[llvm-commits] [llvm] r116143 - in /llvm/trunk/lib/Target/ARM: ARMInstrNEON.td ARMSchedule.td ARMScheduleA8.td ARMScheduleA9.td
Evan Cheng
evan.cheng at apple.com
Fri Oct 8 21:07:58 PDT 2010
Author: evancheng
Date: Fri Oct 8 23:07:58 2010
New Revision: 116143
URL: http://llvm.org/viewvc/llvm-project?rev=116143&view=rev
Log:
Add VLD4 scheduling itineraries.
Modified:
llvm/trunk/lib/Target/ARM/ARMInstrNEON.td
llvm/trunk/lib/Target/ARM/ARMSchedule.td
llvm/trunk/lib/Target/ARM/ARMScheduleA8.td
llvm/trunk/lib/Target/ARM/ARMScheduleA9.td
Modified: llvm/trunk/lib/Target/ARM/ARMInstrNEON.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMInstrNEON.td?rev=116143&r1=116142&r2=116143&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMInstrNEON.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMInstrNEON.td Fri Oct 8 23:07:58 2010
@@ -548,7 +548,7 @@
: NLdSt<1, 0b10, op11_8, op7_4,
(outs DPR:$dst1, DPR:$dst2, DPR:$dst3, DPR:$dst4),
(ins addrmode6:$addr, DPR:$src1, DPR:$src2, DPR:$src3, DPR:$src4,
- nohash_imm:$lane), IIC_VLD4, "vld4", Dt,
+ nohash_imm:$lane), IIC_VLD4ln, "vld4", Dt,
"\\{$dst1[$lane], $dst2[$lane], $dst3[$lane], $dst4[$lane]\\}, $addr",
"$src1 = $dst1, $src2 = $dst2, $src3 = $dst3, $src4 = $dst4", []>;
@@ -556,16 +556,16 @@
def VLD4LNd16 : VLD4LN<0b0111, {?,?,0,?}, "16">;
def VLD4LNd32 : VLD4LN<0b1011, {?,0,?,?}, "32">;
-def VLD4LNd8Pseudo : VLDQQLNPseudo<IIC_VLD4>;
-def VLD4LNd16Pseudo : VLDQQLNPseudo<IIC_VLD4>;
-def VLD4LNd32Pseudo : VLDQQLNPseudo<IIC_VLD4>;
+def VLD4LNd8Pseudo : VLDQQLNPseudo<IIC_VLD4ln>;
+def VLD4LNd16Pseudo : VLDQQLNPseudo<IIC_VLD4ln>;
+def VLD4LNd32Pseudo : VLDQQLNPseudo<IIC_VLD4ln>;
// ...with double-spaced registers:
def VLD4LNq16 : VLD4LN<0b0111, {?,?,1,?}, "16">;
def VLD4LNq32 : VLD4LN<0b1011, {?,1,?,?}, "32">;
-def VLD4LNq16Pseudo : VLDQQQQLNPseudo<IIC_VLD4>;
-def VLD4LNq32Pseudo : VLDQQQQLNPseudo<IIC_VLD4>;
+def VLD4LNq16Pseudo : VLDQQQQLNPseudo<IIC_VLD4ln>;
+def VLD4LNq32Pseudo : VLDQQQQLNPseudo<IIC_VLD4ln>;
// ...with address register writeback:
class VLD4LNWB<bits<4> op11_8, bits<4> op7_4, string Dt>
@@ -573,7 +573,7 @@
(outs DPR:$dst1, DPR:$dst2, DPR:$dst3, DPR:$dst4, GPR:$wb),
(ins addrmode6:$addr, am6offset:$offset,
DPR:$src1, DPR:$src2, DPR:$src3, DPR:$src4, nohash_imm:$lane),
- IIC_VLD4, "vld4", Dt,
+ IIC_VLD4ln, "vld4", Dt,
"\\{$dst1[$lane], $dst2[$lane], $dst3[$lane], $dst4[$lane]\\}, $addr$offset",
"$src1 = $dst1, $src2 = $dst2, $src3 = $dst3, $src4 = $dst4, $addr.addr = $wb",
[]>;
@@ -582,15 +582,15 @@
def VLD4LNd16_UPD : VLD4LNWB<0b0111, {?,?,0,?}, "16">;
def VLD4LNd32_UPD : VLD4LNWB<0b1011, {?,0,?,?}, "32">;
-def VLD4LNd8Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD4>;
-def VLD4LNd16Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD4>;
-def VLD4LNd32Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD4>;
+def VLD4LNd8Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD4lnu>;
+def VLD4LNd16Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD4lnu>;
+def VLD4LNd32Pseudo_UPD : VLDQQLNWBPseudo<IIC_VLD4lnu>;
def VLD4LNq16_UPD : VLD4LNWB<0b0111, {?,?,1,?}, "16">;
def VLD4LNq32_UPD : VLD4LNWB<0b1011, {?,1,?,?}, "32">;
-def VLD4LNq16Pseudo_UPD : VLDQQQQLNWBPseudo<IIC_VLD4>;
-def VLD4LNq32Pseudo_UPD : VLDQQQQLNWBPseudo<IIC_VLD4>;
+def VLD4LNq16Pseudo_UPD : VLDQQQQLNWBPseudo<IIC_VLD4lnu>;
+def VLD4LNq32Pseudo_UPD : VLDQQQQLNWBPseudo<IIC_VLD4lnu>;
// VLD1DUP : Vector Load (single element to all lanes)
// VLD2DUP : Vector Load (single 2-element structure to all lanes)
Modified: llvm/trunk/lib/Target/ARM/ARMSchedule.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMSchedule.td?rev=116143&r1=116142&r2=116143&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMSchedule.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMSchedule.td Fri Oct 8 23:07:58 2010
@@ -145,6 +145,9 @@
def IIC_VLD3u : InstrItinClass;
def IIC_VLD3lnu : InstrItinClass;
def IIC_VLD4 : InstrItinClass;
+def IIC_VLD4ln : InstrItinClass;
+def IIC_VLD4u : InstrItinClass;
+def IIC_VLD4lnu : InstrItinClass;
def IIC_VST : InstrItinClass;
def IIC_VUNAD : InstrItinClass;
def IIC_VUNAQ : InstrItinClass;
Modified: llvm/trunk/lib/Target/ARM/ARMScheduleA8.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMScheduleA8.td?rev=116143&r1=116142&r2=116143&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMScheduleA8.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMScheduleA8.td Fri Oct 8 23:07:58 2010
@@ -390,7 +390,7 @@
//
// VLD1
InstrItinData<IIC_VLD1, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
- InstrStage<1, [A8_NLSPipe]>,
+ InstrStage<1, [A8_NLSPipe], 1>,
InstrStage<1, [A8_LSPipe]>]>,
// VLD1x2
InstrItinData<IIC_VLD1x2, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
@@ -496,8 +496,27 @@
//
// VLD4
InstrItinData<IIC_VLD4, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
- InstrStage<1, [A8_NLSPipe]>,
- InstrStage<1, [A8_LSPipe]>], [2, 2, 2, 2, 1]>,
+ InstrStage<4, [A8_NLSPipe], 1>,
+ InstrStage<4, [A8_LSPipe]>],
+ [3, 3, 4, 4, 1]>,
+ //
+ // VLD4ln
+ InstrItinData<IIC_VLD4ln, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
+ InstrStage<5, [A8_NLSPipe], 1>,
+ InstrStage<5, [A8_LSPipe]>],
+ [4, 4, 5, 5, 1, 1, 1, 1, 2, 2]>,
+ //
+ // VLD4u
+ InstrItinData<IIC_VLD4u, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
+ InstrStage<4, [A8_NLSPipe], 1>,
+ InstrStage<4, [A8_LSPipe]>],
+ [3, 3, 4, 4, 2, 1]>,
+ //
+ // VLD4lnu
+ InstrItinData<IIC_VLD4lnu, [InstrStage<1, [A8_Pipe0, A8_Pipe1]>,
+ InstrStage<5, [A8_NLSPipe], 1>,
+ InstrStage<5, [A8_LSPipe]>],
+ [4, 4, 5, 5, 2, 1, 1, 1, 1, 1, 2, 2]>,
//
// VST
// FIXME: We don't model this instruction properly
Modified: llvm/trunk/lib/Target/ARM/ARMScheduleA9.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/ARM/ARMScheduleA9.td?rev=116143&r1=116142&r2=116143&view=diff
==============================================================================
--- llvm/trunk/lib/Target/ARM/ARMScheduleA9.td (original)
+++ llvm/trunk/lib/Target/ARM/ARMScheduleA9.td Fri Oct 8 23:07:58 2010
@@ -854,14 +854,36 @@
[5, 5, 6, 2, 1, 1, 1, 1, 1, 2]>,
//
// VLD4
- // FIXME: We don't model this instruction properly
InstrItinData<IIC_VLD4, [InstrStage<1, [A9_DRegsN], 0, Required>,
- // Extra latency cycles since wbck is 6 cycles
- InstrStage<7, [A9_DRegsVFP], 0, Reserved>,
+ InstrStage<10, [A9_DRegsVFP], 0, Reserved>,
InstrStage<1, [A9_Issue0, A9_Issue1], 0>,
InstrStage<1, [A9_MUX0], 0>,
- InstrStage<1, [A9_NPipe]>],
- [2, 2, 2, 2, 1]>,
+ InstrStage<4, [A9_NPipe]>],
+ [4, 4, 5, 5, 1]>,
+ //
+ // VLD4ln
+ InstrItinData<IIC_VLD4ln, [InstrStage<1, [A9_DRegsN], 0, Required>,
+ InstrStage<11, [A9_DRegsVFP], 0, Reserved>,
+ InstrStage<1, [A9_Issue0, A9_Issue1], 0>,
+ InstrStage<1, [A9_MUX0], 0>,
+ InstrStage<5, [A9_NPipe]>],
+ [5, 5, 6, 6, 1, 1, 1, 1, 2, 2]>,
+ //
+ // VLD4u
+ InstrItinData<IIC_VLD4u, [InstrStage<1, [A9_DRegsN], 0, Required>,
+ InstrStage<10, [A9_DRegsVFP], 0, Reserved>,
+ InstrStage<1, [A9_Issue0, A9_Issue1], 0>,
+ InstrStage<1, [A9_MUX0], 0>,
+ InstrStage<4, [A9_NPipe]>],
+ [4, 4, 5, 5, 2, 1]>,
+ //
+ // VLD4lnu
+ InstrItinData<IIC_VLD4lnu, [InstrStage<1, [A9_DRegsN], 0, Required>,
+ InstrStage<11, [A9_DRegsVFP], 0, Reserved>,
+ InstrStage<1, [A9_Issue0, A9_Issue1], 0>,
+ InstrStage<1, [A9_MUX0], 0>,
+ InstrStage<5, [A9_NPipe]>],
+ [5, 5, 6, 6, 2, 1, 1, 1, 1, 1, 2, 2]>,
//
// VST
// FIXME: We don't model this instruction properly
More information about the llvm-commits
mailing list