| ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 | |
| ; RUN: llc -mtriple=amdgpu9.00-amd-amdhsa < %s | FileCheck -check-prefix=GFX9 %s | |
| ; RUN: llc -mtriple=amdgpu12.00-amd-amdhsa -mattr=+real-true16 < %s | FileCheck -check-prefixes=GFX12,GFX12-TRUE16 %s | |
| ; RUN: llc -mtriple=amdgpu12.00-amd-amdhsa -mattr=-real-true16 < %s | FileCheck -check-prefixes=GFX12,GFX12-FAKE16 %s | |
| ; Make sure stack use isn't introduced for these bitcasts. | |
| define i160 @bitcast_v5i32_to_i160(<5 x i32> %vec) { | |
| ; GFX9-LABEL: bitcast_v5i32_to_i160: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_v5i32_to_i160: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast <5 x i32> %vec to i160 | |
| ret i160 %bitcast | |
| } | |
| define i192 @bitcast_v6i32_to_i192(<6 x i32> %vec) { | |
| ; GFX9-LABEL: bitcast_v6i32_to_i192: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_v6i32_to_i192: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast <6 x i32> %vec to i192 | |
| ret i192 %bitcast | |
| } | |
| define i224 @bitcast_v7i32_to_i224(<7 x i32> %vec) { | |
| ; GFX9-LABEL: bitcast_v7i32_to_i224: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_v7i32_to_i224: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast <7 x i32> %vec to i224 | |
| ret i224 %bitcast | |
| } | |
| define i256 @bitcast_v8i32_to_i256(<8 x i32> %vec) { | |
| ; GFX9-LABEL: bitcast_v8i32_to_i256: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_v8i32_to_i256: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast <8 x i32> %vec to i256 | |
| ret i256 %bitcast | |
| } | |
| define <5 x i32> @bitcast_i160_to_v5i32(i160 %int) { | |
| ; GFX9-LABEL: bitcast_i160_to_v5i32: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_i160_to_v5i32: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast i160 %int to <5 x i32> | |
| ret <5 x i32> %bitcast | |
| } | |
| define <6 x i32> @bitcast_i192_to_v6i32(i192 %int) { | |
| ; GFX9-LABEL: bitcast_i192_to_v6i32: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_i192_to_v6i32: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast i192 %int to <6 x i32> | |
| ret <6 x i32> %bitcast | |
| } | |
| define <7 x i32> @bitcast_i224_to_v7i32(i224 %int) { | |
| ; GFX9-LABEL: bitcast_i224_to_v7i32: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_i224_to_v7i32: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast i224 %int to <7 x i32> | |
| ret <7 x i32> %bitcast | |
| } | |
| define <8 x i32> @bitcast_i256_to_v8i32(i256 %int) { | |
| ; GFX9-LABEL: bitcast_i256_to_v8i32: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_i256_to_v8i32: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast i256 %int to <8 x i32> | |
| ret <8 x i32> %bitcast | |
| } | |
| define i192 @bitcast_v3i64_to_i192(<3 x i64> %vec) { | |
| ; GFX9-LABEL: bitcast_v3i64_to_i192: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_v3i64_to_i192: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast <3 x i64> %vec to i192 | |
| ret i192 %bitcast | |
| } | |
| define <3 x i64> @bitcast_i192_to_v3i64(i192 %int) { | |
| ; GFX9-LABEL: bitcast_i192_to_v3i64: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_i192_to_v3i64: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast i192 %int to <3 x i64> | |
| ret <3 x i64> %bitcast | |
| } | |
| define <10 x i16> @bitcast_i160_to_v10i16(i160 %int) { | |
| ; GFX9-LABEL: bitcast_i160_to_v10i16: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_mov_b32 s4, 0xffff | |
| ; GFX9-NEXT: v_bfi_b32 v0, s4, v0, v0 | |
| ; GFX9-NEXT: v_bfi_b32 v2, s4, v2, v2 | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-TRUE16-LABEL: bitcast_i160_to_v10i16: | |
| ; GFX12-TRUE16: ; %bb.0: | |
| ; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-FAKE16-LABEL: bitcast_i160_to_v10i16: | |
| ; GFX12-FAKE16: ; %bb.0: | |
| ; GFX12-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: v_bfi_b32 v0, 0xffff, v0, v0 | |
| ; GFX12-FAKE16-NEXT: v_bfi_b32 v2, 0xffff, v2, v2 | |
| ; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast i160 %int to <10 x i16> | |
| ret <10 x i16> %bitcast | |
| } | |
| define i160 @bitcast_v10i16_to_i160(<10 x i16> %vec) { | |
| ; GFX9-LABEL: bitcast_v10i16_to_i160: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_v10i16_to_i160: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast <10 x i16> %vec to i160 | |
| ret i160 %bitcast | |
| } | |
| define i12 @bitcast_v2i6_to_i12(<2 x i6> %vec) { | |
| ; GFX9-LABEL: bitcast_v2i6_to_i12: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: v_lshlrev_b16_e32 v1, 6, v1 | |
| ; GFX9-NEXT: v_and_b32_e32 v0, 63, v0 | |
| ; GFX9-NEXT: v_or_b32_e32 v0, v0, v1 | |
| ; GFX9-NEXT: v_and_b32_e32 v0, 0xfff, v0 | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-TRUE16-LABEL: bitcast_v2i6_to_i12: | |
| ; GFX12-TRUE16: ; %bb.0: | |
| ; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b16 v0.h, 6, v1.l | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v0.l, v0.l, 63 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) | |
| ; GFX12-TRUE16-NEXT: v_or_b16 v0.l, v0.l, v0.h | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v0.l, v0.l, 0xfff | |
| ; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-FAKE16-LABEL: bitcast_v2i6_to_i12: | |
| ; GFX12-FAKE16: ; %bb.0: | |
| ; GFX12-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b16 v1, 6, v1 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v0, 63, v0 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) | |
| ; GFX12-FAKE16-NEXT: v_or_b32_e32 v0, v0, v1 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v0, 0xfff, v0 | |
| ; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast <2 x i6> %vec to i12 | |
| ret i12 %bitcast | |
| } | |
| define <2 x i6> @bitcast_i12_to_v2i6(i12 %int) { | |
| ; GFX9-LABEL: bitcast_i12_to_v2i6: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: v_and_b32_e32 v2, 63, v0 | |
| ; GFX9-NEXT: v_bfe_u32 v1, v0, 6, 6 | |
| ; GFX9-NEXT: v_mov_b32_e32 v0, v2 | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-TRUE16-LABEL: bitcast_i12_to_v2i6: | |
| ; GFX12-TRUE16: ; %bb.0: | |
| ; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: v_lshrrev_b16 v0.h, 6, v0.l | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v0.l, v0.l, 63 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v1.l, v0.h, 63 | |
| ; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-FAKE16-LABEL: bitcast_i12_to_v2i6: | |
| ; GFX12-FAKE16: ; %bb.0: | |
| ; GFX12-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v2, 63, v0 | |
| ; GFX12-FAKE16-NEXT: v_bfe_u32 v1, v0, 6, 6 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | |
| ; GFX12-FAKE16-NEXT: v_mov_b32_e32 v0, v2 | |
| ; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast i12 %int to <2 x i6> | |
| ret <2 x i6> %bitcast | |
| } | |
| define i160 @bitcast_v5f32_to_i160(<5 x float> %vec) { | |
| ; GFX9-LABEL: bitcast_v5f32_to_i160: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_v5f32_to_i160: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast <5 x float> %vec to i160 | |
| ret i160 %bitcast | |
| } | |
| define <5 x float> @bitcast_i160_to_v5f32(i160 %int) { | |
| ; GFX9-LABEL: bitcast_i160_to_v5f32: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_i160_to_v5f32: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast i160 %int to <5 x float> | |
| ret <5 x float> %bitcast | |
| } | |
| define <6 x float> @bitcast_i192_to_v6f32(i192 %int) { | |
| ; GFX9-LABEL: bitcast_i192_to_v6f32: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_i192_to_v6f32: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast i192 %int to <6 x float> | |
| ret <6 x float> %bitcast | |
| } | |
| define i192 @bitcast_v6f32_to_i192(<6 x float> %vec) { | |
| ; GFX9-LABEL: bitcast_v6f32_to_i192: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-LABEL: bitcast_v6f32_to_i192: | |
| ; GFX12: ; %bb.0: | |
| ; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast <6 x float> %vec to i192 | |
| ret i192 %bitcast | |
| } | |
| ; Pack the original bits after promoting each input element. | |
| define <2 x i32> @bitcast_v16i4_to_v2i32(<16 x i4> %vec) { | |
| ; GFX9-LABEL: bitcast_v16i4_to_v2i32: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: v_and_b32_e32 v1, 15, v1 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v1, 4, v1 | |
| ; GFX9-NEXT: v_and_or_b32 v0, v0, 15, v1 | |
| ; GFX9-NEXT: v_and_b32_e32 v1, 15, v3 | |
| ; GFX9-NEXT: v_and_b32_e32 v2, 15, v2 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v1, 12, v1 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v2 | |
| ; GFX9-NEXT: v_or3_b32 v0, v0, v2, v1 | |
| ; GFX9-NEXT: v_and_b32_e32 v1, 15, v5 | |
| ; GFX9-NEXT: v_mov_b32_e32 v2, 15 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v1, 20, v1 | |
| ; GFX9-NEXT: v_and_b32_sdwa v3, v4, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD | |
| ; GFX9-NEXT: v_or3_b32 v0, v0, v3, v1 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v1, 28, v7 | |
| ; GFX9-NEXT: v_and_b32_sdwa v3, v6, v2 dst_sel:BYTE_3 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD | |
| ; GFX9-NEXT: v_or3_b32 v0, v0, v3, v1 | |
| ; GFX9-NEXT: v_and_b32_e32 v1, 15, v9 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v1, 4, v1 | |
| ; GFX9-NEXT: v_and_b32_e32 v3, 15, v11 | |
| ; GFX9-NEXT: v_and_b32_e32 v4, 15, v10 | |
| ; GFX9-NEXT: v_and_or_b32 v1, v8, 15, v1 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v3, 12, v3 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v4, 8, v4 | |
| ; GFX9-NEXT: v_or3_b32 v1, v1, v4, v3 | |
| ; GFX9-NEXT: v_and_b32_e32 v3, 15, v13 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v3, 20, v3 | |
| ; GFX9-NEXT: v_and_b32_sdwa v4, v12, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD | |
| ; GFX9-NEXT: v_or3_b32 v1, v1, v4, v3 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v3, 28, v15 | |
| ; GFX9-NEXT: v_and_b32_sdwa v2, v14, v2 dst_sel:BYTE_3 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD | |
| ; GFX9-NEXT: v_or3_b32 v1, v1, v2, v3 | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-TRUE16-LABEL: bitcast_v16i4_to_v2i32: | |
| ; GFX12-TRUE16: ; %bb.0: | |
| ; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v0.l, v0.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v0.h, v1.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v1.l, v3.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v1.h, v2.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v4.l, v4.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v2, v0.l | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v0, v0.h | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v3, v1.l | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v1, v1.h | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v6.l, v6.l, 15 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) | |
| ; GFX12-TRUE16-NEXT: v_lshl_or_b32 v0, v0, 4, v2 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v2, 12, v3 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_3) | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 8, v1 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v3.l, v5.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v6, 24, v6 | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v0, v0, v1, v2 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_3) | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 20, v3 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v2, 16, v4 | |
| ; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.l, v7.l | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v7.l, v12.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v2, v0, v2, v1 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v0.l, v8.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v0.h, v9.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v1.l, v11.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v1.h, v10.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 28, v3 | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v4, v0.l | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v0, v0.h | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v5, v1.l | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v1, v1.h | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) | |
| ; GFX12-TRUE16-NEXT: v_lshl_or_b32 v0, v0, 4, v4 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 12, v5 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2) | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 8, v1 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v5.l, v13.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v0, v0, v1, v4 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_3) | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 20, v5 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 16, v7 | |
| ; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.l, v15.l | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v7.l, v14.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v1, v0, v4, v1 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 28, v5 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, 24, v7 | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v0, v2, v6, v3 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v1, v1, v5, v4 | |
| ; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-FAKE16-LABEL: bitcast_v16i4_to_v2i32: | |
| ; GFX12-FAKE16: ; %bb.0: | |
| ; GFX12-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v1, 15, v1 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v3, 15, v3 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v2, 15, v2 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v5, 15, v5 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v4, 15, v4 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 4, v1 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v6, 15, v6 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 8, v2 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 20, v5 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4 | |
| ; GFX12-FAKE16-NEXT: v_and_or_b32 v0, v0, 15, v1 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 12, v3 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v3, 15, v9 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v9, 15, v12 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v6, 24, v6 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v0, v0, v2, v1 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 4, v3 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v2, 15, v11 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v3, 15, v10 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v0, v0, v4, v5 | |
| ; GFX12-FAKE16-NEXT: v_and_or_b32 v1, v8, 15, v1 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 12, v2 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 8, v3 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v8, 15, v13 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v4, 15, v14 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 28, v7 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v1, v1, v3, v2 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 20, v8 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 16, v9 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2) | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v0, v0, v6, v5 | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v1, v1, v3, v2 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 28, v15 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 24, v4 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v1, v1, v3, v2 | |
| ; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast <16 x i4> %vec to <2 x i32> | |
| ret <2 x i32> %bitcast | |
| } | |
| define <2 x float> @bitcast_v16i4_to_v2f32(<16 x i4> %vec) { | |
| ; GFX9-LABEL: bitcast_v16i4_to_v2f32: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: v_and_b32_e32 v1, 15, v1 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v1, 4, v1 | |
| ; GFX9-NEXT: v_and_or_b32 v0, v0, 15, v1 | |
| ; GFX9-NEXT: v_and_b32_e32 v1, 15, v3 | |
| ; GFX9-NEXT: v_and_b32_e32 v2, 15, v2 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v1, 12, v1 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v2, 8, v2 | |
| ; GFX9-NEXT: v_or3_b32 v0, v0, v2, v1 | |
| ; GFX9-NEXT: v_and_b32_e32 v1, 15, v5 | |
| ; GFX9-NEXT: v_mov_b32_e32 v2, 15 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v1, 20, v1 | |
| ; GFX9-NEXT: v_and_b32_sdwa v3, v4, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD | |
| ; GFX9-NEXT: v_or3_b32 v0, v0, v3, v1 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v1, 28, v7 | |
| ; GFX9-NEXT: v_and_b32_sdwa v3, v6, v2 dst_sel:BYTE_3 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD | |
| ; GFX9-NEXT: v_or3_b32 v0, v0, v3, v1 | |
| ; GFX9-NEXT: v_and_b32_e32 v1, 15, v9 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v1, 4, v1 | |
| ; GFX9-NEXT: v_and_b32_e32 v3, 15, v11 | |
| ; GFX9-NEXT: v_and_b32_e32 v4, 15, v10 | |
| ; GFX9-NEXT: v_and_or_b32 v1, v8, 15, v1 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v3, 12, v3 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v4, 8, v4 | |
| ; GFX9-NEXT: v_or3_b32 v1, v1, v4, v3 | |
| ; GFX9-NEXT: v_and_b32_e32 v3, 15, v13 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v3, 20, v3 | |
| ; GFX9-NEXT: v_and_b32_sdwa v4, v12, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD | |
| ; GFX9-NEXT: v_or3_b32 v1, v1, v4, v3 | |
| ; GFX9-NEXT: v_lshlrev_b32_e32 v3, 28, v15 | |
| ; GFX9-NEXT: v_and_b32_sdwa v2, v14, v2 dst_sel:BYTE_3 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD | |
| ; GFX9-NEXT: v_or3_b32 v1, v1, v2, v3 | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-TRUE16-LABEL: bitcast_v16i4_to_v2f32: | |
| ; GFX12-TRUE16: ; %bb.0: | |
| ; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v0.l, v0.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v0.h, v1.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v1.l, v3.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v1.h, v2.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v4.l, v4.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v2, v0.l | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v0, v0.h | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v3, v1.l | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v1, v1.h | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v6.l, v6.l, 15 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) | |
| ; GFX12-TRUE16-NEXT: v_lshl_or_b32 v0, v0, 4, v2 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v2, 12, v3 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_3) | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 8, v1 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v3.l, v5.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v6, 24, v6 | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v0, v0, v1, v2 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_3) | instid1(VALU_DEP_3) | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 20, v3 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v2, 16, v4 | |
| ; GFX12-TRUE16-NEXT: v_mov_b16_e32 v3.l, v7.l | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v7.l, v12.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v2, v0, v2, v1 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v0.l, v8.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v0.h, v9.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v1.l, v11.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v1.h, v10.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v3, 28, v3 | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v4, v0.l | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v0, v0.h | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v5, v1.l | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v1, v1.h | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) | |
| ; GFX12-TRUE16-NEXT: v_lshl_or_b32 v0, v0, 4, v4 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 12, v5 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_2) | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 8, v1 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v5.l, v13.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v0, v0, v1, v4 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_3) | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v1, 20, v5 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 16, v7 | |
| ; GFX12-TRUE16-NEXT: v_mov_b16_e32 v5.l, v15.l | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v7.l, v14.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v1, v0, v4, v1 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v4, 28, v5 | |
| ; GFX12-TRUE16-NEXT: v_lshlrev_b32_e32 v5, 24, v7 | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v0, v2, v6, v3 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_2) | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v1, v1, v5, v4 | |
| ; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-FAKE16-LABEL: bitcast_v16i4_to_v2f32: | |
| ; GFX12-FAKE16: ; %bb.0: | |
| ; GFX12-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v1, 15, v1 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v3, 15, v3 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v2, 15, v2 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v5, 15, v5 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v4, 15, v4 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 4, v1 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v6, 15, v6 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 8, v2 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 20, v5 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v4, 16, v4 | |
| ; GFX12-FAKE16-NEXT: v_and_or_b32 v0, v0, 15, v1 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 12, v3 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v3, 15, v9 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v9, 15, v12 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v6, 24, v6 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v0, v0, v2, v1 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v1, 4, v3 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v2, 15, v11 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v3, 15, v10 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v0, v0, v4, v5 | |
| ; GFX12-FAKE16-NEXT: v_and_or_b32 v1, v8, 15, v1 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 12, v2 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 8, v3 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v8, 15, v13 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v4, 15, v14 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v5, 28, v7 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v1, v1, v3, v2 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 20, v8 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 16, v9 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2) | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v0, v0, v6, v5 | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v1, v1, v3, v2 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v2, 28, v15 | |
| ; GFX12-FAKE16-NEXT: v_lshlrev_b32_e32 v3, 24, v4 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_1) | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v1, v1, v3, v2 | |
| ; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31] | |
| %bitcast = bitcast <16 x i4> %vec to <2 x float> | |
| ret <2 x float> %bitcast | |
| } | |
| ; A widened shuffle used to need a temporary store/load for the bitcast. | |
| define <128 x i4> @shuffle_v4i4_to_v128i4(<4 x i4> %arg) { | |
| ; GFX9-LABEL: shuffle_v4i4_to_v128i4: | |
| ; GFX9: ; %bb.0: | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) | |
| ; GFX9-NEXT: v_mov_b32_e32 v2, 15 | |
| ; GFX9-NEXT: v_and_b32_e32 v1, 15, v1 | |
| ; GFX9-NEXT: v_and_b32_sdwa v4, v4, v2 dst_sel:BYTE_3 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD | |
| ; GFX9-NEXT: v_and_b32_sdwa v2, v3, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:DWORD | |
| ; GFX9-NEXT: v_or3_b32 v1, v1, v2, v4 | |
| ; GFX9-NEXT: v_mov_b32_e32 v2, 0 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:60 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:56 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:52 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:48 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:44 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:40 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:36 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:32 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:28 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:24 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:20 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:12 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:8 | |
| ; GFX9-NEXT: buffer_store_dword v2, v0, s[0:3], 0 offen offset:4 | |
| ; GFX9-NEXT: buffer_store_dword v1, v0, s[0:3], 0 offen | |
| ; GFX9-NEXT: s_waitcnt vmcnt(0) | |
| ; GFX9-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-TRUE16-LABEL: shuffle_v4i4_to_v128i4: | |
| ; GFX12-TRUE16: ; %bb.0: | |
| ; GFX12-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v4.l, v4.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v2.l, v1.l, 15 | |
| ; GFX12-TRUE16-NEXT: v_and_b16 v3.l, v3.l, 15 | |
| ; GFX12-TRUE16-NEXT: s_mov_b32 s0, 0 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | |
| ; GFX12-TRUE16-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_lshlrev_b32 v8, 24, v4 | |
| ; GFX12-TRUE16-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-TRUE16-NEXT: s_mov_b32 s1, s0 | |
| ; GFX12-TRUE16-NEXT: v_cvt_u32_u16_e32 v9, v2.l | |
| ; GFX12-TRUE16-NEXT: s_mov_b32 s2, s0 | |
| ; GFX12-TRUE16-NEXT: s_mov_b32 s3, s0 | |
| ; GFX12-TRUE16-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-TRUE16-NEXT: v_dual_mov_b32 v6, s2 :: v_dual_lshlrev_b32 v3, 16, v3 | |
| ; GFX12-TRUE16-NEXT: v_dual_mov_b32 v7, s3 :: v_dual_mov_b32 v4, s0 | |
| ; GFX12-TRUE16-NEXT: v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v2, v1 | |
| ; GFX12-TRUE16-NEXT: s_delay_alu instid0(VALU_DEP_3) | |
| ; GFX12-TRUE16-NEXT: v_or3_b32 v8, v9, v3, v8 | |
| ; GFX12-TRUE16-NEXT: v_dual_mov_b32 v3, v1 :: v_dual_mov_b32 v10, s0 | |
| ; GFX12-TRUE16-NEXT: v_mov_b32_e32 v9, s0 | |
| ; GFX12-TRUE16-NEXT: v_mov_b32_e32 v11, s0 | |
| ; GFX12-TRUE16-NEXT: s_clause 0x3 | |
| ; GFX12-TRUE16-NEXT: scratch_store_b128 v0, v[4:7], off offset:48 | |
| ; GFX12-TRUE16-NEXT: scratch_store_b128 v0, v[4:7], off offset:32 | |
| ; GFX12-TRUE16-NEXT: scratch_store_b128 v0, v[0:3], off offset:16 | |
| ; GFX12-TRUE16-NEXT: scratch_store_b128 v0, v[8:11], off | |
| ; GFX12-TRUE16-NEXT: s_setpc_b64 s[30:31] | |
| ; | |
| ; GFX12-FAKE16-LABEL: shuffle_v4i4_to_v128i4: | |
| ; GFX12-FAKE16: ; %bb.0: | |
| ; GFX12-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_expcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_samplecnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_bvhcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_wait_kmcnt 0x0 | |
| ; GFX12-FAKE16-NEXT: s_mov_b32 s0, 0 | |
| ; GFX12-FAKE16-NEXT: v_and_b32_e32 v2, 15, v4 | |
| ; GFX12-FAKE16-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-FAKE16-NEXT: s_mov_b32 s1, s0 | |
| ; GFX12-FAKE16-NEXT: s_mov_b32 s2, s0 | |
| ; GFX12-FAKE16-NEXT: s_mov_b32 s3, s0 | |
| ; GFX12-FAKE16-NEXT: s_wait_alu depctr_sa_sdst(0) | |
| ; GFX12-FAKE16-NEXT: v_dual_mov_b32 v6, s2 :: v_dual_and_b32 v3, 15, v3 | |
| ; GFX12-FAKE16-NEXT: v_dual_mov_b32 v7, s3 :: v_dual_and_b32 v8, 15, v1 | |
| ; GFX12-FAKE16-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_lshlrev_b32 v2, 24, v2 | |
| ; GFX12-FAKE16-NEXT: v_mov_b32_e32 v5, s1 | |
| ; GFX12-FAKE16-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(SKIP_2) | instid1(VALU_DEP_3) | |
| ; GFX12-FAKE16-NEXT: v_dual_mov_b32 v4, s0 :: v_dual_lshlrev_b32 v3, 16, v3 | |
| ; GFX12-FAKE16-NEXT: v_dual_mov_b32 v9, s0 :: v_dual_mov_b32 v10, s0 | |
| ; GFX12-FAKE16-NEXT: v_mov_b32_e32 v11, s0 | |
| ; GFX12-FAKE16-NEXT: v_or3_b32 v8, v8, v3, v2 | |
| ; GFX12-FAKE16-NEXT: v_dual_mov_b32 v2, v1 :: v_dual_mov_b32 v3, v1 | |
| ; GFX12-FAKE16-NEXT: s_clause 0x3 | |
| ; GFX12-FAKE16-NEXT: scratch_store_b128 v0, v[4:7], off offset:48 | |
| ; GFX12-FAKE16-NEXT: scratch_store_b128 v0, v[4:7], off offset:32 | |
| ; GFX12-FAKE16-NEXT: scratch_store_b128 v0, v[0:3], off offset:16 | |
| ; GFX12-FAKE16-NEXT: scratch_store_b128 v0, v[8:11], off | |
| ; GFX12-FAKE16-NEXT: s_setpc_b64 s[30:31] | |
| %res = shufflevector <4 x i4> %arg, <4 x i4> poison, | |
| <128 x i32> <i32 0 , i32 poison, i32 poison, i32 poison, i32 2 , i32 poison, i32 3 , i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, | |
| i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison> | |
| ret <128 x i4> %res | |
| } |