[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH 27/26] tcg-aarch64: Introduce tcg_out_insn_3312,
From: |
Claudio Fontana |
Subject: |
Re: [Qemu-devel] [PATCH 27/26] tcg-aarch64: Introduce tcg_out_insn_3312, _3310, _3313 |
Date: |
Tue, 8 Apr 2014 11:00:23 +0200 |
User-agent: |
Mozilla/5.0 (Windows NT 6.1; rv:24.0) Gecko/20100101 Thunderbird/24.0.1 |
On 07.04.2014 20:34, Richard Henderson wrote:
> Merge TCGMemOp size, AArch64LdstType type and a few stray opcode bits
> into a single I3312_* argument, eliminating some magic numbers from
> helper functions.
>
> Signed-off-by: Richard Henderson <address@hidden>
> ---
> tcg/aarch64/tcg-target.c | 129
> ++++++++++++++++++++++++++++-------------------
> 1 file changed, 76 insertions(+), 53 deletions(-)
> ---
>
> I'm not really sure how much clearer this is, especially since we do
> have to re-extract the size within tcg_out_ldst. But it does at least
> eliminate some of the magic numbers within the helpers.
>
> Thoughts?
Looks good to me, I'll get to it in more detail later.
C.
>
>
> r~
>
>
>
> diff --git a/tcg/aarch64/tcg-target.c b/tcg/aarch64/tcg-target.c
> index ab4cd25..324a452 100644
> --- a/tcg/aarch64/tcg-target.c
> +++ b/tcg/aarch64/tcg-target.c
> @@ -271,6 +271,28 @@ typedef enum {
> I3207_BLR = 0xd63f0000,
> I3207_RET = 0xd65f0000,
>
> + /* Load/store register. Described here as 3.3.12, but the helper
> + that emits them can transform to 3.3.10 or 3.3.13. */
> + I3312_STRB = 0x38000000 | LDST_ST << 22 | MO_8 << 30,
> + I3312_STRH = 0x38000000 | LDST_ST << 22 | MO_16 << 30,
> + I3312_STRW = 0x38000000 | LDST_ST << 22 | MO_32 << 30,
> + I3312_STRX = 0x38000000 | LDST_ST << 22 | MO_64 << 30,
> +
> + I3312_LDRB = 0x38000000 | LDST_LD << 22 | MO_8 << 30,
> + I3312_LDRH = 0x38000000 | LDST_LD << 22 | MO_16 << 30,
> + I3312_LDRW = 0x38000000 | LDST_LD << 22 | MO_32 << 30,
> + I3312_LDRX = 0x38000000 | LDST_LD << 22 | MO_64 << 30,
> +
> + I3312_LDRSBW = 0x38000000 | LDST_LD_S_W << 22 | MO_8 << 30,
> + I3312_LDRSHW = 0x38000000 | LDST_LD_S_W << 22 | MO_16 << 30,
> +
> + I3312_LDRSBX = 0x38000000 | LDST_LD_S_X << 22 | MO_8 << 30,
> + I3312_LDRSHX = 0x38000000 | LDST_LD_S_X << 22 | MO_16 << 30,
> + I3312_LDRSWX = 0x38000000 | LDST_LD_S_X << 22 | MO_32 << 30,
> +
> + I3312_TO_I3310 = 0x00206800,
> + I3312_TO_I3313 = 0x01000000,
> +
> /* Load/store register pair instructions. */
> I3314_LDP = 0x28400000,
> I3314_STP = 0x28000000,
> @@ -482,21 +504,25 @@ static void tcg_out_insn_3509(TCGContext *s,
> AArch64Insn insn, TCGType ext,
> tcg_out32(s, insn | ext << 31 | rm << 16 | ra << 10 | rn << 5 | rd);
> }
>
> +static void tcg_out_insn_3310(TCGContext *s, AArch64Insn insn,
> + TCGReg rd, TCGReg base, TCGReg regoff)
> +{
> + /* Note the AArch64Insn constants above are for C3.3.12. Adjust. */
> + tcg_out32(s, insn | I3312_TO_I3310 | regoff << 16 | base << 5 | rd);
> +}
> +
>
> -static void tcg_out_ldst_9(TCGContext *s, TCGMemOp size, AArch64LdstType
> type,
> - TCGReg rd, TCGReg rn, intptr_t offset)
> +static void tcg_out_insn_3312(TCGContext *s, AArch64Insn insn,
> + TCGReg rd, TCGReg rn, intptr_t offset)
> {
> - /* use LDUR with BASE register with 9bit signed unscaled offset */
> - tcg_out32(s, 0x38000000 | size << 30 | type << 22
> - | (offset & 0x1ff) << 12 | rn << 5 | rd);
> + tcg_out32(s, insn | (offset & 0x1ff) << 12 | rn << 5 | rd);
> }
>
> -/* tcg_out_ldst_12 expects a scaled unsigned immediate offset */
> -static void tcg_out_ldst_12(TCGContext *s, TCGMemOp size, AArch64LdstType
> type,
> - TCGReg rd, TCGReg rn, tcg_target_ulong
> scaled_uimm)
> +static void tcg_out_insn_3313(TCGContext *s, AArch64Insn insn,
> + TCGReg rd, TCGReg rn, uintptr_t scaled_uimm)
> {
> - tcg_out32(s, 0x39000000 | size << 30 | type << 22
> - | scaled_uimm << 10 | rn << 5 | rd);
> + /* Note the AArch64Insn constants above are for C3.3.12. Adjust. */
> + tcg_out32(s, insn | I3312_TO_I3313 | scaled_uimm << 10 | rn << 5 | rd);
> }
>
> /* Register to register move using ORR (shifted register with no shift). */
> @@ -634,35 +660,32 @@ static void tcg_out_movi(TCGContext *s, TCGType type,
> TCGReg rd,
> }
> }
>
> -static void tcg_out_ldst_r(TCGContext *s, TCGMemOp size, AArch64LdstType
> type,
> - TCGReg rd, TCGReg base, TCGReg regoff)
> -{
> - tcg_out32(s, 0x38206800 | size << 30 | type << 22
> - | regoff << 16 | base << 5 | rd);
> -}
> +/* Define something more legible for general use. */
> +#define tcg_out_ldst_r tcg_out_insn_3310
>
> -/* solve the whole ldst problem */
> -static void tcg_out_ldst(TCGContext *s, TCGMemOp size, AArch64LdstType type,
> +static void tcg_out_ldst(TCGContext *s, AArch64Insn insn,
> TCGReg rd, TCGReg rn, intptr_t offset)
> {
> + TCGMemOp size = (uint32_t)insn >> 30;
> +
> /* If the offset is naturally aligned and in range, then we can
> use the scaled uimm12 encoding */
> if (offset >= 0 && !(offset & ((1 << size) - 1))) {
> - tcg_target_ulong scaled_uimm = offset >> size;
> + uintptr_t scaled_uimm = offset >> size;
> if (scaled_uimm <= 0xfff) {
> - tcg_out_ldst_12(s, size, type, rd, rn, scaled_uimm);
> + tcg_out_insn_3313(s, insn, rd, rn, scaled_uimm);
> return;
> }
> }
>
> if (offset >= -256 && offset < 256) {
> - tcg_out_ldst_9(s, size, type, rd, rn, offset);
> + tcg_out_insn_3312(s, insn, rd, rn, offset);
> return;
> }
>
> /* Worst-case scenario, move offset to temp register, use reg offset. */
> tcg_out_movi(s, TCG_TYPE_I64, TCG_REG_TMP, offset);
> - tcg_out_ldst_r(s, size, type, rd, rn, TCG_REG_TMP);
> + tcg_out_ldst_r(s, insn, rd, rn, TCG_REG_TMP);
> }
>
> static inline void tcg_out_mov(TCGContext *s,
> @@ -676,14 +699,14 @@ static inline void tcg_out_mov(TCGContext *s,
> static inline void tcg_out_ld(TCGContext *s, TCGType type, TCGReg arg,
> TCGReg arg1, intptr_t arg2)
> {
> - tcg_out_ldst(s, type == TCG_TYPE_I64 ? MO_64 : MO_32, LDST_LD,
> + tcg_out_ldst(s, type == TCG_TYPE_I32 ? I3312_LDRW : I3312_LDRX,
> arg, arg1, arg2);
> }
>
> static inline void tcg_out_st(TCGContext *s, TCGType type, TCGReg arg,
> TCGReg arg1, intptr_t arg2)
> {
> - tcg_out_ldst(s, type == TCG_TYPE_I64 ? MO_64 : MO_32, LDST_ST,
> + tcg_out_ldst(s, type == TCG_TYPE_I32 ? I3312_STRW : I3312_STRX,
> arg, arg1, arg2);
> }
>
> @@ -1081,12 +1104,12 @@ static void tcg_out_tlb_read(TCGContext *s, TCGReg
> addr_reg, TCGMemOp s_bits,
>
> /* Merge "low bits" from tlb offset, load the tlb comparator into X0.
> X0 = load [X2 + (tlb_offset & 0x000fff)] */
> - tcg_out_ldst(s, TARGET_LONG_BITS == 64 ? MO_64 : MO_32,
> - LDST_LD, TCG_REG_X0, TCG_REG_X2, tlb_offset & 0xfff);
> + tcg_out_ldst(s, TARGET_LONG_BITS == 32 ? I3312_LDRW : I3312_LDRX,
> + TCG_REG_X0, TCG_REG_X2, tlb_offset & 0xfff);
>
> /* Load the tlb addend. Do that early to avoid stalling.
> X1 = load [X2 + (tlb_offset & 0xfff) + offsetof(addend)] */
> - tcg_out_ldst(s, MO_64, LDST_LD, TCG_REG_X1, TCG_REG_X2,
> + tcg_out_ldst(s, I3312_LDRX, TCG_REG_X1, TCG_REG_X2,
> (tlb_offset & 0xfff) + (offsetof(CPUTLBEntry, addend)) -
> (is_read ? offsetof(CPUTLBEntry, addr_read)
> : offsetof(CPUTLBEntry, addr_write)));
> @@ -1108,43 +1131,43 @@ static void tcg_out_qemu_ld_direct(TCGContext *s,
> TCGMemOp memop,
>
> switch (memop & MO_SSIZE) {
> case MO_UB:
> - tcg_out_ldst_r(s, MO_8, LDST_LD, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_LDRB, data_r, addr_r, off_r);
> break;
> case MO_SB:
> - tcg_out_ldst_r(s, MO_8, LDST_LD_S_X, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_LDRSBX, data_r, addr_r, off_r);
> break;
> case MO_UW:
> - tcg_out_ldst_r(s, MO_16, LDST_LD, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_LDRH, data_r, addr_r, off_r);
> if (bswap) {
> tcg_out_rev16(s, TCG_TYPE_I32, data_r, data_r);
> }
> break;
> case MO_SW:
> if (bswap) {
> - tcg_out_ldst_r(s, MO_16, LDST_LD, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_LDRH, data_r, addr_r, off_r);
> tcg_out_rev16(s, TCG_TYPE_I32, data_r, data_r);
> tcg_out_sxt(s, TCG_TYPE_I64, MO_16, data_r, data_r);
> } else {
> - tcg_out_ldst_r(s, MO_16, LDST_LD_S_X, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_LDRSHX, data_r, addr_r, off_r);
> }
> break;
> case MO_UL:
> - tcg_out_ldst_r(s, MO_32, LDST_LD, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_LDRW, data_r, addr_r, off_r);
> if (bswap) {
> tcg_out_rev(s, TCG_TYPE_I32, data_r, data_r);
> }
> break;
> case MO_SL:
> if (bswap) {
> - tcg_out_ldst_r(s, MO_32, LDST_LD, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_LDRW, data_r, addr_r, off_r);
> tcg_out_rev(s, TCG_TYPE_I32, data_r, data_r);
> tcg_out_sxt(s, TCG_TYPE_I64, MO_32, data_r, data_r);
> } else {
> - tcg_out_ldst_r(s, MO_32, LDST_LD_S_X, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_LDRSWX, data_r, addr_r, off_r);
> }
> break;
> case MO_Q:
> - tcg_out_ldst_r(s, MO_64, LDST_LD, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_LDRX, data_r, addr_r, off_r);
> if (bswap) {
> tcg_out_rev(s, TCG_TYPE_I64, data_r, data_r);
> }
> @@ -1161,28 +1184,28 @@ static void tcg_out_qemu_st_direct(TCGContext *s,
> TCGMemOp memop,
>
> switch (memop & MO_SIZE) {
> case MO_8:
> - tcg_out_ldst_r(s, MO_8, LDST_ST, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_STRB, data_r, addr_r, off_r);
> break;
> case MO_16:
> if (bswap && data_r != TCG_REG_XZR) {
> tcg_out_rev16(s, TCG_TYPE_I32, TCG_REG_TMP, data_r);
> data_r = TCG_REG_TMP;
> }
> - tcg_out_ldst_r(s, MO_16, LDST_ST, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_STRH, data_r, addr_r, off_r);
> break;
> case MO_32:
> if (bswap && data_r != TCG_REG_XZR) {
> tcg_out_rev(s, TCG_TYPE_I32, TCG_REG_TMP, data_r);
> data_r = TCG_REG_TMP;
> }
> - tcg_out_ldst_r(s, MO_32, LDST_ST, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_STRW, data_r, addr_r, off_r);
> break;
> case MO_64:
> if (bswap && data_r != TCG_REG_XZR) {
> tcg_out_rev(s, TCG_TYPE_I64, TCG_REG_TMP, data_r);
> data_r = TCG_REG_TMP;
> }
> - tcg_out_ldst_r(s, MO_64, LDST_ST, data_r, addr_r, off_r);
> + tcg_out_ldst_r(s, I3312_STRX, data_r, addr_r, off_r);
> break;
> default:
> tcg_abort();
> @@ -1275,49 +1298,49 @@ static void tcg_out_op(TCGContext *s, TCGOpcode opc,
>
> case INDEX_op_ld8u_i32:
> case INDEX_op_ld8u_i64:
> - tcg_out_ldst(s, MO_8, LDST_LD, a0, a1, a2);
> + tcg_out_ldst(s, I3312_LDRB, a0, a1, a2);
> break;
> case INDEX_op_ld8s_i32:
> - tcg_out_ldst(s, MO_8, LDST_LD_S_W, a0, a1, a2);
> + tcg_out_ldst(s, I3312_LDRSBW, a0, a1, a2);
> break;
> case INDEX_op_ld8s_i64:
> - tcg_out_ldst(s, MO_8, LDST_LD_S_X, a0, a1, a2);
> + tcg_out_ldst(s, I3312_LDRSBX, a0, a1, a2);
> break;
> case INDEX_op_ld16u_i32:
> case INDEX_op_ld16u_i64:
> - tcg_out_ldst(s, MO_16, LDST_LD, a0, a1, a2);
> + tcg_out_ldst(s, I3312_LDRH, a0, a1, a2);
> break;
> case INDEX_op_ld16s_i32:
> - tcg_out_ldst(s, MO_16, LDST_LD_S_W, a0, a1, a2);
> + tcg_out_ldst(s, I3312_LDRSHW, a0, a1, a2);
> break;
> case INDEX_op_ld16s_i64:
> - tcg_out_ldst(s, MO_16, LDST_LD_S_X, a0, a1, a2);
> + tcg_out_ldst(s, I3312_LDRSHX, a0, a1, a2);
> break;
> case INDEX_op_ld_i32:
> case INDEX_op_ld32u_i64:
> - tcg_out_ldst(s, MO_32, LDST_LD, a0, a1, a2);
> + tcg_out_ldst(s, I3312_LDRW, a0, a1, a2);
> break;
> case INDEX_op_ld32s_i64:
> - tcg_out_ldst(s, MO_32, LDST_LD_S_X, a0, a1, a2);
> + tcg_out_ldst(s, I3312_LDRSWX, a0, a1, a2);
> break;
> case INDEX_op_ld_i64:
> - tcg_out_ldst(s, MO_64, LDST_LD, a0, a1, a2);
> + tcg_out_ldst(s, I3312_LDRX, a0, a1, a2);
> break;
>
> case INDEX_op_st8_i32:
> case INDEX_op_st8_i64:
> - tcg_out_ldst(s, MO_8, LDST_ST, REG0(0), a1, a2);
> + tcg_out_ldst(s, I3312_STRB, REG0(0), a1, a2);
> break;
> case INDEX_op_st16_i32:
> case INDEX_op_st16_i64:
> - tcg_out_ldst(s, MO_16, LDST_ST, REG0(0), a1, a2);
> + tcg_out_ldst(s, I3312_STRH, REG0(0), a1, a2);
> break;
> case INDEX_op_st_i32:
> case INDEX_op_st32_i64:
> - tcg_out_ldst(s, MO_32, LDST_ST, REG0(0), a1, a2);
> + tcg_out_ldst(s, I3312_STRW, REG0(0), a1, a2);
> break;
> case INDEX_op_st_i64:
> - tcg_out_ldst(s, MO_64, LDST_ST, REG0(0), a1, a2);
> + tcg_out_ldst(s, I3312_STRX, REG0(0), a1, a2);
> break;
>
> case INDEX_op_add_i32:
>
--
Claudio Fontana
Server Virtualization Architect
Huawei Technologies Duesseldorf GmbH
Riesstraße 25 - 80992 München
office: +49 89 158834 4135
mobile: +49 15253060158
- Re: [Qemu-devel] [PATCH v3 20/26] tcg-aarch64: Support stores of zero, (continued)
- [Qemu-devel] [PATCH v3 23/26] tcg-aarch64: Replace aarch64_ldst_op_data with TCGMemOp, Richard Henderson, 2014/04/03
- [Qemu-devel] [PATCH v3 22/26] tcg-aarch64: Merge aarch64_ldst_get_data/type into tcg_out_op, Richard Henderson, 2014/04/03
- [Qemu-devel] [PATCH v3 24/26] tcg-aarch64: Replace aarch64_ldst_op_data with AArch64LdstType, Richard Henderson, 2014/04/03
- Re: [Qemu-devel] [PATCH v3 24/26] tcg-aarch64: Replace aarch64_ldst_op_data with AArch64LdstType, Claudio Fontana, 2014/04/07
- [Qemu-devel] [PATCH 27/26] tcg-aarch64: Introduce tcg_out_insn_3312, _3310, _3313, Richard Henderson, 2014/04/07
- Re: [Qemu-devel] [PATCH 27/26] tcg-aarch64: Introduce tcg_out_insn_3312, _3310, _3313,
Claudio Fontana <=
- Re: [Qemu-devel] [PATCH 27/26] tcg-aarch64: Introduce tcg_out_insn_3312, _3310, _3313, Claudio Fontana, 2014/04/11
[Qemu-devel] [PATCH v3 25/26] tcg-aarch64: Prefer unsigned offsets before signed offsets for ldst, Richard Henderson, 2014/04/03
[Qemu-devel] [PATCH v3 26/26] tcg-aarch64: Use tcg_out_mov in preference to tcg_out_movr, Richard Henderson, 2014/04/03