Implement vset* and vector CSRs

spec/std/isa/csr/V/vcsr.yaml

@@ -0,0 +1,30 @@
+# Copyright (c) Qualcomm Technologies, Inc. and/or its subsidiaries.
+# SPDX-License-Identifier: BSD-3-Clause-Clear
+
+# yaml-language-server: $schema=../../schemas/csr_schema.json
+
+$schema: "csr_schema.json#"
+kind: csr
+name: vcsr
+long_name: Vector Control and Status Register
+address: 0x00F
+writable: true
+priv_mode: U
+length: MXLEN
+description: Allows access to vxrm and vxsat CSRs
+definedBy: V
+fields:
+  VXRM:
+    location: 2-1
+    description: See vxrm.
+    type: RW-RH
+    sw_write(csr_value): |
+      return csr_value.VXRM;
+    reset_value: UNDEFINED_LEGAL
+  VXSAT:
+    location: 0
+    description: See vxsat.
+    type: RW-RH
+    sw_write(csr_value): |
+      return csr_value.VXSAT;
+    reset_value: UNDEFINED_LEGAL

spec/std/isa/csr/V/vl.yaml

@@ -0,0 +1,30 @@
+# Copyright (c) Qualcomm Technologies, Inc. and/or its subsidiaries.
+# SPDX-License-Identifier: BSD-3-Clause-Clear
+
+# yaml-language-server: $schema=../../schemas/csr_schema.json
+
+$schema: "csr_schema.json#"
+kind: csr
+name: vl
+long_name: Vector Length
+address: 0xC20
+writable: false
+priv_mode: U
+length: MXLEN
+description: Holds an unsigned integer specifying number of elements to be updated with results from a vector instruction.
+definedBy: V
+fields:
+  VALUE:
+    location_rv32: 31-0
+    location_rv64: 63-0
+    description: |
+      The vl register holds an unsigned integer specifying the number of elements to be updated with
+      results from a vector instruction, as further detailed in Section Section 31.5.4.
+
+      [NOTE]
+      The number of bits implemented in vl depends on the implementation's maximum vector
+      length of the smallest supported type. The smallest vector implementation with VLEN=32
+      and supporting SEW=8 would need at least six bits in vl to hold the values 0-32
+      (VLEN=32, with LMUL=8 and SEW=8, yields VLMAX=32).
+    type: RO-H
+    reset_value: 0

spec/std/isa/csr/V/vlenb.yaml

@@ -0,0 +1,25 @@
+# Copyright (c) Qualcomm Technologies, Inc. and/or its subsidiaries.
+# SPDX-License-Identifier: BSD-3-Clause-Clear
+
+# yaml-language-server: $schema=../../schemas/csr_schema.json
+
+$schema: "csr_schema.json#"
+kind: csr
+name: vlenb
+long_name: Vector Byte Length
+address: 0xC22
+writable: false
+priv_mode: U
+length: MXLEN
+description: Holds the value VLEN/8, the vector register length in bytes.
+definedBy: V
+fields:
+  VALUE:
+    location_rv32: 31-0
+    location_rv64: 63-0
+    description: |
+      The value in vlenb is a design-time constant in any implementation.
+      Without this CSR, several instructions are needed to calculate VLEN in bytes, and the code
+      has to disturb current vl and vtype settings which require them to be saved and restored.
+    type: RO
+    reset_value(): return VLEN / 8;

spec/std/isa/csr/V/vstart.yaml

@@ -0,0 +1,74 @@
+# Copyright (c) Qualcomm Technologies, Inc. and/or its subsidiaries.
+# SPDX-License-Identifier: BSD-3-Clause-Clear
+
+# yaml-language-server: $schema=../../schemas/csr_schema.json
+
+$schema: "csr_schema.json#"
+kind: csr
+name: vstart
+long_name: Vector Start Index
+address: 0x008
+writable: true
+priv_mode: U
+length: MXLEN
+description: Specifies the index of the first element to be executed by a vector instruction.
+definedBy: V
+fields:
+  VALUE:
+    location_rv32: 31-0
+    location_rv64: 63-0
+    description: |
+      Normally, vstart is only written by hardware on a trap on a vector instruction, with the vstart value
+      representing the element on which the trap was taken (either a synchronous exception or an
+      asynchronous interrupt), and at which execution should resume after a resumable trap is handled.
+      All vector instructions are defined to begin execution with the element number given in the vstart
+      CSR, leaving earlier elements in the destination vector undisturbed, and to reset the vstart CSR to
+      zero at the end of execution.
+
+      [NOTE]
+      All vector instructions, including vset{i}vl{i}, reset the vstart CSR to zero.
+      vstart is not modified by vector instructions that raise illegal-instruction exceptions.
+      The vstart CSR is defined to have only enough writable bits to hold the largest element index (one
+      less than the maximum VLMAX).
+
+      [NOTE]
+      The maximum vector length is obtained with the largest LMUL setting (8) and the smallest
+      SEW setting (8), so VLMAX_max = 8*VLEN/8 = VLEN. For example, for VLEN=256,
+      vstart would have 8 bits to represent indices from 0 through 255.
+
+      The use of vstart values greater than the largest element index for the current vtype setting is
+      reserved.
+
+      [NOTE]
+      It is recommended that implementations trap if vstart is out of bounds. It is not required
+      to trap, as a possible future use of upper vstart bits is to store imprecise trap
+      information.
+
+      The vstart CSR is writable by unprivileged code, but non-zero vstart values may cause vector
+      instructions to run substantially slower on some implementations, so vstart should not be used by
+      application programmers. A few vector instructions cannot be executed with a non-zero vstart value
+      and will raise an illegal instruction exception as defined below.
+
+      [NOTE]
+      Making vstart visible to unprivileged code supports user-level threading libraries.
+
+      Implementations are permitted to raise illegal instruction exceptions when attempting to execute a
+      vector instruction with a value of vstart that the implementation can never produce when executing
+      that same instruction with the same vtype setting.
+
+      [NOTE]
+      For example, some implementations will never take interrupts during execution of a vector
+      arithmetic instruction, instead waiting until the instruction completes to take the
+      interrupt. Such implementations are permitted to raise an illegal instruction exception
+      when attempting to execute a vector arithmetic instruction when vstart is nonzero.
+
+      [NOTE]
+      When migrating a software thread between two harts with different microarchitectures,
+      the vstart value might not be supported by the new hart microarchitecture. The runtime
+      on the receiving hart might then have to emulate instruction execution up to the next
+      supported vstart element position. Alternatively, migration events can be constrained to
+      only occur at mutually supported vstart locations.
+    sw_write(csr_value): |
+      return csr_value.VALUE & (VLEN - 1);
+    type: RW-RH
+    reset_value: UNDEFINED_LEGAL

spec/std/isa/csr/V/vtype.yaml

@@ -0,0 +1,118 @@
+# Copyright (c) Qualcomm Technologies, Inc. and/or its subsidiaries.
+# SPDX-License-Identifier: BSD-3-Clause-Clear
+
+# yaml-language-server: $schema=../../schemas/csr_schema.json
+
+$schema: "csr_schema.json#"
+kind: csr
+name: vtype
+long_name: Vector Type
+address: 0xC21
+writable: false
+priv_mode: U
+length: MXLEN
+description: Provides the default type used to interpret the contents of the vector register file.
+definedBy: V
+fields:
+  VILL:
+    location_rv32: 31
+    location_rv64: 63
+    description: |
+      The vill bit is used to encode that a previous vset{i}vl{i} instruction attempted to write an
+      unsupported value to vtype.
+
+      [NOTE]
+      The vill bit is held in bit XLEN-1 of the CSR to support checking for illegal values with a
+      branch on the sign bit.
+
+      If the vill bit is set, then any attempt to execute a vector instruction that depends upon vtype will
+      raise an illegal-instruction exception.
+
+      When the vill bit is set, the other XLEN-1 bits in vtype shall be zero.
+
+      It is recommended that at reset, vill is set.
+    type: RO-H
+    reset_value: 1
+  VMA:
+    location: 7
+    description: |
+      Vector mask agnostic bit. Modifies the behavior of destination inactive masked-off elements during the
+      execution of vector instructions.
+
+      A value of 0 means inactive elements are undisturbed, meaning the corresponding set of destination elements
+      in a vector register group retain the value they previously held.
+
+      A value of 1 means inactive elements are agnostic, meaning the corresponding set of destination elements
+      in any vector destination operand can either retain the value they previously held, or are overwritten with 1s.
+      Within a single vector instruction, each destination element can be either left undisturbed or overwritten
+      with 1s, in any combination, and the pattern of undisturbed or overwritten with 1s is not required to be
+      deterministic when the instruction is executed with the same inputs.
+
+      It is recommended that at reset, vill is set, and the remaining bits in vtype are zero.
+    type: RO-H
+    reset_value: 0
+  VTA:
+    location: 6
+    description: |
+      Vector tail agnostic bit. Modifies the bahavior of destination tail elements during the execution of vector
+      instructions.
+
+      A value of 0 means tail elements are undisturbed, meaning the corresponding set of destination elements
+      in a vector register group retain the value they previously held.
+
+      A value of 1 means tail elements are agnostic, meaning the corresponding set of destination elements
+      in any vector destination operand can either retain the value they previously held, or are overwritten with 1s.
+      Within a single vector instruction, each destination element can be either left undisturbed or overwritten
+      with 1s, in any combination, and the pattern of undisturbed or overwritten with 1s is not required to be
+      deterministic when the instruction is executed with the same inputs.
+
+      It is recommended that at reset, vill is set, and the remaining bits in vtype are zero.
+    type: RO-H
+    reset_value: 0
+  VSEW:
+    location: 5-3
+    description: |
+      The value in vsew sets the dynamic selected element width (SEW).
+
+      [separator="!"]
+      !===
+      ! vsew[2:0] ! SEW ! Elements per vector register
+      ! 000 ! 8 ! 16
+      ! 001 ! 16 ! 8
+      ! 010 ! 32 ! 4
+      ! 011 ! 64 ! 2
+      ! 1XX ! Reserved ! Reserved
+      !===
+
+      It is recommended that at reset, vill is set, and the remaining bits in vtype are zero.
+    type: RO-H
+    reset_value: 0
+  VLMUL:
+    location: 2-0
+    description: |
+      Vector register group multiplier.
+
+      Multiple vector registers can be grouped together, so that a single vector instruction can operate on
+      multiple vector registers. The term vector register group is used herein to refer to one or more vector
+      registers used as a single operand to a vector instruction. Vector register groups can be used to provide
+      greater execution efficiency for longer application vectors, but the main reason for their inclusion is to
+      allow double-width or larger elements to be operated on with the same vector length as single-width
+      elements. The vector length multiplier, LMUL, when greater than 1, represents the default number of
+      vector registers that are combined to form a vector register group. Implementations must support
+      LMUL integer values of 1, 2, 4, and 8.
+
+      [NOTE]
+      The vector architecture includes instructions that take multiple source and destination
+      vector operands with different element widths, but the same number of elements. The
+      effective LMUL (EMUL) of each vector operand is determined by the number of registers
+      required to hold the elements. For example, for a widening add operation, such as add 32-
+      bit values to produce 64-bit results, a double-width result requires twice the LMUL of the
+      single-width inputs.
+
+      LMUL can also be a fractional value, reducing the number of bits used in a single vector register.
+      Fractional LMUL is used to increase the number of effective usable vector register groups when
+      operating on mixed-width values.
+
+      It is recommended that at reset, vill is set, and the remaining bits in vtype are zero.
+    type: RO-H
+    reset_value: 0

spec/std/isa/csr/V/vxrm.yaml

@@ -0,0 +1,47 @@
+# Copyright (c) Qualcomm Technologies, Inc. and/or its subsidiaries.
+# SPDX-License-Identifier: BSD-3-Clause-Clear
+
+# yaml-language-server: $schema=../../schemas/csr_schema.json
+
+$schema: "csr_schema.json#"
+kind: csr
+name: vxrm
+long_name: Vector Fixed-Point Rounding Mode
+address: 0x00A
+writable: true
+priv_mode: U
+length: MXLEN
+description: Holds a 2-bit read-write rounding-mode field in the least-significant bits
+definedBy: V
+sw_read(): |
+  return CSR[vcsr].VXRM;
+fields:
+  VALUE:
+    alias: vcsr.VXRM
+    location_rv32: 31-0
+    location_rv64: 63-0
+    description: |
+      The vector fixed-point rounding-mode register holds a two-bit read-write rounding-mode field in the
+      least-significant bits (vxrm[1:0]). The upper bits, vxrm[XLEN-1:2], should be written as zeros.
+      The vector fixed-point rounding-mode is given a separate CSR address to allow independent access,
+      but is also reflected as a field in vcsr.
+
+      [NOTE]
+      A new rounding mode can be set while saving the original rounding mode using a single csrwi instruction.
+
+      The fixed-point rounding algorithm is specified as follows. Suppose the pre-rounding result is v, and d
+      bits of that result are to be rounded off. Then the rounded result is (v >> d) + r, where r depends on
+      the rounding mode as specified in the following table of vxrm[1:0] values.
+
+      [separator="!"]
+      !===
+      ! vxrm[1:0] ! Abbreviation ! Rounding Mode ! Rounding increment, r
+      ! 00 ! rnu ! round-to-nearest-up (add +0.5 LSB) ! v[d-1]
+      ! 01 ! rne ! round-to-nearest-even ! v[d-1] & (v[d-2:0]\!=0 | v[d])
+      ! 10 ! rdn ! round-down (truncate) ! 0
+      ! 11 ! rod ! round-to-odd (OR bits into LSB, aka "jam") ! \!v[d] & v[d-1:0]\!=0
+
+    sw_write(csr_value): |
+      return csr_value.VALUE & (VLEN - 1);
+    type: RW-H
+    reset_value: UNDEFINED_LEGAL

spec/std/isa/csr/V/vxsat.yaml

@@ -0,0 +1,31 @@
+# Copyright (c) Qualcomm Technologies, Inc. and/or its subsidiaries.
+# SPDX-License-Identifier: BSD-3-Clause-Clear
+
+# yaml-language-server: $schema=../../schemas/csr_schema.json
+
+$schema: "csr_schema.json#"
+kind: csr
+name: vxsat
+long_name: Vector Fixed-Point Saturate Flag
+address: 0x009
+writable: true
+priv_mode: U
+length: MXLEN
+description: Indicates if a fixed-point instruction has had to saturate an output value to fit into a destination format
+definedBy: V
+sw_read(): |
+  return CSR[vcsr].VXSAT;
+fields:
+  VALUE:
+    location_rv32: 31-0
+    location_rv64: 63-0
+    description: |
+      The vxsat CSR has a single read-write least-significant bit (vxsat[0]) that indicates if a fixed-point
+      instruction has had to saturate an output value to fit into a destination format. Bits vxsat[XLEN-1:1]
+      should be written as zeros.
+
+      The vxsat bit is mirrored in vcsr.
+    sw_write(csr_value): |
+      return csr_value.VALUE & (VLEN - 1);
+    type: RW-H
+    reset_value: UNDEFINED_LEGAL

spec/std/isa/ext/V.yaml

@@ -50,3 +50,19 @@ params:
 
       # if HW is writing VS, then Dirty (3) better be a supported value
       assert MSTATUS_VS_LEGAL_VALUES.include?(3) if ext?(:V) && (HW_MSTATUS_VS_DIRTY_UPDATE != "never")
+  RVV_VL_WHEN_AVL_LT_DOUBLE_VLMAX:
+    description: |
+      The value assigned to VL when AVL < 2*VLMAX.
+    schema:
+      type: string
+      enum: ["ceil(AVL/2)", "VLMAX", "custom"]
+  VLEN:
+    description: |
+      The number of bits in a single vector register. VLEN >= ELEN, which must be a power of 2, and must be no greater than 2^16.
+    schema:
+      type: integer
+  ELEN:
+    description: |
+      The maximum size in bits of a vector element that any operation can produce or consume, ELEN >= 8, which must be a power of 2.
+    schema:
+      type: integer