From af7f1821273c45a6101735023736882ec0399e86 Mon Sep 17 00:00:00 2001 From: Taylor Simpson Date: Thu, 8 Apr 2021 20:07:51 -0500 Subject: [PATCH] Hexagon (target/hexagon) bit reverse (brev) addressing The following instructions are added L2_loadrub_pbr Rd32 = memub(Rx32++Mu2:brev) L2_loadrb_pbr Rd32 = memb(Rx32++Mu2:brev) L2_loadruh_pbr Rd32 = memuh(Rx32++Mu2:brev) L2_loadrh_pbr Rd32 = memh(Rx32++Mu2:brev) L2_loadri_pbr Rd32 = memw(Rx32++Mu2:brev) L2_loadrd_pbr Rdd32 = memd(Rx32++Mu2:brev) S2_storerb_pbr memb(Rx32++Mu2:brev).=.Rt32 S2_storerh_pbr memh(Rx32++Mu2:brev).=.Rt32 S2_storerf_pbr memh(Rx32++Mu2:brev).=.Rt.H32 S2_storeri_pbr memw(Rx32++Mu2:brev).=.Rt32 S2_storerd_pbr memd(Rx32++Mu2:brev).=.Rt32 S2_storerinew_pbr memw(Rx32++Mu2:brev).=.Nt8.new S2_storerbnew_pbr memw(Rx32++Mu2:brev).=.Nt8.new S2_storerhnew_pbr memw(Rx32++Mu2:brev).=.Nt8.new Test cases in tests/tcg/hexagon/brev.c Signed-off-by: Taylor Simpson Reviewed-by: Richard Henderson Message-Id: <1617930474-31979-24-git-send-email-tsimpson@quicinc.com> Signed-off-by: Richard Henderson --- target/hexagon/gen_tcg.h | 28 ++++ target/hexagon/helper.h | 1 + target/hexagon/imported/encode_pp.def | 4 + target/hexagon/imported/ldst.idef | 2 + target/hexagon/imported/macros.def | 6 + target/hexagon/macros.h | 1 + target/hexagon/op_helper.c | 8 ++ tests/tcg/hexagon/Makefile.target | 1 + tests/tcg/hexagon/brev.c | 190 ++++++++++++++++++++++++++ 9 files changed, 241 insertions(+) create mode 100644 tests/tcg/hexagon/brev.c diff --git a/target/hexagon/gen_tcg.h b/target/hexagon/gen_tcg.h index 25c228c112..8f0ec01f0a 100644 --- a/target/hexagon/gen_tcg.h +++ b/target/hexagon/gen_tcg.h @@ -37,6 +37,7 @@ * _sp stack pointer relative r0 = memw(r29+#12) * _ap absolute set r0 = memw(r1=##variable) * _pr post increment register r0 = memw(r1++m1) + * _pbr post increment bit reverse r0 = memw(r1++m1:brev) * _pi post increment immediate r0 = memb(r1++#1) * _pci post increment circular immediate r0 = memw(r1++#4:circ(m0)) * _pcr post increment circular register r0 = memw(r1++I:circ(m0)) @@ -53,6 +54,11 @@ fEA_REG(RxV); \ fPM_M(RxV, MuV); \ } while (0) +#define GET_EA_pbr \ + do { \ + gen_helper_fbrev(EA, RxV); \ + tcg_gen_add_tl(RxV, RxV, MuV); \ + } while (0) #define GET_EA_pi \ do { \ fEA_REG(RxV); \ @@ -128,16 +134,22 @@ fGEN_TCG_LOAD_pcr(3, fLOAD(1, 8, u, EA, RddV)) #define fGEN_TCG_L2_loadrub_pr(SHORTCODE) SHORTCODE +#define fGEN_TCG_L2_loadrub_pbr(SHORTCODE) SHORTCODE #define fGEN_TCG_L2_loadrub_pi(SHORTCODE) SHORTCODE #define fGEN_TCG_L2_loadrb_pr(SHORTCODE) SHORTCODE +#define fGEN_TCG_L2_loadrb_pbr(SHORTCODE) SHORTCODE #define fGEN_TCG_L2_loadrb_pi(SHORTCODE) SHORTCODE #define fGEN_TCG_L2_loadruh_pr(SHORTCODE) SHORTCODE +#define fGEN_TCG_L2_loadruh_pbr(SHORTCODE) SHORTCODE #define fGEN_TCG_L2_loadruh_pi(SHORTCODE) SHORTCODE #define fGEN_TCG_L2_loadrh_pr(SHORTCODE) SHORTCODE +#define fGEN_TCG_L2_loadrh_pbr(SHORTCODE) SHORTCODE #define fGEN_TCG_L2_loadrh_pi(SHORTCODE) SHORTCODE #define fGEN_TCG_L2_loadri_pr(SHORTCODE) SHORTCODE +#define fGEN_TCG_L2_loadri_pbr(SHORTCODE) SHORTCODE #define fGEN_TCG_L2_loadri_pi(SHORTCODE) SHORTCODE #define fGEN_TCG_L2_loadrd_pr(SHORTCODE) SHORTCODE +#define fGEN_TCG_L2_loadrd_pbr(SHORTCODE) SHORTCODE #define fGEN_TCG_L2_loadrd_pi(SHORTCODE) SHORTCODE /* @@ -265,41 +277,57 @@ tcg_temp_free(BYTE); \ } while (0) +#define fGEN_TCG_S2_storerb_pbr(SHORTCODE) \ + fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerb_pci(SHORTCODE) \ fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerb_pcr(SHORTCODE) \ fGEN_TCG_STORE_pcr(0, fSTORE(1, 1, EA, fGETBYTE(0, RtV))) +#define fGEN_TCG_S2_storerh_pbr(SHORTCODE) \ + fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerh_pci(SHORTCODE) \ fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerh_pcr(SHORTCODE) \ fGEN_TCG_STORE_pcr(1, fSTORE(1, 2, EA, fGETHALF(0, RtV))) +#define fGEN_TCG_S2_storerf_pbr(SHORTCODE) \ + fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerf_pci(SHORTCODE) \ fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerf_pcr(SHORTCODE) \ fGEN_TCG_STORE_pcr(1, fSTORE(1, 2, EA, fGETHALF(1, RtV))) +#define fGEN_TCG_S2_storeri_pbr(SHORTCODE) \ + fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storeri_pci(SHORTCODE) \ fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storeri_pcr(SHORTCODE) \ fGEN_TCG_STORE_pcr(2, fSTORE(1, 4, EA, RtV)) +#define fGEN_TCG_S2_storerd_pbr(SHORTCODE) \ + fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerd_pci(SHORTCODE) \ fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerd_pcr(SHORTCODE) \ fGEN_TCG_STORE_pcr(3, fSTORE(1, 8, EA, RttV)) +#define fGEN_TCG_S2_storerbnew_pbr(SHORTCODE) \ + fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerbnew_pci(SHORTCODE) \ fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerbnew_pcr(SHORTCODE) \ fGEN_TCG_STORE_pcr(0, fSTORE(1, 1, EA, fGETBYTE(0, NtN))) +#define fGEN_TCG_S2_storerhnew_pbr(SHORTCODE) \ + fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerhnew_pci(SHORTCODE) \ fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerhnew_pcr(SHORTCODE) \ fGEN_TCG_STORE_pcr(1, fSTORE(1, 2, EA, fGETHALF(0, NtN))) +#define fGEN_TCG_S2_storerinew_pbr(SHORTCODE) \ + fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerinew_pci(SHORTCODE) \ fGEN_TCG_STORE(SHORTCODE) #define fGEN_TCG_S2_storerinew_pcr(SHORTCODE) \ diff --git a/target/hexagon/helper.h b/target/hexagon/helper.h index 3824ae01ea..ca201fb680 100644 --- a/target/hexagon/helper.h +++ b/target/hexagon/helper.h @@ -24,6 +24,7 @@ DEF_HELPER_FLAGS_3(debug_check_store_width, TCG_CALL_NO_WG, void, env, int, int) DEF_HELPER_FLAGS_3(debug_commit_end, TCG_CALL_NO_WG, void, env, int, int) DEF_HELPER_2(commit_store, void, env, int) DEF_HELPER_FLAGS_4(fcircadd, TCG_CALL_NO_RWG_SE, s32, s32, s32, s32, s32) +DEF_HELPER_FLAGS_1(fbrev, TCG_CALL_NO_RWG_SE, i32, i32) DEF_HELPER_3(sfrecipa, i64, env, f32, f32) DEF_HELPER_2(sfinvsqrta, i64, env, f32) DEF_HELPER_4(vacsh_val, s64, env, s64, s64, s64) diff --git a/target/hexagon/imported/encode_pp.def b/target/hexagon/imported/encode_pp.def index 68b435ebe7..4464926634 100644 --- a/target/hexagon/imported/encode_pp.def +++ b/target/hexagon/imported/encode_pp.def @@ -301,6 +301,7 @@ DEF_CLASS32(ICLASS_LD" 101- -------- PP1----- --------",LD_ADDR_POST_IMMED_PRED_ DEF_CLASS32(ICLASS_LD" 110- -------- PP-0---- 0-------",LD_ADDR_POST_REG) DEF_CLASS32(ICLASS_LD" 110- -------- PP-1---- --------",LD_ADDR_ABS_PLUS_REG_V4) DEF_CLASS32(ICLASS_LD" 100- -------- PP----1- --------",LD_ADDR_POST_CREG_V2) +DEF_CLASS32(ICLASS_LD" 111- -------- PP------ 0-------",LD_ADDR_POST_BREV_REG) DEF_CLASS32(ICLASS_LD" 111- -------- PP------ 1-------",LD_ADDR_PRED_ABS_V4) DEF_FIELD32(ICLASS_LD" !!!- -------- PP------ --------",LD_Amode,"Amode") @@ -315,6 +316,7 @@ DEF_ENC32(L4_load##TAG##_ap, ICLASS_LD" 1 01 "OPC" eeeee PP01IIII -IIddddd" DEF_ENC32(L2_load##TAG##_pr, ICLASS_LD" 1 10 "OPC" xxxxx PPu0---- 0--ddddd")\ DEF_ENC32(L4_load##TAG##_ur, ICLASS_LD" 1 10 "OPC" ttttt PPi1IIII iIIddddd")\ DEF_ENC32(L2_load##TAG##_pcr, ICLASS_LD" 1 00 "OPC" xxxxx PPu0--1- 0--ddddd")\ +DEF_ENC32(L2_load##TAG##_pbr, ICLASS_LD" 1 11 "OPC" xxxxx PPu0---- 0--ddddd") #define STD_LDX_ENC(TAG,OPC) \ @@ -412,6 +414,7 @@ DEF_CLASS32(ICLASS_ST" 1111 -------- PP------ 1-------",ST_ADDR_PRED_ABS_V4) DEF_CLASS32(ICLASS_ST" 1101 -------- PP------ 0-------",ST_ADDR_POST_REG) DEF_CLASS32(ICLASS_ST" 1101 -------- PP------ 1-------",ST_ADDR_ABS_PLUS_REG_V4) DEF_CLASS32(ICLASS_ST" 1001 -------- PP------ ------1-",ST_ADDR_POST_CREG_V2) +DEF_CLASS32(ICLASS_ST" 1111 -------- PP------ 0-------",ST_ADDR_POST_BREV_REG) DEF_CLASS32(ICLASS_ST" 0--0 1------- PP------ --------",ST_MISC_STORELIKE) DEF_CLASS32(ICLASS_ST" 1--0 0------- PP------ --------",ST_MISC_BUSOP) DEF_CLASS32(ICLASS_ST" 0--0 0------- PP------ --------",ST_MISC_CACHEOP) @@ -425,6 +428,7 @@ DEF_ENC32(S4_store##TAG##_ap, ICLASS_ST" 1 01 "OPC" eeeee PP0"SRC" 1-IIIIII DEF_ENC32(S2_store##TAG##_pr, ICLASS_ST" 1 10 "OPC" xxxxx PPu"SRC" 0-------")\ DEF_ENC32(S4_store##TAG##_ur, ICLASS_ST" 1 10 "OPC" uuuuu PPi"SRC" 1iIIIIII")\ DEF_ENC32(S2_store##TAG##_pcr, ICLASS_ST" 1 00 "OPC" xxxxx PPu"SRC" 0-----1-")\ +DEF_ENC32(S2_store##TAG##_pbr, ICLASS_ST" 1 11 "OPC" xxxxx PPu"SRC" 0-------") #define STD_PST_ENC(TAG,OPC,SRC) \ diff --git a/target/hexagon/imported/ldst.idef b/target/hexagon/imported/ldst.idef index 6ce0635e32..fe7e018cf1 100644 --- a/target/hexagon/imported/ldst.idef +++ b/target/hexagon/imported/ldst.idef @@ -25,6 +25,7 @@ Q6INSN(L2_##TAG##_io, OPER"(Rs32+#s11:"SHFT")", ATTRIB,DESCR,{fIMMEXT( Q6INSN(L4_##TAG##_ur, OPER"(Rt32<<#u2+#U6)", ATTRIB,DESCR,{fMUST_IMMEXT(UiV); fEA_IRs(UiV,RtV,uiV); SEMANTICS;})\ Q6INSN(L4_##TAG##_ap, OPER"(Re32=#U6)", ATTRIB,DESCR,{fMUST_IMMEXT(UiV); fEA_IMM(UiV); SEMANTICS; ReV=UiV; })\ Q6INSN(L2_##TAG##_pr, OPER"(Rx32++Mu2)", ATTRIB,DESCR,{fEA_REG(RxV); fPM_M(RxV,MuV); SEMANTICS;})\ +Q6INSN(L2_##TAG##_pbr, OPER"(Rx32++Mu2:brev)", ATTRIB,DESCR,{fEA_BREVR(RxV); fPM_M(RxV,MuV); SEMANTICS;})\ Q6INSN(L2_##TAG##_pi, OPER"(Rx32++#s4:"SHFT")", ATTRIB,DESCR,{fEA_REG(RxV); fPM_I(RxV,siV); SEMANTICS;})\ Q6INSN(L2_##TAG##_pci, OPER"(Rx32++#s4:"SHFT":circ(Mu2))",ATTRIB,DESCR,{fEA_REG(RxV); fPM_CIRI(RxV,siV,MuV); SEMANTICS;})\ Q6INSN(L2_##TAG##_pcr, OPER"(Rx32++I:circ(Mu2))", ATTRIB,DESCR,{fEA_REG(RxV); fPM_CIRR(RxV,fREAD_IREG(MuV)<>= 1; + } + return result; +} + +int sext8(int x) +{ + return (x << 24) >> 24; +} + +void check(int i, long long result, long long expect) +{ + if (result != expect) { + printf("ERROR(%d): 0x%04llx != 0x%04llx\n", i, result, expect); + err++; + } +} + +#define TEST_BREV_LOAD(SZ, TYPE, BUF, SHIFT, EXP) \ + do { \ + p = BUF; \ + for (i = 0; i < SIZE; i++) { \ + TYPE result; \ + BREV_LOAD_##SZ(result, p, 1 << (SHIFT - NBITS)); \ + check(i, result, EXP); \ + } \ + } while (0) + +#define TEST_BREV_STORE(SZ, TYPE, BUF, VAL, SHIFT) \ + do { \ + p = BUF; \ + memset(BUF, 0xff, sizeof(BUF)); \ + for (i = 0; i < SIZE; i++) { \ + BREV_STORE_##SZ(p, (TYPE)(VAL), 1 << (SHIFT - NBITS)); \ + } \ + for (i = 0; i < SIZE; i++) { \ + check(i, BUF[i], bitreverse(i)); \ + } \ + } while (0) + +#define TEST_BREV_STORE_NEW(SZ, BUF, SHIFT) \ + do { \ + p = BUF; \ + memset(BUF, 0xff, sizeof(BUF)); \ + for (i = 0; i < SIZE; i++) { \ + BREV_STORE_##SZ(p, i, 1 << (SHIFT - NBITS)); \ + } \ + for (i = 0; i < SIZE; i++) { \ + check(i, BUF[i], bitreverse(i)); \ + } \ + } while (0) + +/* + * We'll set high_half[i] = i << 16 for use in the .H form of store + * which stores from the high half of the word. + */ +int high_half[SIZE]; + +int main() +{ + void *p; + int i; + + for (i = 0; i < SIZE; i++) { + bbuf[i] = bitreverse(i); + hbuf[i] = bitreverse(i); + wbuf[i] = bitreverse(i); + dbuf[i] = bitreverse(i); + high_half[i] = i << 16; + } + + TEST_BREV_LOAD(b, int, bbuf, 16, sext8(i)); + TEST_BREV_LOAD(ub, int, bbuf, 16, i); + TEST_BREV_LOAD(h, int, hbuf, 15, i); + TEST_BREV_LOAD(uh, int, hbuf, 15, i); + TEST_BREV_LOAD(w, int, wbuf, 14, i); + TEST_BREV_LOAD(d, long long, dbuf, 13, i); + + TEST_BREV_STORE(b, int, bbuf, i, 16); + TEST_BREV_STORE(h, int, hbuf, i, 15); + TEST_BREV_STORE(f, int, hbuf, high_half[i], 15); + TEST_BREV_STORE(w, int, wbuf, i, 14); + TEST_BREV_STORE(d, long long, dbuf, i, 13); + + TEST_BREV_STORE_NEW(bnew, bbuf, 16); + TEST_BREV_STORE_NEW(hnew, hbuf, 15); + TEST_BREV_STORE_NEW(wnew, wbuf, 14); + + puts(err ? "FAIL" : "PASS"); + return err ? 1 : 0; +}