> %x) #1 { ; SSE: # %bb.0: ; SSE-NEXT: rcpps %xmm0, %xmm1 ; SSE-NEXT: mulps %xmm1, %xmm0 -; SSE-NEXT: movaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm2 = [1,1,1,1] ; SSE-NEXT: subps %xmm0, %xmm2 ; SSE-NEXT: mulps %xmm1, %xmm2 ; SSE-NEXT: addps %xmm1, %xmm2 -; SSE-NEXT: movaps {{.*#+}} xmm0 = [1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm0 = [1,2,3,4] ; SSE-NEXT: mulps %xmm2, %xmm0 ; SSE-NEXT: mulps %xmm2, %xmm0 ; SSE-NEXT: retq @@ -505,7 +505,7 @@ define <4 x float> @v4f32_one_step_2_divs(<4 x float> %x) #1 { ; AVX-RECIP: # %bb.0: ; AVX-RECIP-NEXT: vrcpps %xmm0, %xmm1 ; AVX-RECIP-NEXT: vmulps %xmm1, %xmm0, %xmm0 -; AVX-RECIP-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; AVX-RECIP-NEXT: vmovaps {{.*#+}} xmm2 = [1,1,1,1] ; AVX-RECIP-NEXT: vsubps %xmm0, %xmm2, %xmm0 ; AVX-RECIP-NEXT: vmulps %xmm0, %xmm1, %xmm0 ; AVX-RECIP-NEXT: vaddps %xmm0, %xmm1, %xmm0 @@ -524,7 +524,7 @@ define <4 x float> @v4f32_one_step_2_divs(<4 x float> %x) #1 { ; ; BTVER2-LABEL: v4f32_one_step_2_divs: ; BTVER2: # %bb.0: -; BTVER2-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00] +; BTVER2-NEXT: vmovaps {{.*#+}} xmm2 = [1,1,1,1] sched: [5:1.00] ; BTVER2-NEXT: vrcpps %xmm0, %xmm1 # sched: [2:1.00] ; BTVER2-NEXT: vmulps %xmm1, %xmm0, %xmm0 # sched: [2:1.00] ; BTVER2-NEXT: vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00] @@ -538,7 +538,7 @@ define <4 x float> @v4f32_one_step_2_divs(<4 x float> %x) #1 { ; SANDY: # %bb.0: ; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] ; SANDY-NEXT: vmulps %xmm1, %xmm0, %xmm0 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} xmm2 = [1,1,1,1] sched: [6:0.50] ; SANDY-NEXT: vsubps %xmm0, %xmm2, %xmm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %xmm0, %xmm1, %xmm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %xmm0, %xmm1, %xmm0 # sched: [3:1.00] @@ -597,7 +597,7 @@ define <4 x float> @v4f32_two_step2(<4 x float> %x) #2 { ; SSE-NEXT: rcpps %xmm0, %xmm2 ; SSE-NEXT: movaps %xmm0, %xmm3 ; SSE-NEXT: mulps %xmm2, %xmm3 -; SSE-NEXT: movaps {{.*#+}} xmm1 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm1 = [1,1,1,1] ; SSE-NEXT: movaps %xmm1, %xmm4 ; SSE-NEXT: subps %xmm3, %xmm4 ; SSE-NEXT: mulps %xmm2, %xmm4 @@ -614,7 +614,7 @@ define <4 x float> @v4f32_two_step2(<4 x float> %x) #2 { ; AVX-RECIP: # %bb.0: ; AVX-RECIP-NEXT: vrcpps %xmm0, %xmm1 ; AVX-RECIP-NEXT: vmulps %xmm1, %xmm0, %xmm2 -; AVX-RECIP-NEXT: vmovaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; AVX-RECIP-NEXT: vmovaps {{.*#+}} xmm3 = [1,1,1,1] ; AVX-RECIP-NEXT: vsubps %xmm2, %xmm3, %xmm2 ; AVX-RECIP-NEXT: vmulps %xmm2, %xmm1, %xmm2 ; AVX-RECIP-NEXT: vaddps %xmm2, %xmm1, %xmm1 @@ -628,7 +628,7 @@ define <4 x float> @v4f32_two_step2(<4 x float> %x) #2 { ; FMA-RECIP-LABEL: v4f32_two_step2: ; FMA-RECIP: # %bb.0: ; FMA-RECIP-NEXT: vrcpps %xmm0, %xmm1 -; FMA-RECIP-NEXT: vmovaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; FMA-RECIP-NEXT: vmovaps {{.*#+}} xmm2 = [1,1,1,1] ; FMA-RECIP-NEXT: vmovaps %xmm1, %xmm3 ; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} xmm3 = -(xmm0 * xmm3) + xmm2 ; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} xmm3 = (xmm3 * xmm1) + xmm1 @@ -639,7 +639,7 @@ define <4 x float> @v4f32_two_step2(<4 x float> %x) #2 { ; ; BTVER2-LABEL: v4f32_two_step2: ; BTVER2: # %bb.0: -; BTVER2-NEXT: vmovaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00] +; BTVER2-NEXT: vmovaps {{.*#+}} xmm3 = [1,1,1,1] sched: [5:1.00] ; BTVER2-NEXT: vrcpps %xmm0, %xmm1 # sched: [2:1.00] ; BTVER2-NEXT: vmulps %xmm1, %xmm0, %xmm2 # sched: [2:1.00] ; BTVER2-NEXT: vsubps %xmm2, %xmm3, %xmm2 # sched: [3:1.00] @@ -656,7 +656,7 @@ define <4 x float> @v4f32_two_step2(<4 x float> %x) #2 { ; SANDY: # %bb.0: ; SANDY-NEXT: vrcpps %xmm0, %xmm1 # sched: [5:1.00] ; SANDY-NEXT: vmulps %xmm1, %xmm0, %xmm2 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [6:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} xmm3 = [1,1,1,1] sched: [6:0.50] ; SANDY-NEXT: vsubps %xmm2, %xmm3, %xmm2 # sched: [3:1.00] ; SANDY-NEXT: vmulps %xmm2, %xmm1, %xmm2 # sched: [5:1.00] ; SANDY-NEXT: vaddps %xmm2, %xmm1, %xmm1 # sched: [3:1.00] @@ -726,7 +726,7 @@ define <8 x float> @v8f32_one_step2(<8 x float> %x) #1 { ; SSE: # %bb.0: ; SSE-NEXT: rcpps %xmm1, %xmm4 ; SSE-NEXT: mulps %xmm4, %xmm1 -; SSE-NEXT: movaps {{.*#+}} xmm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm2 = [1,1,1,1] ; SSE-NEXT: movaps %xmm2, %xmm3 ; SSE-NEXT: subps %xmm1, %xmm3 ; SSE-NEXT: mulps %xmm4, %xmm3 @@ -746,7 +746,7 @@ define <8 x float> @v8f32_one_step2(<8 x float> %x) #1 { ; AVX-RECIP: # %bb.0: ; AVX-RECIP-NEXT: vrcpps %ymm0, %ymm1 ; AVX-RECIP-NEXT: vmulps %ymm1, %ymm0, %ymm0 -; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] ; AVX-RECIP-NEXT: vsubps %ymm0, %ymm2, %ymm0 ; AVX-RECIP-NEXT: vmulps %ymm0, %ymm1, %ymm0 ; AVX-RECIP-NEXT: vaddps %ymm0, %ymm1, %ymm0 @@ -763,7 +763,7 @@ define <8 x float> @v8f32_one_step2(<8 x float> %x) #1 { ; ; BTVER2-LABEL: v8f32_one_step2: ; BTVER2: # %bb.0: -; BTVER2-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00] +; BTVER2-NEXT: vmovaps {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00] ; BTVER2-NEXT: vrcpps %ymm0, %ymm1 # sched: [2:2.00] ; BTVER2-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [2:2.00] ; BTVER2-NEXT: vsubps %ymm0, %ymm2, %ymm0 # sched: [3:2.00] @@ -776,7 +776,7 @@ define <8 x float> @v8f32_one_step2(<8 x float> %x) #1 { ; SANDY: # %bb.0: ; SANDY-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] ; SANDY-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [7:0.50] ; SANDY-NEXT: vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] @@ -828,7 +828,7 @@ define <8 x float> @v8f32_one_step_2_divs(<8 x float> %x) #1 { ; SSE: # %bb.0: ; SSE-NEXT: rcpps %xmm0, %xmm2 ; SSE-NEXT: mulps %xmm2, %xmm0 -; SSE-NEXT: movaps {{.*#+}} xmm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm3 = [1,1,1,1] ; SSE-NEXT: movaps %xmm3, %xmm4 ; SSE-NEXT: subps %xmm0, %xmm4 ; SSE-NEXT: mulps %xmm2, %xmm4 @@ -838,9 +838,9 @@ define <8 x float> @v8f32_one_step_2_divs(<8 x float> %x) #1 { ; SSE-NEXT: subps %xmm1, %xmm3 ; SSE-NEXT: mulps %xmm0, %xmm3 ; SSE-NEXT: addps %xmm0, %xmm3 -; SSE-NEXT: movaps {{.*#+}} xmm1 = [5.000000e+00,6.000000e+00,7.000000e+00,8.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm1 = [5,6,7,8] ; SSE-NEXT: mulps %xmm3, %xmm1 -; SSE-NEXT: movaps {{.*#+}} xmm0 = [1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm0 = [1,2,3,4] ; SSE-NEXT: mulps %xmm4, %xmm0 ; SSE-NEXT: mulps %xmm4, %xmm0 ; SSE-NEXT: mulps %xmm3, %xmm1 @@ -850,7 +850,7 @@ define <8 x float> @v8f32_one_step_2_divs(<8 x float> %x) #1 { ; AVX-RECIP: # %bb.0: ; AVX-RECIP-NEXT: vrcpps %ymm0, %ymm1 ; AVX-RECIP-NEXT: vmulps %ymm1, %ymm0, %ymm0 -; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] ; AVX-RECIP-NEXT: vsubps %ymm0, %ymm2, %ymm0 ; AVX-RECIP-NEXT: vmulps %ymm0, %ymm1, %ymm0 ; AVX-RECIP-NEXT: vaddps %ymm0, %ymm1, %ymm0 @@ -869,7 +869,7 @@ define <8 x float> @v8f32_one_step_2_divs(<8 x float> %x) #1 { ; ; BTVER2-LABEL: v8f32_one_step_2_divs: ; BTVER2: # %bb.0: -; BTVER2-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00] +; BTVER2-NEXT: vmovaps {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [5:1.00] ; BTVER2-NEXT: vrcpps %ymm0, %ymm1 # sched: [2:2.00] ; BTVER2-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [2:2.00] ; BTVER2-NEXT: vsubps %ymm0, %ymm2, %ymm0 # sched: [3:2.00] @@ -883,7 +883,7 @@ define <8 x float> @v8f32_one_step_2_divs(<8 x float> %x) #1 { ; SANDY: # %bb.0: ; SANDY-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] ; SANDY-NEXT: vmulps %ymm1, %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] sched: [7:0.50] ; SANDY-NEXT: vsubps %ymm0, %ymm2, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vmulps %ymm0, %ymm1, %ymm0 # sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm0, %ymm1, %ymm0 # sched: [3:1.00] @@ -943,7 +943,7 @@ define <8 x float> @v8f32_two_step2(<8 x float> %x) #2 { ; SSE-NEXT: rcpps %xmm1, %xmm3 ; SSE-NEXT: movaps %xmm1, %xmm4 ; SSE-NEXT: mulps %xmm3, %xmm4 -; SSE-NEXT: movaps {{.*#+}} xmm0 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm0 = [1,1,1,1] ; SSE-NEXT: movaps %xmm0, %xmm5 ; SSE-NEXT: subps %xmm4, %xmm5 ; SSE-NEXT: mulps %xmm3, %xmm5 @@ -973,7 +973,7 @@ define <8 x float> @v8f32_two_step2(<8 x float> %x) #2 { ; AVX-RECIP: # %bb.0: ; AVX-RECIP-NEXT: vrcpps %ymm0, %ymm1 ; AVX-RECIP-NEXT: vmulps %ymm1, %ymm0, %ymm2 -; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] ; AVX-RECIP-NEXT: vsubps %ymm2, %ymm3, %ymm2 ; AVX-RECIP-NEXT: vmulps %ymm2, %ymm1, %ymm2 ; AVX-RECIP-NEXT: vaddps %ymm2, %ymm1, %ymm1 @@ -987,7 +987,7 @@ define <8 x float> @v8f32_two_step2(<8 x float> %x) #2 { ; FMA-RECIP-LABEL: v8f32_two_step2: ; FMA-RECIP: # %bb.0: ; FMA-RECIP-NEXT: vrcpps %ymm0, %ymm1 -; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm2 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm2 = [1,1,1,1,1,1,1,1] ; FMA-RECIP-NEXT: vmovaps %ymm1, %ymm3 ; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm3 = -(ymm0 * ymm3) + ymm2 ; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm3 = (ymm3 * ymm1) + ymm1 @@ -998,7 +998,7 @@ define <8 x float> @v8f32_two_step2(<8 x float> %x) #2 { ; ; BTVER2-LABEL: v8f32_two_step2: ; BTVER2: # %bb.0: -; BTVER2-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00] +; BTVER2-NEXT: vmovaps {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [5:1.00] ; BTVER2-NEXT: vrcpps %ymm0, %ymm1 # sched: [2:2.00] ; BTVER2-NEXT: vmulps %ymm1, %ymm0, %ymm2 # sched: [2:2.00] ; BTVER2-NEXT: vsubps %ymm2, %ymm3, %ymm2 # sched: [3:2.00] @@ -1015,7 +1015,7 @@ define <8 x float> @v8f32_two_step2(<8 x float> %x) #2 { ; SANDY: # %bb.0: ; SANDY-NEXT: vrcpps %ymm0, %ymm1 # sched: [7:2.00] ; SANDY-NEXT: vmulps %ymm1, %ymm0, %ymm2 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [7:0.50] ; SANDY-NEXT: vsubps %ymm2, %ymm3, %ymm2 # sched: [3:1.00] ; SANDY-NEXT: vmulps %ymm2, %ymm1, %ymm2 # sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm2, %ymm1, %ymm1 # sched: [3:1.00] @@ -1198,7 +1198,7 @@ define <16 x float> @v16f32_one_step2(<16 x float> %x) #1 { ; SSE-NEXT: movaps %xmm0, %xmm6 ; SSE-NEXT: rcpps %xmm3, %xmm2 ; SSE-NEXT: mulps %xmm2, %xmm4 -; SSE-NEXT: movaps {{.*#+}} xmm0 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm0 = [1,1,1,1] ; SSE-NEXT: movaps %xmm0, %xmm3 ; SSE-NEXT: subps %xmm4, %xmm3 ; SSE-NEXT: mulps %xmm2, %xmm3 @@ -1231,7 +1231,7 @@ define <16 x float> @v16f32_one_step2(<16 x float> %x) #1 { ; AVX-RECIP: # %bb.0: ; AVX-RECIP-NEXT: vrcpps %ymm1, %ymm2 ; AVX-RECIP-NEXT: vmulps %ymm2, %ymm1, %ymm1 -; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] ; AVX-RECIP-NEXT: vsubps %ymm1, %ymm3, %ymm1 ; AVX-RECIP-NEXT: vmulps %ymm1, %ymm2, %ymm1 ; AVX-RECIP-NEXT: vaddps %ymm1, %ymm2, %ymm1 @@ -1247,7 +1247,7 @@ define <16 x float> @v16f32_one_step2(<16 x float> %x) #1 { ; FMA-RECIP-LABEL: v16f32_one_step2: ; FMA-RECIP: # %bb.0: ; FMA-RECIP-NEXT: vrcpps %ymm1, %ymm2 -; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] ; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm1 = -(ymm2 * ymm1) + ymm3 ; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm1 = (ymm1 * ymm2) + ymm2 ; FMA-RECIP-NEXT: vrcpps %ymm0, %ymm2 @@ -1259,7 +1259,7 @@ define <16 x float> @v16f32_one_step2(<16 x float> %x) #1 { ; ; BTVER2-LABEL: v16f32_one_step2: ; BTVER2: # %bb.0: -; BTVER2-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00] +; BTVER2-NEXT: vmovaps {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [5:1.00] ; BTVER2-NEXT: vrcpps %ymm1, %ymm2 # sched: [2:2.00] ; BTVER2-NEXT: vrcpps %ymm0, %ymm4 # sched: [2:2.00] ; BTVER2-NEXT: vmulps %ymm2, %ymm1, %ymm1 # sched: [2:2.00] @@ -1278,7 +1278,7 @@ define <16 x float> @v16f32_one_step2(<16 x float> %x) #1 { ; SANDY: # %bb.0: ; SANDY-NEXT: vrcpps %ymm1, %ymm2 # sched: [7:2.00] ; SANDY-NEXT: vmulps %ymm2, %ymm1, %ymm1 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [7:0.50] ; SANDY-NEXT: vsubps %ymm1, %ymm3, %ymm1 # sched: [3:1.00] ; SANDY-NEXT: vmulps %ymm1, %ymm2, %ymm1 # sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm1, %ymm2, %ymm1 # sched: [3:1.00] @@ -1345,7 +1345,7 @@ define <16 x float> @v16f32_one_step_2_divs(<16 x float> %x) #1 { ; SSE: # %bb.0: ; SSE-NEXT: rcpps %xmm0, %xmm6 ; SSE-NEXT: mulps %xmm6, %xmm0 -; SSE-NEXT: movaps {{.*#+}} xmm4 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm4 = [1,1,1,1] ; SSE-NEXT: movaps %xmm4, %xmm5 ; SSE-NEXT: subps %xmm0, %xmm5 ; SSE-NEXT: mulps %xmm6, %xmm5 @@ -1367,13 +1367,13 @@ define <16 x float> @v16f32_one_step_2_divs(<16 x float> %x) #1 { ; SSE-NEXT: subps %xmm3, %xmm4 ; SSE-NEXT: mulps %xmm0, %xmm4 ; SSE-NEXT: addps %xmm0, %xmm4 -; SSE-NEXT: movaps {{.*#+}} xmm3 = [1.300000e+01,1.400000e+01,1.500000e+01,1.600000e+01] +; SSE-NEXT: movaps {{.*#+}} xmm3 = [13,14,15,16] ; SSE-NEXT: mulps %xmm4, %xmm3 -; SSE-NEXT: movaps {{.*#+}} xmm2 = [9.000000e+00,1.000000e+01,1.100000e+01,1.200000e+01] +; SSE-NEXT: movaps {{.*#+}} xmm2 = [9,10,11,12] ; SSE-NEXT: mulps %xmm7, %xmm2 -; SSE-NEXT: movaps {{.*#+}} xmm1 = [5.000000e+00,6.000000e+00,7.000000e+00,8.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm1 = [5,6,7,8] ; SSE-NEXT: mulps %xmm6, %xmm1 -; SSE-NEXT: movaps {{.*#+}} xmm0 = [1.000000e+00,2.000000e+00,3.000000e+00,4.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm0 = [1,2,3,4] ; SSE-NEXT: mulps %xmm5, %xmm0 ; SSE-NEXT: mulps %xmm5, %xmm0 ; SSE-NEXT: mulps %xmm6, %xmm1 @@ -1385,7 +1385,7 @@ define <16 x float> @v16f32_one_step_2_divs(<16 x float> %x) #1 { ; AVX-RECIP: # %bb.0: ; AVX-RECIP-NEXT: vrcpps %ymm0, %ymm2 ; AVX-RECIP-NEXT: vmulps %ymm2, %ymm0, %ymm0 -; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] ; AVX-RECIP-NEXT: vsubps %ymm0, %ymm3, %ymm0 ; AVX-RECIP-NEXT: vmulps %ymm0, %ymm2, %ymm0 ; AVX-RECIP-NEXT: vaddps %ymm0, %ymm2, %ymm0 @@ -1403,7 +1403,7 @@ define <16 x float> @v16f32_one_step_2_divs(<16 x float> %x) #1 { ; FMA-RECIP-LABEL: v16f32_one_step_2_divs: ; FMA-RECIP: # %bb.0: ; FMA-RECIP-NEXT: vrcpps %ymm0, %ymm2 -; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] ; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm2 * ymm0) + ymm3 ; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm0 = (ymm0 * ymm2) + ymm2 ; FMA-RECIP-NEXT: vrcpps %ymm1, %ymm2 @@ -1417,7 +1417,7 @@ define <16 x float> @v16f32_one_step_2_divs(<16 x float> %x) #1 { ; ; BTVER2-LABEL: v16f32_one_step_2_divs: ; BTVER2: # %bb.0: -; BTVER2-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00] +; BTVER2-NEXT: vmovaps {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [5:1.00] ; BTVER2-NEXT: vrcpps %ymm0, %ymm2 # sched: [2:2.00] ; BTVER2-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [2:2.00] ; BTVER2-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:2.00] @@ -1438,7 +1438,7 @@ define <16 x float> @v16f32_one_step_2_divs(<16 x float> %x) #1 { ; SANDY: # %bb.0: ; SANDY-NEXT: vrcpps %ymm0, %ymm2 # sched: [7:2.00] ; SANDY-NEXT: vmulps %ymm2, %ymm0, %ymm0 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] sched: [7:0.50] ; SANDY-NEXT: vsubps %ymm0, %ymm3, %ymm0 # sched: [3:1.00] ; SANDY-NEXT: vrcpps %ymm1, %ymm4 # sched: [7:2.00] ; SANDY-NEXT: vmulps %ymm0, %ymm2, %ymm0 # sched: [5:1.00] @@ -1517,7 +1517,7 @@ define <16 x float> @v16f32_two_step2(<16 x float> %x) #2 { ; SSE-NEXT: movaps %xmm0, %xmm4 ; SSE-NEXT: rcpps %xmm3, %xmm2 ; SSE-NEXT: mulps %xmm2, %xmm3 -; SSE-NEXT: movaps {{.*#+}} xmm0 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; SSE-NEXT: movaps {{.*#+}} xmm0 = [1,1,1,1] ; SSE-NEXT: movaps %xmm0, %xmm7 ; SSE-NEXT: subps %xmm3, %xmm7 ; SSE-NEXT: mulps %xmm2, %xmm7 @@ -1573,7 +1573,7 @@ define <16 x float> @v16f32_two_step2(<16 x float> %x) #2 { ; AVX-RECIP: # %bb.0: ; AVX-RECIP-NEXT: vrcpps %ymm1, %ymm2 ; AVX-RECIP-NEXT: vmulps %ymm2, %ymm1, %ymm3 -; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm4 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; AVX-RECIP-NEXT: vmovaps {{.*#+}} ymm4 = [1,1,1,1,1,1,1,1] ; AVX-RECIP-NEXT: vsubps %ymm3, %ymm4, %ymm3 ; AVX-RECIP-NEXT: vmulps %ymm3, %ymm2, %ymm3 ; AVX-RECIP-NEXT: vaddps %ymm3, %ymm2, %ymm2 @@ -1597,7 +1597,7 @@ define <16 x float> @v16f32_two_step2(<16 x float> %x) #2 { ; FMA-RECIP-LABEL: v16f32_two_step2: ; FMA-RECIP: # %bb.0: ; FMA-RECIP-NEXT: vrcpps %ymm1, %ymm2 -; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] +; FMA-RECIP-NEXT: vmovaps {{.*#+}} ymm3 = [1,1,1,1,1,1,1,1] ; FMA-RECIP-NEXT: vmovaps %ymm2, %ymm4 ; FMA-RECIP-NEXT: vfnmadd213ps {{.*#+}} ymm4 = -(ymm1 * ymm4) + ymm3 ; FMA-RECIP-NEXT: vfmadd132ps {{.*#+}} ymm4 = (ymm4 * ymm2) + ymm2 @@ -1615,7 +1615,7 @@ define <16 x float> @v16f32_two_step2(<16 x float> %x) #2 { ; ; BTVER2-LABEL: v16f32_two_step2: ; BTVER2: # %bb.0: -; BTVER2-NEXT: vmovaps {{.*#+}} ymm4 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [5:1.00] +; BTVER2-NEXT: vmovaps {{.*#+}} ymm4 = [1,1,1,1,1,1,1,1] sched: [5:1.00] ; BTVER2-NEXT: vrcpps %ymm1, %ymm2 # sched: [2:2.00] ; BTVER2-NEXT: vmulps %ymm2, %ymm1, %ymm3 # sched: [2:2.00] ; BTVER2-NEXT: vsubps %ymm3, %ymm4, %ymm3 # sched: [3:2.00] @@ -1642,7 +1642,7 @@ define <16 x float> @v16f32_two_step2(<16 x float> %x) #2 { ; SANDY: # %bb.0: ; SANDY-NEXT: vrcpps %ymm1, %ymm2 # sched: [7:2.00] ; SANDY-NEXT: vmulps %ymm2, %ymm1, %ymm3 # sched: [5:1.00] -; SANDY-NEXT: vmovaps {{.*#+}} ymm4 = [1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00,1.000000e+00] sched: [7:0.50] +; SANDY-NEXT: vmovaps {{.*#+}} ymm4 = [1,1,1,1,1,1,1,1] sched: [7:0.50] ; SANDY-NEXT: vsubps %ymm3, %ymm4, %ymm3 # sched: [3:1.00] ; SANDY-NEXT: vmulps %ymm3, %ymm2, %ymm3 # sched: [5:1.00] ; SANDY-NEXT: vaddps %ymm3, %ymm2, %ymm2 # sched: [3:1.00] |