[PATCH][AVX512] Fix execution domains on Gathers and Scatters

[Cameron McInally]

Hey guys,

Attached is a patch to fix the execution domains on AVX512 FP gathers
and scatters.

Thanks,
Cameron
-------------- next part --------------
Index: test/CodeGen/X86/avx512-gather-scatter-intrin.ll
===================================================================
--- test/CodeGen/X86/avx512-gather-scatter-intrin.ll	(revision 204713)
+++ test/CodeGen/X86/avx512-gather-scatter-intrin.ll	(working copy)
@@ -223,3 +223,81 @@
   call void @llvm.x86.avx512.scatter.qpi.512 (i8* %stbuf, <8 x i64>%ind2, <8 x i32> %x, i32 4)
   ret void
 }
+
+;CHECK-LABEL: gather_mask_dpd_execdomain
+;CHECK: vgatherdpd
+;CHECK: vmovapd
+;CHECK: ret
+define void @gather_mask_dpd_execdomain(<8 x i32> %ind, <8 x double> %src, i8 %mask, i8* %base, <8 x double>* %stbuf)  {
+  %x = call <8 x double> @llvm.x86.avx512.gather.dpd.mask.512 (<8 x double> %src, i8 %mask, <8 x i32>%ind, i8* %base, i32 4)
+  store <8 x double> %x, <8 x double>* %stbuf
+  ret void
+}
+
+;CHECK-LABEL: gather_mask_qpd_execdomain
+;CHECK: vgatherqpd
+;CHECK: vmovapd
+;CHECK: ret
+define void @gather_mask_qpd_execdomain(<8 x i64> %ind, <8 x double> %src, i8 %mask, i8* %base, <8 x double>* %stbuf)  {
+  %x = call <8 x double> @llvm.x86.avx512.gather.qpd.mask.512 (<8 x double> %src, i8 %mask, <8 x i64>%ind, i8* %base, i32 4)
+  store <8 x double> %x, <8 x double>* %stbuf
+  ret void
+}
+
+;CHECK-LABEL: gather_mask_dps_execdomain
+;CHECK: vgatherdps
+;CHECK: vmovaps 
+;CHECK: ret
+define <16 x float> @gather_mask_dps_execdomain(<16 x i32> %ind, <16 x float> %src, i16 %mask, i8* %base)  {
+  %res = call <16 x float> @llvm.x86.avx512.gather.dps.mask.512 (<16 x float> %src, i16 %mask, <16 x i32>%ind, i8* %base, i32 4)
+  ret <16 x float> %res;
+}
+
+;CHECK-LABEL: gather_mask_qps_execdomain
+;CHECK: vgatherqps
+;CHECK: vmovaps
+;CHECK: ret
+define <8 x float> @gather_mask_qps_execdomain(<8 x i64> %ind, <8 x float> %src, i8 %mask, i8* %base)  {
+  %res = call <8 x float> @llvm.x86.avx512.gather.qps.mask.512 (<8 x float> %src, i8 %mask, <8 x i64>%ind, i8* %base, i32 4)
+  ret <8 x float> %res;
+}
+
+;CHECK-LABEL: scatter_mask_dpd_execdomain
+;CHECK: vmovapd
+;CHECK: vscatterdpd
+;CHECK: ret
+define void @scatter_mask_dpd_execdomain(<8 x i32> %ind, <8 x double>* %src, i8 %mask, i8* %base, i8* %stbuf)  {
+  %x = load <8 x double>* %src, align 64 
+  call void @llvm.x86.avx512.scatter.dpd.mask.512 (i8* %stbuf, i8 %mask, <8 x i32>%ind, <8 x double> %x, i32 4)
+  ret void
+}
+
+;CHECK-LABEL: scatter_mask_qpd_execdomain
+;CHECK: vmovapd
+;CHECK: vscatterqpd
+;CHECK: ret
+define void @scatter_mask_qpd_execdomain(<8 x i64> %ind, <8 x double>* %src, i8 %mask, i8* %base, i8* %stbuf)  {
+  %x = load <8 x double>* %src, align 64
+  call void @llvm.x86.avx512.scatter.qpd.mask.512 (i8* %stbuf, i8 %mask, <8 x i64>%ind, <8 x double> %x, i32 4)
+  ret void
+}
+
+;CHECK-LABEL: scatter_mask_dps_execdomain
+;CHECK: vmovaps
+;CHECK: vscatterdps
+;CHECK: ret
+define void @scatter_mask_dps_execdomain(<16 x i32> %ind, <16 x float>* %src, i16 %mask, i8* %base, i8* %stbuf)  {
+  %x = load <16 x float>* %src, align 64
+  call void @llvm.x86.avx512.scatter.dps.mask.512 (i8* %stbuf, i16 %mask, <16 x i32>%ind, <16 x float> %x, i32 4)
+  ret void
+}
+
+;CHECK-LABEL: scatter_mask_qps_execdomain
+;CHECK: vmovaps
+;CHECK: vscatterqps
+;CHECK: ret
+define void @scatter_mask_qps_execdomain(<8 x i64> %ind, <8 x float>* %src, i8 %mask, i8* %base, i8* %stbuf)  {
+  %x = load <8 x float>* %src, align 32 
+  call void @llvm.x86.avx512.scatter.qps.mask.512 (i8* %stbuf, i8 %mask, <8 x i64>%ind, <8 x float> %x, i32 4)
+  ret void
+}
Index: lib/Target/X86/X86InstrAVX512.td
===================================================================
--- lib/Target/X86/X86InstrAVX512.td	(revision 204713)
+++ lib/Target/X86/X86InstrAVX512.td	(working copy)
@@ -3809,15 +3809,20 @@
             " \t{$src2, ${dst} {${mask}}|${dst} {${mask}}, $src2}"),
             []>, EVEX, EVEX_K;
 }
+
+let ExeDomain = SSEPackedDouble in {
 defm VGATHERDPDZ : avx512_gather<0x92, "vgatherdpd", VK8WM, VR512, vy64xmem>,
                                  EVEX_V512, VEX_W, EVEX_CD8<64, CD8VT1>;
+defm VGATHERQPDZ : avx512_gather<0x93, "vgatherqpd", VK8WM, VR512, vz64mem>,
+                                 EVEX_V512, VEX_W, EVEX_CD8<64, CD8VT1>;
+}
+
+let ExeDomain = SSEPackedSingle in {
 defm VGATHERDPSZ : avx512_gather<0x92, "vgatherdps", VK16WM, VR512, vz32mem>,
                                  EVEX_V512, EVEX_CD8<32, CD8VT1>;
-
-defm VGATHERQPDZ : avx512_gather<0x93, "vgatherqpd", VK8WM, VR512, vz64mem>,
-                                 EVEX_V512, VEX_W, EVEX_CD8<64, CD8VT1>;
 defm VGATHERQPSZ : avx512_gather<0x93, "vgatherqps", VK8WM, VR256X, vz64mem>,
                                  EVEX_V512, EVEX_CD8<32, CD8VT1>;
+}
   
 defm VPGATHERDQZ : avx512_gather<0x90, "vpgatherdq", VK8WM, VR512,  vy64xmem>,
                                  EVEX_V512, VEX_W, EVEX_CD8<64, CD8VT1>;
@@ -3839,16 +3844,20 @@
             []>, EVEX, EVEX_K;
 }
 
+let ExeDomain = SSEPackedDouble in {
 defm VSCATTERDPDZ : avx512_scatter<0xA2, "vscatterdpd", VK8WM, VR512, vy64xmem>,
                                    EVEX_V512, VEX_W, EVEX_CD8<64, CD8VT1>;
+defm VSCATTERQPDZ : avx512_scatter<0xA3, "vscatterqpd", VK8WM, VR512, vz64mem>,
+                                   EVEX_V512, VEX_W, EVEX_CD8<64, CD8VT1>;
+}
+
+let ExeDomain = SSEPackedSingle in {
 defm VSCATTERDPSZ : avx512_scatter<0xA2, "vscatterdps", VK16WM, VR512, vz32mem>,
                                    EVEX_V512, EVEX_CD8<32, CD8VT1>;
-
-defm VSCATTERQPDZ : avx512_scatter<0xA3, "vscatterqpd", VK8WM, VR512, vz64mem>,
-                                   EVEX_V512, VEX_W, EVEX_CD8<64, CD8VT1>;
 defm VSCATTERQPSZ : avx512_scatter<0xA3, "vscatterqps", VK8WM, VR256X, vz64mem>,
                                    EVEX_V512, EVEX_CD8<32, CD8VT1>;
-  
+}
+
 defm VPSCATTERDQZ : avx512_scatter<0xA0, "vpscatterdq", VK8WM, VR512, vy64xmem>,
                                    EVEX_V512, VEX_W, EVEX_CD8<64, CD8VT1>;
 defm VPSCATTERDDZ : avx512_scatter<0xA0, "vpscatterdd", VK16WM, VR512, vz32mem>,