derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
534 lines
20 KiB
LLVM
534 lines
20 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
|
|
; RUN: llc < %s -mtriple=amdgcn -mcpu=gfx1250 | FileCheck %s -check-prefixes=GFX1250
|
|
; RUN: llc < %s -mtriple=amdgcn -mcpu=gfx1310 | FileCheck %s -check-prefixes=GFX1310
|
|
|
|
define <2 x bfloat> @v_fadd_v2bf16_neg1(<2 x bfloat> %a, <2 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_fadd_v2bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_add_bf16 v0, v0, v1 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fadd_v2bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_add_bf16 v0, v0, v1 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <2 x bfloat> %b
|
|
%add = fadd <2 x bfloat> %a, %neg
|
|
ret <2 x bfloat> %add
|
|
}
|
|
|
|
define <4 x bfloat> @v_fadd_v4bf16_neg1(<4 x bfloat> %a, <4 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_fadd_v4bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_add_bf16 v0, v0, v2 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: v_pk_add_bf16 v1, v1, v3 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fadd_v4bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_add_bf16 v0, v0, v2 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: v_pk_add_bf16 v1, v1, v3 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <4 x bfloat> %b
|
|
%add = fadd <4 x bfloat> %a, %neg
|
|
ret <4 x bfloat> %add
|
|
}
|
|
|
|
define <8 x bfloat> @v_fadd_v8bf16_neg1(<8 x bfloat> %a, <8 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_fadd_v8bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_add_bf16 v0, v0, v4 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: v_pk_add_bf16 v1, v1, v5 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: v_pk_add_bf16 v2, v2, v6 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: v_pk_add_bf16 v3, v3, v7 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fadd_v8bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_add_bf16 v0, v0, v4 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: v_pk_add_bf16 v1, v1, v5 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: v_pk_add_bf16 v2, v2, v6 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: v_pk_add_bf16 v3, v3, v7 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <8 x bfloat> %b
|
|
%add = fadd <8 x bfloat> %a, %neg
|
|
ret <8 x bfloat> %add
|
|
}
|
|
|
|
define <2 x bfloat> @v_fneg_v2bf16(<2 x bfloat> %a) {
|
|
; GFX1250-LABEL: v_fneg_v2bf16:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fneg_v2bf16:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <2 x bfloat> %a
|
|
ret <2 x bfloat> %neg
|
|
}
|
|
|
|
define <4 x bfloat> @v_fneg_v4bf16(<4 x bfloat> %a) {
|
|
; GFX1250-LABEL: v_fneg_v4bf16:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
|
|
; GFX1250-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fneg_v4bf16:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
|
|
; GFX1310-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <4 x bfloat> %a
|
|
ret <4 x bfloat> %neg
|
|
}
|
|
|
|
define <8 x bfloat> @v_fneg_v8bf16(<8 x bfloat> %a) {
|
|
; GFX1250-LABEL: v_fneg_v8bf16:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
|
|
; GFX1250-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
|
|
; GFX1250-NEXT: v_xor_b32_e32 v2, 0x80008000, v2
|
|
; GFX1250-NEXT: v_xor_b32_e32 v3, 0x80008000, v3
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fneg_v8bf16:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_xor_b32_e32 v0, 0x80008000, v0
|
|
; GFX1310-NEXT: v_xor_b32_e32 v1, 0x80008000, v1
|
|
; GFX1310-NEXT: v_xor_b32_e32 v2, 0x80008000, v2
|
|
; GFX1310-NEXT: v_xor_b32_e32 v3, 0x80008000, v3
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <8 x bfloat> %a
|
|
ret <8 x bfloat> %neg
|
|
}
|
|
|
|
define <2 x bfloat> @v_fabs_v2bf16(<2 x bfloat> %a) {
|
|
; GFX1250-LABEL: v_fabs_v2bf16:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fabs_v2bf16:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%abs = call <2 x bfloat> @llvm.fabs.v2bf16(<2 x bfloat> %a)
|
|
ret <2 x bfloat> %abs
|
|
}
|
|
|
|
define <4 x bfloat> @v_fabs_v4bf16(<4 x bfloat> %a) {
|
|
; GFX1250-LABEL: v_fabs_v4bf16:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
|
|
; GFX1250-NEXT: v_and_b32_e32 v1, 0x7fff7fff, v1
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fabs_v4bf16:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
|
|
; GFX1310-NEXT: v_and_b32_e32 v1, 0x7fff7fff, v1
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%abs = call <4 x bfloat> @llvm.fabs.v4bf16(<4 x bfloat> %a)
|
|
ret <4 x bfloat> %abs
|
|
}
|
|
|
|
define <8 x bfloat> @v_fabs_v8bf16(<8 x bfloat> %a) {
|
|
; GFX1250-LABEL: v_fabs_v8bf16:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
|
|
; GFX1250-NEXT: v_and_b32_e32 v1, 0x7fff7fff, v1
|
|
; GFX1250-NEXT: v_and_b32_e32 v2, 0x7fff7fff, v2
|
|
; GFX1250-NEXT: v_and_b32_e32 v3, 0x7fff7fff, v3
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fabs_v8bf16:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_and_b32_e32 v0, 0x7fff7fff, v0
|
|
; GFX1310-NEXT: v_and_b32_e32 v1, 0x7fff7fff, v1
|
|
; GFX1310-NEXT: v_and_b32_e32 v2, 0x7fff7fff, v2
|
|
; GFX1310-NEXT: v_and_b32_e32 v3, 0x7fff7fff, v3
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%abs = call <8 x bfloat> @llvm.fabs.v8bf16(<8 x bfloat> %a)
|
|
ret <8 x bfloat> %abs
|
|
}
|
|
|
|
define <2 x bfloat> @v_fsub_v2bf16_neg1(<2 x bfloat> %a, <2 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_fsub_v2bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_add_bf16 v0, v0, v1
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fsub_v2bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_add_bf16 v0, v0, v1
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <2 x bfloat> %b
|
|
%sub = fsub <2 x bfloat> %a, %neg
|
|
ret <2 x bfloat> %sub
|
|
}
|
|
|
|
define <4 x bfloat> @v_fsub_v4bf16_neg1(<4 x bfloat> %a, <4 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_fsub_v4bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_add_bf16 v0, v0, v2
|
|
; GFX1250-NEXT: v_pk_add_bf16 v1, v1, v3
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fsub_v4bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_add_bf16 v0, v0, v2
|
|
; GFX1310-NEXT: v_pk_add_bf16 v1, v1, v3
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <4 x bfloat> %b
|
|
%sub = fsub <4 x bfloat> %a, %neg
|
|
ret <4 x bfloat> %sub
|
|
}
|
|
|
|
define <8 x bfloat> @v_fsub_v8bf16_neg1(<8 x bfloat> %a, <8 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_fsub_v8bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_add_bf16 v0, v0, v4
|
|
; GFX1250-NEXT: v_pk_add_bf16 v1, v1, v5
|
|
; GFX1250-NEXT: v_pk_add_bf16 v2, v2, v6
|
|
; GFX1250-NEXT: v_pk_add_bf16 v3, v3, v7
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fsub_v8bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_add_bf16 v0, v0, v4
|
|
; GFX1310-NEXT: v_pk_add_bf16 v1, v1, v5
|
|
; GFX1310-NEXT: v_pk_add_bf16 v2, v2, v6
|
|
; GFX1310-NEXT: v_pk_add_bf16 v3, v3, v7
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <8 x bfloat> %b
|
|
%sub = fsub <8 x bfloat> %a, %neg
|
|
ret <8 x bfloat> %sub
|
|
}
|
|
|
|
define <2 x bfloat> @v_fmul_v2bf16_neg1(<2 x bfloat> %a, <2 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_fmul_v2bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_mul_bf16 v0, v0, v1 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fmul_v2bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_mul_bf16 v0, v0, v1 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <2 x bfloat> %b
|
|
%mul = fmul <2 x bfloat> %a, %neg
|
|
ret <2 x bfloat> %mul
|
|
}
|
|
|
|
define <4 x bfloat> @v_fmul_v4bf16_neg1(<4 x bfloat> %a, <4 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_fmul_v4bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_mul_bf16 v0, v0, v2 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: v_pk_mul_bf16 v1, v1, v3 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fmul_v4bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_mul_bf16 v0, v0, v2 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: v_pk_mul_bf16 v1, v1, v3 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <4 x bfloat> %b
|
|
%mul = fmul <4 x bfloat> %a, %neg
|
|
ret <4 x bfloat> %mul
|
|
}
|
|
|
|
define <8 x bfloat> @v_fmul_v8bf16_neg1(<8 x bfloat> %a, <8 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_fmul_v8bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_mul_bf16 v0, v0, v4 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: v_pk_mul_bf16 v1, v1, v5 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: v_pk_mul_bf16 v2, v2, v6 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: v_pk_mul_bf16 v3, v3, v7 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fmul_v8bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_mul_bf16 v0, v0, v4 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: v_pk_mul_bf16 v1, v1, v5 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: v_pk_mul_bf16 v2, v2, v6 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: v_pk_mul_bf16 v3, v3, v7 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <8 x bfloat> %b
|
|
%mul = fmul <8 x bfloat> %a, %neg
|
|
ret <8 x bfloat> %mul
|
|
}
|
|
|
|
define <2 x bfloat> @v_fma_v2bf16_neg1(<2 x bfloat> %a, <2 x bfloat> %b, <2 x bfloat> %c) {
|
|
; GFX1250-LABEL: v_fma_v2bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_fma_bf16 v0, v0, v1, v2 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fma_v2bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_fma_bf16 v0, v0, v1, v2 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <2 x bfloat> %b
|
|
%fma = call <2 x bfloat> @llvm.fma.v2bf16( <2 x bfloat> %a, <2 x bfloat> %neg, <2 x bfloat> %c)
|
|
ret <2 x bfloat> %fma
|
|
}
|
|
|
|
define <4 x bfloat> @v_fma_v4bf16_neg1(<4 x bfloat> %a, <4 x bfloat> %b, <4 x bfloat> %c) {
|
|
; GFX1250-LABEL: v_fma_v4bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_fma_bf16 v0, v0, v2, v4 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1250-NEXT: v_pk_fma_bf16 v1, v1, v3, v5 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fma_v4bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_fma_bf16 v0, v0, v2, v4 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1310-NEXT: v_pk_fma_bf16 v1, v1, v3, v5 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <4 x bfloat> %b
|
|
%fma = call <4 x bfloat> @llvm.fma.v4bf16( <4 x bfloat> %a, <4 x bfloat> %neg, <4 x bfloat> %c)
|
|
ret <4 x bfloat> %fma
|
|
}
|
|
|
|
define <8 x bfloat> @v_fma_v8bf16_neg1(<8 x bfloat> %a, <8 x bfloat> %b, <8 x bfloat> %c) {
|
|
; GFX1250-LABEL: v_fma_v8bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_fma_bf16 v0, v0, v4, v8 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1250-NEXT: v_pk_fma_bf16 v1, v1, v5, v9 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1250-NEXT: v_pk_fma_bf16 v2, v2, v6, v10 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1250-NEXT: v_pk_fma_bf16 v3, v3, v7, v11 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_fma_v8bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_fma_bf16 v0, v0, v4, v8 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1310-NEXT: v_pk_fma_bf16 v1, v1, v5, v9 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1310-NEXT: v_pk_fma_bf16 v2, v2, v6, v10 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1310-NEXT: v_pk_fma_bf16 v3, v3, v7, v11 neg_lo:[0,1,0] neg_hi:[0,1,0]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <8 x bfloat> %b
|
|
%fma = call <8 x bfloat> @llvm.fma.v2bf16( <8 x bfloat> %a, <8 x bfloat> %neg, <8 x bfloat> %c)
|
|
ret <8 x bfloat> %fma
|
|
}
|
|
|
|
define <2 x bfloat> @v_minnum_v2bf16_neg1(<2 x bfloat> %a, <2 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_minnum_v2bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_min_num_bf16 v0, v0, v1 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_minnum_v2bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_min_num_bf16 v0, v0, v1 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <2 x bfloat> %b
|
|
%min = call <2 x bfloat> @llvm.minnum.v2bf16( <2 x bfloat> %a, <2 x bfloat> %neg)
|
|
ret <2 x bfloat> %min
|
|
}
|
|
|
|
define <4 x bfloat> @v_minnum_v4bf16_neg1(<4 x bfloat> %a, <4 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_minnum_v4bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_min_num_bf16 v0, v0, v2 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: v_pk_min_num_bf16 v1, v1, v3 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_minnum_v4bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_min_num_bf16 v0, v0, v2 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: v_pk_min_num_bf16 v1, v1, v3 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <4 x bfloat> %b
|
|
%min = call <4 x bfloat> @llvm.minnum.v4bf16( <4 x bfloat> %a, <4 x bfloat> %neg)
|
|
ret <4 x bfloat> %min
|
|
}
|
|
|
|
define <2 x bfloat> @v_maxnum_v2bf16_neg1(<2 x bfloat> %a, <2 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_maxnum_v2bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_max_num_bf16 v0, v0, v1 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_maxnum_v2bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_max_num_bf16 v0, v0, v1 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <2 x bfloat> %b
|
|
%max = call <2 x bfloat> @llvm.maxnum.v2bf16( <2 x bfloat> %a, <2 x bfloat> %neg)
|
|
ret <2 x bfloat> %max
|
|
}
|
|
|
|
define <4 x bfloat> @v_maxnum_v4bf16_neg1(<4 x bfloat> %a, <4 x bfloat> %b) {
|
|
; GFX1250-LABEL: v_maxnum_v4bf16_neg1:
|
|
; GFX1250: ; %bb.0:
|
|
; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1250-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1250-NEXT: v_pk_max_num_bf16 v0, v0, v2 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: v_pk_max_num_bf16 v1, v1, v3 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1250-NEXT: s_set_pc_i64 s[30:31]
|
|
;
|
|
; GFX1310-LABEL: v_maxnum_v4bf16_neg1:
|
|
; GFX1310: ; %bb.0:
|
|
; GFX1310-NEXT: s_wait_loadcnt_dscnt 0x0
|
|
; GFX1310-NEXT: s_wait_expcnt 0x0
|
|
; GFX1310-NEXT: s_wait_samplecnt 0x0
|
|
; GFX1310-NEXT: s_wait_bvhcnt 0x0
|
|
; GFX1310-NEXT: s_wait_kmcnt 0x0
|
|
; GFX1310-NEXT: v_pk_max_num_bf16 v0, v0, v2 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: v_pk_max_num_bf16 v1, v1, v3 neg_lo:[0,1] neg_hi:[0,1]
|
|
; GFX1310-NEXT: s_set_pc_i64 s[30:31]
|
|
%neg = fneg <4 x bfloat> %b
|
|
%max = call <4 x bfloat> @llvm.maxnum.v4bf16( <4 x bfloat> %a, <4 x bfloat> %neg)
|
|
ret <4 x bfloat> %max
|
|
}
|