[llvm] eb88e79 - AMDGPU: Add some additional test coverage for BFI matching
Matt Arsenault via llvm-commits
llvm-commits at lists.llvm.org
Wed Jan 26 12:06:58 PST 2022
Author: Matt Arsenault
Date: 2022-01-26T15:06:50-05:00
New Revision: eb88e793ff579071e03766970e46a5d60d77cf7c
URL: https://github.com/llvm/llvm-project/commit/eb88e793ff579071e03766970e46a5d60d77cf7c
DIFF: https://github.com/llvm/llvm-project/commit/eb88e793ff579071e03766970e46a5d60d77cf7c.diff
LOG: AMDGPU: Add some additional test coverage for BFI matching
Try to stress constant bus restriction enforcement since some of these
are broken for GlobalISel. Split the r600 test because some of these
cases don't compile (and all the ones using return values are
discarded).
Added:
llvm/test/CodeGen/AMDGPU/bfi_int.r600.ll
Modified:
llvm/test/CodeGen/AMDGPU/bfi_int.ll
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/AMDGPU/bfi_int.ll b/llvm/test/CodeGen/AMDGPU/bfi_int.ll
index a046170f79e3..35b13804990a 100644
--- a/llvm/test/CodeGen/AMDGPU/bfi_int.ll
+++ b/llvm/test/CodeGen/AMDGPU/bfi_int.ll
@@ -2,13 +2,12 @@
; RUN: llc -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GFX7 %s
; RUN: llc -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GFX8 %s
; RUN: llc -march=amdgcn -mcpu=gfx1031 -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GFX10 %s
-; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -enable-var-scope -check-prefixes=R600 %s
; BFI_INT Definition pattern from ISA docs
; (y & x) | (z & ~x)
;
-define amdgpu_kernel void @bfi_def(i32 addrspace(1)* %out, i32 %x, i32 %y, i32 %z) {
-; GFX7-LABEL: bfi_def:
+define amdgpu_kernel void @s_bfi_def_i32(i32 addrspace(1)* %out, i32 %x, i32 %y, i32 %z) {
+; GFX7-LABEL: s_bfi_def_i32:
; GFX7: ; %bb.0: ; %entry
; GFX7-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0xb
; GFX7-NEXT: s_load_dword s6, s[0:1], 0xd
@@ -23,7 +22,7 @@ define amdgpu_kernel void @bfi_def(i32 addrspace(1)* %out, i32 %x, i32 %y, i32 %
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX7-NEXT: s_endpgm
;
-; GFX8-LABEL: bfi_def:
+; GFX8-LABEL: s_bfi_def_i32:
; GFX8: ; %bb.0: ; %entry
; GFX8-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x2c
; GFX8-NEXT: s_load_dword s4, s[0:1], 0x34
@@ -38,7 +37,7 @@ define amdgpu_kernel void @bfi_def(i32 addrspace(1)* %out, i32 %x, i32 %y, i32 %
; GFX8-NEXT: flat_store_dword v[0:1], v2
; GFX8-NEXT: s_endpgm
;
-; GFX10-LABEL: bfi_def:
+; GFX10-LABEL: s_bfi_def_i32:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_clause 0x2
; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x2c
@@ -52,17 +51,6 @@ define amdgpu_kernel void @bfi_def(i32 addrspace(1)* %out, i32 %x, i32 %y, i32 %
; GFX10-NEXT: v_mov_b32_e32 v1, s2
; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
; GFX10-NEXT: s_endpgm
-;
-; R600-LABEL: bfi_def:
-; R600: ; %bb.0: ; %entry
-; R600-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
-; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
-; R600-NEXT: CF_END
-; R600-NEXT: PAD
-; R600-NEXT: ALU clause starting at 4:
-; R600-NEXT: LSHR * T0.X, KC0[2].Y, literal.x,
-; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
-; R600-NEXT: BFI_INT * T1.X, KC0[2].Z, KC0[2].W, KC0[3].X,
entry:
%0 = xor i32 %x, -1
%1 = and i32 %z, %0
@@ -72,10 +60,37 @@ entry:
ret void
}
+define i32 @v_bfi_def_i32(i32 %x, i32 %y, i32 %z) {
+; GFX7-LABEL: v_bfi_def_i32:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX7-NEXT: v_bfi_b32 v0, v0, v1, v2
+; GFX7-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX8-LABEL: v_bfi_def_i32:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX8-NEXT: v_bfi_b32 v0, v0, v1, v2
+; GFX8-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: v_bfi_def_i32:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_bfi_b32 v0, v0, v1, v2
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+entry:
+ %0 = xor i32 %x, -1
+ %1 = and i32 %z, %0
+ %2 = and i32 %y, %x
+ %3 = or i32 %1, %2
+ ret i32 %3
+}
+
; SHA-256 Ch function
; z ^ (x & (y ^ z))
-define amdgpu_kernel void @bfi_sha256_ch(i32 addrspace(1)* %out, i32 %x, i32 %y, i32 %z) {
-; GFX7-LABEL: bfi_sha256_ch:
+define amdgpu_kernel void @s_bfi_sha256_ch(i32 addrspace(1)* %out, i32 %x, i32 %y, i32 %z) {
+; GFX7-LABEL: s_bfi_sha256_ch:
; GFX7: ; %bb.0: ; %entry
; GFX7-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0xb
; GFX7-NEXT: s_load_dword s6, s[0:1], 0xd
@@ -90,7 +105,7 @@ define amdgpu_kernel void @bfi_sha256_ch(i32 addrspace(1)* %out, i32 %x, i32 %y,
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX7-NEXT: s_endpgm
;
-; GFX8-LABEL: bfi_sha256_ch:
+; GFX8-LABEL: s_bfi_sha256_ch:
; GFX8: ; %bb.0: ; %entry
; GFX8-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x2c
; GFX8-NEXT: s_load_dword s4, s[0:1], 0x34
@@ -105,7 +120,7 @@ define amdgpu_kernel void @bfi_sha256_ch(i32 addrspace(1)* %out, i32 %x, i32 %y,
; GFX8-NEXT: flat_store_dword v[0:1], v2
; GFX8-NEXT: s_endpgm
;
-; GFX10-LABEL: bfi_sha256_ch:
+; GFX10-LABEL: s_bfi_sha256_ch:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_clause 0x2
; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x2c
@@ -119,17 +134,6 @@ define amdgpu_kernel void @bfi_sha256_ch(i32 addrspace(1)* %out, i32 %x, i32 %y,
; GFX10-NEXT: v_mov_b32_e32 v1, s2
; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
; GFX10-NEXT: s_endpgm
-;
-; R600-LABEL: bfi_sha256_ch:
-; R600: ; %bb.0: ; %entry
-; R600-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
-; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
-; R600-NEXT: CF_END
-; R600-NEXT: PAD
-; R600-NEXT: ALU clause starting at 4:
-; R600-NEXT: LSHR * T0.X, KC0[2].Y, literal.x,
-; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
-; R600-NEXT: BFI_INT * T1.X, KC0[2].Z, KC0[2].W, KC0[3].X,
entry:
%0 = xor i32 %y, %z
%1 = and i32 %x, %0
@@ -138,10 +142,180 @@ entry:
ret void
}
+define i32 @v_bfi_sha256_ch(i32 %x, i32 %y, i32 %z) {
+; GFX7-LABEL: v_bfi_sha256_ch:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX7-NEXT: v_bfi_b32 v0, v0, v1, v2
+; GFX7-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX8-LABEL: v_bfi_sha256_ch:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX8-NEXT: v_bfi_b32 v0, v0, v1, v2
+; GFX8-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: v_bfi_sha256_ch:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_bfi_b32 v0, v0, v1, v2
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+entry:
+ %0 = xor i32 %y, %z
+ %1 = and i32 %x, %0
+ %2 = xor i32 %z, %1
+ ret i32 %2
+}
+
+define amdgpu_ps float @v_s_s_bfi_sha256_ch(i32 %x, i32 inreg %y, i32 inreg %z) {
+; GFX7-LABEL: v_s_s_bfi_sha256_ch:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: v_mov_b32_e32 v1, s1
+; GFX7-NEXT: v_bfi_b32 v0, v0, s0, v1
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: v_s_s_bfi_sha256_ch:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: v_mov_b32_e32 v1, s1
+; GFX8-NEXT: v_bfi_b32 v0, v0, s0, v1
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: v_s_s_bfi_sha256_ch:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: v_bfi_b32 v0, v0, s0, s1
+; GFX10-NEXT: ; return to shader part epilog
+entry:
+ %xor0 = xor i32 %y, %z
+ %and = and i32 %x, %xor0
+ %xor1 = xor i32 %z, %and
+ %cast = bitcast i32 %xor1 to float
+ ret float %cast
+}
+
+define amdgpu_ps float @s_v_s_bfi_sha256_ch(i32 inreg %x, i32 %y, i32 inreg %z) {
+; GFX7-LABEL: s_v_s_bfi_sha256_ch:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: v_mov_b32_e32 v1, s1
+; GFX7-NEXT: v_bfi_b32 v0, s0, v0, v1
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: s_v_s_bfi_sha256_ch:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: v_mov_b32_e32 v1, s1
+; GFX8-NEXT: v_bfi_b32 v0, s0, v0, v1
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: s_v_s_bfi_sha256_ch:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: v_bfi_b32 v0, s0, v0, s1
+; GFX10-NEXT: ; return to shader part epilog
+entry:
+ %xor0 = xor i32 %y, %z
+ %and = and i32 %x, %xor0
+ %xor1 = xor i32 %z, %and
+ %cast = bitcast i32 %xor1 to float
+ ret float %cast
+}
+
+define amdgpu_ps float @s_s_v_bfi_sha256_ch(i32 inreg %x, i32 inreg %y, i32 %z) {
+; GFX7-LABEL: s_s_v_bfi_sha256_ch:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: v_mov_b32_e32 v1, s1
+; GFX7-NEXT: v_bfi_b32 v0, s0, v1, v0
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: s_s_v_bfi_sha256_ch:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: v_mov_b32_e32 v1, s1
+; GFX8-NEXT: v_bfi_b32 v0, s0, v1, v0
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: s_s_v_bfi_sha256_ch:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: v_bfi_b32 v0, s0, s1, v0
+; GFX10-NEXT: ; return to shader part epilog
+entry:
+ %xor0 = xor i32 %y, %z
+ %and = and i32 %x, %xor0
+ %xor1 = xor i32 %z, %and
+ %cast = bitcast i32 %xor1 to float
+ ret float %cast
+}
+
+define amdgpu_ps float @s_v_v_bfi_sha256_ch(i32 inreg %x, i32 %y, i32 %z) {
+; GFX7-LABEL: s_v_v_bfi_sha256_ch:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: v_bfi_b32 v0, s0, v0, v1
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: s_v_v_bfi_sha256_ch:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: v_bfi_b32 v0, s0, v0, v1
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: s_v_v_bfi_sha256_ch:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: v_bfi_b32 v0, s0, v0, v1
+; GFX10-NEXT: ; return to shader part epilog
+entry:
+ %xor0 = xor i32 %y, %z
+ %and = and i32 %x, %xor0
+ %xor1 = xor i32 %z, %and
+ %cast = bitcast i32 %xor1 to float
+ ret float %cast
+}
+
+define amdgpu_ps float @v_s_v_bfi_sha256_ch(i32 %x, i32 inreg %y, i32 %z) {
+; GFX7-LABEL: v_s_v_bfi_sha256_ch:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: v_bfi_b32 v0, v0, s0, v1
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: v_s_v_bfi_sha256_ch:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: v_bfi_b32 v0, v0, s0, v1
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: v_s_v_bfi_sha256_ch:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: v_bfi_b32 v0, v0, s0, v1
+; GFX10-NEXT: ; return to shader part epilog
+entry:
+ %xor0 = xor i32 %y, %z
+ %and = and i32 %x, %xor0
+ %xor1 = xor i32 %z, %and
+ %cast = bitcast i32 %xor1 to float
+ ret float %cast
+}
+
+define amdgpu_ps float @v_v_s_bfi_sha256_ch(i32 %x, i32 %y, i32 inreg %z) {
+; GFX7-LABEL: v_v_s_bfi_sha256_ch:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: v_bfi_b32 v0, v0, v1, s0
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: v_v_s_bfi_sha256_ch:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: v_bfi_b32 v0, v0, v1, s0
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: v_v_s_bfi_sha256_ch:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: v_bfi_b32 v0, v0, v1, s0
+; GFX10-NEXT: ; return to shader part epilog
+entry:
+ %xor0 = xor i32 %y, %z
+ %and = and i32 %x, %xor0
+ %xor1 = xor i32 %z, %and
+ %cast = bitcast i32 %xor1 to float
+ ret float %cast
+}
+
; SHA-256 Ma function
; ((x & z) | (y & (x | z)))
-define amdgpu_kernel void @bfi_sha256_ma(i32 addrspace(1)* %out, i32 %x, i32 %y, i32 %z) {
-; GFX7-LABEL: bfi_sha256_ma:
+define amdgpu_kernel void @s_bfi_sha256_ma(i32 addrspace(1)* %out, i32 %x, i32 %y, i32 %z) {
+; GFX7-LABEL: s_bfi_sha256_ma:
; GFX7: ; %bb.0: ; %entry
; GFX7-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0xb
; GFX7-NEXT: s_load_dword s6, s[0:1], 0xd
@@ -157,7 +331,7 @@ define amdgpu_kernel void @bfi_sha256_ma(i32 addrspace(1)* %out, i32 %x, i32 %y,
; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0
; GFX7-NEXT: s_endpgm
;
-; GFX8-LABEL: bfi_sha256_ma:
+; GFX8-LABEL: s_bfi_sha256_ma:
; GFX8: ; %bb.0: ; %entry
; GFX8-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x2c
; GFX8-NEXT: s_load_dword s4, s[0:1], 0x34
@@ -173,7 +347,7 @@ define amdgpu_kernel void @bfi_sha256_ma(i32 addrspace(1)* %out, i32 %x, i32 %y,
; GFX8-NEXT: flat_store_dword v[0:1], v2
; GFX8-NEXT: s_endpgm
;
-; GFX10-LABEL: bfi_sha256_ma:
+; GFX10-LABEL: s_bfi_sha256_ma:
; GFX10: ; %bb.0: ; %entry
; GFX10-NEXT: s_clause 0x2
; GFX10-NEXT: s_load_dwordx2 s[2:3], s[0:1], 0x2c
@@ -188,18 +362,6 @@ define amdgpu_kernel void @bfi_sha256_ma(i32 addrspace(1)* %out, i32 %x, i32 %y,
; GFX10-NEXT: v_mov_b32_e32 v1, s2
; GFX10-NEXT: global_store_dword v0, v1, s[0:1]
; GFX10-NEXT: s_endpgm
-;
-; R600-LABEL: bfi_sha256_ma:
-; R600: ; %bb.0: ; %entry
-; R600-NEXT: ALU 3, @4, KC0[CB0:0-32], KC1[]
-; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
-; R600-NEXT: CF_END
-; R600-NEXT: PAD
-; R600-NEXT: ALU clause starting at 4:
-; R600-NEXT: XOR_INT * T0.W, KC0[2].Z, KC0[2].W,
-; R600-NEXT: BFI_INT * T0.X, PV.W, KC0[3].X, KC0[2].W,
-; R600-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
-; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
entry:
%0 = and i32 %x, %z
%1 = or i32 %x, %z
@@ -209,6 +371,36 @@ entry:
ret void
}
+define i32 @v_bfi_sha256_ma(i32 %x, i32 %y, i32 %z) {
+; GFX7-LABEL: v_bfi_sha256_ma:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX7-NEXT: v_xor_b32_e32 v0, v0, v1
+; GFX7-NEXT: v_bfi_b32 v0, v0, v2, v1
+; GFX7-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX8-LABEL: v_bfi_sha256_ma:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX8-NEXT: v_xor_b32_e32 v0, v0, v1
+; GFX8-NEXT: v_bfi_b32 v0, v0, v2, v1
+; GFX8-NEXT: s_setpc_b64 s[30:31]
+;
+; GFX10-LABEL: v_bfi_sha256_ma:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX10-NEXT: s_waitcnt_vscnt null, 0x0
+; GFX10-NEXT: v_xor_b32_e32 v0, v0, v1
+; GFX10-NEXT: v_bfi_b32 v0, v0, v2, v1
+; GFX10-NEXT: s_setpc_b64 s[30:31]
+entry:
+ %0 = and i32 %x, %z
+ %1 = or i32 %x, %z
+ %2 = and i32 %y, %1
+ %3 = or i32 %0, %2
+ ret i32 %3
+}
+
define <2 x i32> @v_bitselect_v2i32_pat1(<2 x i32> %a, <2 x i32> %b, <2 x i32> %mask) {
; GFX7-LABEL: v_bitselect_v2i32_pat1:
; GFX7: ; %bb.0:
@@ -231,11 +423,6 @@ define <2 x i32> @v_bitselect_v2i32_pat1(<2 x i32> %a, <2 x i32> %b, <2 x i32> %
; GFX10-NEXT: v_bfi_b32 v0, v2, v0, v4
; GFX10-NEXT: v_bfi_b32 v1, v3, v1, v5
; GFX10-NEXT: s_setpc_b64 s[30:31]
-;
-; R600-LABEL: v_bitselect_v2i32_pat1:
-; R600: ; %bb.0:
-; R600-NEXT: CF_END
-; R600-NEXT: PAD
%xor.0 = xor <2 x i32> %a, %mask
%and = and <2 x i32> %xor.0, %b
%bitselect = xor <2 x i32> %and, %mask
@@ -264,11 +451,6 @@ define i64 @v_bitselect_i64_pat_0(i64 %a, i64 %b, i64 %mask) {
; GFX10-NEXT: v_bfi_b32 v0, v0, v2, v4
; GFX10-NEXT: v_bfi_b32 v1, v1, v3, v5
; GFX10-NEXT: s_setpc_b64 s[30:31]
-;
-; R600-LABEL: v_bitselect_i64_pat_0:
-; R600: ; %bb.0:
-; R600-NEXT: CF_END
-; R600-NEXT: PAD
%and0 = and i64 %a, %b
%not.a = xor i64 %a, -1
%and1 = and i64 %not.a, %mask
@@ -276,6 +458,174 @@ define i64 @v_bitselect_i64_pat_0(i64 %a, i64 %b, i64 %mask) {
ret i64 %bitselect
}
+define amdgpu_ps <2 x float> @v_s_s_bitselect_i64_pat_0(i64 %a, i64 inreg %b, i64 inreg %mask) {
+; GFX7-LABEL: v_s_s_bitselect_i64_pat_0:
+; GFX7: ; %bb.0:
+; GFX7-NEXT: v_mov_b32_e32 v2, s3
+; GFX7-NEXT: v_bfi_b32 v1, v1, s1, v2
+; GFX7-NEXT: v_mov_b32_e32 v2, s2
+; GFX7-NEXT: v_bfi_b32 v0, v0, s0, v2
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: v_s_s_bitselect_i64_pat_0:
+; GFX8: ; %bb.0:
+; GFX8-NEXT: v_mov_b32_e32 v2, s3
+; GFX8-NEXT: v_bfi_b32 v1, v1, s1, v2
+; GFX8-NEXT: v_mov_b32_e32 v2, s2
+; GFX8-NEXT: v_bfi_b32 v0, v0, s0, v2
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: v_s_s_bitselect_i64_pat_0:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: v_bfi_b32 v0, v0, s0, s2
+; GFX10-NEXT: v_bfi_b32 v1, v1, s1, s3
+; GFX10-NEXT: ; return to shader part epilog
+ %and0 = and i64 %a, %b
+ %not.a = xor i64 %a, -1
+ %and1 = and i64 %not.a, %mask
+ %bitselect = or i64 %and0, %and1
+ %cast = bitcast i64 %bitselect to <2 x float>
+ ret <2 x float> %cast
+}
+
+define amdgpu_ps <2 x float> @s_v_s_bitselect_i64_pat_0(i64 inreg %a, i64 %b, i64 inreg %mask) {
+; GFX7-LABEL: s_v_s_bitselect_i64_pat_0:
+; GFX7: ; %bb.0:
+; GFX7-NEXT: v_mov_b32_e32 v2, s3
+; GFX7-NEXT: v_bfi_b32 v1, s1, v1, v2
+; GFX7-NEXT: v_mov_b32_e32 v2, s2
+; GFX7-NEXT: v_bfi_b32 v0, s0, v0, v2
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: s_v_s_bitselect_i64_pat_0:
+; GFX8: ; %bb.0:
+; GFX8-NEXT: v_mov_b32_e32 v2, s3
+; GFX8-NEXT: v_bfi_b32 v1, s1, v1, v2
+; GFX8-NEXT: v_mov_b32_e32 v2, s2
+; GFX8-NEXT: v_bfi_b32 v0, s0, v0, v2
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: s_v_s_bitselect_i64_pat_0:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: v_bfi_b32 v0, s0, v0, s2
+; GFX10-NEXT: v_bfi_b32 v1, s1, v1, s3
+; GFX10-NEXT: ; return to shader part epilog
+ %and0 = and i64 %a, %b
+ %not.a = xor i64 %a, -1
+ %and1 = and i64 %not.a, %mask
+ %bitselect = or i64 %and0, %and1
+ %cast = bitcast i64 %bitselect to <2 x float>
+ ret <2 x float> %cast
+}
+
+define amdgpu_ps <2 x float> @s_s_v_bitselect_i64_pat_0(i64 inreg %a, i64 inreg %b, i64 %mask) {
+; GFX7-LABEL: s_s_v_bitselect_i64_pat_0:
+; GFX7: ; %bb.0:
+; GFX7-NEXT: v_mov_b32_e32 v2, s3
+; GFX7-NEXT: v_bfi_b32 v1, s1, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v2, s2
+; GFX7-NEXT: v_bfi_b32 v0, s0, v2, v0
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: s_s_v_bitselect_i64_pat_0:
+; GFX8: ; %bb.0:
+; GFX8-NEXT: v_mov_b32_e32 v2, s3
+; GFX8-NEXT: v_bfi_b32 v1, s1, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v2, s2
+; GFX8-NEXT: v_bfi_b32 v0, s0, v2, v0
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: s_s_v_bitselect_i64_pat_0:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: v_bfi_b32 v0, s0, s2, v0
+; GFX10-NEXT: v_bfi_b32 v1, s1, s3, v1
+; GFX10-NEXT: ; return to shader part epilog
+ %and0 = and i64 %a, %b
+ %not.a = xor i64 %a, -1
+ %and1 = and i64 %not.a, %mask
+ %bitselect = or i64 %and0, %and1
+ %cast = bitcast i64 %bitselect to <2 x float>
+ ret <2 x float> %cast
+}
+
+define amdgpu_ps <2 x float> @v_v_s_bitselect_i64_pat_0(i64 %a, i64 %b, i64 inreg %mask) {
+; GFX7-LABEL: v_v_s_bitselect_i64_pat_0:
+; GFX7: ; %bb.0:
+; GFX7-NEXT: v_bfi_b32 v1, v1, v3, s1
+; GFX7-NEXT: v_bfi_b32 v0, v0, v2, s0
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: v_v_s_bitselect_i64_pat_0:
+; GFX8: ; %bb.0:
+; GFX8-NEXT: v_bfi_b32 v1, v1, v3, s1
+; GFX8-NEXT: v_bfi_b32 v0, v0, v2, s0
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: v_v_s_bitselect_i64_pat_0:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: v_bfi_b32 v0, v0, v2, s0
+; GFX10-NEXT: v_bfi_b32 v1, v1, v3, s1
+; GFX10-NEXT: ; return to shader part epilog
+ %and0 = and i64 %a, %b
+ %not.a = xor i64 %a, -1
+ %and1 = and i64 %not.a, %mask
+ %bitselect = or i64 %and0, %and1
+ %cast = bitcast i64 %bitselect to <2 x float>
+ ret <2 x float> %cast
+}
+
+define amdgpu_ps <2 x float> @v_s_v_bitselect_i64_pat_0(i64 %a, i64 inreg %b, i64 %mask) {
+; GFX7-LABEL: v_s_v_bitselect_i64_pat_0:
+; GFX7: ; %bb.0:
+; GFX7-NEXT: v_bfi_b32 v1, v1, s1, v3
+; GFX7-NEXT: v_bfi_b32 v0, v0, s0, v2
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: v_s_v_bitselect_i64_pat_0:
+; GFX8: ; %bb.0:
+; GFX8-NEXT: v_bfi_b32 v1, v1, s1, v3
+; GFX8-NEXT: v_bfi_b32 v0, v0, s0, v2
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: v_s_v_bitselect_i64_pat_0:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: v_bfi_b32 v0, v0, s0, v2
+; GFX10-NEXT: v_bfi_b32 v1, v1, s1, v3
+; GFX10-NEXT: ; return to shader part epilog
+ %and0 = and i64 %a, %b
+ %not.a = xor i64 %a, -1
+ %and1 = and i64 %not.a, %mask
+ %bitselect = or i64 %and0, %and1
+ %cast = bitcast i64 %bitselect to <2 x float>
+ ret <2 x float> %cast
+}
+
+define amdgpu_ps <2 x float> @s_v_v_bitselect_i64_pat_0(i64 inreg %a, i64 %b, i64 %mask) {
+; GFX7-LABEL: s_v_v_bitselect_i64_pat_0:
+; GFX7: ; %bb.0:
+; GFX7-NEXT: v_bfi_b32 v1, s1, v1, v3
+; GFX7-NEXT: v_bfi_b32 v0, s0, v0, v2
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: s_v_v_bitselect_i64_pat_0:
+; GFX8: ; %bb.0:
+; GFX8-NEXT: v_bfi_b32 v1, s1, v1, v3
+; GFX8-NEXT: v_bfi_b32 v0, s0, v0, v2
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: s_v_v_bitselect_i64_pat_0:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: v_bfi_b32 v0, s0, v0, v2
+; GFX10-NEXT: v_bfi_b32 v1, s1, v1, v3
+; GFX10-NEXT: ; return to shader part epilog
+ %and0 = and i64 %a, %b
+ %not.a = xor i64 %a, -1
+ %and1 = and i64 %not.a, %mask
+ %bitselect = or i64 %and0, %and1
+ %cast = bitcast i64 %bitselect to <2 x float>
+ ret <2 x float> %cast
+}
+
define i64 @v_bitselect_i64_pat_1(i64 %a, i64 %b, i64 %mask) {
; GFX7-LABEL: v_bitselect_i64_pat_1:
; GFX7: ; %bb.0:
@@ -298,17 +648,99 @@ define i64 @v_bitselect_i64_pat_1(i64 %a, i64 %b, i64 %mask) {
; GFX10-NEXT: v_bfi_b32 v0, v2, v0, v4
; GFX10-NEXT: v_bfi_b32 v1, v3, v1, v5
; GFX10-NEXT: s_setpc_b64 s[30:31]
-;
-; R600-LABEL: v_bitselect_i64_pat_1:
-; R600: ; %bb.0:
-; R600-NEXT: CF_END
-; R600-NEXT: PAD
%xor.0 = xor i64 %a, %mask
%and = and i64 %xor.0, %b
%bitselect = xor i64 %and, %mask
ret i64 %bitselect
}
+define amdgpu_ps <2 x float> @v_s_s_bitselect_i64_pat_1(i64 %a, i64 inreg %b, i64 inreg %mask) {
+; GFX7-LABEL: v_s_s_bitselect_i64_pat_1:
+; GFX7: ; %bb.0:
+; GFX7-NEXT: v_mov_b32_e32 v2, s3
+; GFX7-NEXT: v_bfi_b32 v1, s1, v1, v2
+; GFX7-NEXT: v_mov_b32_e32 v2, s2
+; GFX7-NEXT: v_bfi_b32 v0, s0, v0, v2
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: v_s_s_bitselect_i64_pat_1:
+; GFX8: ; %bb.0:
+; GFX8-NEXT: v_mov_b32_e32 v2, s3
+; GFX8-NEXT: v_bfi_b32 v1, s1, v1, v2
+; GFX8-NEXT: v_mov_b32_e32 v2, s2
+; GFX8-NEXT: v_bfi_b32 v0, s0, v0, v2
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: v_s_s_bitselect_i64_pat_1:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: v_bfi_b32 v0, s0, v0, s2
+; GFX10-NEXT: v_bfi_b32 v1, s1, v1, s3
+; GFX10-NEXT: ; return to shader part epilog
+ %xor.0 = xor i64 %a, %mask
+ %and = and i64 %xor.0, %b
+ %bitselect = xor i64 %and, %mask
+ %cast = bitcast i64 %bitselect to <2 x float>
+ ret <2 x float> %cast
+}
+
+define amdgpu_ps <2 x float> @s_s_v_bitselect_i64_pat_1(i64 inreg %a, i64 inreg %b, i64 %mask) {
+; GFX7-LABEL: s_s_v_bitselect_i64_pat_1:
+; GFX7: ; %bb.0:
+; GFX7-NEXT: v_mov_b32_e32 v2, s1
+; GFX7-NEXT: v_bfi_b32 v1, s3, v2, v1
+; GFX7-NEXT: v_mov_b32_e32 v2, s0
+; GFX7-NEXT: v_bfi_b32 v0, s2, v2, v0
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: s_s_v_bitselect_i64_pat_1:
+; GFX8: ; %bb.0:
+; GFX8-NEXT: v_mov_b32_e32 v2, s1
+; GFX8-NEXT: v_bfi_b32 v1, s3, v2, v1
+; GFX8-NEXT: v_mov_b32_e32 v2, s0
+; GFX8-NEXT: v_bfi_b32 v0, s2, v2, v0
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: s_s_v_bitselect_i64_pat_1:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: v_bfi_b32 v0, s2, s0, v0
+; GFX10-NEXT: v_bfi_b32 v1, s3, s1, v1
+; GFX10-NEXT: ; return to shader part epilog
+ %xor.0 = xor i64 %a, %mask
+ %and = and i64 %xor.0, %b
+ %bitselect = xor i64 %and, %mask
+ %cast = bitcast i64 %bitselect to <2 x float>
+ ret <2 x float> %cast
+}
+
+define amdgpu_ps <2 x float> @s_v_s_bitselect_i64_pat_1(i64 inreg %a, i64 %b, i64 inreg %mask) {
+; GFX7-LABEL: s_v_s_bitselect_i64_pat_1:
+; GFX7: ; %bb.0:
+; GFX7-NEXT: v_mov_b32_e32 v2, s3
+; GFX7-NEXT: v_bfi_b32 v1, v1, s1, v2
+; GFX7-NEXT: v_mov_b32_e32 v2, s2
+; GFX7-NEXT: v_bfi_b32 v0, v0, s0, v2
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: s_v_s_bitselect_i64_pat_1:
+; GFX8: ; %bb.0:
+; GFX8-NEXT: v_mov_b32_e32 v2, s3
+; GFX8-NEXT: v_bfi_b32 v1, v1, s1, v2
+; GFX8-NEXT: v_mov_b32_e32 v2, s2
+; GFX8-NEXT: v_bfi_b32 v0, v0, s0, v2
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: s_v_s_bitselect_i64_pat_1:
+; GFX10: ; %bb.0:
+; GFX10-NEXT: v_bfi_b32 v0, v0, s0, s2
+; GFX10-NEXT: v_bfi_b32 v1, v1, s1, s3
+; GFX10-NEXT: ; return to shader part epilog
+ %xor.0 = xor i64 %a, %mask
+ %and = and i64 %xor.0, %b
+ %bitselect = xor i64 %and, %mask
+ %cast = bitcast i64 %bitselect to <2 x float>
+ ret <2 x float> %cast
+}
+
define i64 @v_bitselect_i64_pat_2(i64 %a, i64 %b, i64 %mask) {
; GFX7-LABEL: v_bitselect_i64_pat_2:
; GFX7: ; %bb.0:
@@ -331,11 +763,6 @@ define i64 @v_bitselect_i64_pat_2(i64 %a, i64 %b, i64 %mask) {
; GFX10-NEXT: v_bfi_b32 v0, v2, v0, v4
; GFX10-NEXT: v_bfi_b32 v1, v3, v1, v5
; GFX10-NEXT: s_setpc_b64 s[30:31]
-;
-; R600-LABEL: v_bitselect_i64_pat_2:
-; R600: ; %bb.0:
-; R600-NEXT: CF_END
-; R600-NEXT: PAD
%xor.0 = xor i64 %a, %mask
%and = and i64 %xor.0, %b
%bitselect = xor i64 %and, %mask
@@ -370,11 +797,6 @@ define i64 @v_bfi_sha256_ma_i64(i64 %x, i64 %y, i64 %z) {
; GFX10-NEXT: v_bfi_b32 v0, v0, v4, v2
; GFX10-NEXT: v_bfi_b32 v1, v1, v5, v3
; GFX10-NEXT: s_setpc_b64 s[30:31]
-;
-; R600-LABEL: v_bfi_sha256_ma_i64:
-; R600: ; %bb.0: ; %entry
-; R600-NEXT: CF_END
-; R600-NEXT: PAD
entry:
%and0 = and i64 %x, %z
%or0 = or i64 %x, %z
@@ -383,6 +805,146 @@ entry:
ret i64 %or1
}
+define amdgpu_ps <2 x float> @v_s_s_bfi_sha256_ma_i64(i64 %x, i64 inreg %y, i64 inreg %z) {
+; GFX7-LABEL: v_s_s_bfi_sha256_ma_i64:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: v_xor_b32_e32 v1, s1, v1
+; GFX7-NEXT: v_mov_b32_e32 v2, s1
+; GFX7-NEXT: v_bfi_b32 v1, v1, s3, v2
+; GFX7-NEXT: v_xor_b32_e32 v0, s0, v0
+; GFX7-NEXT: v_mov_b32_e32 v2, s0
+; GFX7-NEXT: v_bfi_b32 v0, v0, s2, v2
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: v_s_s_bfi_sha256_ma_i64:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: v_xor_b32_e32 v1, s1, v1
+; GFX8-NEXT: v_mov_b32_e32 v2, s1
+; GFX8-NEXT: v_bfi_b32 v1, v1, s3, v2
+; GFX8-NEXT: v_xor_b32_e32 v0, s0, v0
+; GFX8-NEXT: v_mov_b32_e32 v2, s0
+; GFX8-NEXT: v_bfi_b32 v0, v0, s2, v2
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: v_s_s_bfi_sha256_ma_i64:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: v_xor_b32_e32 v0, s0, v0
+; GFX10-NEXT: v_xor_b32_e32 v1, s1, v1
+; GFX10-NEXT: v_bfi_b32 v0, v0, s2, s0
+; GFX10-NEXT: v_bfi_b32 v1, v1, s3, s1
+; GFX10-NEXT: ; return to shader part epilog
+entry:
+ %and0 = and i64 %x, %z
+ %or0 = or i64 %x, %z
+ %and1 = and i64 %y, %or0
+ %or1 = or i64 %and0, %and1
+ %cast = bitcast i64 %or1 to <2 x float>
+ ret <2 x float> %cast
+}
+
+define amdgpu_ps <2 x float> @s_v_s_bfi_sha256_ma_i64(i64 inreg %x, i64 %y, i64 inreg %z) {
+; GFX7-LABEL: s_v_s_bfi_sha256_ma_i64:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: v_xor_b32_e32 v2, s1, v1
+; GFX7-NEXT: v_bfi_b32 v1, v2, s3, v1
+; GFX7-NEXT: v_xor_b32_e32 v2, s0, v0
+; GFX7-NEXT: v_bfi_b32 v0, v2, s2, v0
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: s_v_s_bfi_sha256_ma_i64:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: v_xor_b32_e32 v2, s1, v1
+; GFX8-NEXT: v_bfi_b32 v1, v2, s3, v1
+; GFX8-NEXT: v_xor_b32_e32 v2, s0, v0
+; GFX8-NEXT: v_bfi_b32 v0, v2, s2, v0
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: s_v_s_bfi_sha256_ma_i64:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: v_xor_b32_e32 v2, s0, v0
+; GFX10-NEXT: v_xor_b32_e32 v3, s1, v1
+; GFX10-NEXT: v_bfi_b32 v0, v2, s2, v0
+; GFX10-NEXT: v_bfi_b32 v1, v3, s3, v1
+; GFX10-NEXT: ; return to shader part epilog
+entry:
+ %and0 = and i64 %x, %z
+ %or0 = or i64 %x, %z
+ %and1 = and i64 %y, %or0
+ %or1 = or i64 %and0, %and1
+ %cast = bitcast i64 %or1 to <2 x float>
+ ret <2 x float> %cast
+}
+
+define amdgpu_ps <2 x float> @s_s_v_bfi_sha256_ma_i64(i64 inreg %x, i64 inreg %y, i64 %z) {
+; GFX7-LABEL: s_s_v_bfi_sha256_ma_i64:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: v_mov_b32_e32 v2, s3
+; GFX7-NEXT: v_xor_b32_e32 v2, s1, v2
+; GFX7-NEXT: v_bfi_b32 v1, v2, v1, s3
+; GFX7-NEXT: v_mov_b32_e32 v2, s2
+; GFX7-NEXT: v_xor_b32_e32 v2, s0, v2
+; GFX7-NEXT: v_bfi_b32 v0, v2, v0, s2
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: s_s_v_bfi_sha256_ma_i64:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: v_mov_b32_e32 v2, s3
+; GFX8-NEXT: v_xor_b32_e32 v2, s1, v2
+; GFX8-NEXT: v_bfi_b32 v1, v2, v1, s3
+; GFX8-NEXT: v_mov_b32_e32 v2, s2
+; GFX8-NEXT: v_xor_b32_e32 v2, s0, v2
+; GFX8-NEXT: v_bfi_b32 v0, v2, v0, s2
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: s_s_v_bfi_sha256_ma_i64:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: v_xor_b32_e64 v2, s0, s2
+; GFX10-NEXT: v_xor_b32_e64 v3, s1, s3
+; GFX10-NEXT: v_bfi_b32 v0, v2, v0, s2
+; GFX10-NEXT: v_bfi_b32 v1, v3, v1, s3
+; GFX10-NEXT: ; return to shader part epilog
+entry:
+ %and0 = and i64 %x, %z
+ %or0 = or i64 %x, %z
+ %and1 = and i64 %y, %or0
+ %or1 = or i64 %and0, %and1
+ %cast = bitcast i64 %or1 to <2 x float>
+ ret <2 x float> %cast
+}
+
+define amdgpu_ps <2 x float> @v_s_v_bfi_sha256_ma_i64(i64 %x, i64 inreg %y, i64 %z) {
+; GFX7-LABEL: v_s_v_bfi_sha256_ma_i64:
+; GFX7: ; %bb.0: ; %entry
+; GFX7-NEXT: v_xor_b32_e32 v1, s1, v1
+; GFX7-NEXT: v_xor_b32_e32 v0, s0, v0
+; GFX7-NEXT: v_bfi_b32 v1, v1, v3, s1
+; GFX7-NEXT: v_bfi_b32 v0, v0, v2, s0
+; GFX7-NEXT: ; return to shader part epilog
+;
+; GFX8-LABEL: v_s_v_bfi_sha256_ma_i64:
+; GFX8: ; %bb.0: ; %entry
+; GFX8-NEXT: v_xor_b32_e32 v1, s1, v1
+; GFX8-NEXT: v_xor_b32_e32 v0, s0, v0
+; GFX8-NEXT: v_bfi_b32 v1, v1, v3, s1
+; GFX8-NEXT: v_bfi_b32 v0, v0, v2, s0
+; GFX8-NEXT: ; return to shader part epilog
+;
+; GFX10-LABEL: v_s_v_bfi_sha256_ma_i64:
+; GFX10: ; %bb.0: ; %entry
+; GFX10-NEXT: v_xor_b32_e32 v0, s0, v0
+; GFX10-NEXT: v_xor_b32_e32 v1, s1, v1
+; GFX10-NEXT: v_bfi_b32 v0, v0, v2, s0
+; GFX10-NEXT: v_bfi_b32 v1, v1, v3, s1
+; GFX10-NEXT: ; return to shader part epilog
+entry:
+ %and0 = and i64 %x, %z
+ %or0 = or i64 %x, %z
+ %and1 = and i64 %y, %or0
+ %or1 = or i64 %and0, %and1
+ %cast = bitcast i64 %or1 to <2 x float>
+ ret <2 x float> %cast
+}
+
define amdgpu_kernel void @s_bitselect_i64_pat_0(i64 %a, i64 %b, i64 %mask) {
; GFX7-LABEL: s_bitselect_i64_pat_0:
; GFX7: ; %bb.0:
@@ -431,24 +993,6 @@ define amdgpu_kernel void @s_bitselect_i64_pat_0(i64 %a, i64 %b, i64 %mask) {
; GFX10-NEXT: v_mov_b32_e32 v1, s1
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_endpgm
-;
-; R600-LABEL: s_bitselect_i64_pat_0:
-; R600: ; %bb.0:
-; R600-NEXT: ALU 9, @4, KC0[CB0:0-32], KC1[]
-; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
-; R600-NEXT: CF_END
-; R600-NEXT: PAD
-; R600-NEXT: ALU clause starting at 4:
-; R600-NEXT: MOV * T0.W, KC0[3].Y,
-; R600-NEXT: BFI_INT * T0.W, KC0[2].Y, KC0[2].W, PV.W,
-; R600-NEXT: MOV * T1.W, KC0[3].Z,
-; R600-NEXT: BFI_INT T1.W, KC0[2].Z, KC0[3].X, PV.W,
-; R600-NEXT: ADDC_UINT * T2.W, T0.W, literal.x,
-; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
-; R600-NEXT: ADD_INT * T0.Y, PV.W, PS,
-; R600-NEXT: ADD_INT T0.X, T0.W, literal.x,
-; R600-NEXT: MOV * T1.X, literal.y,
-; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
%and0 = and i64 %a, %b
%not.a = xor i64 %a, -1
%and1 = and i64 %not.a, %mask
@@ -506,24 +1050,6 @@ define amdgpu_kernel void @s_bitselect_i64_pat_1(i64 %a, i64 %b, i64 %mask) {
; GFX10-NEXT: v_mov_b32_e32 v1, s1
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_endpgm
-;
-; R600-LABEL: s_bitselect_i64_pat_1:
-; R600: ; %bb.0:
-; R600-NEXT: ALU 9, @4, KC0[CB0:0-32], KC1[]
-; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
-; R600-NEXT: CF_END
-; R600-NEXT: PAD
-; R600-NEXT: ALU clause starting at 4:
-; R600-NEXT: MOV * T0.W, KC0[3].Y,
-; R600-NEXT: BFI_INT * T0.W, KC0[2].W, KC0[2].Y, PV.W,
-; R600-NEXT: MOV * T1.W, KC0[3].Z,
-; R600-NEXT: BFI_INT T1.W, KC0[3].X, KC0[2].Z, PV.W,
-; R600-NEXT: ADDC_UINT * T2.W, T0.W, literal.x,
-; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
-; R600-NEXT: ADD_INT * T0.Y, PV.W, PS,
-; R600-NEXT: ADD_INT T0.X, T0.W, literal.x,
-; R600-NEXT: MOV * T1.X, literal.y,
-; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
%xor.0 = xor i64 %a, %mask
%and = and i64 %xor.0, %b
%bitselect = xor i64 %and, %mask
@@ -581,24 +1107,6 @@ define amdgpu_kernel void @s_bitselect_i64_pat_2(i64 %a, i64 %b, i64 %mask) {
; GFX10-NEXT: v_mov_b32_e32 v1, s1
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_endpgm
-;
-; R600-LABEL: s_bitselect_i64_pat_2:
-; R600: ; %bb.0:
-; R600-NEXT: ALU 9, @4, KC0[CB0:0-32], KC1[]
-; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
-; R600-NEXT: CF_END
-; R600-NEXT: PAD
-; R600-NEXT: ALU clause starting at 4:
-; R600-NEXT: MOV * T0.W, KC0[3].Y,
-; R600-NEXT: BFI_INT * T0.W, KC0[2].W, KC0[2].Y, PV.W,
-; R600-NEXT: MOV * T1.W, KC0[3].Z,
-; R600-NEXT: BFI_INT T1.W, KC0[3].X, KC0[2].Z, PV.W,
-; R600-NEXT: ADDC_UINT * T2.W, T0.W, literal.x,
-; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
-; R600-NEXT: ADD_INT * T0.Y, PV.W, PS,
-; R600-NEXT: ADD_INT T0.X, T0.W, literal.x,
-; R600-NEXT: MOV * T1.X, literal.y,
-; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
%xor.0 = xor i64 %a, %mask
%and = and i64 %xor.0, %b
%bitselect = xor i64 %and, %mask
@@ -659,24 +1167,6 @@ define amdgpu_kernel void @s_bfi_sha256_ma_i64(i64 %x, i64 %y, i64 %z) {
; GFX10-NEXT: v_mov_b32_e32 v1, s1
; GFX10-NEXT: global_store_dwordx2 v[0:1], v[0:1], off
; GFX10-NEXT: s_endpgm
-;
-; R600-LABEL: s_bfi_sha256_ma_i64:
-; R600: ; %bb.0: ; %entry
-; R600-NEXT: ALU 9, @4, KC0[CB0:0-32], KC1[]
-; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
-; R600-NEXT: CF_END
-; R600-NEXT: PAD
-; R600-NEXT: ALU clause starting at 4:
-; R600-NEXT: XOR_INT * T0.W, KC0[2].Y, KC0[2].W,
-; R600-NEXT: BFI_INT T0.W, PV.W, KC0[3].Y, KC0[2].W,
-; R600-NEXT: XOR_INT * T1.W, KC0[2].Z, KC0[3].X,
-; R600-NEXT: BFI_INT T1.W, PS, KC0[3].Z, KC0[3].X,
-; R600-NEXT: ADDC_UINT * T2.W, PV.W, literal.x,
-; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
-; R600-NEXT: ADD_INT * T0.Y, PV.W, PS,
-; R600-NEXT: ADD_INT T0.X, T0.W, literal.x,
-; R600-NEXT: MOV * T1.X, literal.y,
-; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
entry:
%and0 = and i64 %x, %z
%or0 = or i64 %x, %z
diff --git a/llvm/test/CodeGen/AMDGPU/bfi_int.r600.ll b/llvm/test/CodeGen/AMDGPU/bfi_int.r600.ll
new file mode 100644
index 000000000000..34eb088b16f4
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/bfi_int.r600.ll
@@ -0,0 +1,237 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -enable-var-scope -check-prefixes=R600 %s
+
+; BFI_INT Definition pattern from ISA docs
+; (y & x) | (z & ~x)
+;
+define amdgpu_kernel void @bfi_def(i32 addrspace(1)* %out, i32 %x, i32 %y, i32 %z) {
+; R600-LABEL: bfi_def:
+; R600: ; %bb.0: ; %entry
+; R600-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; R600-NEXT: CF_END
+; R600-NEXT: PAD
+; R600-NEXT: ALU clause starting at 4:
+; R600-NEXT: LSHR * T0.X, KC0[2].Y, literal.x,
+; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; R600-NEXT: BFI_INT * T1.X, KC0[2].Z, KC0[2].W, KC0[3].X,
+entry:
+ %0 = xor i32 %x, -1
+ %1 = and i32 %z, %0
+ %2 = and i32 %y, %x
+ %3 = or i32 %1, %2
+ store i32 %3, i32 addrspace(1)* %out
+ ret void
+}
+
+; SHA-256 Ch function
+; z ^ (x & (y ^ z))
+define amdgpu_kernel void @bfi_sha256_ch(i32 addrspace(1)* %out, i32 %x, i32 %y, i32 %z) {
+; R600-LABEL: bfi_sha256_ch:
+; R600: ; %bb.0: ; %entry
+; R600-NEXT: ALU 2, @4, KC0[CB0:0-32], KC1[]
+; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T1.X, T0.X, 1
+; R600-NEXT: CF_END
+; R600-NEXT: PAD
+; R600-NEXT: ALU clause starting at 4:
+; R600-NEXT: LSHR * T0.X, KC0[2].Y, literal.x,
+; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+; R600-NEXT: BFI_INT * T1.X, KC0[2].Z, KC0[2].W, KC0[3].X,
+entry:
+ %0 = xor i32 %y, %z
+ %1 = and i32 %x, %0
+ %2 = xor i32 %z, %1
+ store i32 %2, i32 addrspace(1)* %out
+ ret void
+}
+
+; SHA-256 Ma function
+; ((x & z) | (y & (x | z)))
+define amdgpu_kernel void @bfi_sha256_ma(i32 addrspace(1)* %out, i32 %x, i32 %y, i32 %z) {
+; R600-LABEL: bfi_sha256_ma:
+; R600: ; %bb.0: ; %entry
+; R600-NEXT: ALU 3, @4, KC0[CB0:0-32], KC1[]
+; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.X, T1.X, 1
+; R600-NEXT: CF_END
+; R600-NEXT: PAD
+; R600-NEXT: ALU clause starting at 4:
+; R600-NEXT: XOR_INT * T0.W, KC0[2].Z, KC0[2].W,
+; R600-NEXT: BFI_INT * T0.X, PV.W, KC0[3].X, KC0[2].W,
+; R600-NEXT: LSHR * T1.X, KC0[2].Y, literal.x,
+; R600-NEXT: 2(2.802597e-45), 0(0.000000e+00)
+entry:
+ %0 = and i32 %x, %z
+ %1 = or i32 %x, %z
+ %2 = and i32 %y, %1
+ %3 = or i32 %0, %2
+ store i32 %3, i32 addrspace(1)* %out
+ ret void
+}
+
+define <2 x i32> @v_bitselect_v2i32_pat1(<2 x i32> %a, <2 x i32> %b, <2 x i32> %mask) {
+; R600-LABEL: v_bitselect_v2i32_pat1:
+; R600: ; %bb.0:
+; R600-NEXT: CF_END
+; R600-NEXT: PAD
+ %xor.0 = xor <2 x i32> %a, %mask
+ %and = and <2 x i32> %xor.0, %b
+ %bitselect = xor <2 x i32> %and, %mask
+ ret <2 x i32> %bitselect
+}
+
+define i64 @v_bitselect_i64_pat_0(i64 %a, i64 %b, i64 %mask) {
+; R600-LABEL: v_bitselect_i64_pat_0:
+; R600: ; %bb.0:
+; R600-NEXT: CF_END
+; R600-NEXT: PAD
+ %and0 = and i64 %a, %b
+ %not.a = xor i64 %a, -1
+ %and1 = and i64 %not.a, %mask
+ %bitselect = or i64 %and0, %and1
+ ret i64 %bitselect
+}
+
+define i64 @v_bitselect_i64_pat_1(i64 %a, i64 %b, i64 %mask) {
+; R600-LABEL: v_bitselect_i64_pat_1:
+; R600: ; %bb.0:
+; R600-NEXT: CF_END
+; R600-NEXT: PAD
+ %xor.0 = xor i64 %a, %mask
+ %and = and i64 %xor.0, %b
+ %bitselect = xor i64 %and, %mask
+ ret i64 %bitselect
+}
+
+define i64 @v_bitselect_i64_pat_2(i64 %a, i64 %b, i64 %mask) {
+; R600-LABEL: v_bitselect_i64_pat_2:
+; R600: ; %bb.0:
+; R600-NEXT: CF_END
+; R600-NEXT: PAD
+ %xor.0 = xor i64 %a, %mask
+ %and = and i64 %xor.0, %b
+ %bitselect = xor i64 %and, %mask
+ ret i64 %bitselect
+}
+
+define i64 @v_bfi_sha256_ma_i64(i64 %x, i64 %y, i64 %z) {
+; R600-LABEL: v_bfi_sha256_ma_i64:
+; R600: ; %bb.0: ; %entry
+; R600-NEXT: CF_END
+; R600-NEXT: PAD
+entry:
+ %and0 = and i64 %x, %z
+ %or0 = or i64 %x, %z
+ %and1 = and i64 %y, %or0
+ %or1 = or i64 %and0, %and1
+ ret i64 %or1
+}
+
+define amdgpu_kernel void @s_bitselect_i64_pat_0(i64 %a, i64 %b, i64 %mask) {
+; R600-LABEL: s_bitselect_i64_pat_0:
+; R600: ; %bb.0:
+; R600-NEXT: ALU 9, @4, KC0[CB0:0-32], KC1[]
+; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; R600-NEXT: CF_END
+; R600-NEXT: PAD
+; R600-NEXT: ALU clause starting at 4:
+; R600-NEXT: MOV * T0.W, KC0[3].Y,
+; R600-NEXT: BFI_INT * T0.W, KC0[2].Y, KC0[2].W, PV.W,
+; R600-NEXT: MOV * T1.W, KC0[3].Z,
+; R600-NEXT: BFI_INT T1.W, KC0[2].Z, KC0[3].X, PV.W,
+; R600-NEXT: ADDC_UINT * T2.W, T0.W, literal.x,
+; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
+; R600-NEXT: ADD_INT * T0.Y, PV.W, PS,
+; R600-NEXT: ADD_INT T0.X, T0.W, literal.x,
+; R600-NEXT: MOV * T1.X, literal.y,
+; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
+ %and0 = and i64 %a, %b
+ %not.a = xor i64 %a, -1
+ %and1 = and i64 %not.a, %mask
+ %bitselect = or i64 %and0, %and1
+ %scalar.use = add i64 %bitselect, 10
+ store i64 %scalar.use, i64 addrspace(1)* undef
+ ret void
+}
+
+define amdgpu_kernel void @s_bitselect_i64_pat_1(i64 %a, i64 %b, i64 %mask) {
+; R600-LABEL: s_bitselect_i64_pat_1:
+; R600: ; %bb.0:
+; R600-NEXT: ALU 9, @4, KC0[CB0:0-32], KC1[]
+; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; R600-NEXT: CF_END
+; R600-NEXT: PAD
+; R600-NEXT: ALU clause starting at 4:
+; R600-NEXT: MOV * T0.W, KC0[3].Y,
+; R600-NEXT: BFI_INT * T0.W, KC0[2].W, KC0[2].Y, PV.W,
+; R600-NEXT: MOV * T1.W, KC0[3].Z,
+; R600-NEXT: BFI_INT T1.W, KC0[3].X, KC0[2].Z, PV.W,
+; R600-NEXT: ADDC_UINT * T2.W, T0.W, literal.x,
+; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
+; R600-NEXT: ADD_INT * T0.Y, PV.W, PS,
+; R600-NEXT: ADD_INT T0.X, T0.W, literal.x,
+; R600-NEXT: MOV * T1.X, literal.y,
+; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
+ %xor.0 = xor i64 %a, %mask
+ %and = and i64 %xor.0, %b
+ %bitselect = xor i64 %and, %mask
+
+ %scalar.use = add i64 %bitselect, 10
+ store i64 %scalar.use, i64 addrspace(1)* undef
+ ret void
+}
+
+define amdgpu_kernel void @s_bitselect_i64_pat_2(i64 %a, i64 %b, i64 %mask) {
+; R600-LABEL: s_bitselect_i64_pat_2:
+; R600: ; %bb.0:
+; R600-NEXT: ALU 9, @4, KC0[CB0:0-32], KC1[]
+; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; R600-NEXT: CF_END
+; R600-NEXT: PAD
+; R600-NEXT: ALU clause starting at 4:
+; R600-NEXT: MOV * T0.W, KC0[3].Y,
+; R600-NEXT: BFI_INT * T0.W, KC0[2].W, KC0[2].Y, PV.W,
+; R600-NEXT: MOV * T1.W, KC0[3].Z,
+; R600-NEXT: BFI_INT T1.W, KC0[3].X, KC0[2].Z, PV.W,
+; R600-NEXT: ADDC_UINT * T2.W, T0.W, literal.x,
+; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
+; R600-NEXT: ADD_INT * T0.Y, PV.W, PS,
+; R600-NEXT: ADD_INT T0.X, T0.W, literal.x,
+; R600-NEXT: MOV * T1.X, literal.y,
+; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
+ %xor.0 = xor i64 %a, %mask
+ %and = and i64 %xor.0, %b
+ %bitselect = xor i64 %and, %mask
+
+ %scalar.use = add i64 %bitselect, 10
+ store i64 %scalar.use, i64 addrspace(1)* undef
+ ret void
+}
+
+define amdgpu_kernel void @s_bfi_sha256_ma_i64(i64 %x, i64 %y, i64 %z) {
+; R600-LABEL: s_bfi_sha256_ma_i64:
+; R600: ; %bb.0: ; %entry
+; R600-NEXT: ALU 9, @4, KC0[CB0:0-32], KC1[]
+; R600-NEXT: MEM_RAT_CACHELESS STORE_RAW T0.XY, T1.X, 1
+; R600-NEXT: CF_END
+; R600-NEXT: PAD
+; R600-NEXT: ALU clause starting at 4:
+; R600-NEXT: XOR_INT * T0.W, KC0[2].Y, KC0[2].W,
+; R600-NEXT: BFI_INT T0.W, PV.W, KC0[3].Y, KC0[2].W,
+; R600-NEXT: XOR_INT * T1.W, KC0[2].Z, KC0[3].X,
+; R600-NEXT: BFI_INT T1.W, PS, KC0[3].Z, KC0[3].X,
+; R600-NEXT: ADDC_UINT * T2.W, PV.W, literal.x,
+; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
+; R600-NEXT: ADD_INT * T0.Y, PV.W, PS,
+; R600-NEXT: ADD_INT T0.X, T0.W, literal.x,
+; R600-NEXT: MOV * T1.X, literal.y,
+; R600-NEXT: 10(1.401298e-44), 0(0.000000e+00)
+entry:
+ %and0 = and i64 %x, %z
+ %or0 = or i64 %x, %z
+ %and1 = and i64 %y, %or0
+ %or1 = or i64 %and0, %and1
+
+ %scalar.use = add i64 %or1, 10
+ store i64 %scalar.use, i64 addrspace(1)* undef
+ ret void
+}
More information about the llvm-commits
mailing list