[llvm] r296293 - [X86] Fix execution domain for cmpss/sd instructions.
Craig Topper via llvm-commits
llvm-commits at lists.llvm.org
Sat Feb 25 22:45:59 PST 2017
Author: ctopper
Date: Sun Feb 26 00:45:59 2017
New Revision: 296293
URL: http://llvm.org/viewvc/llvm-project?rev=296293&view=rev
Log:
[X86] Fix execution domain for cmpss/sd instructions.
Modified:
llvm/trunk/lib/Target/X86/X86InstrSSE.td
llvm/trunk/test/CodeGen/X86/fast-isel-select-sse.ll
llvm/trunk/test/CodeGen/X86/fp-select-cmp-and.ll
llvm/trunk/test/CodeGen/X86/logical-load-fold.ll
llvm/trunk/test/CodeGen/X86/sse-minmax.ll
Modified: llvm/trunk/lib/Target/X86/X86InstrSSE.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrSSE.td?rev=296293&r1=296292&r2=296293&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrSSE.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrSSE.td Sun Feb 26 00:45:59 2017
@@ -2308,10 +2308,12 @@ multiclass sse12_cmp_scalar<RegisterClas
}
}
+let ExeDomain = SSEPackedSingle in
defm VCMPSS : sse12_cmp_scalar<FR32, f32mem, AVXCC, X86cmps, f32, loadf32,
"cmp${cc}ss\t{$src2, $src1, $dst|$dst, $src1, $src2}",
"cmpss\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
SSE_ALU_F32S, i8immZExt5>, XS, VEX_4V, VEX_LIG, VEX_WIG;
+let ExeDomain = SSEPackedDouble in
defm VCMPSD : sse12_cmp_scalar<FR64, f64mem, AVXCC, X86cmps, f64, loadf64,
"cmp${cc}sd\t{$src2, $src1, $dst|$dst, $src1, $src2}",
"cmpsd\t{$cc, $src2, $src1, $dst|$dst, $src1, $src2, $cc}",
@@ -2319,10 +2321,12 @@ defm VCMPSD : sse12_cmp_scalar<FR64, f64
XD, VEX_4V, VEX_LIG, VEX_WIG;
let Constraints = "$src1 = $dst" in {
+ let ExeDomain = SSEPackedSingle in
defm CMPSS : sse12_cmp_scalar<FR32, f32mem, SSECC, X86cmps, f32, loadf32,
"cmp${cc}ss\t{$src2, $dst|$dst, $src2}",
"cmpss\t{$cc, $src2, $dst|$dst, $src2, $cc}", SSE_ALU_F32S,
i8immZExt3>, XS;
+ let ExeDomain = SSEPackedDouble in
defm CMPSD : sse12_cmp_scalar<FR64, f64mem, SSECC, X86cmps, f64, loadf64,
"cmp${cc}sd\t{$src2, $dst|$dst, $src2}",
"cmpsd\t{$cc, $src2, $dst|$dst, $src2, $cc}",
@@ -2348,18 +2352,22 @@ multiclass sse12_cmp_scalar_int<Operand
let isCodeGenOnly = 1 in {
// Aliases to match intrinsics which expect XMM operand(s).
+ let ExeDomain = SSEPackedSingle in
defm Int_VCMPSS : sse12_cmp_scalar_int<ssmem, AVXCC, int_x86_sse_cmp_ss,
"cmp${cc}ss\t{$src, $src1, $dst|$dst, $src1, $src}",
SSE_ALU_F32S, i8immZExt5, sse_load_f32>,
XS, VEX_4V;
+ let ExeDomain = SSEPackedDouble in
defm Int_VCMPSD : sse12_cmp_scalar_int<sdmem, AVXCC, int_x86_sse2_cmp_sd,
"cmp${cc}sd\t{$src, $src1, $dst|$dst, $src1, $src}",
SSE_ALU_F32S, i8immZExt5, sse_load_f64>, // same latency as f32
XD, VEX_4V;
let Constraints = "$src1 = $dst" in {
+ let ExeDomain = SSEPackedSingle in
defm Int_CMPSS : sse12_cmp_scalar_int<ssmem, SSECC, int_x86_sse_cmp_ss,
"cmp${cc}ss\t{$src, $dst|$dst, $src}",
SSE_ALU_F32S, i8immZExt3, sse_load_f32>, XS;
+ let ExeDomain = SSEPackedDouble in
defm Int_CMPSD : sse12_cmp_scalar_int<sdmem, SSECC, int_x86_sse2_cmp_sd,
"cmp${cc}sd\t{$src, $dst|$dst, $src}",
SSE_ALU_F64S, i8immZExt3, sse_load_f64>,
Modified: llvm/trunk/test/CodeGen/X86/fast-isel-select-sse.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/fast-isel-select-sse.ll?rev=296293&r1=296292&r2=296293&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/fast-isel-select-sse.ll (original)
+++ llvm/trunk/test/CodeGen/X86/fast-isel-select-sse.ll Sun Feb 26 00:45:59 2017
@@ -39,9 +39,9 @@ define double @select_fcmp_oeq_f64(doubl
; SSE-LABEL: select_fcmp_oeq_f64:
; SSE: # BB#0:
; SSE-NEXT: cmpeqsd %xmm1, %xmm0
-; SSE-NEXT: andps %xmm0, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm0
-; SSE-NEXT: orps %xmm2, %xmm0
+; SSE-NEXT: andpd %xmm0, %xmm2
+; SSE-NEXT: andnpd %xmm3, %xmm0
+; SSE-NEXT: orpd %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: select_fcmp_oeq_f64:
@@ -94,10 +94,10 @@ define double @select_fcmp_ogt_f64(doubl
; SSE-LABEL: select_fcmp_ogt_f64:
; SSE: # BB#0:
; SSE-NEXT: cmpltsd %xmm0, %xmm1
-; SSE-NEXT: andps %xmm1, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm1
-; SSE-NEXT: orps %xmm2, %xmm1
-; SSE-NEXT: movaps %xmm1, %xmm0
+; SSE-NEXT: andpd %xmm1, %xmm2
+; SSE-NEXT: andnpd %xmm3, %xmm1
+; SSE-NEXT: orpd %xmm2, %xmm1
+; SSE-NEXT: movapd %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: select_fcmp_ogt_f64:
@@ -150,10 +150,10 @@ define double @select_fcmp_oge_f64(doubl
; SSE-LABEL: select_fcmp_oge_f64:
; SSE: # BB#0:
; SSE-NEXT: cmplesd %xmm0, %xmm1
-; SSE-NEXT: andps %xmm1, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm1
-; SSE-NEXT: orps %xmm2, %xmm1
-; SSE-NEXT: movaps %xmm1, %xmm0
+; SSE-NEXT: andpd %xmm1, %xmm2
+; SSE-NEXT: andnpd %xmm3, %xmm1
+; SSE-NEXT: orpd %xmm2, %xmm1
+; SSE-NEXT: movapd %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: select_fcmp_oge_f64:
@@ -205,9 +205,9 @@ define double @select_fcmp_olt_f64(doubl
; SSE-LABEL: select_fcmp_olt_f64:
; SSE: # BB#0:
; SSE-NEXT: cmpltsd %xmm1, %xmm0
-; SSE-NEXT: andps %xmm0, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm0
-; SSE-NEXT: orps %xmm2, %xmm0
+; SSE-NEXT: andpd %xmm0, %xmm2
+; SSE-NEXT: andnpd %xmm3, %xmm0
+; SSE-NEXT: orpd %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: select_fcmp_olt_f64:
@@ -259,9 +259,9 @@ define double @select_fcmp_ole_f64(doubl
; SSE-LABEL: select_fcmp_ole_f64:
; SSE: # BB#0:
; SSE-NEXT: cmplesd %xmm1, %xmm0
-; SSE-NEXT: andps %xmm0, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm0
-; SSE-NEXT: orps %xmm2, %xmm0
+; SSE-NEXT: andpd %xmm0, %xmm2
+; SSE-NEXT: andnpd %xmm3, %xmm0
+; SSE-NEXT: orpd %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: select_fcmp_ole_f64:
@@ -313,9 +313,9 @@ define double @select_fcmp_ord_f64(doubl
; SSE-LABEL: select_fcmp_ord_f64:
; SSE: # BB#0:
; SSE-NEXT: cmpordsd %xmm1, %xmm0
-; SSE-NEXT: andps %xmm0, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm0
-; SSE-NEXT: orps %xmm2, %xmm0
+; SSE-NEXT: andpd %xmm0, %xmm2
+; SSE-NEXT: andnpd %xmm3, %xmm0
+; SSE-NEXT: orpd %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: select_fcmp_ord_f64:
@@ -367,9 +367,9 @@ define double @select_fcmp_uno_f64(doubl
; SSE-LABEL: select_fcmp_uno_f64:
; SSE: # BB#0:
; SSE-NEXT: cmpunordsd %xmm1, %xmm0
-; SSE-NEXT: andps %xmm0, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm0
-; SSE-NEXT: orps %xmm2, %xmm0
+; SSE-NEXT: andpd %xmm0, %xmm2
+; SSE-NEXT: andnpd %xmm3, %xmm0
+; SSE-NEXT: orpd %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: select_fcmp_uno_f64:
@@ -421,9 +421,9 @@ define double @select_fcmp_ugt_f64(doubl
; SSE-LABEL: select_fcmp_ugt_f64:
; SSE: # BB#0:
; SSE-NEXT: cmpnlesd %xmm1, %xmm0
-; SSE-NEXT: andps %xmm0, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm0
-; SSE-NEXT: orps %xmm2, %xmm0
+; SSE-NEXT: andpd %xmm0, %xmm2
+; SSE-NEXT: andnpd %xmm3, %xmm0
+; SSE-NEXT: orpd %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: select_fcmp_ugt_f64:
@@ -475,9 +475,9 @@ define double @select_fcmp_uge_f64(doubl
; SSE-LABEL: select_fcmp_uge_f64:
; SSE: # BB#0:
; SSE-NEXT: cmpnltsd %xmm1, %xmm0
-; SSE-NEXT: andps %xmm0, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm0
-; SSE-NEXT: orps %xmm2, %xmm0
+; SSE-NEXT: andpd %xmm0, %xmm2
+; SSE-NEXT: andnpd %xmm3, %xmm0
+; SSE-NEXT: orpd %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: select_fcmp_uge_f64:
@@ -530,10 +530,10 @@ define double @select_fcmp_ult_f64(doubl
; SSE-LABEL: select_fcmp_ult_f64:
; SSE: # BB#0:
; SSE-NEXT: cmpnlesd %xmm0, %xmm1
-; SSE-NEXT: andps %xmm1, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm1
-; SSE-NEXT: orps %xmm2, %xmm1
-; SSE-NEXT: movaps %xmm1, %xmm0
+; SSE-NEXT: andpd %xmm1, %xmm2
+; SSE-NEXT: andnpd %xmm3, %xmm1
+; SSE-NEXT: orpd %xmm2, %xmm1
+; SSE-NEXT: movapd %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: select_fcmp_ult_f64:
@@ -586,10 +586,10 @@ define double @select_fcmp_ule_f64(doubl
; SSE-LABEL: select_fcmp_ule_f64:
; SSE: # BB#0:
; SSE-NEXT: cmpnltsd %xmm0, %xmm1
-; SSE-NEXT: andps %xmm1, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm1
-; SSE-NEXT: orps %xmm2, %xmm1
-; SSE-NEXT: movaps %xmm1, %xmm0
+; SSE-NEXT: andpd %xmm1, %xmm2
+; SSE-NEXT: andnpd %xmm3, %xmm1
+; SSE-NEXT: orpd %xmm2, %xmm1
+; SSE-NEXT: movapd %xmm1, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: select_fcmp_ule_f64:
@@ -641,9 +641,9 @@ define double @select_fcmp_une_f64(doubl
; SSE-LABEL: select_fcmp_une_f64:
; SSE: # BB#0:
; SSE-NEXT: cmpneqsd %xmm1, %xmm0
-; SSE-NEXT: andps %xmm0, %xmm2
-; SSE-NEXT: andnps %xmm3, %xmm0
-; SSE-NEXT: orps %xmm2, %xmm0
+; SSE-NEXT: andpd %xmm0, %xmm2
+; SSE-NEXT: andnpd %xmm3, %xmm0
+; SSE-NEXT: orpd %xmm2, %xmm0
; SSE-NEXT: retq
;
; AVX-LABEL: select_fcmp_une_f64:
Modified: llvm/trunk/test/CodeGen/X86/fp-select-cmp-and.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/fp-select-cmp-and.ll?rev=296293&r1=296292&r2=296293&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/fp-select-cmp-and.ll (original)
+++ llvm/trunk/test/CodeGen/X86/fp-select-cmp-and.ll Sun Feb 26 00:45:59 2017
@@ -5,7 +5,7 @@ define double @test1(double %a, double %
; CHECK-LABEL: test1:
; CHECK: # BB#0:
; CHECK-NEXT: cmpltsd %xmm2, %xmm0
-; CHECK-NEXT: andps %xmm1, %xmm0
+; CHECK-NEXT: andpd %xmm1, %xmm0
; CHECK-NEXT: retq
;
%cmp = fcmp olt double %a, %eps
@@ -17,7 +17,7 @@ define double @test2(double %a, double %
; CHECK-LABEL: test2:
; CHECK: # BB#0:
; CHECK-NEXT: cmplesd %xmm2, %xmm0
-; CHECK-NEXT: andps %xmm1, %xmm0
+; CHECK-NEXT: andpd %xmm1, %xmm0
; CHECK-NEXT: retq
;
%cmp = fcmp ole double %a, %eps
@@ -29,8 +29,8 @@ define double @test3(double %a, double %
; CHECK-LABEL: test3:
; CHECK: # BB#0:
; CHECK-NEXT: cmpltsd %xmm0, %xmm2
-; CHECK-NEXT: andps %xmm1, %xmm2
-; CHECK-NEXT: movaps %xmm2, %xmm0
+; CHECK-NEXT: andpd %xmm1, %xmm2
+; CHECK-NEXT: movapd %xmm2, %xmm0
; CHECK-NEXT: retq
;
%cmp = fcmp ogt double %a, %eps
@@ -42,8 +42,8 @@ define double @test4(double %a, double %
; CHECK-LABEL: test4:
; CHECK: # BB#0:
; CHECK-NEXT: cmplesd %xmm0, %xmm2
-; CHECK-NEXT: andps %xmm1, %xmm2
-; CHECK-NEXT: movaps %xmm2, %xmm0
+; CHECK-NEXT: andpd %xmm1, %xmm2
+; CHECK-NEXT: movapd %xmm2, %xmm0
; CHECK-NEXT: retq
;
%cmp = fcmp oge double %a, %eps
@@ -55,7 +55,7 @@ define double @test5(double %a, double %
; CHECK-LABEL: test5:
; CHECK: # BB#0:
; CHECK-NEXT: cmpltsd %xmm2, %xmm0
-; CHECK-NEXT: andnps %xmm1, %xmm0
+; CHECK-NEXT: andnpd %xmm1, %xmm0
; CHECK-NEXT: retq
;
%cmp = fcmp olt double %a, %eps
@@ -67,7 +67,7 @@ define double @test6(double %a, double %
; CHECK-LABEL: test6:
; CHECK: # BB#0:
; CHECK-NEXT: cmplesd %xmm2, %xmm0
-; CHECK-NEXT: andnps %xmm1, %xmm0
+; CHECK-NEXT: andnpd %xmm1, %xmm0
; CHECK-NEXT: retq
;
%cmp = fcmp ole double %a, %eps
@@ -79,8 +79,8 @@ define double @test7(double %a, double %
; CHECK-LABEL: test7:
; CHECK: # BB#0:
; CHECK-NEXT: cmpltsd %xmm0, %xmm2
-; CHECK-NEXT: andnps %xmm1, %xmm2
-; CHECK-NEXT: movaps %xmm2, %xmm0
+; CHECK-NEXT: andnpd %xmm1, %xmm2
+; CHECK-NEXT: movapd %xmm2, %xmm0
; CHECK-NEXT: retq
;
%cmp = fcmp ogt double %a, %eps
@@ -92,8 +92,8 @@ define double @test8(double %a, double %
; CHECK-LABEL: test8:
; CHECK: # BB#0:
; CHECK-NEXT: cmplesd %xmm0, %xmm2
-; CHECK-NEXT: andnps %xmm1, %xmm2
-; CHECK-NEXT: movaps %xmm2, %xmm0
+; CHECK-NEXT: andnpd %xmm1, %xmm2
+; CHECK-NEXT: movapd %xmm2, %xmm0
; CHECK-NEXT: retq
;
%cmp = fcmp oge double %a, %eps
@@ -220,10 +220,10 @@ define double @test18(double %a, double
; CHECK-LABEL: test18:
; CHECK: # BB#0:
; CHECK-NEXT: cmplesd %xmm0, %xmm3
-; CHECK-NEXT: andps %xmm3, %xmm2
-; CHECK-NEXT: andnps %xmm1, %xmm3
-; CHECK-NEXT: orps %xmm2, %xmm3
-; CHECK-NEXT: movaps %xmm3, %xmm0
+; CHECK-NEXT: andpd %xmm3, %xmm2
+; CHECK-NEXT: andnpd %xmm1, %xmm3
+; CHECK-NEXT: orpd %xmm2, %xmm3
+; CHECK-NEXT: movapd %xmm3, %xmm0
; CHECK-NEXT: retq
;
%cmp = fcmp oge double %a, %eps
Modified: llvm/trunk/test/CodeGen/X86/logical-load-fold.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/logical-load-fold.ll?rev=296293&r1=296292&r2=296293&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/logical-load-fold.ll (original)
+++ llvm/trunk/test/CodeGen/X86/logical-load-fold.ll Sun Feb 26 00:45:59 2017
@@ -15,14 +15,14 @@ define double @load_double_no_fold(doubl
; SSE2: # BB#0:
; SSE2-NEXT: cmplesd %xmm0, %xmm1
; SSE2-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
-; SSE2-NEXT: andps %xmm1, %xmm0
+; SSE2-NEXT: andpd %xmm1, %xmm0
; SSE2-NEXT: retq
;
; AVX-LABEL: load_double_no_fold:
; AVX: # BB#0:
; AVX-NEXT: vcmplesd %xmm0, %xmm1, %xmm0
; AVX-NEXT: vmovsd {{.*#+}} xmm1 = mem[0],zero
-; AVX-NEXT: vandps %xmm1, %xmm0, %xmm0
+; AVX-NEXT: vandpd %xmm1, %xmm0, %xmm0
; AVX-NEXT: retq
%cmp = fcmp oge double %x, %y
Modified: llvm/trunk/test/CodeGen/X86/sse-minmax.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/sse-minmax.ll?rev=296293&r1=296292&r2=296293&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/sse-minmax.ll (original)
+++ llvm/trunk/test/CodeGen/X86/sse-minmax.ll Sun Feb 26 00:45:59 2017
@@ -80,11 +80,11 @@ define double @olt_inverse(double %x, do
define double @oge(double %x, double %y) {
; STRICT-LABEL: oge:
; STRICT: # BB#0:
-; STRICT-NEXT: movaps %xmm1, %xmm2
+; STRICT-NEXT: movapd %xmm1, %xmm2
; STRICT-NEXT: cmplesd %xmm0, %xmm2
-; STRICT-NEXT: andps %xmm2, %xmm0
-; STRICT-NEXT: andnps %xmm1, %xmm2
-; STRICT-NEXT: orps %xmm2, %xmm0
+; STRICT-NEXT: andpd %xmm2, %xmm0
+; STRICT-NEXT: andnpd %xmm1, %xmm2
+; STRICT-NEXT: orpd %xmm2, %xmm0
; STRICT-NEXT: retq
;
; RELAX-LABEL: oge:
@@ -99,12 +99,12 @@ define double @oge(double %x, double %y)
define double @ole(double %x, double %y) {
; STRICT-LABEL: ole:
; STRICT: # BB#0:
-; STRICT-NEXT: movaps %xmm0, %xmm2
+; STRICT-NEXT: movapd %xmm0, %xmm2
; STRICT-NEXT: cmplesd %xmm1, %xmm2
-; STRICT-NEXT: andps %xmm2, %xmm0
-; STRICT-NEXT: andnps %xmm1, %xmm2
-; STRICT-NEXT: orps %xmm0, %xmm2
-; STRICT-NEXT: movaps %xmm2, %xmm0
+; STRICT-NEXT: andpd %xmm2, %xmm0
+; STRICT-NEXT: andnpd %xmm1, %xmm2
+; STRICT-NEXT: orpd %xmm0, %xmm2
+; STRICT-NEXT: movapd %xmm2, %xmm0
; STRICT-NEXT: retq
;
; RELAX-LABEL: ole:
@@ -119,12 +119,12 @@ define double @ole(double %x, double %y)
define double @oge_inverse(double %x, double %y) {
; STRICT-LABEL: oge_inverse:
; STRICT: # BB#0:
-; STRICT-NEXT: movaps %xmm1, %xmm2
+; STRICT-NEXT: movapd %xmm1, %xmm2
; STRICT-NEXT: cmplesd %xmm0, %xmm2
-; STRICT-NEXT: andps %xmm2, %xmm1
-; STRICT-NEXT: andnps %xmm0, %xmm2
-; STRICT-NEXT: orps %xmm1, %xmm2
-; STRICT-NEXT: movaps %xmm2, %xmm0
+; STRICT-NEXT: andpd %xmm2, %xmm1
+; STRICT-NEXT: andnpd %xmm0, %xmm2
+; STRICT-NEXT: orpd %xmm1, %xmm2
+; STRICT-NEXT: movapd %xmm2, %xmm0
; STRICT-NEXT: retq
;
; UNSAFE-LABEL: oge_inverse:
@@ -145,12 +145,12 @@ define double @oge_inverse(double %x, do
define double @ole_inverse(double %x, double %y) {
; STRICT-LABEL: ole_inverse:
; STRICT: # BB#0:
-; STRICT-NEXT: movaps %xmm0, %xmm2
+; STRICT-NEXT: movapd %xmm0, %xmm2
; STRICT-NEXT: cmplesd %xmm1, %xmm2
-; STRICT-NEXT: andps %xmm2, %xmm1
-; STRICT-NEXT: andnps %xmm0, %xmm2
-; STRICT-NEXT: orps %xmm1, %xmm2
-; STRICT-NEXT: movaps %xmm2, %xmm0
+; STRICT-NEXT: andpd %xmm2, %xmm1
+; STRICT-NEXT: andnpd %xmm0, %xmm2
+; STRICT-NEXT: orpd %xmm1, %xmm2
+; STRICT-NEXT: movapd %xmm2, %xmm0
; STRICT-NEXT: retq
;
; UNSAFE-LABEL: ole_inverse:
@@ -243,9 +243,9 @@ define double @olt_inverse_x(double %x)
define double @oge_x(double %x) {
; STRICT-LABEL: oge_x:
; STRICT: # BB#0:
-; STRICT-NEXT: xorps %xmm1, %xmm1
+; STRICT-NEXT: xorpd %xmm1, %xmm1
; STRICT-NEXT: cmplesd %xmm0, %xmm1
-; STRICT-NEXT: andps %xmm1, %xmm0
+; STRICT-NEXT: andpd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; RELAX-LABEL: oge_x:
@@ -261,11 +261,11 @@ define double @oge_x(double %x) {
define double @ole_x(double %x) {
; STRICT-LABEL: ole_x:
; STRICT: # BB#0:
-; STRICT-NEXT: xorps %xmm2, %xmm2
-; STRICT-NEXT: movaps %xmm0, %xmm1
+; STRICT-NEXT: xorpd %xmm2, %xmm2
+; STRICT-NEXT: movapd %xmm0, %xmm1
; STRICT-NEXT: cmplesd %xmm2, %xmm1
-; STRICT-NEXT: andps %xmm0, %xmm1
-; STRICT-NEXT: movaps %xmm1, %xmm0
+; STRICT-NEXT: andpd %xmm0, %xmm1
+; STRICT-NEXT: movapd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; RELAX-LABEL: ole_x:
@@ -281,10 +281,10 @@ define double @ole_x(double %x) {
define double @oge_inverse_x(double %x) {
; STRICT-LABEL: oge_inverse_x:
; STRICT: # BB#0:
-; STRICT-NEXT: xorps %xmm1, %xmm1
+; STRICT-NEXT: xorpd %xmm1, %xmm1
; STRICT-NEXT: cmplesd %xmm0, %xmm1
-; STRICT-NEXT: andnps %xmm0, %xmm1
-; STRICT-NEXT: movaps %xmm1, %xmm0
+; STRICT-NEXT: andnpd %xmm0, %xmm1
+; STRICT-NEXT: movapd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; UNSAFE-LABEL: oge_inverse_x:
@@ -307,11 +307,11 @@ define double @oge_inverse_x(double %x)
define double @ole_inverse_x(double %x) {
; STRICT-LABEL: ole_inverse_x:
; STRICT: # BB#0:
-; STRICT-NEXT: xorps %xmm2, %xmm2
-; STRICT-NEXT: movaps %xmm0, %xmm1
+; STRICT-NEXT: xorpd %xmm2, %xmm2
+; STRICT-NEXT: movapd %xmm0, %xmm1
; STRICT-NEXT: cmplesd %xmm2, %xmm1
-; STRICT-NEXT: andnps %xmm0, %xmm1
-; STRICT-NEXT: movaps %xmm1, %xmm0
+; STRICT-NEXT: andnpd %xmm0, %xmm1
+; STRICT-NEXT: movapd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; UNSAFE-LABEL: ole_inverse_x:
@@ -334,12 +334,12 @@ define double @ole_inverse_x(double %x)
define double @ugt(double %x, double %y) {
; STRICT-LABEL: ugt:
; STRICT: # BB#0:
-; STRICT-NEXT: movaps %xmm0, %xmm2
+; STRICT-NEXT: movapd %xmm0, %xmm2
; STRICT-NEXT: cmpnlesd %xmm1, %xmm2
-; STRICT-NEXT: andps %xmm2, %xmm0
-; STRICT-NEXT: andnps %xmm1, %xmm2
-; STRICT-NEXT: orps %xmm0, %xmm2
-; STRICT-NEXT: movaps %xmm2, %xmm0
+; STRICT-NEXT: andpd %xmm2, %xmm0
+; STRICT-NEXT: andnpd %xmm1, %xmm2
+; STRICT-NEXT: orpd %xmm0, %xmm2
+; STRICT-NEXT: movapd %xmm2, %xmm0
; STRICT-NEXT: retq
;
; RELAX-LABEL: ugt:
@@ -354,11 +354,11 @@ define double @ugt(double %x, double %y)
define double @ult(double %x, double %y) {
; STRICT-LABEL: ult:
; STRICT: # BB#0:
-; STRICT-NEXT: movaps %xmm1, %xmm2
+; STRICT-NEXT: movapd %xmm1, %xmm2
; STRICT-NEXT: cmpnlesd %xmm0, %xmm2
-; STRICT-NEXT: andps %xmm2, %xmm0
-; STRICT-NEXT: andnps %xmm1, %xmm2
-; STRICT-NEXT: orps %xmm2, %xmm0
+; STRICT-NEXT: andpd %xmm2, %xmm0
+; STRICT-NEXT: andnpd %xmm1, %xmm2
+; STRICT-NEXT: orpd %xmm2, %xmm0
; STRICT-NEXT: retq
;
; RELAX-LABEL: ult:
@@ -373,12 +373,12 @@ define double @ult(double %x, double %y)
define double @ugt_inverse(double %x, double %y) {
; STRICT-LABEL: ugt_inverse:
; STRICT: # BB#0:
-; STRICT-NEXT: movaps %xmm0, %xmm2
+; STRICT-NEXT: movapd %xmm0, %xmm2
; STRICT-NEXT: cmpnlesd %xmm1, %xmm2
-; STRICT-NEXT: andps %xmm2, %xmm1
-; STRICT-NEXT: andnps %xmm0, %xmm2
-; STRICT-NEXT: orps %xmm1, %xmm2
-; STRICT-NEXT: movaps %xmm2, %xmm0
+; STRICT-NEXT: andpd %xmm2, %xmm1
+; STRICT-NEXT: andnpd %xmm0, %xmm2
+; STRICT-NEXT: orpd %xmm1, %xmm2
+; STRICT-NEXT: movapd %xmm2, %xmm0
; STRICT-NEXT: retq
;
; UNSAFE-LABEL: ugt_inverse:
@@ -399,12 +399,12 @@ define double @ugt_inverse(double %x, do
define double @ult_inverse(double %x, double %y) {
; STRICT-LABEL: ult_inverse:
; STRICT: # BB#0:
-; STRICT-NEXT: movaps %xmm1, %xmm2
+; STRICT-NEXT: movapd %xmm1, %xmm2
; STRICT-NEXT: cmpnlesd %xmm0, %xmm2
-; STRICT-NEXT: andps %xmm2, %xmm1
-; STRICT-NEXT: andnps %xmm0, %xmm2
-; STRICT-NEXT: orps %xmm1, %xmm2
-; STRICT-NEXT: movaps %xmm2, %xmm0
+; STRICT-NEXT: andpd %xmm2, %xmm1
+; STRICT-NEXT: andnpd %xmm0, %xmm2
+; STRICT-NEXT: orpd %xmm1, %xmm2
+; STRICT-NEXT: movapd %xmm2, %xmm0
; STRICT-NEXT: retq
;
; UNSAFE-LABEL: ult_inverse:
@@ -499,11 +499,11 @@ define double @ule_inverse(double %x, do
define double @ugt_x(double %x) {
; STRICT-LABEL: ugt_x:
; STRICT: # BB#0:
-; STRICT-NEXT: xorps %xmm2, %xmm2
-; STRICT-NEXT: movaps %xmm0, %xmm1
+; STRICT-NEXT: xorpd %xmm2, %xmm2
+; STRICT-NEXT: movapd %xmm0, %xmm1
; STRICT-NEXT: cmpnlesd %xmm2, %xmm1
-; STRICT-NEXT: andps %xmm0, %xmm1
-; STRICT-NEXT: movaps %xmm1, %xmm0
+; STRICT-NEXT: andpd %xmm0, %xmm1
+; STRICT-NEXT: movapd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; RELAX-LABEL: ugt_x:
@@ -519,9 +519,9 @@ define double @ugt_x(double %x) {
define double @ult_x(double %x) {
; STRICT-LABEL: ult_x:
; STRICT: # BB#0:
-; STRICT-NEXT: xorps %xmm1, %xmm1
+; STRICT-NEXT: xorpd %xmm1, %xmm1
; STRICT-NEXT: cmpnlesd %xmm0, %xmm1
-; STRICT-NEXT: andps %xmm1, %xmm0
+; STRICT-NEXT: andpd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; RELAX-LABEL: ult_x:
@@ -537,11 +537,11 @@ define double @ult_x(double %x) {
define double @ugt_inverse_x(double %x) {
; STRICT-LABEL: ugt_inverse_x:
; STRICT: # BB#0:
-; STRICT-NEXT: xorps %xmm2, %xmm2
-; STRICT-NEXT: movaps %xmm0, %xmm1
+; STRICT-NEXT: xorpd %xmm2, %xmm2
+; STRICT-NEXT: movapd %xmm0, %xmm1
; STRICT-NEXT: cmpnlesd %xmm2, %xmm1
-; STRICT-NEXT: andnps %xmm0, %xmm1
-; STRICT-NEXT: movaps %xmm1, %xmm0
+; STRICT-NEXT: andnpd %xmm0, %xmm1
+; STRICT-NEXT: movapd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; UNSAFE-LABEL: ugt_inverse_x:
@@ -564,10 +564,10 @@ define double @ugt_inverse_x(double %x)
define double @ult_inverse_x(double %x) {
; STRICT-LABEL: ult_inverse_x:
; STRICT: # BB#0:
-; STRICT-NEXT: xorps %xmm1, %xmm1
+; STRICT-NEXT: xorpd %xmm1, %xmm1
; STRICT-NEXT: cmpnlesd %xmm0, %xmm1
-; STRICT-NEXT: andnps %xmm0, %xmm1
-; STRICT-NEXT: movaps %xmm1, %xmm0
+; STRICT-NEXT: andnpd %xmm0, %xmm1
+; STRICT-NEXT: movapd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; UNSAFE-LABEL: ult_inverse_x:
@@ -743,11 +743,11 @@ define double @oge_y(double %x) {
; STRICT-LABEL: oge_y:
; STRICT: # BB#0:
; STRICT-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
-; STRICT-NEXT: movaps %xmm1, %xmm2
+; STRICT-NEXT: movapd %xmm1, %xmm2
; STRICT-NEXT: cmplesd %xmm0, %xmm2
-; STRICT-NEXT: andps %xmm2, %xmm0
-; STRICT-NEXT: andnps %xmm1, %xmm2
-; STRICT-NEXT: orps %xmm2, %xmm0
+; STRICT-NEXT: andpd %xmm2, %xmm0
+; STRICT-NEXT: andnpd %xmm1, %xmm2
+; STRICT-NEXT: orpd %xmm2, %xmm0
; STRICT-NEXT: retq
;
; RELAX-LABEL: oge_y:
@@ -763,12 +763,12 @@ define double @ole_y(double %x) {
; STRICT-LABEL: ole_y:
; STRICT: # BB#0:
; STRICT-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
-; STRICT-NEXT: movaps %xmm0, %xmm1
+; STRICT-NEXT: movapd %xmm0, %xmm1
; STRICT-NEXT: cmplesd %xmm2, %xmm1
-; STRICT-NEXT: andps %xmm1, %xmm0
-; STRICT-NEXT: andnps %xmm2, %xmm1
-; STRICT-NEXT: orps %xmm0, %xmm1
-; STRICT-NEXT: movaps %xmm1, %xmm0
+; STRICT-NEXT: andpd %xmm1, %xmm0
+; STRICT-NEXT: andnpd %xmm2, %xmm1
+; STRICT-NEXT: orpd %xmm0, %xmm1
+; STRICT-NEXT: movapd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; RELAX-LABEL: ole_y:
@@ -784,12 +784,12 @@ define double @oge_inverse_y(double %x)
; STRICT-LABEL: oge_inverse_y:
; STRICT: # BB#0:
; STRICT-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
-; STRICT-NEXT: movaps %xmm2, %xmm1
+; STRICT-NEXT: movapd %xmm2, %xmm1
; STRICT-NEXT: cmplesd %xmm0, %xmm1
-; STRICT-NEXT: andps %xmm1, %xmm2
-; STRICT-NEXT: andnps %xmm0, %xmm1
-; STRICT-NEXT: orps %xmm2, %xmm1
-; STRICT-NEXT: movaps %xmm1, %xmm0
+; STRICT-NEXT: andpd %xmm1, %xmm2
+; STRICT-NEXT: andnpd %xmm0, %xmm1
+; STRICT-NEXT: orpd %xmm2, %xmm1
+; STRICT-NEXT: movapd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; UNSAFE-LABEL: oge_inverse_y:
@@ -812,12 +812,12 @@ define double @ole_inverse_y(double %x)
; STRICT-LABEL: ole_inverse_y:
; STRICT: # BB#0:
; STRICT-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
-; STRICT-NEXT: movaps %xmm0, %xmm1
+; STRICT-NEXT: movapd %xmm0, %xmm1
; STRICT-NEXT: cmplesd %xmm2, %xmm1
-; STRICT-NEXT: andps %xmm1, %xmm2
-; STRICT-NEXT: andnps %xmm0, %xmm1
-; STRICT-NEXT: orps %xmm2, %xmm1
-; STRICT-NEXT: movaps %xmm1, %xmm0
+; STRICT-NEXT: andpd %xmm1, %xmm2
+; STRICT-NEXT: andnpd %xmm0, %xmm1
+; STRICT-NEXT: orpd %xmm2, %xmm1
+; STRICT-NEXT: movapd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; UNSAFE-LABEL: ole_inverse_y:
@@ -840,12 +840,12 @@ define double @ugt_y(double %x) {
; STRICT-LABEL: ugt_y:
; STRICT: # BB#0:
; STRICT-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
-; STRICT-NEXT: movaps %xmm0, %xmm1
+; STRICT-NEXT: movapd %xmm0, %xmm1
; STRICT-NEXT: cmpnlesd %xmm2, %xmm1
-; STRICT-NEXT: andps %xmm1, %xmm0
-; STRICT-NEXT: andnps %xmm2, %xmm1
-; STRICT-NEXT: orps %xmm0, %xmm1
-; STRICT-NEXT: movaps %xmm1, %xmm0
+; STRICT-NEXT: andpd %xmm1, %xmm0
+; STRICT-NEXT: andnpd %xmm2, %xmm1
+; STRICT-NEXT: orpd %xmm0, %xmm1
+; STRICT-NEXT: movapd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; RELAX-LABEL: ugt_y:
@@ -861,11 +861,11 @@ define double @ult_y(double %x) {
; STRICT-LABEL: ult_y:
; STRICT: # BB#0:
; STRICT-NEXT: movsd {{.*#+}} xmm1 = mem[0],zero
-; STRICT-NEXT: movaps %xmm1, %xmm2
+; STRICT-NEXT: movapd %xmm1, %xmm2
; STRICT-NEXT: cmpnlesd %xmm0, %xmm2
-; STRICT-NEXT: andps %xmm2, %xmm0
-; STRICT-NEXT: andnps %xmm1, %xmm2
-; STRICT-NEXT: orps %xmm2, %xmm0
+; STRICT-NEXT: andpd %xmm2, %xmm0
+; STRICT-NEXT: andnpd %xmm1, %xmm2
+; STRICT-NEXT: orpd %xmm2, %xmm0
; STRICT-NEXT: retq
;
; RELAX-LABEL: ult_y:
@@ -881,12 +881,12 @@ define double @ugt_inverse_y(double %x)
; STRICT-LABEL: ugt_inverse_y:
; STRICT: # BB#0:
; STRICT-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
-; STRICT-NEXT: movaps %xmm0, %xmm1
+; STRICT-NEXT: movapd %xmm0, %xmm1
; STRICT-NEXT: cmpnlesd %xmm2, %xmm1
-; STRICT-NEXT: andps %xmm1, %xmm2
-; STRICT-NEXT: andnps %xmm0, %xmm1
-; STRICT-NEXT: orps %xmm2, %xmm1
-; STRICT-NEXT: movaps %xmm1, %xmm0
+; STRICT-NEXT: andpd %xmm1, %xmm2
+; STRICT-NEXT: andnpd %xmm0, %xmm1
+; STRICT-NEXT: orpd %xmm2, %xmm1
+; STRICT-NEXT: movapd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; UNSAFE-LABEL: ugt_inverse_y:
@@ -909,12 +909,12 @@ define double @ult_inverse_y(double %x)
; STRICT-LABEL: ult_inverse_y:
; STRICT: # BB#0:
; STRICT-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
-; STRICT-NEXT: movaps %xmm2, %xmm1
+; STRICT-NEXT: movapd %xmm2, %xmm1
; STRICT-NEXT: cmpnlesd %xmm0, %xmm1
-; STRICT-NEXT: andps %xmm1, %xmm2
-; STRICT-NEXT: andnps %xmm0, %xmm1
-; STRICT-NEXT: orps %xmm2, %xmm1
-; STRICT-NEXT: movaps %xmm1, %xmm0
+; STRICT-NEXT: andpd %xmm1, %xmm2
+; STRICT-NEXT: andnpd %xmm0, %xmm1
+; STRICT-NEXT: orpd %xmm2, %xmm1
+; STRICT-NEXT: movapd %xmm1, %xmm0
; STRICT-NEXT: retq
;
; UNSAFE-LABEL: ult_inverse_y:
More information about the llvm-commits
mailing list